/[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 1063 by chpe, Tue Oct 16 15:54:09 2012 UTC revision 1608 by ph10, Fri Nov 20 17:34:16 2015 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) 1997-2014 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 47  supporting internal functions that are n Line 47  supporting internal functions that are n
47  #endif  #endif
48    
49  #define NLBLOCK cd             /* Block containing newline information */  #define NLBLOCK cd             /* Block containing newline information */
50  #define PSSTART start_pattern  /* Field containing processed string start */  #define PSSTART start_pattern  /* Field containing pattern start */
51  #define PSEND   end_pattern    /* Field containing processed string end */  #define PSEND   end_pattern    /* Field containing pattern end */
52    
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
# Line 80  to check them every time. */ Line 80  to check them every time. */
80  /* Definitions to allow mutual recursion */  /* Definitions to allow mutual recursion */
81    
82  static int  static int
83    add_list_to_class(pcre_uint8 *, pcre_uchar **, int, compile_data *,    add_list_to_class(pcre_uint8 *, pcre_uchar **, int, compile_data *,
84      const pcre_uint32 *, unsigned int);      const pcre_uint32 *, unsigned int);
85    
86  static BOOL  static BOOL
87    compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL,    compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL, int, int,
88      int, int, int *, int *, branch_chain *, compile_data *, int *);      pcre_uint32 *, pcre_int32 *, pcre_uint32 *, pcre_int32 *, branch_chain *,
89        compile_data *, int *);
90    
91    
92    
# Line 114  kicks in at the same number of forward r Line 115  kicks in at the same number of forward r
115  #define COMPILE_WORK_SIZE (2048*LINK_SIZE)  #define COMPILE_WORK_SIZE (2048*LINK_SIZE)
116  #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)  #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)
117    
118    /* This value determines the size of the initial vector that is used for
119    remembering named groups during the pre-compile. It is allocated on the stack,
120    but if it is too small, it is expanded using malloc(), in a similar way to the
121    workspace. The value is the number of slots in the list. */
122    
123    #define NAMED_GROUP_LIST_SIZE  20
124    
125  /* The overrun tests check for a slightly smaller size so that they detect the  /* The overrun tests check for a slightly smaller size so that they detect the
126  overrun before it actually does run off the end of the data block. */  overrun before it actually does run off the end of the data block. */
127    
# Line 121  overrun before it actually does run off Line 129  overrun before it actually does run off
129    
130  /* Private flags added to firstchar and reqchar. */  /* Private flags added to firstchar and reqchar. */
131    
132  #define REQ_CASELESS   0x10000000l      /* Indicates caselessness */  #define REQ_CASELESS    (1 << 0)        /* Indicates caselessness */
133  #define REQ_VARY       0x20000000l      /* Reqchar followed non-literal item */  #define REQ_VARY        (1 << 1)        /* Reqchar followed non-literal item */
134  #define REQ_MASK       (REQ_CASELESS | REQ_VARY)  /* Negative values for the firstchar and reqchar flags */
135    #define REQ_UNSET       (-2)
136    #define REQ_NONE        (-1)
137    
138  /* Repeated character flags. */  /* Repeated character flags. */
139    
# Line 164  static const short int escapes[] = { Line 174  static const short int escapes[] = {
174       -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,       -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,
175       CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,       CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,
176       CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,       CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,
177       CHAR_GRAVE_ACCENT,       7,       CHAR_GRAVE_ACCENT,       ESC_a,
178       -ESC_b,                  0,       -ESC_b,                  0,
179       -ESC_d,                  ESC_e,       -ESC_d,                  ESC_e,
180       ESC_f,                   0,       ESC_f,                   0,
# Line 192  static const short int escapes[] = { Line 202  static const short int escapes[] = {
202  /*  68 */     0,     0,    '|',     ',',    '%',   '_',    '>',    '?',  /*  68 */     0,     0,    '|',     ',',    '%',   '_',    '>',    '?',
203  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,
204  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',
205  /*  80 */     0,     7, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,  /*  80 */     0, ESC_a, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,
206  /*  88 */-ESC_h,     0,      0,     '{',      0,     0,      0,      0,  /*  88 */-ESC_h,     0,      0,     '{',      0,     0,      0,      0,
207  /*  90 */     0,     0, -ESC_k,     'l',      0, ESC_n,      0, -ESC_p,  /*  90 */     0,     0, -ESC_k,       0,      0, ESC_n,      0, -ESC_p,
208  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,
209  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,-ESC_v, -ESC_w,      0,  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,-ESC_v, -ESC_w,      0,
210  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,
# Line 209  static const short int escapes[] = { Line 219  static const short int escapes[] = {
219  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,
220  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0
221  };  };
222    
223    /* We also need a table of characters that may follow \c in an EBCDIC
224    environment for characters 0-31. */
225    
226    static unsigned char ebcdic_escape_c[] = "@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\\]^_";
227    
228  #endif  #endif
229    
230    
# Line 250  static const verbitem verbs[] = { Line 266  static const verbitem verbs[] = {
266  static const int verbcount = sizeof(verbs)/sizeof(verbitem);  static const int verbcount = sizeof(verbs)/sizeof(verbitem);
267    
268    
269    /* Substitutes for [[:<:]] and [[:>:]], which mean start and end of word in
270    another regex library. */
271    
272    static const pcre_uchar sub_start_of_word[] = {
273      CHAR_BACKSLASH, CHAR_b, CHAR_LEFT_PARENTHESIS, CHAR_QUESTION_MARK,
274      CHAR_EQUALS_SIGN, CHAR_BACKSLASH, CHAR_w, CHAR_RIGHT_PARENTHESIS, '\0' };
275    
276    static const pcre_uchar sub_end_of_word[] = {
277      CHAR_BACKSLASH, CHAR_b, CHAR_LEFT_PARENTHESIS, CHAR_QUESTION_MARK,
278      CHAR_LESS_THAN_SIGN, CHAR_EQUALS_SIGN, CHAR_BACKSLASH, CHAR_w,
279      CHAR_RIGHT_PARENTHESIS, '\0' };
280    
281    
282  /* Tables of names of POSIX character classes and their lengths. The names are  /* Tables of names of POSIX character classes and their lengths. The names are
283  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
284  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
285  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
286  for handling case independence. */  for handling case independence. The indices for graph, print, and punct are
287    needed, so identify them. */
288    
289  static const char posix_names[] =  static const char posix_names[] =
290    STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0    STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
# Line 265  static const char posix_names[] = Line 295  static const char posix_names[] =
295  static const pcre_uint8 posix_name_lengths[] = {  static const pcre_uint8 posix_name_lengths[] = {
296    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 };
297    
298    #define PC_GRAPH  8
299    #define PC_PRINT  9
300    #define PC_PUNCT 10
301    
302    
303  /* 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
304  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
305  classes, there is some additional tweaking: for [:blank:] the vertical space  classes, there is some additional tweaking: for [:blank:] the vertical space
# Line 292  static const int posix_class_maps[] = { Line 327  static const int posix_class_maps[] = {
327    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
328  };  };
329    
330  /* 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
331  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. */  
332    
333  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
334  static const pcre_uchar string_PNd[]  = {  static const pcre_uchar string_PNd[]  = {
# Line 319  static const pcre_uchar string_pXwd[] = Line 353  static const pcre_uchar string_pXwd[] =
353  static const pcre_uchar *substitutes[] = {  static const pcre_uchar *substitutes[] = {
354    string_PNd,           /* \D */    string_PNd,           /* \D */
355    string_pNd,           /* \d */    string_pNd,           /* \d */
356    string_PXsp,          /* \S */       /* NOTE: Xsp is Perl space */    string_PXsp,          /* \S */   /* Xsp is Perl space, but from 8.34, Perl */
357    string_pXsp,          /* \s */    string_pXsp,          /* \s */   /* space and POSIX space are the same. */
358    string_PXwd,          /* \W */    string_PXwd,          /* \W */
359    string_pXwd           /* \w */    string_pXwd           /* \w */
360  };  };
361    
362    /* The POSIX class substitutes must be in the order of the POSIX class names,
363    defined above, and there are both positive and negative cases. NULL means no
364    general substitute of a Unicode property escape (\p or \P). However, for some
365    POSIX classes (e.g. graph, print, punct) a special property code is compiled
366    directly. */
367    
368  static const pcre_uchar string_pL[] =   {  static const pcre_uchar string_pL[] =   {
369    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
370    CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };    CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
# Line 372  static const pcre_uchar *posix_substitut Line 412  static const pcre_uchar *posix_substitut
412    NULL,                 /* graph */    NULL,                 /* graph */
413    NULL,                 /* print */    NULL,                 /* print */
414    NULL,                 /* punct */    NULL,                 /* punct */
415    string_pXps,          /* space */    /* NOTE: Xps is POSIX space */    string_pXps,          /* space */   /* Xps is POSIX space, but from 8.34 */
416    string_pXwd,          /* word */    string_pXwd,          /* word  */   /* Perl and POSIX space are the same */
417    NULL,                 /* xdigit */    NULL,                 /* xdigit */
418    /* Negated cases */    /* Negated cases */
419    string_PL,            /* ^alpha */    string_PL,            /* ^alpha */
# Line 387  static const pcre_uchar *posix_substitut Line 427  static const pcre_uchar *posix_substitut
427    NULL,                 /* ^graph */    NULL,                 /* ^graph */
428    NULL,                 /* ^print */    NULL,                 /* ^print */
429    NULL,                 /* ^punct */    NULL,                 /* ^punct */
430    string_PXps,          /* ^space */   /* NOTE: Xps is POSIX space */    string_PXps,          /* ^space */  /* Xps is POSIX space, but from 8.34 */
