/[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 994 by ph10, Tue Jul 10 14:29:26 2012 UTC revision 1394 by ph10, Sat Nov 9 09:17:20 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 492  static const char error_texts[] = Line 525  static const char error_texts[] =
525    /* 75 */    /* 75 */
526    "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"    "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"
527    "character value in \\u.... sequence is too large\0"    "character value in \\u.... sequence is too large\0"
528      "invalid UTF-32 string\0"
529      "setting UTF is disabled by the application\0"
530      "non-hex character in \\x{} (closing brace missing?)\0"
531      /* 80 */
532      "non-octal character in \\o{} (closing brace missing?)\0"
533      "missing opening brace after \\o\0"
534      "parentheses are too deeply nested\0"
535      "invalid range in character class\0"
536      "group name must start with a non-digit\0"
537    ;    ;
538    
539  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
# Line 631  static const pcre_uint8 ebcdic_chartab[] Line 673  static const pcre_uint8 ebcdic_chartab[]
673  #endif  #endif
674    
675    
676  /* Definition to allow mutual recursion */  /* This table is used to check whether auto-possessification is possible
677    between adjacent character-type opcodes. The left-hand (repeated) opcode is
678    used to select the row, and the right-hand opcode is use to select the column.
679    A value of 1 means that auto-possessification is OK. For example, the second
680    value in the first row means that \D+\d can be turned into \D++\d.
681    
682    The Unicode property types (\P and \p) have to be present to fill out the table
683    because of what their opcode values are, but the table values should always be
684    zero because property types are handled separately in the code. The last four
685    columns apply to items that cannot be repeated, so there is no need to have
686    rows for them. Note that OP_DIGIT etc. are generated only when PCRE_UCP is
687    *not* set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
688    
689    #define APTROWS (LAST_AUTOTAB_LEFT_OP - FIRST_AUTOTAB_OP + 1)
690    #define APTCOLS (LAST_AUTOTAB_RIGHT_OP - FIRST_AUTOTAB_OP + 1)
691    
692    static const pcre_uint8 autoposstab[APTROWS][APTCOLS] = {
693    /* \D \d \S \s \W \w  . .+ \C \P \p \R \H \h \V \v \X \Z \z  $ $M */
694      { 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \D */
695      { 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \d */
696      { 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \S */
697      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \s */
698      { 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \W */
699      { 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \w */
700      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .  */
701      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .+ */
702      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \C */
703      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \P */
704      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \p */
705      { 0, 1, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \R */
706      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \H */
707      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \h */
708      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \V */
709      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0 },  /* \v */
710      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }   /* \X */
711    };
712    
713  static BOOL  
714    compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL, int, int,  /* This table is used to check whether auto-possessification is possible
715      int *, int *, branch_chain *, compile_data *, int *);  between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP). The
716    left-hand (repeated) opcode is used to select the row, and the right-hand
717    opcode is used to select the column. The values are as follows:
718    
719      0   Always return FALSE (never auto-possessify)
720      1   Character groups are distinct (possessify if both are OP_PROP)
721      2   Check character categories in the same group (general or particular)
722      3   TRUE if the two opcodes are not the same (PROP vs NOTPROP)
723    
724      4   Check left general category vs right particular category
725      5   Check right general category vs left particular category
726    
727      6   Left alphanum vs right general category
728      7   Left space vs right general category
729      8   Left word vs right general category
730    
731      9   Right alphanum vs left general category
732     10   Right space vs left general category
733     11   Right word vs left general category
734    
735     12   Left alphanum vs right particular category
736     13   Left space vs right particular category
737     14   Left word vs right particular category
738    
739     15   Right alphanum vs left particular category
740     16   Right space vs left particular category
741     17   Right word vs left particular category
742    */
743    
744    static const pcre_uint8 propposstab[PT_TABSIZE][PT_TABSIZE] = {
745    /* ANY LAMP GC  PC  SC ALNUM SPACE PXSPACE WORD CLIST UCNC */
746      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_ANY */
747      { 0,  3,  0,  0,  0,    3,    1,      1,   0,    0,   0 },  /* PT_LAMP */
748      { 0,  0,  2,  4,  0,    9,   10,     10,  11,    0,   0 },  /* PT_GC */
749      { 0,  0,  5,  2,  0,   15,   16,     16,  17,    0,   0 },  /* PT_PC */
750      { 0,  0,  0,  0,  2,    0,    0,      0,   0,    0,   0 },  /* PT_SC */
751      { 0,  3,  6, 12,  0,    3,    1,      1,   0,    0,   0 },  /* PT_ALNUM */
752      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_SPACE */
753      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_PXSPACE */
754      { 0,  0,  8, 14,  0,    0,    1,      1,   3,    0,   0 },  /* PT_WORD */
755      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_CLIST */
756      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   3 }   /* PT_UCNC */
757    };
758    
759    /* This table is used to check whether auto-possessification is possible
760    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP) when one
761    specifies a general category and the other specifies a particular category. The
762    row is selected by the general category and the column by the particular
763    category. The value is 1 if the particular category is not part of the general
764    category. */
765    
766    static const pcre_uint8 catposstab[7][30] = {
767    /* Cc Cf Cn Co Cs Ll Lm Lo Lt Lu Mc Me Mn Nd Nl No Pc Pd Pe Pf Pi Po Ps Sc Sk Sm So Zl Zp Zs */
768      { 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* C */
769      { 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* L */
770      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* M */
771      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* N */
772      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1 },  /* P */
773      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 1, 1, 1 },  /* S */
774      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0 }   /* Z */
775    };
776    
777    /* This table is used when checking ALNUM, (PX)SPACE, SPACE, and WORD against
778    a general or particular category. The properties in each row are those
779    that apply to the character set in question. Duplication means that a little
780    unnecessary work is done when checking, but this keeps things much simpler
781    because they can all use the same code. For more details see the comment where
782    this table is used.
783    
784    Note: SPACE and PXSPACE used to be different because Perl excluded VT from
785    "space", but from Perl 5.18 it's included, so both categories are treated the
786    same here. */
787    
788    static const pcre_uint8 posspropstab[3][4] = {
789      { ucp_L, ucp_N, ucp_N, ucp_Nl },  /* ALNUM, 3rd and 4th values redundant */
790      { ucp_Z, ucp_Z, ucp_C, ucp_Cc },  /* SPACE and PXSPACE, 2nd value redundant */
791      { ucp_L, ucp_N, ucp_P, ucp_Po }   /* WORD */
792    };
793    
794    /* This table is used when converting repeating opcodes into possessified
795    versions as a result of an explicit possessive quantifier such as ++. A zero
796    value means there is no possessified version - in those cases the item in
797    question must be wrapped in ONCE brackets. The table is truncated at OP_CALLOUT
798    because all relevant opcodes are less than that. */
799    
800    static const pcre_uint8 opcode_possessify[] = {
801      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 0 - 15  */
802      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 16 - 31 */
803    
804      0,                       /* NOTI */
805      OP_POSSTAR, 0,           /* STAR, MINSTAR */
806      OP_POSPLUS, 0,           /* PLUS, MINPLUS */
807      OP_POSQUERY, 0,          /* QUERY, MINQUERY */
808      OP_POSUPTO, 0,           /* UPTO, MINUPTO */
809      0,                       /* EXACT */
810      0, 0, 0, 0,              /* POS{STAR,PLUS,QUERY,UPTO} */
811    
812      OP_POSSTARI, 0,          /* STARI, MINSTARI */
813      OP_POSPLUSI, 0,          /* PLUSI, MINPLUSI */
814      OP_POSQUERYI, 0,         /* QUERYI, MINQUERYI */
815      OP_POSUPTOI, 0,          /* UPTOI, MINUPTOI */
816      0,                       /* EXACTI */
817      0, 0, 0, 0,              /* POS{STARI,PLUSI,QUERYI,UPTOI} */
818    
819      OP_NOTPOSSTAR, 0,        /* NOTSTAR, NOTMINSTAR */
820      OP_NOTPOSPLUS, 0,        /* NOTPLUS, NOTMINPLUS */
821      OP_NOTPOSQUERY, 0,       /* NOTQUERY, NOTMINQUERY */
822      OP_NOTPOSUPTO, 0,        /* NOTUPTO, NOTMINUPTO */
823      0,                       /* NOTEXACT */
824      0, 0, 0, 0,              /* NOTPOS{STAR,PLUS,QUERY,UPTO} */
825    
826      OP_NOTPOSSTARI, 0,       /* NOTSTARI, NOTMINSTARI */
827      OP_NOTPOSPLUSI, 0,       /* NOTPLUSI, NOTMINPLUSI */
828      OP_NOTPOSQUERYI, 0,      /* NOTQUERYI, NOTMINQUERYI */
829      OP_NOTPOSUPTOI, 0,       /* NOTUPTOI, NOTMINUPTOI */
830      0,                       /* NOTEXACTI */
831      0, 0, 0, 0,              /* NOTPOS{STARI,PLUSI,QUERYI,UPTOI} */
832    
833      OP_TYPEPOSSTAR, 0,       /* TYPESTAR, TYPEMINSTAR */
834      OP_TYPEPOSPLUS, 0,       /* TYPEPLUS, TYPEMINPLUS */
835      OP_TYPEPOSQUERY, 0,      /* TYPEQUERY, TYPEMINQUERY */
836      OP_TYPEPOSUPTO, 0,       /* TYPEUPTO, TYPEMINUPTO */
837      0,                       /* TYPEEXACT */
838      0, 0, 0, 0,              /* TYPEPOS{STAR,PLUS,QUERY,UPTO} */
839    
840      OP_CRPOSSTAR, 0,         /* CRSTAR, CRMINSTAR */
841      OP_CRPOSPLUS, 0,         /* CRPLUS, CRMINPLUS */
842      OP_CRPOSQUERY, 0,        /* CRQUERY, CRMINQUERY */
843      OP_CRPOSRANGE, 0,        /* CRRANGE, CRMINRANGE */
844      0, 0, 0, 0,              /* CRPOS{STAR,PLUS,QUERY,RANGE} */
845    
846      0, 0, 0,                 /* CLASS, NCLASS, XCLASS */
847      0, 0,                    /* REF, REFI */
848      0, 0,                    /* DNREF, DNREFI */
849      0, 0                     /* RECURSE, CALLOUT */
850    };
851    
852    
853    
# Line 658  find_error_text(int n) Line 870  find_error_text(int n)
870  const char *s = error_texts;  const char *s = error_texts;
871  for (; n > 0; n--)  for (; n > 0; n--)
872    {    {
873    while (*s++ != 0) {};    while (*s++ != CHAR_NULL) {};
874    if (*s == 0) return "Error text not found (please report)";    if (*s == CHAR_NULL) return "Error text not found (please report)";
875    }    }
876  return s;  return s;
877  }  }
878    
879    
880    
881  /*************************************************  /*************************************************
882  *           Expand the workspace                 *  *           Expand the workspace                 *
883  *************************************************/  *************************************************/
# Line 742  return (*p == CHAR_RIGHT_CURLY_BRACKET); Line 955  return (*p == CHAR_RIGHT_CURLY_BRACKET);
955  *************************************************/  *************************************************/
956    
957  /* This function is called when a \ has been encountered. It either returns a  /* This function is called when a \ has been encountered. It either returns a
958  positive value for a simple escape such as \n, or a negative value which  positive value for a simple escape such as \n, or 0 for a data character which
959  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.
960  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
961  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
962  ptr is pointing at the \. On exit, it is on the final character of the escape  character of the escape sequence.
 sequence.  
963    
964  Arguments:  Arguments:
965    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
966      chptr          points to a returned data character
967    errorcodeptr   points to the errorcode variable    errorcodeptr   points to the errorcode variable
968    bracount       number of previous extracting brackets    bracount       number of previous extracting brackets
969    options        the options bits    options        the options bits
970    isclass        TRUE if inside a character class    isclass        TRUE if inside a character class
971    
972  Returns:         zero or positive => a data character  Returns:         zero => a data character
973                   negative => a special escape sequence                   positive => a special escape sequence
974                     negative => a back reference
975                   on error, errorcodeptr is set                   on error, errorcodeptr is set
976  */  */
977    
978  static int  static int
979  check_escape(const pcre_uchar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,
980    int options, BOOL isclass)    int bracount, int options, BOOL isclass)
981  {  {
982  /* PCRE_UTF16 has the same value as PCRE_UTF8. */  /* PCRE_UTF16 has the same value as PCRE_UTF8. */
983  BOOL utf = (options & PCRE_UTF8) != 0;  BOOL utf = (options & PCRE_UTF8) != 0;
984  const pcre_uchar *ptr = *ptrptr + 1;  const pcre_uchar *ptr = *ptrptr + 1;
985  pcre_int32 c;  pcre_uint32 c;
986    int escape = 0;
987  int i;  int i;
988    
989  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
# Line 776  ptr--;                            /* Set Line 991  ptr--;                            /* Set
991    
992  /* If backslash is at the end of the pattern, it's an error. */  /* If backslash is at the end of the pattern, it's an error. */
993    
994  if (c == 0) *errorcodeptr = ERR1;  if (c == CHAR_NULL) *errorcodeptr = ERR1;
995    
996  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
997  in a table. A non-zero result is something that can be returned immediately.  in a table. A non-zero result is something that can be returned immediately.
# Line 785  Otherwise further processing may be requ Line 1000  Otherwise further processing may be requ
1000  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1001  /* Not alphanumeric */  /* Not alphanumeric */
1002  else if (c < CHAR_0 || c > CHAR_z) {}  else if (c < CHAR_0 || c > CHAR_z) {}
1003  else if ((i = escapes[c - CHAR_0]) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0)
1004      { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1005    
1006  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1007  /* Not alphanumeric */  /* Not alphanumeric */
1008  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)) {}
1009  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; }
1010  #endif  #endif
1011    
1012  /* Escapes that need further processing, or are illegal. */  /* Escapes that need further processing, or are illegal. */
# Line 798  else if ((i = escapes[c - 0x48]) != 0) Line 1014  else if ((i = escapes[c - 0x48]) != 0)
1014  else  else
1015    {    {
1016    const pcre_uchar *oldptr;    const pcre_uchar *oldptr;
1017    BOOL braced, negated;    BOOL braced, negated, overflow;
1018      int s;
1019    
1020    switch (c)    switch (c)
1021      {      {
# Line 823  else Line 1040  else
1040          c = 0;          c = 0;
1041          for (i = 0; i < 4; ++i)          for (i = 0; i < 4; ++i)
1042            {            {
1043            register int cc = *(++ptr);            register pcre_uint32 cc = *(++ptr);
1044  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1045            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1046            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 833  else Line 1050  else
1050  #endif  #endif
1051            }            }
1052    
1053  #ifdef COMPILE_PCRE8  #if defined COMPILE_PCRE8
1054          if (c > (utf ? 0x10ffff : 0xff))          if (c > (utf ? 0x10ffffU : 0xffU))
1055  #else  #elif defined COMPILE_PCRE16
1056  #ifdef COMPILE_PCRE16          if (c > (utf ? 0x10ffffU : 0xffffU))
1057          if (c > (utf ? 0x10ffff : 0xffff))  #elif defined COMPILE_PCRE32
1058  #endif          if (utf && c > 0x10ffffU)
1059  #endif  #endif
1060            {            {
1061            *errorcodeptr = ERR76;            *errorcodeptr = ERR76;
# Line 870  else Line 1087  else
1087      (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
1088      number either in angle brackets or in single quotes. However, these are      number either in angle brackets or in single quotes. However, these are
1089      (possibly recursive) subroutine calls, _not_ backreferences. Just return      (possibly recursive) subroutine calls, _not_ backreferences. Just return
1090      the -ESC_g code (cf \k). */      the ESC_g code (cf \k). */
1091    
1092      case CHAR_g:      case CHAR_g:
1093      if (isclass) break;      if (isclass) break;
1094      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
1095        {        {
1096        c = -ESC_g;        escape = ESC_g;
1097        break;        break;
1098        }        }
1099    
# Line 885  else Line 1102  else
1102      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1103        {        {
1104        const pcre_uchar *p;        const pcre_uchar *p;
1105        for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)        for (p = ptr+2; *p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
1106          if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;          if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
1107        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)        if (*p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET)
1108          {          {
1109          c = -ESC_k;          escape = ESC_k;
1110          break;          break;
1111          }          }
1112        braced = TRUE;        braced = TRUE;
# Line 905  else Line 1122  else
1122      else negated = FALSE;      else negated = FALSE;
1123    
1124      /* The integer range is limited by the machine's int representation. */      /* The integer range is limited by the machine's int representation. */
1125      c = 0;      s = 0;
1126        overflow = FALSE;
1127      while (IS_DIGIT(ptr[1]))      while (IS_DIGIT(ptr[1]))
1128        {        {
1129        if (((unsigned int)c) > INT_MAX / 10) /* Integer overflow */        if (s > INT_MAX / 10 - 1) /* Integer overflow */
1130          {          {
1131          c = -1;          overflow = TRUE;
1132          break;          break;
1133          }          }
1134        c = c * 10 + *(++ptr) - CHAR_0;        s = s * 10 + (int)(*(++ptr) - CHAR_0);
1135        }        }
1136      if (((unsigned int)c) > INT_MAX) /* Integer overflow */      if (overflow) /* Integer overflow */
1137        {        {
1138        while (IS_DIGIT(ptr[1]))        while (IS_DIGIT(ptr[1]))
1139          ptr++;          ptr++;
# Line 929  else Line 1147  else
1147        break;        break;
1148        }        }
1149    
1150      if (c == 0)      if (s == 0)
1151        {        {
1152        *errorcodeptr = ERR58;        *errorcodeptr = ERR58;
1153        break;        break;
# Line 937  else Line 1155  else
1155    
1156      if (negated)      if (negated)
1157        {        {
1158        if (c > bracount)        if (s > bracount)
1159          {          {
1160          *errorcodeptr = ERR15;          *errorcodeptr = ERR15;
1161          break;          break;
1162          }          }
1163        c = bracount - (c - 1);        s = bracount - (s - 1);
1164        }        }
1165    
1166      c = -(ESC_REF + c);      escape = -s;
1167      break;      break;
1168    
1169      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
1170      starting with one that is not zero is not straightforward. By experiment,      starting with one that is not zero is not straightforward. Perl has changed
1171      the way Perl works seems to be as follows:      over the years. Nowadays \g{} for backreferences and \o{} for octal are
1172        recommended to avoid the ambiguities in the old syntax.
1173    
1174      Outside a character class, the digits are read as a decimal number. If the      Outside a character class, the digits are read as a decimal number. If the
1175      number is less than 10, or if there are that many previous extracting      number is less than 8 (used to be 10), or if there are that many previous
1176      left brackets, then it is a back reference. Otherwise, up to three octal      extracting left brackets, then it is a back reference. Otherwise, up to
1177      digits are read to form an escaped byte. Thus \123 is likely to be octal      three octal digits are read to form an escaped byte. Thus \123 is likely to
1178      123 (cf \0123, which is octal 012 followed by the literal 3). If the octal      be octal 123 (cf \0123, which is octal 012 followed by the literal 3). If
1179      value is greater than 377, the least significant 8 bits are taken. Inside a      the octal value is greater than 377, the least significant 8 bits are
1180      character class, \ followed by a digit is always an octal number. */      taken. \8 and \9 are treated as the literal characters 8 and 9.
1181    
1182        Inside a character class, \ followed by a digit is always either a literal
1183        8 or 9 or an octal number. */
1184    
1185      case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:      case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:
1186      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
# Line 967  else Line 1189  else
1189        {        {
1190        oldptr = ptr;        oldptr = ptr;
1191        /* The integer range is limited by the machine's int representation. */        /* The integer range is limited by the machine's int representation. */
1192        c -= CHAR_0;        s = (int)(c -CHAR_0);
1193          overflow = FALSE;
1194        while (IS_DIGIT(ptr[1]))        while (IS_DIGIT(ptr[1]))
1195          {          {
1196          if (((unsigned int)c) > INT_MAX / 10) /* Integer overflow */          if (s > INT_MAX / 10 - 1) /* Integer overflow */
1197            {            {
1198            c = -1;            overflow = TRUE;
1199            break;            break;
1200            }            }
1201          c = c * 10 + *(++ptr) - CHAR_0;          s = s * 10 + (int)(*(++ptr) - CHAR_0);
1202          }          }
1203        if (((unsigned int)c) > INT_MAX) /* Integer overflow */        if (overflow) /* Integer overflow */
1204          {          {
1205          while (IS_DIGIT(ptr[1]))          while (IS_DIGIT(ptr[1]))
1206            ptr++;            ptr++;
1207          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
1208          break;          break;
1209          }          }
1210        if (c < 10 || c <= bracount)        if (s < 8 || s <= bracount)  /* Check for back reference */
1211          {          {
1212          c = -(ESC_REF + c);          escape = -s;
1213          break;          break;
1214          }          }
1215        ptr = oldptr;      /* Put the pointer back and fall through */        ptr = oldptr;      /* Put the pointer back and fall through */
1216        }        }
1217    
1218      /* Handle an octal number following \. If the first digit is 8 or 9, Perl      /* Handle a digit following \ when the number is not a back reference. If
1219      generates a binary zero byte and treats the digit as a following literal.      the first digit is 8 or 9, Perl used to generate a binary zero byte and
1220      Thus we have to pull back the pointer by one. */      then treat the digit as a following literal. At least by Perl 5.18 this
1221        changed so as not to insert the binary zero. */
1222    
1223      if ((c = *ptr) >= CHAR_8)      if ((c = *ptr) >= CHAR_8) break;
1224        {  
1225        ptr--;      /* Fall through with a digit less than 8 */
       c = 0;  
       break;  
       }  
1226    
1227      /* \0 always starts an octal number, but we may drop through to here with a      /* \0 always starts an octal number, but we may drop through to here with a
1228      larger first octal digit. The original code used just to take the least      larger first octal digit. The original code used just to take the least
# Line 1018  else Line 1239  else
1239  #endif  #endif
1240      break;      break;
1241    
1242      /* \x is complicated. \x{ddd} is a character number which can be greater      /* \o is a relatively new Perl feature, supporting a more general way of
1243      than 0xff in utf or non-8bit mode, but only if the ddd are hex digits.      specifying character codes in octal. The only supported form is \o{ddd}. */
1244      If not, { is treated as a data character. */  
1245        case CHAR_o:
1246        if (ptr[1] != CHAR_LEFT_CURLY_BRACKET) *errorcodeptr = ERR81; else
1247          {
1248          ptr += 2;
1249          c = 0;
1250          overflow = FALSE;
1251          while (*ptr >= CHAR_0 && *ptr <= CHAR_7)
1252            {
1253            register pcre_uint32 cc = *ptr++;
1254            if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1255    #ifdef COMPILE_PCRE32
1256            if (c >= 0x20000000l) { overflow = TRUE; break; }
1257    #endif
1258            c = (c << 3) + cc - CHAR_0 ;
1259    #if defined COMPILE_PCRE8
1260            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1261    #elif defined COMPILE_PCRE16
1262            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1263    #elif defined COMPILE_PCRE32
1264            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1265    #endif
1266            }
1267          if (overflow)
1268            {
1269            while (*ptr >= CHAR_0 && *ptr <= CHAR_7) ptr++;
1270            *errorcodeptr = ERR34;
1271            }
1272          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1273            {
1274            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1275            }
1276          else *errorcodeptr = ERR80;
1277          }
1278        break;
1279    
1280        /* \x is complicated. In JavaScript, \x must be followed by two hexadecimal
1281        numbers. Otherwise it is a lowercase x letter. */
1282    
1283      case CHAR_x:      case CHAR_x:
1284      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1285        {        {
       /* In JavaScript, \x must be followed by two hexadecimal numbers.  
       Otherwise it is a lowercase x letter. */  
1286        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1287          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1288          {          {
1289          c = 0;          c = 0;
1290          for (i = 0; i < 2; ++i)          for (i = 0; i < 2; ++i)
1291            {            {
1292            register int cc = *(++ptr);            register pcre_uint32 cc = *(++ptr);
1293  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1294            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1295            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 1043  else Line 1299  else
1299  #endif  #endif
1300            }            }
1301          }          }
1302        break;        }    /* End JavaScript handling */
       }  
1303    
1304      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      /* Handle \x in Perl's style. \x{ddd} is a character number which can be
1305        {      greater than 0xff in utf or non-8bit mode, but only if the ddd are hex
1306        const pcre_uchar *pt = ptr + 2;      digits. If not, { used to be treated as a data character. However, Perl
1307        seems to read hex digits up to the first non-such, and ignore the rest, so
1308        that, for example \x{zz} matches a binary zero. This seems crazy, so PCRE
1309        now gives an error. */
1310    
1311        c = 0;      else
1312        while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0)        {
1313          if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1314          {          {
1315          register int cc = *pt++;          ptr += 2;
1316          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */          c = 0;
1317            overflow = FALSE;
1318            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0)
1319              {
1320              register pcre_uint32 cc = *ptr++;
1321              if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1322    
1323    #ifdef COMPILE_PCRE32
1324              if (c >= 0x10000000l) { overflow = TRUE; break; }
1325    #endif
1326    
1327  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1328          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1329          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1330  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1331          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */            if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1332          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1333  #endif  #endif
1334    
1335  #ifdef COMPILE_PCRE8  #if defined COMPILE_PCRE8
1336          if (c > (utf ? 0x10ffff : 0xff)) { c = -1; break; }            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1337  #else  #elif defined COMPILE_PCRE16
1338  #ifdef COMPILE_PCRE16            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1339          if (c > (utf ? 0x10ffff : 0xffff)) { c = -1; break; }  #elif defined COMPILE_PCRE32
1340  #endif            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1341  #endif  #endif
1342          }            }
1343    
1344        if (c < 0)          if (overflow)
1345          {            {
1346          while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0) pt++;            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0) ptr++;
1347          *errorcodeptr = ERR34;            *errorcodeptr = ERR34;
1348          }            }
1349    
1350        if (*pt == CHAR_RIGHT_CURLY_BRACKET)          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1351          {            {
1352          if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1353          ptr = pt;            }
         break;  
         }  
1354    
1355        /* If the sequence of hex digits does not end with '}', then we don't          /* If the sequence of hex digits does not end with '}', give an error.
1356        recognize this construct; fall through to the normal \x handling. */          We used just to recognize this construct and fall through to the normal
1357        }          \x handling, but nowadays Perl gives an error, which seems much more
1358            sensible, so we do too. */
1359    
1360      /* Read just a single-byte hex-defined char */          else *errorcodeptr = ERR79;
1361            }   /* End of \x{} processing */
1362    
1363      c = 0;        /* Read a single-byte hex-defined char (up to two hex digits after \x) */
1364      while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)  
1365        {        else
1366        int cc;                                  /* Some compilers don't like */          {
1367        cc = *(++ptr);                           /* ++ in initializers */          c = 0;
1368            while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1369              {
1370              pcre_uint32 cc;                          /* Some compilers don't like */
1371              cc = *(++ptr);                           /* ++ in initializers */
1372  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1373        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1374        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1375  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1376        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */            if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1377        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1378  #endif  #endif
1379        }            }
1380            }     /* End of \xdd handling */
1381          }       /* End of Perl-style \x handling */
1382      break;      break;
1383    
1384      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
# Line 1114  else Line 1388  else
1388    
1389      case CHAR_c:      case CHAR_c:
1390      c = *(++ptr);      c = *(++ptr);
1391      if (c == 0)      if (c == CHAR_NULL)
1392        {        {
1393        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
1394        break;        break;
# Line 1154  else Line 1428  else
1428  newline". PCRE does not support \N{name}. However, it does support  newline". PCRE does not support \N{name}. However, it does support
1429  quantification such as \N{2,3}. */  quantification such as \N{2,3}. */
1430    
1431  if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&  if (escape == ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1432       !is_counted_repeat(ptr+2))       !is_counted_repeat(ptr+2))
1433    *errorcodeptr = ERR37;    *errorcodeptr = ERR37;
1434    
1435  /* If PCRE_UCP is set, we change the values for \d etc. */  /* If PCRE_UCP is set, we change the values for \d etc. */
1436    
1437  if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)  if ((options & PCRE_UCP) != 0 && escape >= ESC_D && escape <= ESC_w)
1438    c -= (ESC_DU - ESC_D);    escape += (ESC_DU - ESC_D);
1439    
1440  /* Set the pointer to the final character before returning. */  /* Set the pointer to the final character before returning. */
1441    
1442  *ptrptr = ptr;  *ptrptr = ptr;
1443  return c;  *chptr = c;
1444    return escape;
1445  }  }
1446    
1447    
# Line 1184  escape sequence. Line 1459  escape sequence.
1459  Argument:  Argument:
1460    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
1461    negptr         points to a boolean that is set TRUE for negation else FALSE    negptr         points to a boolean that is set TRUE for negation else FALSE
1462    dptr           points to an int that is set to the detailed property value    ptypeptr       points to an unsigned int that is set to the type value
1463      pdataptr       points to an unsigned int that is set to the detailed property value
1464    errorcodeptr   points to the error code variable    errorcodeptr   points to the error code variable
1465    
1466  Returns:         type value from ucp_type_table, or -1 for an invalid type  Returns:         TRUE if the type value was found, or FALSE for an invalid type
1467  */  */
1468    
1469  static int  static BOOL
1470  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, unsigned int *ptypeptr,
1471      unsigned int *pdataptr, int *errorcodeptr)
1472  {  {
1473  int c, i, bot, top;  pcre_uchar c;
1474    int i, bot, top;
1475  const pcre_uchar *ptr = *ptrptr;  const pcre_uchar *ptr = *ptrptr;
1476  pcre_uchar name[32];  pcre_uchar name[32];
1477    
1478  c = *(++ptr);  c = *(++ptr);
1479  if (c == 0) goto ERROR_RETURN;  if (c == CHAR_NULL) goto ERROR_RETURN;
1480    
1481  *negptr = FALSE;  *negptr = FALSE;
1482    
# Line 1215  if (c == CHAR_LEFT_CURLY_BRACKET) Line 1493  if (c == CHAR_LEFT_CURLY_BRACKET)
1493    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1494      {      {
1495      c = *(++ptr);      c = *(++ptr);
1496      if (c == 0) goto ERROR_RETURN;      if (c == CHAR_NULL) goto ERROR_RETURN;
1497      if (c == CHAR_RIGHT_CURLY_BRACKET) break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1498      name[i] = c;      name[i] = c;
1499      }      }
# Line 1240  top = PRIV(utt_size); Line 1518  top = PRIV(utt_size);
1518    
1519  while (bot < top)  while (bot < top)
1520    {    {
1521      int r;
1522    i = (bot + top) >> 1;    i = (bot + top) >> 1;
1523    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);
1524    if (c == 0)    if (r == 0)
1525      {      {
1526      *dptr = PRIV(utt)[i].value;      *ptypeptr = PRIV(utt)[i].type;
1527      return PRIV(utt)[i].type;      *pdataptr = PRIV(utt)[i].value;
1528        return TRUE;
1529      }      }
1530    if (c > 0) bot = i + 1; else top = i;    if (r > 0) bot = i + 1; else top = i;
1531    }    }
1532    
1533  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
1534  *ptrptr = ptr;  *ptrptr = ptr;
1535  return -1;  return FALSE;
1536    
1537  ERROR_RETURN:  ERROR_RETURN:
1538  *errorcodeptr = ERR46;  *errorcodeptr = ERR46;
1539  *ptrptr = ptr;  *ptrptr = ptr;
1540  return -1;  return FALSE;
1541  }  }
1542  #endif  #endif
1543    
1544    
1545    
   