431    string_PXwd,          /* ^word */    string_PXwd,          /* ^word */   /* Perl and POSIX space are the same */
432    NULL                  /* ^xdigit */    NULL                  /* ^xdigit */
433  };  };
434  #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))  #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))
# Line 424  static const char error_texts[] = Line 464  static const char error_texts[] =
464    "range out of order in character class\0"    "range out of order in character class\0"
465    "nothing to repeat\0"    "nothing to repeat\0"
466    /* 10 */    /* 10 */
467    "operand of unlimited repeat could match the empty string\0"  /** DEAD **/    "internal error: invalid forward reference offset\0"
468    "internal error: unexpected repeat\0"    "internal error: unexpected repeat\0"
469    "unrecognized character after (? or (?-\0"    "unrecognized character after (? or (?-\0"
470    "POSIX named classes are supported only within a class\0"    "POSIX named classes are supported only within a class\0"
# Line 452  static const char error_texts[] = Line 492  static const char error_texts[] =
492    "POSIX collating elements are not supported\0"    "POSIX collating elements are not supported\0"
493    "this version of PCRE is compiled without UTF support\0"    "this version of PCRE is compiled without UTF support\0"
494    "spare error\0"  /** DEAD **/    "spare error\0"  /** DEAD **/
495    "character value in \\x{...} sequence is too large\0"    "character value in \\x{} or \\o{} is too large\0"
496    /* 35 */    /* 35 */
497    "invalid condition (?(0)\0"    "invalid condition (?(0)\0"
498    "\\C not allowed in lookbehind assertion\0"    "\\C not allowed in lookbehind assertion\0"
# Line 484  static const char error_texts[] = Line 524  static const char error_texts[] =
524    "a numbered reference must not be zero\0"    "a numbered reference must not be zero\0"
525    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
526    /* 60 */    /* 60 */
527    "(*VERB) not recognized\0"    "(*VERB) not recognized or malformed\0"
528    "number is too big\0"    "number is too big\0"
529    "subpattern name expected\0"    "subpattern name expected\0"
530    "digit expected after (?+\0"    "digit expected after (?+\0"
# Line 493  static const char error_texts[] = Line 533  static const char error_texts[] =
533    "different names for subpatterns of the same number are not allowed\0"    "different names for subpatterns of the same number are not allowed\0"
534    "(*MARK) must have an argument\0"    "(*MARK) must have an argument\0"
535    "this version of PCRE is not compiled with Unicode property support\0"    "this version of PCRE is not compiled with Unicode property support\0"
536    #ifndef EBCDIC
537    "\\c must be followed by an ASCII character\0"    "\\c must be followed by an ASCII character\0"
538    #else
539      "\\c must be followed by a letter or one of [\\]^_?\0"
540    #endif
541    "\\k is not followed by a braced, angle-bracketed, or quoted name\0"    "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
542    /* 70 */    /* 70 */
543    "internal error: unknown opcode in find_fixedlength()\0"    "internal error: unknown opcode in find_fixedlength()\0"
# Line 505  static const char error_texts[] = Line 549  static const char error_texts[] =
549    "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"    "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"
550    "character value in \\u.... sequence is too large\0"    "character value in \\u.... sequence is too large\0"
551    "invalid UTF-32 string\0"    "invalid UTF-32 string\0"
552      "setting UTF is disabled by the application\0"
553      "non-hex character in \\x{} (closing brace missing?)\0"
554      /* 80 */
555      "non-octal character in \\o{} (closing brace missing?)\0"
556      "missing opening brace after \\o\0"
557      "parentheses are too deeply nested\0"
558      "invalid range in character class\0"
559      "group name must start with a non-digit\0"
560      /* 85 */
561      "parentheses are too deeply nested (stack check)\0"
562      "digits missing in \\x{} or \\o{}\0"
563    ;    ;
564    
565  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
# Line 644  static const pcre_uint8 ebcdic_chartab[] Line 699  static const pcre_uint8 ebcdic_chartab[]
699  #endif  #endif
700    
701    
702    /* This table is used to check whether auto-possessification is possible
703    between adjacent character-type opcodes. The left-hand (repeated) opcode is
704    used to select the row, and the right-hand opcode is use to select the column.
705    A value of 1 means that auto-possessification is OK. For example, the second
706    value in the first row means that \D+\d can be turned into \D++\d.
707    
708    The Unicode property types (\P and \p) have to be present to fill out the table
709    because of what their opcode values are, but the table values should always be
710    zero because property types are handled separately in the code. The last four
711    columns apply to items that cannot be repeated, so there is no need to have
712    rows for them. Note that OP_DIGIT etc. are generated only when PCRE_UCP is
713    *not* set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
714    
715    #define APTROWS (LAST_AUTOTAB_LEFT_OP - FIRST_AUTOTAB_OP + 1)
716    #define APTCOLS (LAST_AUTOTAB_RIGHT_OP - FIRST_AUTOTAB_OP + 1)
717    
718    static const pcre_uint8 autoposstab[APTROWS][APTCOLS] = {
719    /* \D \d \S \s \W \w  . .+ \C \P \p \R \H \h \V \v \X \Z \z  $ $M */
720      { 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \D */
721      { 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \d */
722      { 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \S */
723      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \s */
724      { 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \W */
725      { 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \w */
726      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .  */
727      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .+ */
728      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \C */
729      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \P */
730      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \p */
731      { 0, 1, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \R */
732      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \H */
733      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \h */
734      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \V */
735      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0 },  /* \v */
736      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }   /* \X */
737    };
738    
739    
740    /* This table is used to check whether auto-possessification is possible
741    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP). The
742    left-hand (repeated) opcode is used to select the row, and the right-hand
743    opcode is used to select the column. The values are as follows:
744    
745      0   Always return FALSE (never auto-possessify)
746      1   Character groups are distinct (possessify if both are OP_PROP)
747      2   Check character categories in the same group (general or particular)
748      3   TRUE if the two opcodes are not the same (PROP vs NOTPROP)
749    
750      4   Check left general category vs right particular category
751      5   Check right general category vs left particular category
752    
753      6   Left alphanum vs right general category
754      7   Left space vs right general category
755      8   Left word vs right general category
756    
757      9   Right alphanum vs left general category
758     10   Right space vs left general category
759     11   Right word vs left general category
760    
761     12   Left alphanum vs right particular category
762     13   Left space vs right particular category
763     14   Left word vs right particular category
764    
765     15   Right alphanum vs left particular category
766     16   Right space vs left particular category
767     17   Right word vs left particular category
768    */
769    
770    static const pcre_uint8 propposstab[PT_TABSIZE][PT_TABSIZE] = {
771    /* ANY LAMP GC  PC  SC ALNUM SPACE PXSPACE WORD CLIST UCNC */
772      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_ANY */
773      { 0,  3,  0,  0,  0,    3,    1,      1,   0,    0,   0 },  /* PT_LAMP */
774      { 0,  0,  2,  4,  0,    9,   10,     10,  11,    0,   0 },  /* PT_GC */
775      { 0,  0,  5,  2,  0,   15,   16,     16,  17,    0,   0 },  /* PT_PC */
776      { 0,  0,  0,  0,  2,    0,    0,      0,   0,    0,   0 },  /* PT_SC */
777      { 0,  3,  6, 12,  0,    3,    1,      1,   0,    0,   0 },  /* PT_ALNUM */
778      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_SPACE */
779      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_PXSPACE */
780      { 0,  0,  8, 14,  0,    0,    1,      1,   3,    0,   0 },  /* PT_WORD */
781      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_CLIST */
782      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   3 }   /* PT_UCNC */
783    };
784    
785    /* This table is used to check whether auto-possessification is possible
786    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP) when one
787    specifies a general category and the other specifies a particular category. The
788    row is selected by the general category and the column by the particular
789    category. The value is 1 if the particular category is not part of the general
790    category. */
791    
792    static const pcre_uint8 catposstab[7][30] = {
793    /* 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 */
794      { 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 */
795      { 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 */
796      { 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 */
797      { 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 */
798      { 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 */
799      { 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 */
800      { 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 */
801    };
802    
803    /* This table is used when checking ALNUM, (PX)SPACE, SPACE, and WORD against
804    a general or particular category. The properties in each row are those
805    that apply to the character set in question. Duplication means that a little
806    unnecessary work is done when checking, but this keeps things much simpler
807    because they can all use the same code. For more details see the comment where
808    this table is used.
809    
810    Note: SPACE and PXSPACE used to be different because Perl excluded VT from
811    "space", but from Perl 5.18 it's included, so both categories are treated the
812    same here. */
813    
814    static const pcre_uint8 posspropstab[3][4] = {
815      { ucp_L, ucp_N, ucp_N, ucp_Nl },  /* ALNUM, 3rd and 4th values redundant */
816      { ucp_Z, ucp_Z, ucp_C, ucp_Cc },  /* SPACE and PXSPACE, 2nd value redundant */
817      { ucp_L, ucp_N, ucp_P, ucp_Po }   /* WORD */
818    };
819    
820    /* This table is used when converting repeating opcodes into possessified
821    versions as a result of an explicit possessive quantifier such as ++. A zero
822    value means there is no possessified version - in those cases the item in
823    question must be wrapped in ONCE brackets. The table is truncated at OP_CALLOUT
824    because all relevant opcodes are less than that. */
825    
826    static const pcre_uint8 opcode_possessify[] = {
827      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 0 - 15  */
828      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 16 - 31 */
829    
830      0,                       /* NOTI */
831      OP_POSSTAR, 0,           /* STAR, MINSTAR */
832      OP_POSPLUS, 0,           /* PLUS, MINPLUS */
833      OP_POSQUERY, 0,          /* QUERY, MINQUERY */
834      OP_POSUPTO, 0,           /* UPTO, MINUPTO */
835      0,                       /* EXACT */
836      0, 0, 0, 0,              /* POS{STAR,PLUS,QUERY,UPTO} */
837    
838      OP_POSSTARI, 0,          /* STARI, MINSTARI */
839      OP_POSPLUSI, 0,          /* PLUSI, MINPLUSI */
840      OP_POSQUERYI, 0,         /* QUERYI, MINQUERYI */
841      OP_POSUPTOI, 0,          /* UPTOI, MINUPTOI */
842      0,                       /* EXACTI */
843      0, 0, 0, 0,              /* POS{STARI,PLUSI,QUERYI,UPTOI} */
844    
845      OP_NOTPOSSTAR, 0,        /* NOTSTAR, NOTMINSTAR */
846      OP_NOTPOSPLUS, 0,        /* NOTPLUS, NOTMINPLUS */
847      OP_NOTPOSQUERY, 0,       /* NOTQUERY, NOTMINQUERY */
848      OP_NOTPOSUPTO, 0,        /* NOTUPTO, NOTMINUPTO */
849      0,                       /* NOTEXACT */
850      0, 0, 0, 0,              /* NOTPOS{STAR,PLUS,QUERY,UPTO} */
851    
852      OP_NOTPOSSTARI, 0,       /* NOTSTARI, NOTMINSTARI */
853      OP_NOTPOSPLUSI, 0,       /* NOTPLUSI, NOTMINPLUSI */
854      OP_NOTPOSQUERYI, 0,      /* NOTQUERYI, NOTMINQUERYI */
855      OP_NOTPOSUPTOI, 0,       /* NOTUPTOI, NOTMINUPTOI */
856      0,                       /* NOTEXACTI */
857      0, 0, 0, 0,              /* NOTPOS{STARI,PLUSI,QUERYI,UPTOI} */
858    
859      OP_TYPEPOSSTAR, 0,       /* TYPESTAR, TYPEMINSTAR */
860      OP_TYPEPOSPLUS, 0,       /* TYPEPLUS, TYPEMINPLUS */
861      OP_TYPEPOSQUERY, 0,      /* TYPEQUERY, TYPEMINQUERY */
862      OP_TYPEPOSUPTO, 0,       /* TYPEUPTO, TYPEMINUPTO */
863      0,                       /* TYPEEXACT */
864      0, 0, 0, 0,              /* TYPEPOS{STAR,PLUS,QUERY,UPTO} */
865    
866      OP_CRPOSSTAR, 0,         /* CRSTAR, CRMINSTAR */
867      OP_CRPOSPLUS, 0,         /* CRPLUS, CRMINPLUS */
868      OP_CRPOSQUERY, 0,        /* CRQUERY, CRMINQUERY */
869      OP_CRPOSRANGE, 0,        /* CRRANGE, CRMINRANGE */
870      0, 0, 0, 0,              /* CRPOS{STAR,PLUS,QUERY,RANGE} */
871    
872      0, 0, 0,                 /* CLASS, NCLASS, XCLASS */
873      0, 0,                    /* REF, REFI */
874      0, 0,                    /* DNREF, DNREFI */
875      0, 0                     /* RECURSE, CALLOUT */
876    };
877    
878    
879    
880  /*************************************************  /*************************************************
# Line 665  find_error_text(int n) Line 896  find_error_text(int n)
896  const char *s = error_texts;  const char *s = error_texts;
897  for (; n > 0; n--)  for (; n > 0; n--)
898    {    {
899    while (*s++ != 0) {};    while (*s++ != CHAR_NULL) {};
900    if (*s == 0) return "Error text not found (please report)";    if (*s == CHAR_NULL) return "Error text not found (please report)";
901    }    }
902  return s;  return s;
903  }  }
904    
905    
906    
907  /*************************************************  /*************************************************
908  *           Expand the workspace                 *  *           Expand the workspace                 *
909  *************************************************/  *************************************************/
# Line 749  return (*p == CHAR_RIGHT_CURLY_BRACKET); Line 981  return (*p == CHAR_RIGHT_CURLY_BRACKET);
981  *************************************************/  *************************************************/
982    
983  /* 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
984  positive value for a simple escape such as \n, or 0 for a data character  positive value for a simple escape such as \n, or 0 for a data character which
985  which will be placed in chptr. A backreference to group n is returned as  will be placed in chptr. A backreference to group n is returned as negative n.
986  negative n. When UTF-8 is enabled, a positive value greater than 255 may  When UTF-8 is enabled, a positive value greater than 255 may be returned in
987  be returned in chptr.  chptr. On entry, ptr is pointing at the \. On exit, it is on the final
988  On entry,ptr is pointing at the \. On exit, it is on the final character of the  character of the escape sequence.
 escape sequence.  
989    
990  Arguments:  Arguments:
991    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
992    chptr          points to the data character    chptr          points to a returned data character
993    errorcodeptr   points to the errorcode variable    errorcodeptr   points to the errorcode variable
994    bracount       number of previous extracting brackets    bracount       number of previous extracting brackets
995    options        the options bits    options        the options bits
# Line 771  Returns:         zero => a data characte Line 1002  Returns:         zero => a data characte
1002  */  */
1003    
1004  static int  static int
1005  check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,  check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,
1006    int bracount, int options, BOOL isclass)    int bracount, int options, BOOL isclass)
1007  {  {
1008  /* PCRE_UTF16 has the same value as PCRE_UTF8. */  /* PCRE_UTF16 has the same value as PCRE_UTF8. */
# Line 786  ptr--;                            /* Set Line 1017  ptr--;                            /* Set
1017    
1018  /* 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. */
1019    
1020  if (c == 0) *errorcodeptr = ERR1;  if (c == CHAR_NULL) *errorcodeptr = ERR1;
1021    
1022  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
1023  in a table. A non-zero result is something that can be returned immediately.  in a table. A non-zero result is something that can be returned immediately.
# Line 795  Otherwise further processing may be requ Line 1026  Otherwise further processing may be requ
1026  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1027  /* Not alphanumeric */  /* Not alphanumeric */
1028  else if (c < CHAR_0 || c > CHAR_z) {}  else if (c < CHAR_0 || c > CHAR_z) {}
1029  else if ((i = escapes[c - CHAR_0]) != 0) { if (i > 0) c = (pcre_uint32)i; else escape = -i; }  else if ((i = escapes[c - CHAR_0]) != 0)
1030      { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1031    
1032  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1033  /* Not alphanumeric */  /* Not alphanumeric */
# Line 845  else Line 1077  else
1077            }            }
1078    
1079  #if defined COMPILE_PCRE8  #if defined COMPILE_PCRE8
1080          if (c > (utf ? 0x10ffff : 0xff))          if (c > (utf ? 0x10ffffU : 0xffU))
1081  #elif defined COMPILE_PCRE16  #elif defined COMPILE_PCRE16
1082          if (c > (utf ? 0x10ffff : 0xffff))          if (c > (utf ? 0x10ffffU : 0xffffU))
1083  #elif defined COMPILE_PCRE32  #elif defined COMPILE_PCRE32
1084          if (utf && c > 0x10ffff)          if (utf && c > 0x10ffffU)
1085  #endif  #endif
1086            {            {
1087            *errorcodeptr = ERR76;            *errorcodeptr = ERR76;
# Line 896  else Line 1128  else
1128      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1129        {        {
1130        const pcre_uchar *p;        const pcre_uchar *p;
1131        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++)
1132          if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;          if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
1133        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)        if (*p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET)
1134          {          {
1135          escape = ESC_k;          escape = ESC_k;
1136          break;          break;
# Line 961  else Line 1193  else
1193      break;      break;
1194    
1195      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
1196      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
1197      the way Perl works seems to be as follows:      over the years. Nowadays \g{} for backreferences and \o{} for octal are
1198        recommended to avoid the ambiguities in the old syntax.
1199    
1200      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
1201      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
1202      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
1203      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
1204      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
1205      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
1206      character class, \ followed by a digit is always an octal number. */      taken. \8 and \9 are treated as the literal characters 8 and 9.
1207    
1208        Inside a character class, \ followed by a digit is always either a literal
1209        8 or 9 or an octal number. */
1210    
1211      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:
1212      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
# Line 997  else Line 1233  else
1233          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
1234          break;          break;
1235          }          }
1236        if (s < 10 || s <= bracount)        if (s < 8 || s <= bracount)  /* Check for back reference */
1237          {          {
1238          escape = -s;          escape = -s;
1239          break;          break;
# Line 1005  else Line 1241  else
1241        ptr = oldptr;      /* Put the pointer back and fall through */        ptr = oldptr;      /* Put the pointer back and fall through */
1242        }        }
1243    
1244      /* 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
1245      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
1246      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
1247        changed so as not to insert the binary zero. */
1248    
1249      if ((c = *ptr) >= CHAR_8)      if ((c = *ptr) >= CHAR_8) break;
1250        {  
1251        ptr--;      /* Fall through with a digit less than 8 */
       c = 0;  
       break;  
       }  
1252    
1253      /* \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
1254      larger first octal digit. The original code used just to take the least      larger first octal digit. The original code used just to take the least
# Line 1031  else Line 1265  else
1265  #endif  #endif
1266      break;      break;
1267    
1268      /* \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
1269      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}. */
1270      If not, { is treated as a data character. */  
1271        case CHAR_o:
1272        if (ptr[1] != CHAR_LEFT_CURLY_BRACKET) *errorcodeptr = ERR81; else