1546  /*************************************************  /*************************************************
1547  *         Read repeat counts                     *  *         Read repeat counts                     *
1548  *************************************************/  *************************************************/
# Line 1292  int max = -1; Line 1571  int max = -1;
1571  /* 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
1572  an integer overflow. */  an integer overflow. */
1573    
1574  while (IS_DIGIT(*p)) min = min * 10 + *p++ - CHAR_0;  while (IS_DIGIT(*p)) min = min * 10 + (int)(*p++ - CHAR_0);
1575  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1576    {    {
1577    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 1307  if (*p == CHAR_RIGHT_CURLY_BRACKET) max Line 1586  if (*p == CHAR_RIGHT_CURLY_BRACKET) max
1586    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1587      {      {
1588      max = 0;      max = 0;
1589      while(IS_DIGIT(*p)) max = max * 10 + *p++ - CHAR_0;      while(IS_DIGIT(*p)) max = max * 10 + (int)(*p++ - CHAR_0);
1590      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1591        {        {
1592        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 1332  return p; Line 1611  return p;
1611    
1612    
1613  /*************************************************  /*************************************************
 *  Subroutine for finding forward reference      *  
 *************************************************/  
   
 /* This recursive function is called only from find_parens() below. The  
 top-level call starts at the beginning of the pattern. All other calls must  
 start at a parenthesis. It scans along a pattern's text looking for capturing  
 subpatterns, and counting them. If it finds a named pattern that matches the  
 name it is given, it returns its number. Alternatively, if the name is NULL, it  
 returns when it reaches a given numbered subpattern. Recursion is used to keep  
 track of subpatterns that reset the capturing group numbers - the (?| feature.  
   
 This function was originally called only from the second pass, in which we know  
 that if (?< or (?' or (?P< is encountered, the name will be correctly  
 terminated because that is checked in the first pass. There is now one call to  
 this function in the first pass, to check for a recursive back reference by  
 name (so that we can make the whole group atomic). In this case, we need check  
 only up to the current position in the pattern, and that is still OK because  
 and previous occurrences will have been checked. To make this work, the test  
 for "end of pattern" is a check against cd->end_pattern in the main loop,  
 instead of looking for a binary zero. This means that the special first-pass  
 call can adjust cd->end_pattern temporarily. (Checks for binary zero while  
 processing items within the loop are OK, because afterwards the main loop will  
 terminate.)  
   
 Arguments:  
   ptrptr       address of the current character pointer (updated)  
   cd           compile background data  
   name         name to seek, or NULL if seeking a numbered subpattern  
   lorn         name length, or subpattern number if name is NULL  
   xmode        TRUE if we are in /x mode  
   utf          TRUE if we are in UTF-8 / UTF-16 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;  
 }  
   
   
   
   
 /*************************************************  
1614  *      Find first significant op code            *  *      Find first significant op code            *
1615  *************************************************/  *************************************************/
1616    
# Line 1665  for (;;) Line 1649  for (;;)
1649    
1650      case OP_CALLOUT:      case OP_CALLOUT:
1651      case OP_CREF:      case OP_CREF:
1652      case OP_NCREF:      case OP_DNCREF:
1653      case OP_RREF:      case OP_RREF:
1654      case OP_NRREF:      case OP_DNRREF:
1655      case OP_DEF:      case OP_DEF:
1656      code += PRIV(OP_lengths)[*code];      code += PRIV(OP_lengths)[*code];
1657      break;      break;
# Line 1681  for (;;) Line 1665  for (;;)
1665    
1666    
1667    
   
1668  /*************************************************  /*************************************************
1669  *        Find the fixed length of a branch       *  *        Find the fixed length of a branch       *
1670  *************************************************/  *************************************************/
# Line 1699  and doing the check at the end; a flag s Line 1682  and doing the check at the end; a flag s
1682    
1683  Arguments:  Arguments:
1684    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1685    utf      TRUE in UTF-8 / UTF-16 mode    utf      TRUE in UTF-8 / UTF-16 / UTF-32 mode
1686    atend    TRUE if called when the pattern is complete    atend    TRUE if called when the pattern is complete
1687    cd       the "compile data" structure    cd       the "compile data" structure
1688    
# Line 1725  for (;;) Line 1708  for (;;)
1708    {    {
1709    int d;    int d;
1710    pcre_uchar *ce, *cs;    pcre_uchar *ce, *cs;
1711    register int op = *cc;    register pcre_uchar op = *cc;
1712    
1713    switch (op)    switch (op)
1714      {      {
# Line 1805  for (;;) Line 1788  for (;;)
1788      case OP_COMMIT:      case OP_COMMIT:
1789      case OP_CREF:      case OP_CREF:
1790      case OP_DEF:      case OP_DEF:
1791        case OP_DNCREF:
1792        case OP_DNRREF:
1793      case OP_DOLL:      case OP_DOLL:
1794      case OP_DOLLM:      case OP_DOLLM:
1795      case OP_EOD:      case OP_EOD:
1796      case OP_EODN:      case OP_EODN:
1797      case OP_FAIL:      case OP_FAIL:
     case OP_NCREF:  
     case OP_NRREF:  
1798      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1799      case OP_PRUNE:      case OP_PRUNE:
1800      case OP_REVERSE:      case OP_REVERSE:
# Line 1845  for (;;) Line 1828  for (;;)
1828      case OP_EXACTI:      case OP_EXACTI:
1829      case OP_NOTEXACT:      case OP_NOTEXACT:
1830      case OP_NOTEXACTI:      case OP_NOTEXACTI:
1831      branchlength += GET2(cc,1);      branchlength += (int)GET2(cc,1);
1832      cc += 2 + IMM2_SIZE;      cc += 2 + IMM2_SIZE;
1833  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
1834      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
# Line 1854  for (;;) Line 1837  for (;;)
1837    
1838      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1839      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1840      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP) cc += 2;      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)
1841          cc += 2;
1842      cc += 1 + IMM2_SIZE + 1;      cc += 1 + IMM2_SIZE + 1;
1843      break;      break;
1844    
# Line 1889  for (;;) Line 1873  for (;;)
1873    
1874      /* Check a class for variable quantification */      /* Check a class for variable quantification */
1875    
 #if defined SUPPORT_UTF || defined COMPILE_PCRE16  
     case OP_XCLASS:  
     cc += GET(cc, 1) - PRIV(OP_lengths)[OP_CLASS];  
     /* Fall through */  
 #endif  
   
1876      case OP_CLASS:      case OP_CLASS:
1877      case OP_NCLASS:      case OP_NCLASS:
1878    #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1879        case OP_XCLASS:
1880        /* The original code caused an unsigned overflow in 64 bit systems,
1881        so now we use a conditional statement. */
1882        if (op == OP_XCLASS)
1883          cc += GET(cc, 1);
1884        else
1885          cc += PRIV(OP_lengths)[OP_CLASS];
1886    #else
1887      cc += PRIV(OP_lengths)[OP_CLASS];      cc += PRIV(OP_lengths)[OP_CLASS];
1888    #endif
1889    
1890      switch (*cc)      switch (*cc)
1891        {        {
       case OP_CRPLUS:  
       case OP_CRMINPLUS:  
1892        case OP_CRSTAR:        case OP_CRSTAR:
1893        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1894          case OP_CRPLUS:
1895          case OP_CRMINPLUS:
1896        case OP_CRQUERY:        case OP_CRQUERY:
1897        case OP_CRMINQUERY:        case OP_CRMINQUERY:
1898          case OP_CRPOSSTAR:
1899          case OP_CRPOSPLUS:
1900          case OP_CRPOSQUERY:
1901        return -1;        return -1;
1902    
1903        case OP_CRRANGE:        case OP_CRRANGE:
1904        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1905          case OP_CRPOSRANGE:
1906        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1907        branchlength += GET2(cc,1);        branchlength += (int)GET2(cc,1);
1908        cc += 1 + 2 * IMM2_SIZE;        cc += 1 + 2 * IMM2_SIZE;
1909        break;        break;
1910    
# Line 1979  for (;;) Line 1971  for (;;)
1971      case OP_QUERYI:      case OP_QUERYI:
1972      case OP_REF:      case OP_REF:
1973      case OP_REFI:      case OP_REFI:
1974        case OP_DNREF:
1975        case OP_DNREFI:
1976      case OP_SBRA:      case OP_SBRA:
1977      case OP_SBRAPOS:      case OP_SBRAPOS:
1978      case OP_SCBRA:      case OP_SCBRA:
# Line 2015  for (;;) Line 2009  for (;;)
2009    
2010    
2011    
   
2012  /*************************************************  /*************************************************
2013  *    Scan compiled regex for specific bracket    *  *    Scan compiled regex for specific bracket    *
2014  *************************************************/  *************************************************/
# Line 2028  length. Line 2021  length.
2021    
2022  Arguments:  Arguments:
2023    code        points to start of expression    code        points to start of expression
2024    utf         TRUE in UTF-8 / UTF-16 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2025    number      the required bracket number or negative to find a lookbehind    number      the required bracket number or negative to find a lookbehind
2026    
2027  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 2039  PRIV(find_bracket)(const pcre_uchar *cod Line 2032  PRIV(find_bracket)(const pcre_uchar *cod
2032  {  {
2033  for (;;)  for (;;)
2034    {    {
2035    register int c = *code;    register pcre_uchar c = *code;
2036    
2037    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2038    
# Line 2062  for (;;) Line 2055  for (;;)
2055    else if (c == OP_CBRA || c == OP_SCBRA ||    else if (c == OP_CBRA || c == OP_SCBRA ||
2056             c == OP_CBRAPOS || c == OP_SCBRAPOS)             c == OP_CBRAPOS || c == OP_SCBRAPOS)
2057      {      {
2058      int n = GET2(code, 1+LINK_SIZE);      int n = (int)GET2(code, 1+LINK_SIZE);
2059      if (n == number) return (pcre_uchar *)code;      if (n == number) return (pcre_uchar *)code;
2060      code += PRIV(OP_lengths)[c];      code += PRIV(OP_lengths)[c];
2061      }      }
# Line 2092  for (;;) Line 2085  for (;;)
2085        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2086        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2087        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
2088        if (code[1 + IMM2_SIZE] == OP_PROP        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2089          || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;          code += 2;
2090        break;        break;
2091    
2092        case OP_MARK:        case OP_MARK:
2093        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
2094        case OP_SKIP_ARG:        case OP_SKIP_ARG:
       code += code[1];  
       break;  
   