1273        if (ptr[2] == CHAR_RIGHT_CURLY_BRACKET) *errorcodeptr = ERR86; else
1274          {
1275          ptr += 2;
1276          c = 0;
1277          overflow = FALSE;
1278          while (*ptr >= CHAR_0 && *ptr <= CHAR_7)
1279            {
1280            register pcre_uint32 cc = *ptr++;
1281            if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1282    #ifdef COMPILE_PCRE32
1283            if (c >= 0x20000000l) { overflow = TRUE; break; }
1284    #endif
1285            c = (c << 3) + cc - CHAR_0 ;
1286    #if defined COMPILE_PCRE8
1287            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1288    #elif defined COMPILE_PCRE16
1289            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1290    #elif defined COMPILE_PCRE32
1291            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1292    #endif
1293            }
1294          if (overflow)
1295            {
1296            while (*ptr >= CHAR_0 && *ptr <= CHAR_7) ptr++;
1297            *errorcodeptr = ERR34;
1298            }
1299          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1300            {
1301            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1302            }
1303          else *errorcodeptr = ERR80;
1304          }
1305        break;
1306    
1307        /* \x is complicated. In JavaScript, \x must be followed by two hexadecimal
1308        numbers. Otherwise it is a lowercase x letter. */
1309    
1310      case CHAR_x:      case CHAR_x:
1311      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1312        {        {
       /* In JavaScript, \x must be followed by two hexadecimal numbers.  
       Otherwise it is a lowercase x letter. */  
1313        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1314          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1315          {          {
# Line 1056  else Line 1326  else
1326  #endif  #endif
1327            }            }
1328          }          }
1329        break;        }    /* End JavaScript handling */
       }  
1330    
1331      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      /* Handle \x in Perl's style. \x{ddd} is a character number which can be
1332        {      greater than 0xff in utf or non-8bit mode, but only if the ddd are hex
1333        const pcre_uchar *pt = ptr + 2;      digits. If not, { used to be treated as a data character. However, Perl
1334        seems to read hex digits up to the first non-such, and ignore the rest, so
1335        that, for example \x{zz} matches a binary zero. This seems crazy, so PCRE
1336        now gives an error. */
1337    
1338        c = 0;      else
1339        overflow = FALSE;        {
1340        while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0)        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1341          {          {
1342          register pcre_uint32 cc = *pt++;          ptr += 2;
1343          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */          if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1344              {
1345              *errorcodeptr = ERR86;
1346              break;
1347              }
1348            c = 0;
1349            overflow = FALSE;
1350            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0)
1351              {
1352              register pcre_uint32 cc = *ptr++;
1353              if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1354    
1355  #ifdef COMPILE_PCRE32  #ifdef COMPILE_PCRE32
1356          if (c >= 0x10000000l) { overflow = TRUE; break; }            if (c >= 0x10000000l) { overflow = TRUE; break; }
1357  #endif  #endif
1358    
1359  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1360          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1361          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1362  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1363          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 */
1364          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1365  #endif  #endif
1366    
1367  #if defined COMPILE_PCRE8  #if defined COMPILE_PCRE8
1368          if (c > (utf ? 0x10ffff : 0xff)) { overflow = TRUE; break; }            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1369  #elif defined COMPILE_PCRE16  #elif defined COMPILE_PCRE16
1370          if (c > (utf ? 0x10ffff : 0xffff)) { overflow = TRUE; break; }            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1371  #elif defined COMPILE_PCRE32  #elif defined COMPILE_PCRE32
1372          if (utf && c > 0x10ffff) { overflow = TRUE; break; }            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1373  #endif  #endif
1374          }            }
1375    
1376        if (overflow)          if (overflow)
1377          {            {
1378          while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0) pt++;            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0) ptr++;
1379          *errorcodeptr = ERR34;            *errorcodeptr = ERR34;
1380          }            }
1381    
1382        if (*pt == CHAR_RIGHT_CURLY_BRACKET)          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1383          {            {
1384          if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1385          ptr = pt;            }
         break;  
         }  
1386    
1387        /* 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.
1388        recognize this construct; fall through to the normal \x handling. */          We used just to recognize this construct and fall through to the normal
1389        }          \x handling, but nowadays Perl gives an error, which seems much more
1390            sensible, so we do too. */
1391    
1392      /* Read just a single-byte hex-defined char */          else *errorcodeptr = ERR79;
1393            }   /* End of \x{} processing */
1394    
1395      c = 0;        /* Read a single-byte hex-defined char (up to two hex digits after \x) */
1396      while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)  
1397        {        else
1398        pcre_uint32 cc;                          /* Some compilers don't like */          {
1399        cc = *(++ptr);                           /* ++ in initializers */          c = 0;
1400            while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1401              {
1402              pcre_uint32 cc;                          /* Some compilers don't like */
1403              cc = *(++ptr);                           /* ++ in initializers */
1404  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1405        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1406        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1407  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1408        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */            if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1409        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1410  #endif  #endif
1411        }            }
1412            }     /* End of \xdd handling */
1413          }       /* End of Perl-style \x handling */
1414      break;      break;
1415    
1416      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
# Line 1132  else Line 1420  else
1420    
1421      case CHAR_c:      case CHAR_c:
1422      c = *(++ptr);      c = *(++ptr);
1423      if (c == 0)      if (c == CHAR_NULL)
1424        {        {
1425        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
1426        break;        break;
# Line 1147  else Line 1435  else
1435      c ^= 0x40;      c ^= 0x40;
1436  #else             /* EBCDIC coding */  #else             /* EBCDIC coding */
1437      if (c >= CHAR_a && c <= CHAR_z) c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
1438      c ^= 0xC0;      if (c == CHAR_QUESTION_MARK)
1439          c = ('\\' == 188 && '`' == 74)? 0x5f : 0xff;
1440        else
1441          {
1442          for (i = 0; i < 32; i++)
1443            {
1444            if (c == ebcdic_escape_c[i]) break;
1445            }
1446          if (i < 32) c = i; else *errorcodeptr = ERR68;
1447          }
1448  #endif  #endif
1449      break;      break;
1450    
# Line 1188  if ((options & PCRE_UCP) != 0 && escape Line 1485  if ((options & PCRE_UCP) != 0 && escape
1485  return escape;  return escape;
1486  }  }
1487    
1488    
1489    
1490  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
1491  /*************************************************  /*************************************************
1492  *               Handle \P and \p                 *  *               Handle \P and \p                 *
# Line 1201  escape sequence. Line 1500  escape sequence.
1500  Argument:  Argument:
1501    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
1502    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
1503    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
1504      pdataptr       points to an unsigned int that is set to the detailed property value
1505    errorcodeptr   points to the error code variable    errorcodeptr   points to the error code variable
1506    
1507  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
1508  */  */
1509    
1510  static int  static BOOL
1511  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, unsigned int *ptypeptr,
1512      unsigned int *pdataptr, int *errorcodeptr)
1513  {  {
1514  int c, i, bot, top;  pcre_uchar c;
1515    int i, bot, top;
1516  const pcre_uchar *ptr = *ptrptr;  const pcre_uchar *ptr = *ptrptr;
1517  pcre_uchar name[32];  pcre_uchar name[32];
1518    
1519  c = *(++ptr);  c = *(++ptr);
1520  if (c == 0) goto ERROR_RETURN;  if (c == CHAR_NULL) goto ERROR_RETURN;
1521    
1522  *negptr = FALSE;  *negptr = FALSE;
1523    
# Line 1232  if (c == CHAR_LEFT_CURLY_BRACKET) Line 1534  if (c == CHAR_LEFT_CURLY_BRACKET)
1534    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1535      {      {
1536      c = *(++ptr);      c = *(++ptr);
1537      if (c == 0) goto ERROR_RETURN;      if (c == CHAR_NULL) goto ERROR_RETURN;
1538      if (c == CHAR_RIGHT_CURLY_BRACKET) break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1539      name[i] = c;      name[i] = c;
1540      }      }
# Line 1257  top = PRIV(utt_size); Line 1559  top = PRIV(utt_size);
1559    
1560  while (bot < top)  while (bot < top)
1561    {    {
1562      int r;
1563    i = (bot + top) >> 1;    i = (bot + top) >> 1;
1564    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);
1565    if (c == 0)    if (r == 0)
1566      {      {
1567      *dptr = PRIV(utt)[i].value;      *ptypeptr = PRIV(utt)[i].type;
1568      return PRIV(utt)[i].type;      *pdataptr = PRIV(utt)[i].value;
1569        return TRUE;
1570      }      }
1571    if (c > 0) bot = i + 1; else top = i;    if (r > 0) bot = i + 1; else top = i;
1572    }    }
1573    
1574  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
1575  *ptrptr = ptr;  *ptrptr = ptr;
1576  return -1;  return FALSE;
1577    
1578  ERROR_RETURN:  ERROR_RETURN:
1579  *errorcodeptr = ERR46;  *errorcodeptr = ERR46;
1580  *ptrptr = ptr;  *ptrptr = ptr;
1581  return -1;  return FALSE;
1582  }  }
1583  #endif  #endif
1584    
1585    
1586    
   
1587  /*************************************************  /*************************************************
1588  *         Read repeat counts                     *  *         Read repeat counts                     *
1589  *************************************************/  *************************************************/
# Line 1306  read_repeat_counts(const pcre_uchar *p, Line 1609  read_repeat_counts(const pcre_uchar *p,
1609  int min = 0;  int min = 0;
1610  int max = -1;  int max = -1;
1611    
1612  /* Read the minimum value and do a paranoid check: a negative value indicates  while (IS_DIGIT(*p))
 an integer overflow. */  
   
 while (IS_DIGIT(*p)) min = min * 10 + *p++ - CHAR_0;  
 if (min < 0 || min > 65535)  
1613    {    {
1614    *errorcodeptr = ERR5;    min = min * 10 + (int)(*p++ - CHAR_0);
1615    return p;    if (min > 65535)
1616        {
1617        *errorcodeptr = ERR5;
1618        return p;
1619        }
1620    }    }
1621    
 /* Read the maximum value if there is one, and again do a paranoid on its size.  
 Also, max must not be less than min. */  
   
1622  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1623    {    {
1624    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1625      {      {
1626      max = 0;      max = 0;
1627      while(IS_DIGIT(*p)) max = max * 10 + *p++ - CHAR_0;      while(IS_DIGIT(*p))
     if (max < 0 || max > 65535)  
1628        {        {
1629        *errorcodeptr = ERR5;        max = max * 10 + (int)(*p++ - CHAR_0);
1630        return p;        if (max > 65535)
1631            {
1632            *errorcodeptr = ERR5;
1633            return p;
1634            }
1635        }        }
1636      if (max < min)      if (max < min)
1637        {        {
# Line 1338  if (*p == CHAR_RIGHT_CURLY_BRACKET) max Line 1641  if (*p == CHAR_RIGHT_CURLY_BRACKET) max
1641      }      }
1642    }    }
1643    
 /* Fill in the required variables, and pass back the pointer to the terminating  
 '}'. */  
   