2095        case OP_THEN_ARG:        case OP_THEN_ARG:
2096        code += code[1];        code += code[1];
2097        break;        break;
# Line 2115  for (;;) Line 2105  for (;;)
2105    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
2106    arrange to skip the extra bytes. */    arrange to skip the extra bytes. */
2107    
2108  #ifdef SUPPORT_UTF  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2109      if (utf) switch(c)      if (utf) switch(c)
2110        {        {
2111        case OP_CHAR:        case OP_CHAR:
# Line 2167  instance of OP_RECURSE. Line 2157  instance of OP_RECURSE.
2157    
2158  Arguments:  Arguments:
2159    code        points to start of expression    code        points to start of expression
2160    utf         TRUE in UTF-8 / UTF-16 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2161    
2162  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
2163  */  */
# Line 2177  find_recurse(const pcre_uchar *code, BOO Line 2167  find_recurse(const pcre_uchar *code, BOO
2167  {  {
2168  for (;;)  for (;;)
2169    {    {
2170    register int c = *code;    register pcre_uchar c = *code;
2171    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2172    if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
2173    
# Line 2212  for (;;) Line 2202  for (;;)
2202        case OP_TYPEUPTO:        case OP_TYPEUPTO:
2203        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2204        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2205        if (code[1 + IMM2_SIZE] == OP_PROP        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2206          || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;          code += 2;
2207        break;        break;
2208    
2209        case OP_MARK:        case OP_MARK:
2210        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
2211        case OP_SKIP_ARG:        case OP_SKIP_ARG:
2212          case OP_THEN_ARG:
2213        code += code[1];        code += code[1];
2214        break;        break;
2215          }
       case OP_THEN_ARG:  
       code += code[1];  
       break;  
       }  
2216    
2217      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
2218    
# Line 2235  for (;;) Line 2222  for (;;)
2222      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
2223      to arrange to skip the extra bytes. */      to arrange to skip the extra bytes. */
2224    
2225  #ifdef SUPPORT_UTF  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2226      if (utf) switch(c)      if (utf) switch(c)
2227        {        {
2228        case OP_CHAR:        case OP_CHAR:
# Line 2321  bracket whose current branch will alread Line 2308  bracket whose current branch will alread
2308  Arguments:  Arguments:
2309    code        points to start of search    code        points to start of search
2310    endcode     points to where to stop    endcode     points to where to stop
2311    utf         TRUE if in UTF-8 / UTF-16 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2312    cd          contains pointers to tables etc.    cd          contains pointers to tables etc.
2313      recurses    chain of recurse_check to catch mutual recursion
2314    
2315  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2316  */  */
2317    
2318    typedef struct recurse_check {
2319      struct recurse_check *prev;
2320      const pcre_uchar *group;
2321    } recurse_check;
2322    
2323  static BOOL  static BOOL
2324  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2325    BOOL utf, compile_data *cd)    BOOL utf, compile_data *cd, recurse_check *recurses)
2326  {  {
2327  register int c;  register pcre_uchar c;
2328    recurse_check this_recurse;
2329    
2330  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2331       code < endcode;       code < endcode;
2332       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
# Line 2359  for (code = first_significant_code(code Line 2354  for (code = first_significant_code(code
2354    
2355    if (c == OP_RECURSE)    if (c == OP_RECURSE)
2356      {      {
2357      const pcre_uchar *scode;      const pcre_uchar *scode = cd->start_code + GET(code, 1);
2358      BOOL empty_branch;      BOOL empty_branch;
2359    
2360      /* Test for forward reference */      /* Test for forward reference or uncompleted reference. This is disabled
2361        when called to scan a completed pattern by setting cd->start_workspace to
2362        NULL. */
2363    
2364        if (cd->start_workspace != NULL)
2365          {
2366          const pcre_uchar *tcode;
2367          for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)
2368            if ((int)GET(tcode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;
2369          if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2370          }
2371    
2372        /* If we are scanning a completed pattern, there are no forward references
2373        and all groups are complete. We need to detect whether this is a recursive
2374        call, as otherwise there will be an infinite loop. If it is a recursion,
2375        just skip over it. Simple recursions are easily detected. For mutual
2376        recursions we keep a chain on the stack. */
2377    
2378      for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)      else
2379        if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;        {
2380          recurse_check *r = recurses;
2381          const pcre_uchar *endgroup = scode;
2382    
2383      /* Not a forward reference, test for completed backward reference */        do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);
2384          if (code >= scode && code <= endgroup) continue;  /* Simple recursion */
2385    
2386      empty_branch = FALSE;        for (r = recurses; r != NULL; r = r->prev)
2387      scode = cd->start_code + GET(code, 1);          if (r->group == scode) break;
2388      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */        if (r != NULL) continue;   /* Mutual recursion */
2389          }
2390    
2391      /* Completed backwards reference */      /* Completed reference; scan the referenced group, remembering it on the
2392        stack chain to detect mutual recursions. */
2393    
2394        empty_branch = FALSE;
2395        this_recurse.prev = recurses;
2396        this_recurse.group = scode;
2397    
2398      do      do
2399        {        {
2400        if (could_be_empty_branch(scode, endcode, utf, cd))        if (could_be_empty_branch(scode, endcode, utf, cd, &this_recurse))
2401          {          {
2402          empty_branch = TRUE;          empty_branch = TRUE;
2403          break;          break;
# Line 2433  for (code = first_significant_code(code Line 2453  for (code = first_significant_code(code
2453        empty_branch = FALSE;        empty_branch = FALSE;
2454        do        do
2455          {          {
2456          if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd))          if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd, NULL))
2457            empty_branch = TRUE;            empty_branch = TRUE;
2458          code += GET(code, 1);          code += GET(code, 1);
2459          }          }
# Line 2475  for (code = first_significant_code(code Line 2495  for (code = first_significant_code(code
2495        case OP_CRMINSTAR:        case OP_CRMINSTAR:
2496        case OP_CRQUERY:        case OP_CRQUERY:
2497        case OP_CRMINQUERY:        case OP_CRMINQUERY:
2498          case OP_CRPOSSTAR:
2499          case OP_CRPOSQUERY:
2500        break;        break;
2501    
2502        default:                   /* Non-repeat => class must match */        default:                   /* Non-repeat => class must match */
2503        case OP_CRPLUS:            /* These repeats aren't empty */        case OP_CRPLUS:            /* These repeats aren't empty */
2504        case OP_CRMINPLUS:        case OP_CRMINPLUS:
2505          case OP_CRPOSPLUS:
2506        return FALSE;        return FALSE;
2507    
2508        case OP_CRRANGE:        case OP_CRRANGE:
2509        case OP_CRMINRANGE:        case OP_CRMINRANGE:
2510          case OP_CRPOSRANGE:
2511        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */
2512        break;        break;
2513        }        }
# Line 2491  for (code = first_significant_code(code Line 2515  for (code = first_significant_code(code
2515    
2516      /* Opcodes that must match a character */      /* Opcodes that must match a character */
2517    
2518        case OP_ANY:
2519        case OP_ALLANY:
2520        case OP_ANYBYTE:
2521    
2522      case OP_PROP:      case OP_PROP:
2523      case OP_NOTPROP:      case OP_NOTPROP:
2524        case OP_ANYNL:
2525    
2526        case OP_NOT_HSPACE:
2527        case OP_HSPACE:
2528        case OP_NOT_VSPACE:
2529        case OP_VSPACE:
2530      case OP_EXTUNI:      case OP_EXTUNI:
2531    
2532      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2533      case OP_DIGIT:      case OP_DIGIT:
2534      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2535      case OP_WHITESPACE:      case OP_WHITESPACE:
2536      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2537      case OP_WORDCHAR:      case OP_WORDCHAR:
2538      case OP_ANY:  
     case OP_ALLANY:  
     case OP_ANYBYTE:  
2539      case OP_CHAR:      case OP_CHAR:
2540      case OP_CHARI:      case OP_CHARI:
2541      case OP_NOT:      case OP_NOT:
2542      case OP_NOTI:      case OP_NOTI:
2543    
2544      case OP_PLUS:      case OP_PLUS:
2545        case OP_PLUSI:
2546      case OP_MINPLUS:      case OP_MINPLUS:
2547      case OP_POSPLUS:      case OP_MINPLUSI:
2548      case OP_EXACT:  
2549      case OP_NOTPLUS:      case OP_NOTPLUS:
2550        case OP_NOTPLUSI:
2551      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
2552        case OP_NOTMINPLUSI:
2553    
2554        case OP_POSPLUS:
2555        case OP_POSPLUSI:
2556      case OP_NOTPOSPLUS:      case OP_NOTPOSPLUS:
2557        case OP_NOTPOSPLUSI:
2558    
2559        case OP_EXACT:
2560        case OP_EXACTI:
2561      case OP_NOTEXACT:      case OP_NOTEXACT:
2562        case OP_NOTEXACTI:
2563    
2564      case OP_TYPEPLUS:      case OP_TYPEPLUS:
2565      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
2566      case OP_TYPEPOSPLUS:      case OP_TYPEPOSPLUS:
2567      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2568    
2569      return FALSE;      return FALSE;
2570    
2571      /* These are going to continue, as they may be empty, but we have to      /* These are going to continue, as they may be empty, but we have to
# Line 2538  for (code = first_significant_code(code Line 2585  for (code = first_significant_code(code
2585      case OP_TYPEUPTO:      case OP_TYPEUPTO:
2586      case OP_TYPEMINUPTO:      case OP_TYPEMINUPTO:
2587      case OP_TYPEPOSUPTO:      case OP_TYPEPOSUPTO:
2588      if (code[1 + IMM2_SIZE] == OP_PROP      if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2589        || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;        code += 2;
2590      break;      break;
2591    
2592      /* End of branch */      /* End of branch */
# Line 2552  for (code = first_significant_code(code Line 2599  for (code = first_significant_code(code
2599      return TRUE;      return TRUE;
2600    
2601      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2602      MINUPTO, and POSUPTO may be followed by a multibyte character */      MINUPTO, and POSUPTO and their caseless and negative versions may be
2603        followed by a multibyte character. */
2604    
2605  #ifdef SUPPORT_UTF  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2606      case OP_STAR:      case OP_STAR:
2607      case OP_STARI:      case OP_STARI:
2608        case OP_NOTSTAR:
2609        case OP_NOTSTARI:
2610    
2611      case OP_MINSTAR:      case OP_MINSTAR:
2612      case OP_MINSTARI:      case OP_MINSTARI:
2613        case OP_NOTMINSTAR:
2614        case OP_NOTMINSTARI:
2615    
2616      case OP_POSSTAR:      case OP_POSSTAR:
2617      case OP_POSSTARI:      case OP_POSSTARI:
2618        case OP_NOTPOSSTAR:
2619        case OP_NOTPOSSTARI:
2620    
2621      case OP_QUERY:      case OP_QUERY:
2622      case OP_QUERYI:      case OP_QUERYI:
2623        case OP_NOTQUERY:
2624        case OP_NOTQUERYI:
2625    
2626      case OP_MINQUERY:      case OP_MINQUERY:
2627      case OP_MINQUERYI:      case OP_MINQUERYI:
2628        case OP_NOTMINQUERY:
2629        case OP_NOTMINQUERYI:
2630    
2631      case OP_POSQUERY:      case OP_POSQUERY:
2632      case OP_POSQUERYI:      case OP_POSQUERYI:
2633        case OP_NOTPOSQUERY:
2634        case OP_NOTPOSQUERYI:
2635    
2636      if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);      if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2637      break;      break;
2638    
2639      case OP_UPTO:      case OP_UPTO:
2640      case OP_UPTOI:      case OP_UPTOI:
2641        case OP_NOTUPTO:
2642        case OP_NOTUPTOI:
2643    
2644      case OP_MINUPTO:      case OP_MINUPTO:
2645      case OP_MINUPTOI:      case OP_MINUPTOI:
2646        case OP_NOTMINUPTO:
2647        case OP_NOTMINUPTOI:
2648    
2649      case OP_POSUPTO:      case OP_POSUPTO:
2650      case OP_POSUPTOI:      case OP_POSUPTOI:
2651        case OP_NOTPOSUPTO:
2652        case OP_NOTPOSUPTOI:
2653    
2654      if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);      if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2655      break;      break;
2656  #endif  #endif
# Line 2586  for (code = first_significant_code(code Line 2661  for (code = first_significant_code(code
2661      case OP_MARK:      case OP_MARK:
2662      case OP_PRUNE_ARG:      case OP_PRUNE_ARG:
2663      case OP_SKIP_ARG:      case OP_SKIP_ARG:
     code += code[1];  
     break;  
   
2664      case OP_THEN_ARG:      case OP_THEN_ARG:
2665      code += code[1];      code += code[1];
2666      break;      break;
# Line 2620  Arguments: Line 2692  Arguments:
2692    code        points to start of the recursion    code        points to start of the recursion
2693    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2694    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2695    utf         TRUE if in UTF-8 / UTF-16 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2696    cd          pointers to tables etc    cd          pointers to tables etc
2697    
2698  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
# Line 2632  could_be_empty(const pcre_uchar *code, c Line 2704  could_be_empty(const pcre_uchar *code, c
2704  {  {
2705  while (bcptr != NULL && bcptr->current_branch >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2706    {    {
2707    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd))    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd, NULL))
2708      return FALSE;      return FALSE;
2709    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2710    }    }
# Line 2642  return TRUE; Line 2714  return TRUE;
2714    
2715    
2716  /*************************************************  /*************************************************
2717  *           Check for POSIX class syntax         *  *        Base opcode of repeated opcodes         *
2718  *************************************************/  *************************************************/
2719    
2720  /* This function is called when the sequence "[:" or "[." or "[=" is  /* Returns the base opcode for repeated single character type opcodes. If the
2721  encountered in a character class. It checks whether this is followed by a  opcode is not a repeated character type, it returns with the original value.
 sequence of characters terminated by a matching ":]" or ".]" or "=]". If we  
 reach an unescaped ']' without the special preceding character, return FALSE.  
   
 Originally, this function only recognized a sequence of letters between the  
 terminators, but it seems that Perl recognizes any sequence of characters,  
 though of course unknown POSIX names are subsequently rejected. Perl gives an  
 "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE  
 didn't consider this to be a POSIX class. Likewise for [:1234:].  
   
 The problem in trying to be exactly like Perl is in the handling of escapes. We  
 have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX  
 class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code  
 below handles the special case of \], but does not try to do any other escape  
 processing. This makes it different from Perl for cases such as [:l\ower:]  
 where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize  
 "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,  
 I think.  
   
 A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.  
 It seems that the appearance of a nested POSIX class supersedes an apparent  
 external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or  
 a digit.  
   
 In Perl, unescaped square brackets may also appear as part of class names. For  
 example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for  
 [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not  
 seem right at all. PCRE does not allow closing square brackets in POSIX class  
 names.  
   