1644  *minp = min;  *minp = min;
1645  *maxp = max;  *maxp = max;
1646  return p;  return p;
# Line 1349  return p; Line 1649  return p;
1649    
1650    
1651  /*************************************************  /*************************************************
 *  Subroutine for finding forward reference      *  
 *************************************************/  
   
 /* This recursive function is called only from find_parens() below. The  
 top-level call starts at the beginning of the pattern. All other calls must  
 start at a parenthesis. It scans along a pattern's text looking for capturing  
 subpatterns, and counting them. If it finds a named pattern that matches the  
 name it is given, it returns its number. Alternatively, if the name is NULL, it  
 returns when it reaches a given numbered subpattern. Recursion is used to keep  
 track of subpatterns that reset the capturing group numbers - the (?| feature.  
   
 This function was originally called only from the second pass, in which we know  
 that if (?< or (?' or (?P< is encountered, the name will be correctly  
 terminated because that is checked in the first pass. There is now one call to  
 this function in the first pass, to check for a recursive back reference by  
 name (so that we can make the whole group atomic). In this case, we need check  
 only up to the current position in the pattern, and that is still OK because  
 and previous occurrences will have been checked. To make this work, the test  
 for "end of pattern" is a check against cd->end_pattern in the main loop,  
 instead of looking for a binary zero. This means that the special first-pass  
 call can adjust cd->end_pattern temporarily. (Checks for binary zero while  
 processing items within the loop are OK, because afterwards the main loop will  
 terminate.)  
   
 Arguments:  
   ptrptr       address of the current character pointer (updated)  
   cd           compile background data  
   name         name to seek, or NULL if seeking a numbered subpattern  
   lorn         name length, or subpattern number if name is NULL  
   xmode        TRUE if we are in /x mode  
   utf          TRUE if we are in UTF-8 / UTF-16 / UTF-32 mode  
   count        pointer to the current capturing subpattern number (updated)  
   
 Returns:       the number of the named subpattern, or -1 if not found  
 */  
   
 static int  
 find_parens_sub(pcre_uchar **ptrptr, compile_data *cd, const pcre_uchar *name, int lorn,  
   BOOL xmode, BOOL utf, int *count)  
 {  
 pcre_uchar *ptr = *ptrptr;  
 int start_count = *count;  
 int hwm_count = start_count;  
 BOOL dup_parens = FALSE;  
   
 /* If the first character is a parenthesis, check on the type of group we are  
 dealing with. The very first call may not start with a parenthesis. */  
   
 if (ptr[0] == CHAR_LEFT_PARENTHESIS)  
   {  
   /* Handle specials such as (*SKIP) or (*UTF8) etc. */  
   
   if (ptr[1] == CHAR_ASTERISK) ptr += 2;  
   
   /* Handle a normal, unnamed capturing parenthesis. */  
   
   else if (ptr[1] != CHAR_QUESTION_MARK)  
     {  
     *count += 1;  
     if (name == NULL && *count == lorn) return *count;  
     ptr++;  
     }  
   
   /* All cases now have (? at the start. Remember when we are in a group  
   where the parenthesis numbers are duplicated. */  
   
   else if (ptr[2] == CHAR_VERTICAL_LINE)  
     {  
     ptr += 3;  
     dup_parens = TRUE;  
     }  
   
   /* Handle comments; all characters are allowed until a ket is reached. */  
   
   else if (ptr[2] == CHAR_NUMBER_SIGN)  
     {  
     for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;  
     goto FAIL_EXIT;  
     }  
   
   /* Handle a condition. If it is an assertion, just carry on so that it  
   is processed as normal. If not, skip to the closing parenthesis of the  
   condition (there can't be any nested parens). */  
   
   else if (ptr[2] == CHAR_LEFT_PARENTHESIS)  
     {  
     ptr += 2;  
     if (ptr[1] != CHAR_QUESTION_MARK)  
       {  
       while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;  
       if (*ptr != 0) ptr++;  
       }  
     }  
   
   /* Start with (? but not a condition. */  
   
   else  
     {  
     ptr += 2;  
     if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */  
   
     /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */  
   
     if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&  
         ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)  
       {  
       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 / UTF-32 mode  
   
 Returns:       the number of the found subpattern, or -1 if not found  
 */  
   
 static int  
 find_parens(compile_data *cd, const pcre_uchar *name, int lorn, BOOL xmode,  
   BOOL utf)  
 {  
 pcre_uchar *ptr = (pcre_uchar *)cd->start_pattern;  
 int count = 0;  
 int rc;  
   
 /* If the pattern does not start with an opening parenthesis, the first call  
 to find_parens_sub() will scan right to the end (if necessary). However, if it  
 does start with a parenthesis, find_parens_sub() will return when it hits the  
 matching closing parens. That is why we have to have a loop. */  
   
 for (;;)  
   {  
   rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, &count);  
   if (rc > 0 || *ptr++ == 0) break;  
   }  
   
 return rc;  
 }  
   
   
   
   
 /*************************************************  
1652  *      Find first significant op code            *  *      Find first significant op code            *
1653  *************************************************/  *************************************************/
1654    
# Line 1682  for (;;) Line 1687  for (;;)
1687    
1688      case OP_CALLOUT:      case OP_CALLOUT:
1689      case OP_CREF:      case OP_CREF:
1690      case OP_NCREF:      case OP_DNCREF:
1691      case OP_RREF:      case OP_RREF:
1692      case OP_NRREF:      case OP_DNRREF:
1693      case OP_DEF:      case OP_DEF:
1694      code += PRIV(OP_lengths)[*code];      code += PRIV(OP_lengths)[*code];
1695      break;      break;
# Line 1698  for (;;) Line 1703  for (;;)
1703    
1704    
1705    
   