 Arguments:  
   ptr      pointer to the initial [  
   endptr   where to return the end pointer  
2722    
2723  Returns:   TRUE or FALSE  Arguments:  c opcode
2724    Returns:    base opcode for the type
2725  */  */
2726    
2727  static BOOL  static pcre_uchar
2728  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)  get_repeat_base(pcre_uchar c)
2729  {  {
2730  int terminator;          /* Don't combine these lines; the Solaris cc */  return (c > OP_TYPEPOSUPTO)? c :
2731  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */         (c >= OP_TYPESTAR)?   OP_TYPESTAR :
2732  for (++ptr; *ptr != 0; ptr++)         (c >= OP_NOTSTARI)?   OP_NOTSTARI :
2733    {         (c >= OP_NOTSTAR)?    OP_NOTSTAR :
2734    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)         (c >= OP_STARI)?      OP_STARI :
2735      ptr++;                               OP_STAR;
   else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;  
   else  
     {  
     if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)  
       {  
       *endptr = ptr;  
       return TRUE;  
       }  
     if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&  
          (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||  
           ptr[1] == CHAR_EQUALS_SIGN) &&  
         check_posix_syntax(ptr, endptr))  
       return FALSE;  
     }  
   }  
 return FALSE;  
2736  }  }
2737    
2738    
2739    
2740    #ifdef SUPPORT_UCP
2741  /*************************************************  /*************************************************
2742  *          Check POSIX class name                *  *        Check a character and a property        *
2743  *************************************************/  *************************************************/
2744    
2745  /* This function is called to check the name given in a POSIX-style class entry  /* This function is called by check_auto_possessive() when a property item
2746  such as [:alnum:].  is adjacent to a fixed character.
2747    
2748  Arguments:  Arguments:
2749    ptr        points to the first letter    c            the character
2750    len        the length of the name    ptype        the property type
2751      pdata        the data for the type
2752      negated      TRUE if it's a negated property (\P or \p{^)
2753    
2754  Returns:     a value representing the name, or -1 if unknown  Returns:       TRUE if auto-possessifying is OK
2755  */  */
2756    
2757  static int  static BOOL
2758  check_posix_name(const pcre_uchar *ptr, int len)  check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata,
2759      BOOL negated)
2760  {  {
2761  const char *pn = posix_names;  const pcre_uint32 *p;
2762  register int yield = 0;  const ucd_record *prop = GET_UCD(c);
 while (posix_name_lengths[yield] != 0)  
   {  
   if (len == posix_name_lengths[yield] &&  
     STRNCMP_UC_C8(ptr, pn, len) == 0) return yield;  
   pn += posix_name_lengths[yield] + 1;  
   yield++;  
   }  
 return -1;  
 }  
   
   
 /*************************************************  
 *    Adjust OP_RECURSE items in repeated group   *  
 *************************************************/  
2763    
2764  /* OP_RECURSE items contain an offset from the start of the regex to the group  switch(ptype)
2765  that is referenced. This means that groups can be replicated for fixed    {
2766  repetition simply by copying (because the recursion is allowed to refer to    case PT_LAMP:
2767  earlier groups that are outside the current group). However, when a group is    return (prop->chartype == ucp_Lu ||
2768  optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is            prop->chartype == ucp_Ll ||
2769  inserted before it, after it has been compiled. This means that any OP_RECURSE            prop->chartype == ucp_Lt) == negated;
 items within it that refer to the group itself or any contained groups have to  
 have their offsets adjusted. That one of the jobs of this function. Before it  
 is called, the partially compiled regex must be temporarily terminated with  
 OP_END.  
2770    
2771  This function has been extended with the possibility of forward references for    case PT_GC:
2772  recursions and subroutine calls. It must also check the list of such references    return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
 for the group we are dealing with. If it finds that one of the recursions in  
 the current group is on this list, it adjusts the offset in the list, not the  
 value in the reference (which is a group number).  
2773    
2774  Arguments:    case PT_PC:
2775    group      points to the start of the group    return (pdata == prop->chartype) == negated;
   adjust     the amount by which the group is to be moved  
   utf        TRUE in UTF-8 / UTF-16 mode  
   cd         contains pointers to tables etc.  
   save_hwm   the hwm forward reference pointer at the start of the group  
2776    
2777  Returns:     nothing    case PT_SC:
2778  */    return (pdata == prop->script) == negated;
2779    
2780  static void    /* These are specials */
 adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,  
   pcre_uchar *save_hwm)  
 {  
 pcre_uchar *ptr = group;  
2781    
2782  while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)    case PT_ALNUM:
2783    {    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2784    int offset;            PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
   pcre_uchar *hc;  
2785    
2786    /* See if this recursion is on the forward reference list. If so, adjust the    /* Perl space used to exclude VT, but from Perl 5.18 it is included, which
2787    reference. */    means that Perl space and POSIX space are now identical. PCRE was changed
2788      at release 8.34. */
2789    
2790    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)    case PT_SPACE:    /* Perl space */
2791      case PT_PXSPACE:  /* POSIX space */
2792      switch(c)
2793      {      {
2794      offset = GET(hc, 0);      HSPACE_CASES:
2795      if (cd->start_code + offset == ptr + 1)      VSPACE_CASES:
2796        {      return negated;
2797        PUT(hc, 0, offset + adjust);  
2798        break;      default:
2799        }      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z) == negated;
2800      }      }
2801      break;  /* Control never reaches here */
2802    
2803    /* Otherwise, adjust the recursion offset if it's after the start of this    case PT_WORD:
2804    group. */    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2805              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2806              c == CHAR_UNDERSCORE) == negated;
2807    
2808    if (hc >= cd->hwm)    case PT_CLIST:
2809      p = PRIV(ucd_caseless_sets) + prop->caseset;
2810      for (;;)
2811      {      {
2812      offset = GET(ptr, 1);      if (c < *p) return !negated;
2813      if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);      if (c == *p++) return negated;
2814      }      }
2815      break;  /* Control never reaches here */
   ptr += 1 + LINK_SIZE;  
2816    }    }
2817    
2818    return FALSE;
2819  }  }
2820    #endif  /* SUPPORT_UCP */
2821    
2822    
2823    
2824  /*************************************************  /*************************************************
2825  *        Insert an automatic callout point       *  *        Fill the character property list        *
2826  *************************************************/  *************************************************/
2827    
2828  /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert  /* Checks whether the code points to an opcode that can take part in auto-
2829  callout points before each pattern item.  possessification, and if so, fills a list with its properties.
2830    
2831  Arguments:  Arguments:
2832    code           current code pointer    code        points to start of expression
2833    ptr            current pattern pointer    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2834    cd             pointers to tables etc    fcc         points to case-flipping table
2835      list        points to output list
2836                  list[0] will be filled with the opcode
2837                  list[1] will be non-zero if this opcode
2838                    can match an empty character string
2839                  list[2..7] depends on the opcode
2840    
2841  Returns:         new code pointer  Returns:      points to the start of the next opcode if *code is accepted
2842                  NULL if *code is not accepted
2843  */  */
2844    
2845  static pcre_uchar *  static const pcre_uchar *
2846  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)  get_chr_property_list(const pcre_uchar *code, BOOL utf,
2847      const pcre_uint8 *fcc, pcre_uint32 *list)
2848  {  {
2849  *code++ = OP_CALLOUT;  pcre_uchar c = *code;
2850    const pcre_uchar *end;
2851    const pcre_uint32 *clist_src;
2852    pcre_uint32 *clist_dest;
2853    pcre_uint32 chr;
2854    pcre_uchar base;
2855    
2856    list[0] = c;
2857    list[1] = FALSE;
2858    code++;
2859    
2860    if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
2861      {
2862      base = get_repeat_base(c);
2863      c -= (base - OP_STAR);
2864    
2865      if (c == OP_UPTO || c == OP_MINUPTO || c == OP_EXACT || c == OP_POSUPTO)
2866        code += IMM2_SIZE;
2867    
2868      list[1] = (c != OP_PLUS && c != OP_MINPLUS && c != OP_EXACT && c != OP_POSPLUS);
2869    
2870      switch(base)
2871        {
2872        case OP_STAR:
2873        list[0] = OP_CHAR;
2874        break;
2875    
2876        case OP_STARI:
2877        list[0] = OP_CHARI;
2878        break;
2879    
2880        case OP_NOTSTAR:
2881        list[0] = OP_NOT;
2882        break;
2883    
2884        case OP_NOTSTARI:
2885        list[0] = OP_NOTI;
2886        break;
2887    
2888        case OP_TYPESTAR:
2889        list[0] = *code;
2890        code++;
2891        break;
2892        }
2893      c = list[0];
2894      }
2895    
2896    switch(c)
2897      {
2898      case OP_NOT_DIGIT:
2899      case OP_DIGIT:
2900      case OP_NOT_WHITESPACE:
2901      case OP_WHITESPACE:
2902      case OP_NOT_WORDCHAR:
2903      case OP_WORDCHAR:
2904      case OP_ANY:
2905      case OP_ALLANY:
2906      case OP_ANYNL:
2907      case OP_NOT_HSPACE:
2908      case OP_HSPACE:
2909      case OP_NOT_VSPACE:
2910      case OP_VSPACE:
2911      case OP_EXTUNI:
2912      case OP_EODN:
2913      case OP_EOD:
2914      case OP_DOLL:
2915      case OP_DOLLM:
2916      return code;
2917    
2918      case OP_CHAR:
2919      case OP_NOT:
2920      GETCHARINCTEST(chr, code);
2921      list[2] = chr;
2922      list[3] = NOTACHAR;
2923      return code;
2924    
2925      case OP_CHARI:
2926      case OP_NOTI:
2927      list[0] = (c == OP_CHARI) ? OP_CHAR : OP_NOT;
2928      GETCHARINCTEST(chr, code);
2929      list[2] = chr;
2930    
2931    #ifdef SUPPORT_UCP
2932      if (chr < 128 || (chr < 256 && !utf))
2933        list[3] = fcc[chr];
2934      else
2935        list[3] = UCD_OTHERCASE(chr);
2936    #elif defined SUPPORT_UTF || !defined COMPILE_PCRE8
2937      list[3] = (chr < 256) ? fcc[chr] : chr;
2938    #else
2939      list[3] = fcc[chr];
2940    #endif
2941    
2942      /* The othercase might be the same value. */
2943    
2944      if (chr == list[3])
2945        list[3] = NOTACHAR;
2946      else
2947        list[4] = NOTACHAR;
2948      return code;
2949    
2950    #ifdef SUPPORT_UCP
2951      case OP_PROP:
2952      case OP_NOTPROP:
2953      if (code[0] != PT_CLIST)
2954        {
2955        list[2] = code[0];
2956        list[3] = code[1];
2957        return code + 2;
2958        }
2959    
2960      /* Convert only if we have enough space. */
2961    
2962      clist_src = PRIV(ucd_caseless_sets) + code[1];
2963      clist_dest = list + 2;
2964      code += 2;
2965    
2966      do {
2967         if (clist_dest >= list + 8)
2968           {
2969           /* Early return if there is not enough space. This should never
2970           happen, since all clists are shorter than 5 character now. */
2971           list[2] = code[0];
2972           list[3] = code[1];
2973           return code;
2974           }
2975         *clist_dest++ = *clist_src;
2976         }
2977      while(*clist_src++ != NOTACHAR);
2978    
2979      /* All characters are stored. The terminating NOTACHAR
2980      is copied form the clist itself. */
2981    
2982      list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
2983      return code;
2984    #endif
2985    
2986      case OP_NCLASS:
2987      case OP_CLASS:
2988    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2989      case OP_XCLASS:
2990      if (c == OP_XCLASS)
2991        end = code + GET(code, 0) - 1;
2992      else
2993    #endif
2994        end = code + 32 / sizeof(pcre_uchar);
2995    
2996      switch(*end)
2997        {
2998        case OP_CRSTAR:
2999        case OP_CRMINSTAR:
3000        case OP_CRQUERY:
3001        case OP_CRMINQUERY:
3002        case OP_CRPOSSTAR:
3003        case OP_CRPOSQUERY:
3004        list[1] = TRUE;
3005        end++;
3006        break;
3007    
3008        case OP_CRPLUS:
3009        case OP_CRMINPLUS:
3010        case OP_CRPOSPLUS:
3011        end++;
3012        break;
3013    
3014        case OP_CRRANGE:
3015        case OP_CRMINRANGE:
3016        case OP_CRPOSRANGE:
3017        list[1] = (GET2(end, 1) == 0);
3018        end += 1 + 2 * IMM2_SIZE;
3019        break;
3020        }
3021      list[2] = end - code;
3022      return end;
3023      }
3024    return NULL;    /* Opcode not accepted */
3025    }
3026    
3027    
3028    
3029    /*************************************************
3030    *    Scan further character sets for match       *
3031    *************************************************/
3032    
3033    /* Checks whether the base and the current opcode have a common character, in
3034    which case the base cannot be possessified.
3035    
3036    Arguments:
3037      code        points to the byte code
3038      utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3039      cd          static compile data
3040      base_list   the data list of the base opcode
3041    
3042    Returns:      TRUE if the auto-possessification is possible
3043    */
3044    
3045    static BOOL
3046    compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
3047      const pcre_uint32 *base_list, const pcre_uchar *base_end)
3048    {
3049    pcre_uchar c;
3050    pcre_uint32 list[8];
3051    const pcre_uint32 *chr_ptr;
3052    const pcre_uint32 *ochr_ptr;
3053    const pcre_uint32 *list_ptr;
3054    const pcre_uchar *next_code;
3055    const pcre_uint8 *class_bitset;
3056    const pcre_uint32 *set1, *set2, *set_end;
3057    pcre_uint32 chr;
3058    BOOL accepted, invert_bits;
3059    
3060    /* Note: the base_list[1] contains whether the current opcode has greedy
3061    (represented by a non-zero value) quantifier. This is a different from
3062    other character type lists, which stores here that the character iterator
3063    matches to an empty string (also represented by a non-zero value). */
3064    
3065    for(;;)
3066      {
3067      /* All operations move the code pointer forward.
3068      Therefore infinite recursions are not possible. */
3069    
3070      c = *code;
3071    
3072      /* Skip over callouts */
3073    
3074      if (c == OP_CALLOUT)
3075        {
3076        code += PRIV(OP_lengths)[c];
3077        continue;
3078        }
3079    
3080      if (c == OP_ALT)
3081        {
3082        do code += GET(code, 1); while (*code == OP_ALT);
3083        c = *code;
3084        }
3085    
3086      switch(c)
3087        {
3088        case OP_END:
3089        case OP_KETRPOS:
3090        /* TRUE only in greedy case. The non-greedy case could be replaced by
3091        an OP_EXACT, but it is probably not worth it. (And note that OP_EXACT
3092        uses more memory, which we cannot get at this stage.) */
3093    
3094        return base_list[1] != 0;
3095    
3096        case OP_KET:
3097        /* If the bracket is capturing, and referenced by an OP_RECURSE, or
3098        it is an atomic sub-pattern (assert, once, etc.) the non-greedy case
3099        cannot be converted to a possessive form. */
3100    
3101        if (base_list[1] == 0) return FALSE;
3102    
3103        switch(*(code - GET(code, 1)))
3104          {
3105          case OP_ASSERT:
3106          case OP_ASSERT_NOT:
3107          case OP_ASSERTBACK:
3108          case OP_ASSERTBACK_NOT:
3109          case OP_ONCE:
3110          case OP_ONCE_NC:
3111          /* Atomic sub-patterns and assertions can always auto-possessify their
3112          last iterator. */
3113          return TRUE;
3114          }
3115    
3116        code += PRIV(OP_lengths)[c];
3117        continue;
3118    
3119        case OP_ONCE:
3120        case OP_ONCE_NC:
3121        case OP_BRA:
3122        case OP_CBRA:
3123        next_code = code + GET(code, 1);
3124        code += PRIV(OP_lengths)[c];
3125    
3126        while (*next_code == OP_ALT)
3127          {
3128          if (!compare_opcodes(code, utf, cd, base_list, base_end)) return FALSE;
3129          code = next_code + 1 + LINK_SIZE;
3130          next_code += GET(next_code, 1);
3131          }
3132        continue;
3133    
3134        case OP_BRAZERO:
3135        case OP_BRAMINZERO:
3136    
3137        next_code = code + 1;
3138        if (*next_code != OP_BRA && *next_code != OP_CBRA
3139            && *next_code != OP_ONCE && *next_code != OP_ONCE_NC) return FALSE;
3140    
3141        do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
3142    
3143        /* The bracket content will be checked by the
3144        OP_BRA/OP_CBRA case above. */
3145        next_code += 1 + LINK_SIZE;
3146        if (!compare_opcodes(next_code, utf, cd, base_list, base_end))
3147          return FALSE;
3148    
3149        code += PRIV(OP_lengths)[c];
3150        continue;
3151        }
3152    
3153      /* Check for a supported opcode, and load its properties. */
3154    
3155      code = get_chr_property_list(code, utf, cd->fcc, list);
3156      if (code == NULL) return FALSE;    /* Unsupported */
3157    
3158      /* If either opcode is a small character list, set pointers for comparing
3159      characters from that list with another list, or with a property. */
3160    
3161      if (base_list[0] == OP_CHAR)
3162        {
3163        chr_ptr = base_list + 2;
3164        list_ptr = list;
3165        }
3166      else if (list[0] == OP_CHAR)
3167        {
3168        chr_ptr = list + 2;
3169        list_ptr = base_list;
3170        }
3171    
3172      /* Character bitsets can also be compared to certain opcodes. */
3173    
3174      else if (base_list[0] == OP_CLASS || list[0] == OP_CLASS
3175    #ifdef COMPILE_PCRE8
3176          /* In 8 bit, non-UTF mode, OP_CLASS and OP_NCLASS are the same. */
3177          || (!utf && (base_list[0] == OP_NCLASS || list[0] == OP_NCLASS))
3178    #endif
3179          )
3180        {
3181    #ifdef COMPILE_PCRE8
3182        if (base_list[0] == OP_CLASS || (!utf && base_list[0] == OP_NCLASS))
3183    #else
3184        if (base_list[0] == OP_CLASS)
3185    #endif
3186          {
3187          set1 = (pcre_uint32 *)(base_end - base_list[2]);
3188          list_ptr = list;
3189          }
3190        else
3191          {
3192          set1 = (pcre_uint32 *)(code - list[2]);
3193          list_ptr = base_list;
3194          }
3195    
3196        invert_bits = FALSE;
3197        switch(list_ptr[0])
3198          {
3199          case OP_CLASS:
3200          case OP_NCLASS:
3201          set2 = (pcre_uint32 *)
3202            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3203          break;
3204    
3205          /* OP_XCLASS cannot be supported here, because its bitset
3206          is not necessarily complete. E.g: [a-\0x{200}] is stored
3207          as a character range, and the appropriate bits are not set. */
3208    
3209          case OP_NOT_DIGIT:
3210            invert_bits = TRUE;
3211            /* Fall through */
3212          case OP_DIGIT:
3213            set2 = (pcre_uint32 *)(cd->cbits + cbit_digit);
3214            break;
3215    
3216          case OP_NOT_WHITESPACE:
3217            invert_bits = TRUE;
3218            /* Fall through */
3219          case OP_WHITESPACE:
3220            set2 = (pcre_uint32 *)(cd->cbits + cbit_space);
3221            break;
3222    
3223          case OP_NOT_WORDCHAR:
3224            invert_bits = TRUE;
3225            /* Fall through */
3226          case OP_WORDCHAR:
3227            set2 = (pcre_uint32 *)(cd->cbits + cbit_word);
3228            break;
3229    