1706  /*************************************************  /*************************************************
1707  *        Find the fixed length of a branch       *  *        Find the fixed length of a branch       *
1708  *************************************************/  *************************************************/
# Line 1719  Arguments: Line 1723  Arguments:
1723    utf      TRUE in UTF-8 / UTF-16 / UTF-32 mode    utf      TRUE in UTF-8 / UTF-16 / UTF-32 mode
1724    atend    TRUE if called when the pattern is complete    atend    TRUE if called when the pattern is complete
1725    cd       the "compile data" structure    cd       the "compile data" structure
1726      recurses    chain of recurse_check to catch mutual recursion
1727    
1728  Returns:   the fixed length,  Returns:   the fixed length,
1729               or -1 if there is no fixed length,               or -1 if there is no fixed length,
# Line 1728  Returns:   the fixed length, Line 1733  Returns:   the fixed length,
1733  */  */
1734    
1735  static int  static int
1736  find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd)  find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd,
1737      recurse_check *recurses)
1738  {  {
1739  int length = -1;  int length = -1;
1740    recurse_check this_recurse;
1741  register int branchlength = 0;  register int branchlength = 0;
1742  register pcre_uchar *cc = code + 1 + LINK_SIZE;  register pcre_uchar *cc = code + 1 + LINK_SIZE;
1743    
# Line 1742  for (;;) Line 1748  for (;;)
1748    {    {
1749    int d;    int d;
1750    pcre_uchar *ce, *cs;    pcre_uchar *ce, *cs;
1751    register int op = *cc;    register pcre_uchar op = *cc;
1752    
1753    switch (op)    switch (op)
1754      {      {
# Line 1756  for (;;) Line 1762  for (;;)
1762      case OP_ONCE:      case OP_ONCE:
1763      case OP_ONCE_NC:      case OP_ONCE_NC:
1764      case OP_COND:      case OP_COND:
1765      d = find_fixedlength(cc + ((op == OP_CBRA)? IMM2_SIZE : 0), utf, atend, cd);      d = find_fixedlength(cc + ((op == OP_CBRA)? IMM2_SIZE : 0), utf, atend, cd,
1766          recurses);
1767      if (d < 0) return d;      if (d < 0) return d;
1768      branchlength += d;      branchlength += d;
1769      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1790  for (;;) Line 1797  for (;;)
1797      cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1);  /* Start subpattern */      cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1798      do ce += GET(ce, 1); while (*ce == OP_ALT);           /* End subpattern */      do ce += GET(ce, 1); while (*ce == OP_ALT);           /* End subpattern */
1799      if (cc > cs && cc < ce) return -1;                    /* Recursion */      if (cc > cs && cc < ce) return -1;                    /* Recursion */
1800      d = find_fixedlength(cs + IMM2_SIZE, utf, atend, cd);      else   /* Check for mutual recursion */
1801          {
1802          recurse_check *r = recurses;
1803          for (r = recurses; r != NULL; r = r->prev) if (r->group == cs) break;
1804          if (r != NULL) return -1;   /* Mutual recursion */
1805          }
1806        this_recurse.prev = recurses;
1807        this_recurse.group = cs;
1808        d = find_fixedlength(cs + IMM2_SIZE, utf, atend, cd, &this_recurse);
1809      if (d < 0) return d;      if (d < 0) return d;
1810      branchlength += d;      branchlength += d;
1811      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
# Line 1803  for (;;) Line 1818  for (;;)
1818      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1819      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1820      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1821      cc += PRIV(OP_lengths)[*cc];      cc += 1 + LINK_SIZE;
1822      break;      break;
1823    
1824      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
# Line 1822  for (;;) Line 1837  for (;;)
1837      case OP_COMMIT:      case OP_COMMIT:
1838      case OP_CREF:      case OP_CREF:
1839      case OP_DEF:      case OP_DEF:
1840        case OP_DNCREF:
1841        case OP_DNRREF:
1842      case OP_DOLL:      case OP_DOLL:
1843      case OP_DOLLM:      case OP_DOLLM:
1844      case OP_EOD:      case OP_EOD:
1845      case OP_EODN:      case OP_EODN:
1846      case OP_FAIL:      case OP_FAIL:
     case OP_NCREF:  
     case OP_NRREF:  
1847      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1848      case OP_PRUNE:      case OP_PRUNE:
1849      case OP_REVERSE:      case OP_REVERSE:
# Line 1850  for (;;) Line 1865  for (;;)
1865      case OP_NOTI:      case OP_NOTI:
1866      branchlength++;      branchlength++;
1867      cc += 2;      cc += 2;
1868  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32  #ifdef SUPPORT_UTF
1869      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1870  #endif  #endif
1871      break;      break;
# Line 1862  for (;;) Line 1877  for (;;)
1877      case OP_EXACTI:      case OP_EXACTI:
1878      case OP_NOTEXACT:      case OP_NOTEXACT:
1879      case OP_NOTEXACTI:      case OP_NOTEXACTI:
1880      branchlength += GET2(cc,1);      branchlength += (int)GET2(cc,1);
1881      cc += 2 + IMM2_SIZE;      cc += 2 + IMM2_SIZE;
1882  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32  #ifdef SUPPORT_UTF
1883      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1884  #endif  #endif
1885      break;      break;
1886    
1887      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1888      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1889      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)
1890        cc += 2;        cc += 2;
1891      cc += 1 + IMM2_SIZE + 1;      cc += 1 + IMM2_SIZE + 1;
1892      break;      break;
# Line 1907  for (;;) Line 1922  for (;;)
1922    
1923      /* Check a class for variable quantification */      /* Check a class for variable quantification */
1924    
 #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32  
     case OP_XCLASS:  
     cc += GET(cc, 1) - PRIV(OP_lengths)[OP_CLASS];  
     /* Fall through */  
 #endif  
   
1925      case OP_CLASS:      case OP_CLASS:
1926      case OP_NCLASS:      case OP_NCLASS:
1927    #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1928        case OP_XCLASS:
1929        /* The original code caused an unsigned overflow in 64 bit systems,
1930        so now we use a conditional statement. */
1931        if (op == OP_XCLASS)
1932          cc += GET(cc, 1);
1933        else
1934          cc += PRIV(OP_lengths)[OP_CLASS];
1935    #else
1936      cc += PRIV(OP_lengths)[OP_CLASS];      cc += PRIV(OP_lengths)[OP_CLASS];
1937    #endif
1938    
1939      switch (*cc)      switch (*cc)
1940        {        {
       case OP_CRPLUS:  
       case OP_CRMINPLUS:  
1941        case OP_CRSTAR:        case OP_CRSTAR:
1942        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1943          case OP_CRPLUS:
1944          case OP_CRMINPLUS:
1945        case OP_CRQUERY:        case OP_CRQUERY:
1946        case OP_CRMINQUERY:        case OP_CRMINQUERY:
1947          case OP_CRPOSSTAR:
1948          case OP_CRPOSPLUS:
1949          case OP_CRPOSQUERY:
1950        return -1;        return -1;
1951    
1952        case OP_CRRANGE:        case OP_CRRANGE:
1953        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1954          case OP_CRPOSRANGE:
1955        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1956        branchlength += GET2(cc,1);        branchlength += (int)GET2(cc,1);
1957        cc += 1 + 2 * IMM2_SIZE;        cc += 1 + 2 * IMM2_SIZE;
1958        break;        break;
1959    
# Line 1997  for (;;) Line 2020  for (;;)
2020      case OP_QUERYI:      case OP_QUERYI:
2021      case OP_REF:      case OP_REF:
2022      case OP_REFI:      case OP_REFI:
2023        case OP_DNREF:
2024        case OP_DNREFI:
2025      case OP_SBRA:      case OP_SBRA:
2026      case OP_SBRAPOS:      case OP_SBRAPOS:
2027      case OP_SCBRA:      case OP_SCBRA:
# Line 2033  for (;;) Line 2058  for (;;)
2058    
2059    
2060    
   