3230          default:
3231          return FALSE;
3232          }
3233    
3234        /* Compare 4 bytes to improve speed. */
3235        set_end = set1 + (32 / 4);
3236        if (invert_bits)
3237          {
3238          do
3239            {
3240            if ((*set1++ & ~(*set2++)) != 0) return FALSE;
3241            }
3242          while (set1 < set_end);
3243          }
3244        else
3245          {
3246          do
3247            {
3248            if ((*set1++ & *set2++) != 0) return FALSE;
3249            }
3250          while (set1 < set_end);
3251          }
3252    
3253        if (list[1] == 0) return TRUE;
3254        /* Might be an empty repeat. */
3255        continue;
3256        }
3257    
3258      /* Some property combinations also acceptable. Unicode property opcodes are
3259      processed specially; the rest can be handled with a lookup table. */
3260    
3261      else
3262        {
3263        pcre_uint32 leftop, rightop;
3264    
3265        leftop = base_list[0];
3266        rightop = list[0];
3267    
3268    #ifdef SUPPORT_UCP
3269        accepted = FALSE; /* Always set in non-unicode case. */
3270        if (leftop == OP_PROP || leftop == OP_NOTPROP)
3271          {
3272          if (rightop == OP_EOD)
3273            accepted = TRUE;
3274          else if (rightop == OP_PROP || rightop == OP_NOTPROP)
3275            {
3276            int n;
3277            const pcre_uint8 *p;
3278            BOOL same = leftop == rightop;
3279            BOOL lisprop = leftop == OP_PROP;
3280            BOOL risprop = rightop == OP_PROP;
3281            BOOL bothprop = lisprop && risprop;
3282    
3283            /* There's a table that specifies how each combination is to be
3284            processed:
3285              0   Always return FALSE (never auto-possessify)
3286              1   Character groups are distinct (possessify if both are OP_PROP)
3287              2   Check character categories in the same group (general or particular)
3288              3   Return TRUE if the two opcodes are not the same
3289              ... see comments below
3290            */
3291    
3292            n = propposstab[base_list[2]][list[2]];
3293            switch(n)
3294              {
3295              case 0: break;
3296              case 1: accepted = bothprop; break;
3297              case 2: accepted = (base_list[3] == list[3]) != same; break;
3298              case 3: accepted = !same; break;
3299    
3300              case 4:  /* Left general category, right particular category */
3301              accepted = risprop && catposstab[base_list[3]][list[3]] == same;
3302              break;
3303    
3304              case 5:  /* Right general category, left particular category */
3305              accepted = lisprop && catposstab[list[3]][base_list[3]] == same;
3306              break;
3307    
3308              /* This code is logically tricky. Think hard before fiddling with it.
3309              The posspropstab table has four entries per row. Each row relates to
3310              one of PCRE's special properties such as ALNUM or SPACE or WORD.
3311              Only WORD actually needs all four entries, but using repeats for the
3312              others means they can all use the same code below.
3313    
3314              The first two entries in each row are Unicode general categories, and
3315              apply always, because all the characters they include are part of the
3316              PCRE character set. The third and fourth entries are a general and a
3317              particular category, respectively, that include one or more relevant
3318              characters. One or the other is used, depending on whether the check
3319              is for a general or a particular category. However, in both cases the
3320              category contains more characters than the specials that are defined
3321              for the property being tested against. Therefore, it cannot be used
3322              in a NOTPROP case.
3323    
3324              Example: the row for WORD contains ucp_L, ucp_N, ucp_P, ucp_Po.
3325              Underscore is covered by ucp_P or ucp_Po. */
3326    
3327              case 6:  /* Left alphanum vs right general category */
3328              case 7:  /* Left space vs right general category */
3329              case 8:  /* Left word vs right general category */
3330              p = posspropstab[n-6];
3331              accepted = risprop && lisprop ==
3332                (list[3] != p[0] &&
3333                 list[3] != p[1] &&
3334                (list[3] != p[2] || !lisprop));
3335              break;
3336    
3337              case 9:   /* Right alphanum vs left general category */
3338              case 10:  /* Right space vs left general category */
3339              case 11:  /* Right word vs left general category */
3340              p = posspropstab[n-9];
3341              accepted = lisprop && risprop ==
3342                (base_list[3] != p[0] &&
3343                 base_list[3] != p[1] &&
3344                (base_list[3] != p[2] || !risprop));
3345              break;
3346    
3347              case 12:  /* Left alphanum vs right particular category */
3348              case 13:  /* Left space vs right particular category */
3349              case 14:  /* Left word vs right particular category */
3350              p = posspropstab[n-12];
3351              accepted = risprop && lisprop ==
3352                (catposstab[p[0]][list[3]] &&
3353                 catposstab[p[1]][list[3]] &&
3354                (list[3] != p[3] || !lisprop));
3355              break;
3356    
3357              case 15:  /* Right alphanum vs left particular category */
3358              case 16:  /* Right space vs left particular category */
3359              case 17:  /* Right word vs left particular category */
3360              p = posspropstab[n-15];
3361              accepted = lisprop && risprop ==
3362                (catposstab[p[0]][base_list[3]] &&
3363                 catposstab[p[1]][base_list[3]] &&
3364                (base_list[3] != p[3] || !risprop));
3365              break;
3366              }
3367            }
3368          }
3369    
3370        else
3371    #endif  /* SUPPORT_UCP */
3372    
3373        accepted = leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&
3374               rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&
3375               autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];
3376    
3377        if (!accepted)
3378          return FALSE;
3379    
3380        if (list[1] == 0) return TRUE;
3381        /* Might be an empty repeat. */
3382        continue;
3383        }
3384    
3385      /* Control reaches here only if one of the items is a small character list.
3386      All characters are checked against the other side. */
3387    
3388      do
3389        {
3390        chr = *chr_ptr;
3391    
3392        switch(list_ptr[0])
3393          {
3394          case OP_CHAR:
3395          ochr_ptr = list_ptr + 2;
3396          do
3397            {
3398            if (chr == *ochr_ptr) return FALSE;
3399            ochr_ptr++;
3400            }
3401          while(*ochr_ptr != NOTACHAR);
3402          break;
3403    
3404          case OP_NOT:
3405          ochr_ptr = list_ptr + 2;
3406          do
3407            {
3408            if (chr == *ochr_ptr)
3409              break;
3410            ochr_ptr++;
3411            }
3412          while(*ochr_ptr != NOTACHAR);
3413          if (*ochr_ptr == NOTACHAR) return FALSE;   /* Not found */
3414          break;
3415    
3416          /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not*
3417          set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3418    
3419          case OP_DIGIT:
3420          if (chr < 256 && (cd->ctypes[chr] & ctype_digit) != 0) return FALSE;
3421          break;
3422    
3423          case OP_NOT_DIGIT:
3424          if (chr > 255 || (cd->ctypes[chr] & ctype_digit) == 0) return FALSE;
3425          break;
3426    
3427          case OP_WHITESPACE:
3428          if (chr < 256 && (cd->ctypes[chr] & ctype_space) != 0) return FALSE;
3429          break;
3430    
3431          case OP_NOT_WHITESPACE:
3432          if (chr > 255 || (cd->ctypes[chr] & ctype_space) == 0) return FALSE;
3433          break;
3434    
3435          case OP_WORDCHAR:
3436          if (chr < 255 && (cd->ctypes[chr] & ctype_word) != 0) return FALSE;
3437          break;
3438    
3439          case OP_NOT_WORDCHAR:
3440          if (chr > 255 || (cd->ctypes[chr] & ctype_word) == 0) return FALSE;
3441          break;
3442    
3443          case OP_HSPACE:
3444          switch(chr)
3445            {
3446            HSPACE_CASES: return FALSE;
3447            default: break;
3448            }
3449          break;
3450    
3451          case OP_NOT_HSPACE:
3452          switch(chr)
3453            {
3454            HSPACE_CASES: break;
3455            default: return FALSE;
3456            }
3457          break;
3458    
3459          case OP_ANYNL:
3460          case OP_VSPACE:
3461          switch(chr)
3462            {
3463            VSPACE_CASES: return FALSE;
3464            default: break;
3465            }
3466          break;
3467    
3468          case OP_NOT_VSPACE:
3469          switch(chr)
3470            {
3471            VSPACE_CASES: break;
3472            default: return FALSE;
3473            }
3474          break;
3475    
3476          case OP_DOLL:
3477          case OP_EODN:
3478          switch (chr)
3479            {
3480            case CHAR_CR:
3481            case CHAR_LF:
3482            case CHAR_VT:
3483            case CHAR_FF:
3484            case CHAR_NEL:
3485    #ifndef EBCDIC
3486            case 0x2028:
3487            case 0x2029:
3488    #endif  /* Not EBCDIC */
3489            return FALSE;
3490            }
3491          break;
3492    
3493          case OP_EOD:    /* Can always possessify before \z */
3494          break;
3495    
3496    #ifdef SUPPORT_UCP
3497          case OP_PROP:
3498          case OP_NOTPROP:
3499          if (!check_char_prop(chr, list_ptr[2], list_ptr[3],
3500                list_ptr[0] == OP_NOTPROP))
3501            return FALSE;
3502          break;
3503    #endif
3504    
3505          case OP_NCLASS:
3506          if (chr > 255) return FALSE;
3507          /* Fall through */
3508    
3509          case OP_CLASS:
3510          if (chr > 255) break;
3511          class_bitset = (pcre_uint8 *)
3512            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3513          if ((class_bitset[chr >> 3] & (1 << (chr & 7))) != 0) return FALSE;
3514          break;
3515    
3516    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3517          case OP_XCLASS:
3518          if (PRIV(xclass)(chr, (list_ptr == list ? code : base_end) -
3519              list_ptr[2] + LINK_SIZE, utf)) return FALSE;
3520          break;
3521    #endif
3522    
3523          default:
3524          return FALSE;
3525          }
3526    
3527        chr_ptr++;
3528        }
3529      while(*chr_ptr != NOTACHAR);
3530    
3531      /* At least one character must be matched from this opcode. */
3532    
3533      if (list[1] == 0) return TRUE;
3534      }
3535    
3536    return FALSE;
3537    }
3538    
3539    
3540    
3541    /*************************************************
3542    *    Scan compiled regex for auto-possession     *
3543    *************************************************/
3544    
3545    /* Replaces single character iterations with their possessive alternatives
3546    if appropriate. This function modifies the compiled opcode!
3547    
3548    Arguments:
3549      code        points to start of the byte code
3550      utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3551      cd          static compile data
3552    
3553    Returns:      nothing
3554    */
3555    
3556    static void
3557    auto_possessify(pcre_uchar *code, BOOL utf, const compile_data *cd)
3558    {
3559    register pcre_uchar c;
3560    const pcre_uchar *end;
3561    pcre_uchar *repeat_opcode;
3562    pcre_uint32 list[8];
3563    
3564    for (;;)
3565      {
3566      c = *code;
3567    
3568      if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
3569        {
3570        c -= get_repeat_base(c) - OP_STAR;
3571        end = (c <= OP_MINUPTO) ?
3572          get_chr_property_list(code, utf, cd->fcc, list) : NULL;
3573        list[1] = c == OP_STAR || c == OP_PLUS || c == OP_QUERY || c == OP_UPTO;
3574    
3575        if (end != NULL && compare_opcodes(end, utf, cd, list, end))
3576          {
3577          switch(c)
3578            {
3579            case OP_STAR:
3580            *code += OP_POSSTAR - OP_STAR;
3581            break;
3582    
3583            case OP_MINSTAR:
3584            *code += OP_POSSTAR - OP_MINSTAR;
3585            break;
3586    
3587            case OP_PLUS:
3588            *code += OP_POSPLUS - OP_PLUS;
3589            break;
3590    
3591            case OP_MINPLUS:
3592            *code += OP_POSPLUS - OP_MINPLUS;
3593            break;
3594    
3595            case OP_QUERY:
3596            *code += OP_POSQUERY - OP_QUERY;
3597            break;
3598    
3599            case OP_MINQUERY:
3600            *code += OP_POSQUERY - OP_MINQUERY;
3601            break;
3602    
3603            case OP_UPTO:
3604            *code += OP_POSUPTO - OP_UPTO;
3605            break;
3606    
3607            case OP_MINUPTO:
3608            *code += OP_MINUPTO - OP_UPTO;
3609            break;
3610            }
3611          }
3612        c = *code;
3613        }
3614      else if (c == OP_CLASS || c == OP_NCLASS || c == OP_XCLASS)
3615        {
3616    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3617        if (c == OP_XCLASS)
3618          repeat_opcode = code + GET(code, 1);
3619        else
3620    #endif
3621          repeat_opcode = code + 1 + (32 / sizeof(pcre_uchar));
3622    
3623        c = *repeat_opcode;
3624        if (c >= OP_CRSTAR && c <= OP_CRMINRANGE)
3625          {
3626          /* end must not be NULL. */
3627          end = get_chr_property_list(code, utf, cd->fcc, list);
3628    
3629          list[1] = (c & 1) == 0;
3630    
3631          if (compare_opcodes(end, utf, cd, list, end))
3632            {
3633            switch (c)
3634              {
3635              case OP_CRSTAR:
3636              case OP_CRMINSTAR:
3637              *repeat_opcode = OP_CRPOSSTAR;
3638              break;
3639    
3640              case OP_CRPLUS:
3641              case OP_CRMINPLUS:
3642              *repeat_opcode = OP_CRPOSPLUS;
3643              break;
3644    
3645              case OP_CRQUERY:
3646              case OP_CRMINQUERY:
3647              *repeat_opcode = OP_CRPOSQUERY;
3648              break;
3649    
3650              case OP_CRRANGE:
3651              case OP_CRMINRANGE:
3652              *repeat_opcode = OP_CRPOSRANGE;
3653              break;
3654              }
3655            }
3656          }
3657        c = *code;
3658        }
3659    
3660      switch(c)
3661        {
3662        case OP_END:
3663        return;
3664    
3665        case OP_TYPESTAR:
3666        case OP_TYPEMINSTAR:
3667        case OP_TYPEPLUS:
3668        case OP_TYPEMINPLUS:
3669        case OP_TYPEQUERY:
3670        case OP_TYPEMINQUERY:
3671        case OP_TYPEPOSSTAR:
3672        case OP_TYPEPOSPLUS:
3673        case OP_TYPEPOSQUERY:
3674        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
3675        break;
3676    
3677        case OP_TYPEUPTO:
3678        case OP_TYPEMINUPTO:
3679        case OP_TYPEEXACT:
3680        case OP_TYPEPOSUPTO:
3681        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
3682          code += 2;
3683        break;
3684    
3685    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3686        case OP_XCLASS:
3687        code += GET(code, 1);
3688        break;
3689    #endif
3690    
3691        case OP_MARK:
3692        case OP_PRUNE_ARG:
3693        case OP_SKIP_ARG:
3694        case OP_THEN_ARG:
3695        code += code[1];
3696        break;
3697        }
3698    
3699      /* Add in the fixed length from the table */
3700    
3701      code += PRIV(OP_lengths)[c];
3702    
3703      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
3704      a multi-byte character. The length in the table is a minimum, so we have to
3705      arrange to skip the extra bytes. */
3706    
3707    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3708      if (utf) switch(c)
3709        {
3710        case OP_CHAR:
3711        case OP_CHARI:
3712        case OP_NOT:
3713        case OP_NOTI:
3714        case OP_STAR:
3715        case OP_MINSTAR:
3716        case OP_PLUS:
3717        case OP_MINPLUS:
3718        case OP_QUERY:
3719        case OP_MINQUERY:
3720        case OP_UPTO:
3721        case OP_MINUPTO:
3722        case OP_EXACT:
3723        case OP_POSSTAR:
3724        case OP_POSPLUS:
3725        case OP_POSQUERY:
3726        case OP_POSUPTO:
3727        case OP_STARI:
3728        case OP_MINSTARI:
3729        case OP_PLUSI:
3730        case OP_MINPLUSI:
3731        case OP_QUERYI:
3732        case OP_MINQUERYI:
3733        case OP_UPTOI:
3734        case OP_MINUPTOI:
3735        case OP_EXACTI:
3736        case OP_POSSTARI:
3737        case OP_POSPLUSI:
3738        case OP_POSQUERYI:
3739        case OP_POSUPTOI:
3740        case OP_NOTSTAR:
3741        case OP_NOTMINSTAR:
3742        case OP_NOTPLUS:
3743        case OP_NOTMINPLUS:
3744        case OP_NOTQUERY:
3745        case OP_NOTMINQUERY:
3746        case OP_NOTUPTO:
3747        case OP_NOTMINUPTO:
3748        case OP_NOTEXACT:
3749        case OP_NOTPOSSTAR:
3750        case OP_NOTPOSPLUS:
3751        case OP_NOTPOSQUERY:
3752        case OP_NOTPOSUPTO:
3753        case OP_NOTSTARI:
3754        case OP_NOTMINSTARI:
3755        case OP_NOTPLUSI:
3756        case OP_NOTMINPLUSI:
3757        case OP_NOTQUERYI:
3758        case OP_NOTMINQUERYI:
3759        case OP_NOTUPTOI:
3760        case OP_NOTMINUPTOI:
3761        case OP_NOTEXACTI:
3762        case OP_NOTPOSSTARI:
3763        case OP_NOTPOSPLUSI:
3764        case OP_NOTPOSQUERYI:
3765        case OP_NOTPOSUPTOI:
3766        if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
3767        break;
3768        }
3769    #else
3770      (void)(utf);  /* Keep compiler happy by referencing function argument */
3771    #endif
3772      }
3773    }
3774    
3775    
3776    
3777    /*************************************************
3778    *           Check for POSIX class syntax         *
3779    *************************************************/
3780    
3781    /* This function is called when the sequence "[:" or "[." or "[=" is
3782    encountered in a character class. It checks whether this is followed by a
3783    sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
3784    reach an unescaped ']' without the special preceding character, return FALSE.
3785    
3786    Originally, this function only recognized a sequence of letters between the
3787    terminators, but it seems that Perl recognizes any sequence of characters,
3788    though of course unknown POSIX names are subsequently rejected. Perl gives an
3789    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
3790    didn't consider this to be a POSIX class. Likewise for [:1234:].
3791    
3792    The problem in trying to be exactly like Perl is in the handling of escapes. We
3793    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
3794    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
3795    below handles the special case of \], but does not try to do any other escape
3796    processing. This makes it different from Perl for cases such as [:l\ower:]
3797    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
3798    "l\ower". This is a lesser evil than not diagnosing bad classes when Perl does,
3799    I think.
3800    
3801    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
3802    It seems that the appearance of a nested POSIX class supersedes an apparent
3803    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
3804    a digit.
3805    
3806    In Perl, unescaped square brackets may also appear as part of class names. For
3807    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
3808    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
3809    seem right at all. PCRE does not allow closing square brackets in POSIX class
3810    names.
3811    
3812    Arguments:
3813      ptr      pointer to the initial [
3814      endptr   where to return the end pointer
3815    
3816    Returns:   TRUE or FALSE
3817    */
3818    
3819    static BOOL
3820    check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
3821    {
3822    pcre_uchar terminator;          /* Don't combine these lines; the Solaris cc */
3823    terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
3824    for (++ptr; *ptr != CHAR_NULL; ptr++)
3825      {
3826      if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3827        ptr++;
3828      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
3829      else
3830        {
3831        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3832          {
3833          *endptr = ptr;
3834          return TRUE;
3835          }
3836        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
3837             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3838              ptr[1] == CHAR_EQUALS_SIGN) &&
3839            check_posix_syntax(ptr, endptr))
3840          return FALSE;
3841        }
3842      }
3843    return FALSE;
3844    }
3845    
3846    
3847    
3848    
3849    /*************************************************
3850    *          Check POSIX class name                *
3851    *************************************************/
3852    
3853    /* This function is called to check the name given in a POSIX-style class entry
3854    such as [:alnum:].
3855    
3856    Arguments:
3857      ptr        points to the first letter
3858      len        the length of the name
3859    
3860    Returns:     a value representing the name, or -1 if unknown
3861    */
3862    
3863    static int
3864    check_posix_name(const pcre_uchar *ptr, int len)
3865    {
3866    const char *pn = posix_names;
3867    register int yield = 0;
3868    while (posix_name_lengths[yield] != 0)
3869      {
3870      if (len == posix_name_lengths[yield] &&
3871        STRNCMP_UC_C8(ptr, pn, (unsigned int)len) == 0) return yield;
3872      pn += posix_name_lengths[yield] + 1;
3873      yield++;
3874      }
3875    return -1;
3876    }
3877    
3878    
3879    /*************************************************
3880    *    Adjust OP_RECURSE items in repeated group   *
3881    *************************************************/
3882    
3883    /* OP_RECURSE items contain an offset from the start of the regex to the group
3884    that is referenced. This means that groups can be replicated for fixed
3885    repetition simply by copying (because the recursion is allowed to refer to
3886    earlier groups that are outside the current group). However, when a group is
3887    optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is
3888    inserted before it, after it has been compiled. This means that any OP_RECURSE
3889    items within it that refer to the group itself or any contained groups have to
3890    have their offsets adjusted. That one of the jobs of this function. Before it
3891    is called, the partially compiled regex must be temporarily terminated with
3892    OP_END.
3893    
3894    This function has been extended with the possibility of forward references for
3895    recursions and subroutine calls. It must also check the list of such references
3896    for the group we are dealing with. If it finds that one of the recursions in
3897    the current group is on this list, it adjusts the offset in the list, not the
3898    value in the reference (which is a group number).
3899    
3900    Arguments:
3901      group      points to the start of the group
3902      adjust     the amount by which the group is to be moved
3903      utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode
3904      cd         contains pointers to tables etc.
3905      save_hwm   the hwm forward reference pointer at the start of the group
3906    
3907    Returns:     nothing
3908    */
3909    
3910    static void
3911    adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
3912      pcre_uchar *save_hwm)
3913    {
3914    pcre_uchar *ptr = group;
3915    
3916    while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
3917      {
3918      int offset;
3919      pcre_uchar *hc;
3920    
3921      /* See if this recursion is on the forward reference list. If so, adjust the
3922      reference. */
3923    
3924      for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
3925        {
3926        offset = (int)GET(hc, 0);
3927        if (cd->start_code + offset == ptr + 1)
3928          {
3929          PUT(hc, 0, offset + adjust);
3930          break;
3931          }
3932        }
3933    
3934      /* Otherwise, adjust the recursion offset if it's after the start of this
3935      group. */
3936    
3937      if (hc >= cd->hwm)
3938        {
3939        offset = (int)GET(ptr, 1);
3940        if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
3941        }
3942    
3943      ptr += 1 + LINK_SIZE;
3944      }
3945    }
3946    
3947    
3948    
3949    /*************************************************
3950    *        Insert an automatic callout point       *
3951    *************************************************/
3952    
3953    /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert
3954    callout points before each pattern item.
3955    
3956    Arguments:
3957      code           current code pointer
3958      ptr            current pattern pointer
3959      cd             pointers to tables etc
3960    
3961    Returns:         new code pointer
3962    */
3963    
3964    static pcre_uchar *
3965    auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
3966    {
3967    *code++ = OP_CALLOUT;
3968  *code++ = 255;  *code++ = 255;
3969  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
3970  PUT(code, LINK_SIZE, 0);                       /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
# Line 2871  PUT(previous_callout, 2 + LINK_SIZE, len Line 4004  PUT(previous_callout, 2 + LINK_SIZE, len
4004  *************************************************/  *************************************************/
4005    
4006  /* 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
4007  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
4008  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
4009  start address.  start address. A character with multiple other cases is returned on its own
4010    with a special return value.
4011    
4012  Arguments:  Arguments:
4013    cptr        points to starting character value; updated    cptr        points to starting character value; updated
# Line 2881  Arguments: Line 4015  Arguments:
4015    ocptr       where to put start of othercase range    ocptr       where to put start of othercase range
4016    odptr       where to put end of othercase range    odptr       where to put end of othercase range
4017    
4018  Yield:        TRUE when range returned; FALSE when no more  Yield:        -1 when no more
4019                   0 when a range is returned
4020                  >0 the CASESET offset for char with multiple other cases
4021                    in this case, ocptr contains the original
4022  */  */
4023    
4024  static BOOL  static int
4025  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,  get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,
4026    unsigned int *odptr)    pcre_uint32 *odptr)
4027  {  {
4028  unsigned int c, othercase, next;  pcre_uint32 c, othercase, next;
4029    unsigned int co;
4030    
4031    /* Find the first character that has an other case. If it has multiple other
4032    cases, return its case offset value. */
4033    
4034  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
4035    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }    {
4036      if ((co = UCD_CASESET(c)) != 0)
4037        {
4038        *ocptr = c++;   /* Character that has the set */
4039        *cptr = c;      /* Rest of input range */
4040        return (int)co;
4041        }
4042      if ((othercase = UCD_OTHERCASE(c)) != c) break;
4043      }
4044    
4045  if (c > d) return FALSE;  if (c > d) return -1;  /* Reached end of range */
4046    
4047  *ocptr = othercase;  *ocptr = othercase;
4048  next = othercase + 1;  next = othercase + 1;
# Line 2904  for (++c; c <= d; c++) Line 4053  for (++c; c <= d; c++)
4053    next++;    next++;
4054    }    }
4055    
4056  *odptr = next - 1;  *odptr = next - 1;     /* End of othercase range */
4057  *cptr = c;  *cptr = c;             /* Rest of input range */
4058    return 0;
 return TRUE;  
4059  }  }
4060    #endif  /* SUPPORT_UCP */
4061    
4062    
4063    
4064  /*************************************************  /*************************************************
4065  *        Check a character and a property        *  *        Add a character or range to a class     *
4066  *************************************************/  *************************************************/
4067    
4068  /* This function is called by check_auto_possessive() when a property item  /* This function packages up the logic of adding a character or range of
4069  is adjacent to a fixed character.  characters to a class. The character values in the arguments will be within the
4070    valid values for the current mode (8-bit, 16-bit, UTF, etc). This function is
4071    mutually recursive with the function immediately below.
4072    
4073  Arguments:  Arguments:
4074    c            the character    classbits     the bit map for characters < 256
4075    ptype        the property type    uchardptr     points to the pointer for extra data
4076    pdata        the data for the type    options       the options word
4077    negated      TRUE if it's a negated property (\P or \p{^)    cd            contains pointers to tables etc.
4078      start         start of range character
4079      end           end of range character
4080    
4081  Returns:       TRUE if auto-possessifying is OK  Returns:        the number of < 256 characters added
4082                    the pointer to extra data is updated
4083  */  */
4084    
4085  static BOOL  static int
4086  check_char_prop(int c, int ptype, int pdata, BOOL negated)  add_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
4087      compile_data *cd, pcre_uint32 start, pcre_uint32 end)
4088  {  {
4089  const ucd_record *prop = GET_UCD(c);  pcre_uint32 c;
4090  switch(ptype)  int n8 = 0;
   {  
   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;  
4091    
4092    case PT_PXSPACE:  /* POSIX space */  /* If caseless matching is required, scan the range and process alternate
4093    return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||  cases. In Unicode, there are 8-bit characters that have alternate cases that
4094            c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||  are greater than 255 and vice-versa. Sometimes we can just extend the original
4095            c == CHAR_FF || c == CHAR_CR)  range. */
           == negated;  
4096    
4097    case PT_WORD:  if ((options & PCRE_CASELESS) != 0)
4098    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||    {
4099            PRIV(ucp_gentype)[prop->chartype] == ucp_N ||  #ifdef SUPPORT_UCP
4100            c == CHAR_UNDERSCORE) == negated;    if ((options & PCRE_UTF8) != 0)
4101    }      {
4102  return FALSE;      int rc;
4103  }      pcre_uint32 oc, od;
 #endif  /* SUPPORT_UCP */  
4104    
4105        options &= ~PCRE_CASELESS;   /* Remove for recursive calls */
4106        c = start;
4107    
4108        while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)
4109          {
4110          /* Handle a single character that has more than one other case. */
4111    
4112  /*************************************************        if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,
4113  *     Check if auto-possessifying is possible    *          PRIV(ucd_caseless_sets) + rc, oc);
 *************************************************/  
4114    
4115  /* This function is called for unlimited repeats of certain items, to see        /* Do nothing if the other case range is within the original range. */
 whether the next thing could possibly match the repeated item. If not, it makes  
 sense to automatically possessify the repeated item.  
4116    
4117  Arguments:        else if (oc >= start && od <= end) continue;
   previous      pointer to the repeated opcode  
   utf           TRUE in UTF-8 / UTF-16 mode  
   ptr           next character in pattern  
   options       options bits  
   cd            contains pointers to tables etc.  
4118    
4119  Returns:        TRUE if possessifying is wanted        /* Extend the original range if there is overlap, noting that if oc < c, we
4120  */        can't have od > end because a subrange is always shorter than the basic
4121          range. Otherwise, use a recursive call to add the additional range. */
4122    
4123  static BOOL        else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */
4124  check_auto_possessive(const pcre_uchar *previous, BOOL utf,        else if (od > end && oc <= end + 1) end = od;       /* Extend upwards */
4125    const pcre_uchar *ptr, int options, compile_data *cd)        else n8 += add_to_class(classbits, uchardptr, options, cd, oc, od);
4126  {        }
4127  pcre_int32 c, next;      }
4128  int op_code = *previous++;    else
4129    #endif  /* SUPPORT_UCP */
4130    
4131  /* Skip whitespace and comments in extended mode */    /* Not UTF-mode, or no UCP */
4132    
4133  if ((options & PCRE_EXTENDED) != 0)    for (c = start; c <= end && c < 256; c++)
   {  
   for (;;)  
4134      {      {
4135      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      SETBIT(classbits, cd->fcc[c]);
4136      if (*ptr == CHAR_NUMBER_SIGN)      n8++;
       {  
       ptr++;  
       while (*ptr != 0)  
         {  
         if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }  
         ptr++;  
 #ifdef SUPPORT_UTF  
         if (utf) FORWARDCHAR(ptr);  
 #endif  
         }  
       }  
     else break;  
4137      }      }
4138    }    }
4139    
4140  /* If the next item is one that we can handle, get its value. A non-negative  /* Now handle the original range. Adjust the final value according to the bit
4141  value is a character, a negative value is an escape value. */  length - this means that the same lists of (e.g.) horizontal spaces can be used
4142    in all cases. */
4143    
4144  if (*ptr == CHAR_BACKSLASH)  #if defined COMPILE_PCRE8
   {  
   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)  
   {  
4145  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4146    if (utf) { GETCHARINC(next, ptr); } else    if ((options & PCRE_UTF8) == 0)
4147  #endif  #endif
4148    next = *ptr++;    if (end > 0xff) end = 0xff;
   }  
 else return FALSE;  
4149    
4150  /* Skip whitespace and comments in extended mode */  #elif defined COMPILE_PCRE16
   
 if ((options & PCRE_EXTENDED) != 0)  
   {  
   for (;;)  
     {  
     while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;  
     if (*ptr == CHAR_NUMBER_SIGN)  
       {  
       ptr++;  
       while (*ptr != 0)  
         {  
         if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }  
         ptr++;  
4151  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4152          if (utf) FORWARDCHAR(ptr);    if ((options & PCRE_UTF16) == 0)
4153  #endif  #endif
4154          }    if (end > 0xffff) end = 0xffff;
       }  
     else break;  
     }  
   }  
   
 /* If the next thing is itself optional, we have to give up. */  
4155    
4156  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||  #endif /* COMPILE_PCRE[8|16] */
   STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)  
     return FALSE;  
4157    
4158  /* Now compare the next item with the previous opcode. First, handle cases when  /* If all characters are less than 256, use the bit map. Otherwise use extra
4159  the next item is a character. */  data. */
4160    
4161  if (next >= 0) switch(op_code)  if (end < 0x100)
4162    {    {
4163    case OP_CHAR:    for (c = start; c <= end; c++)
 #ifdef SUPPORT_UTF  
   GETCHARTEST(c, previous);  
 #else  
   c = *previous;  
 #endif  
   return c != next;  
   
   /* For CHARI (caseless character) we must check the other case. If we have  
   Unicode property support, we can use it to test the other case of  
   high-valued characters. */  
   
   case OP_CHARI:  
 #ifdef SUPPORT_UTF  
   GETCHARTEST(c, previous);  
 #else  
   c = *previous;  
 #endif  
   if (c == next) return FALSE;  
 #ifdef SUPPORT_UTF  
   if (utf)  
4164      {      {
4165      unsigned int othercase;      n8++;
4166      if (next < 128) othercase = cd->fcc[next]; else      SETBIT(classbits, c);
 #ifdef SUPPORT_UCP  
     othercase = UCD_OTHERCASE((unsigned int)next);  
 #else  
     othercase = NOTACHAR;  
 #endif  
     return (unsigned int)c != othercase;  
4167      }      }
4168    else    }
 #endif  /* SUPPORT_UTF */  
   return (c != TABLE_GET((unsigned int)next, cd->fcc, next));  /* Non-UTF-8 mode */  
4169    
4170    case OP_NOT:  else
4171  #ifdef SUPPORT_UTF    {
4172    GETCHARTEST(c, previous);    pcre_uchar *uchardata = *uchardptr;
 #else  
   c = *previous;  
 #endif  
   return c == next;  
4173    
   case OP_NOTI:  
 #ifdef SUPPORT_UTF  
   GETCHARTEST(c, previous);  
 #else  
   c = *previous;  
 #endif  
   if (c == next) return TRUE;  
4174  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4175    if (utf)    if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */
4176      {      {
4177      unsigned int othercase;      if (start < end)
4178      if (next < 128) othercase = cd->fcc[next]; else        {
4179  #ifdef SUPPORT_UCP        *uchardata++ = XCL_RANGE;
4180      othercase = UCD_OTHERCASE((unsigned int)next);        uchardata += PRIV(ord2utf)(start, uchardata);
4181  #else        uchardata += PRIV(ord2utf)(end, uchardata);
4182      othercase = NOTACHAR;        }
4183  #endif      else if (start == end)
4184      return (unsigned int)c == othercase;        {
4185          *uchardata++ = XCL_SINGLE;
4186          uchardata += PRIV(ord2utf)(start, uchardata);
4187          }
4188      }      }
4189    else    else
4190  #endif  /* SUPPORT_UTF */  #endif  /* SUPPORT_UTF */
   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;  
4191    
4192    case OP_WHITESPACE:    /* Without UTF support, character values are constrained by the bit length,
4193    return next > 255 || (cd->ctypes[next] & ctype_space) == 0;    and can only be > 256 for 16-bit and 32-bit libraries. */
   
   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;  
4194    
4195    case OP_HSPACE:  #ifdef COMPILE_PCRE8
4196    case OP_NOT_HSPACE:      {}
4197    switch(next)  #else
4198      if (start < end)
4199      {      {
4200      case 0x09:      *uchardata++ = XCL_RANGE;
4201      case 0x20:      *uchardata++ = start;
4202      case 0xa0:      *uchardata++ = end;
     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;  
4203      }      }
4204      else if (start == end)
   case OP_ANYNL:  
   case OP_VSPACE:  
   case OP_NOT_VSPACE:  
   switch(next)  
4205      {      {
4206      case 0x0a:      *uchardata++ = XCL_SINGLE;
4207      case 0x0b:      *uchardata++ = start;
     case 0x0c:  
     case 0x0d:  
     case 0x85:  
     case 0x2028:  
     case 0x2029:  
     return op_code == OP_NOT_VSPACE;  
     default:  
     return op_code != OP_NOT_VSPACE;  
4208      }      }
   
 #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);  
4209  #endif  #endif
4210    
4211    default:    *uchardptr = uchardata;   /* Updata extra data pointer */
   return FALSE;  
4212    }    }
4213    
4214    return n8;    /* Number of 8-bit characters */
4215    }
4216    
 /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP  
 is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are  
 generated only when PCRE_UCP is *not* set, that is, when only ASCII  
 characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are  
 replaced by OP_PROP codes when PCRE_UCP is set. */  
   
 switch(op_code)  
   {  
   case OP_CHAR:  
   case OP_CHARI:  
 #ifdef SUPPORT_UTF  
   GETCHARTEST(c, previous);  
 #else  
   c = *previous;  
 #endif  
   switch(-next)  
     {  
     case ESC_d:  
     return c > 255 || (cd->ctypes[c] & ctype_digit) == 0;  
   
     case ESC_D:  
     return c <= 255 && (cd->ctypes[c] & ctype_digit) != 0;  
   
     case ESC_s:  
     return c > 255 || (cd->ctypes[c] & ctype_space) == 0;  
   
     case ESC_S:  
     return c <= 255 && (cd->ctypes[c] & ctype_space) != 0;  
   
     case ESC_w:  
     return c > 255 || (cd->ctypes[c] & ctype_word) == 0;  
   
     case ESC_W:  
     return c <= 255 && (cd->ctypes[c] & ctype_word) != 0;  
   
     case ESC_h:  
     case ESC_H:  
     switch(c)  
       {  
       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;  
       }  
   
     case ESC_v:  
     case ESC_V:  
     switch(c)  
       {  
       case 0x0a:  
       case 0x0b:  
       case 0x0c:  
       case 0x0d:  
       case 0x85:  
       case 0x2028:  
       case 0x2029:  
       return -next != ESC_v;  
       default:  
       return -next == ESC_v;  
       }  
   
     /* When PCRE_UCP is set, these values get generated for \d etc. Find  
     their substitutions and process them. The result will always be either  
     -ESC_p or -ESC_P. Then fall through to process those values. */  
   
 #ifdef SUPPORT_UCP  
     case ESC_du:  
     case ESC_DU:  
     case ESC_wu:  
     case ESC_WU:  
     case ESC_su:  
     case ESC_SU:  
       {  
       int temperrorcode = 0;  
       ptr = substitutes[-next - ESC_DU];  
       next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);  
       if (temperrorcode != 0) return FALSE;  
       ptr++;    /* For compatibility */  
       }  
     /* Fall through */  
   
     case ESC_p:  
     case ESC_P:  
       {  
       int ptype, pdata, errorcodeptr;  
       BOOL negated;  
   
       ptr--;      /* Make ptr point at the p or P */  
       ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);  
       if (ptype < 0) return FALSE;  
       ptr++;      /* Point past the final curly ket */  
   
       /* If the property item is optional, we have to give up. (When generated  
       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. */  
   
       if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||  
         STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)  
           return FALSE;  
   
       /* Do the property check. */  
4217    
       return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);  
       }  
 #endif  
4218    
     default:  
     return FALSE;  
     }  
4219    
4220    /* In principle, support for Unicode properties should be integrated here as  /*************************************************
4221    well. It means re-organizing the above code so as to get hold of the property  *        Add a list of characters to a class     *
4222    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.) */  
4223    
4224    case OP_DIGIT:  /* This function is used for adding a list of case-equivalent characters to a
4225    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||  class, and also for adding a list of horizontal or vertical whitespace. If the
4226           next == -ESC_h || next == -ESC_v || next == -ESC_R;  list is in order (which it should be), ranges of characters are detected and
4227    handled appropriately. This function is mutually recursive with the function
4228    above.
4229    
4230    case OP_NOT_DIGIT:  Arguments:
4231    return next == -ESC_d;    classbits     the bit map for characters < 256
4232      uchardptr     points to the pointer for extra data
4233      options       the options word
4234      cd            contains pointers to tables etc.
4235      p             points to row of 32-bit values, terminated by NOTACHAR
4236      except        character to omit; this is used when adding lists of
4237                      case-equivalent characters to avoid including the one we
4238                      already know about
4239    
4240    case OP_WHITESPACE:  Returns:        the number of < 256 characters added
4241    return next == -ESC_S || next == -ESC_d || next == -ESC_w;                  the pointer to extra data is updated
4242    */
4243    
4244    case OP_NOT_WHITESPACE:  static int
4245    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,
4246      compile_data *cd, const pcre_uint32 *p, unsigned int except)
4247    {
4248    int n8 = 0;
4249    while (p[0] < NOTACHAR)
4250      {
4251      int n = 0;
4252      if (p[0] != except)
4253        {
4254        while(p[n+1] == p[0] + n + 1) n++;
4255        n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);
4256        }
4257      p += n + 1;
4258      }
4259    return n8;
4260    }
4261    
   case OP_HSPACE:  
   return next == -ESC_S || next == -ESC_H || next == -ESC_d ||  
          next == -ESC_w || next == -ESC_v || next == -ESC_R;  
4262    
   case OP_NOT_HSPACE:  
   return next == -ESC_h;  
4263    
4264    /* Can't have \S in here because VT matches \S (Perl anomaly) */  /*************************************************
4265    case OP_ANYNL:  *    Add characters not in a list to a class     *
4266