2061  /*************************************************  /*************************************************
2062  *    Scan compiled regex for specific bracket    *  *    Scan compiled regex for specific bracket    *
2063  *************************************************/  *************************************************/
# Line 2057  PRIV(find_bracket)(const pcre_uchar *cod Line 2081  PRIV(find_bracket)(const pcre_uchar *cod
2081  {  {
2082  for (;;)  for (;;)
2083    {    {
2084    register int c = *code;    register pcre_uchar c = *code;
2085    
2086    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2087    
# Line 2080  for (;;) Line 2104  for (;;)
2104    else if (c == OP_CBRA || c == OP_SCBRA ||    else if (c == OP_CBRA || c == OP_SCBRA ||
2105             c == OP_CBRAPOS || c == OP_SCBRAPOS)             c == OP_CBRAPOS || c == OP_SCBRAPOS)
2106      {      {
2107      int n = GET2(code, 1+LINK_SIZE);      int n = (int)GET2(code, 1+LINK_SIZE);
2108      if (n == number) return (pcre_uchar *)code;      if (n == number) return (pcre_uchar *)code;
2109      code += PRIV(OP_lengths)[c];      code += PRIV(OP_lengths)[c];
2110      }      }
# Line 2110  for (;;) Line 2134  for (;;)
2134        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2135        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2136        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
2137        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2138          code += 2;          code += 2;
2139        break;        break;
2140    
2141        case OP_MARK:        case OP_MARK:
2142        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
2143        case OP_SKIP_ARG:        case OP_SKIP_ARG:
       code += code[1];  
       break;  
   
2144        case OP_THEN_ARG:        case OP_THEN_ARG:
2145        code += code[1];        code += code[1];
2146        break;        break;
# Line 2138  for (;;) Line 2159  for (;;)
2159        {        {
2160        case OP_CHAR:        case OP_CHAR:
2161        case OP_CHARI:        case OP_CHARI:
2162          case OP_NOT:
2163          case OP_NOTI:
2164        case OP_EXACT:        case OP_EXACT:
2165        case OP_EXACTI:        case OP_EXACTI:
2166          case OP_NOTEXACT:
2167          case OP_NOTEXACTI:
2168        case OP_UPTO:        case OP_UPTO:
2169        case OP_UPTOI:        case OP_UPTOI:
2170          case OP_NOTUPTO:
2171          case OP_NOTUPTOI:
2172        case OP_MINUPTO:        case OP_MINUPTO:
2173        case OP_MINUPTOI:        case OP_MINUPTOI:
2174          case OP_NOTMINUPTO:
2175          case OP_NOTMINUPTOI:
2176        case OP_POSUPTO:        case OP_POSUPTO:
2177        case OP_POSUPTOI:        case OP_POSUPTOI:
2178          case OP_NOTPOSUPTO:
2179          case OP_NOTPOSUPTOI:
2180        case OP_STAR:        case OP_STAR:
2181        case OP_STARI:        case OP_STARI:
2182          case OP_NOTSTAR:
2183          case OP_NOTSTARI:
2184        case OP_MINSTAR:        case OP_MINSTAR:
2185        case OP_MINSTARI:        case OP_MINSTARI:
2186          case OP_NOTMINSTAR:
2187          case OP_NOTMINSTARI:
2188        case OP_POSSTAR:        case OP_POSSTAR:
2189        case OP_POSSTARI:        case OP_POSSTARI:
2190          case OP_NOTPOSSTAR:
2191          case OP_NOTPOSSTARI:
2192        case OP_PLUS:        case OP_PLUS:
2193        case OP_PLUSI:        case OP_PLUSI:
2194          case OP_NOTPLUS:
2195          case OP_NOTPLUSI:
2196        case OP_MINPLUS:        case OP_MINPLUS:
2197        case OP_MINPLUSI:        case OP_MINPLUSI:
2198          case OP_NOTMINPLUS:
2199          case OP_NOTMINPLUSI:
2200        case OP_POSPLUS:        case OP_POSPLUS:
2201        case OP_POSPLUSI:        case OP_POSPLUSI:
2202          case OP_NOTPOSPLUS:
2203          case OP_NOTPOSPLUSI:
2204        case OP_QUERY:        case OP_QUERY:
2205        case OP_QUERYI:        case OP_QUERYI:
2206          case OP_NOTQUERY:
2207          case OP_NOTQUERYI:
2208        case OP_MINQUERY:        case OP_MINQUERY:
2209        case OP_MINQUERYI:        case OP_MINQUERYI:
2210          case OP_NOTMINQUERY:
2211          case OP_NOTMINQUERYI:
2212        case OP_POSQUERY:        case OP_POSQUERY:
2213        case OP_POSQUERYI:        case OP_POSQUERYI:
2214          case OP_NOTPOSQUERY:
2215          case OP_NOTPOSQUERYI:
2216        if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);        if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2217        break;        break;
2218        }        }
# Line 2195  find_recurse(const pcre_uchar *code, BOO Line 2244  find_recurse(const pcre_uchar *code, BOO
2244  {  {
2245  for (;;)  for (;;)
2246    {    {
2247    register int c = *code;    register pcre_uchar c = *code;
2248    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2249    if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
2250    
# Line 2230  for (;;) Line 2279  for (;;)
2279        case OP_TYPEUPTO:        case OP_TYPEUPTO:
2280        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2281        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2282        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2283          code += 2;          code += 2;
2284        break;        break;
2285    
2286        case OP_MARK:        case OP_MARK:
2287        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
2288        case OP_SKIP_ARG:        case OP_SKIP_ARG:
       code += code[1];  
       break;  
   
2289        case OP_THEN_ARG:        case OP_THEN_ARG:
2290        code += code[1];        code += code[1];
2291        break;        break;
# Line 2341  Arguments: Line 2387  Arguments:
2387    endcode     points to where to stop    endcode     points to where to stop
2388    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2389    cd          contains pointers to tables etc.    cd          contains pointers to tables etc.
2390      recurses    chain of recurse_check to catch mutual recursion
2391    
2392  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2393  */  */
2394    
2395  static BOOL  static BOOL
2396  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2397    BOOL utf, compile_data *cd)    BOOL utf, compile_data *cd, recurse_check *recurses)
2398  {  {
2399  register int c;  register pcre_uchar c;
2400    recurse_check this_recurse;
2401    
2402  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2403       code < endcode;       code < endcode;
2404       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
# Line 2377  for (code = first_significant_code(code Line 2426  for (code = first_significant_code(code
2426    
2427    if (c == OP_RECURSE)    if (c == OP_RECURSE)
2428      {      {
2429      const pcre_uchar *scode;      const pcre_uchar *scode = cd->start_code + GET(code, 1);
2430        const pcre_uchar *endgroup = scode;
2431      BOOL empty_branch;      BOOL empty_branch;
2432    
2433      /* Test for forward reference */      /* Test for forward reference or uncompleted reference. This is disabled
2434        when called to scan a completed pattern by setting cd->start_workspace to
2435        NULL. */
2436    
2437      for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)      if (cd->start_workspace != NULL)
2438        if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;        {
2439          const pcre_uchar *tcode;
2440          for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)
2441            if ((int)GET(tcode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;
2442          if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2443          }
2444    
2445      /* Not a forward reference, test for completed backward reference */      /* If the reference is to a completed group, we need to detect whether this
2446        is a recursive call, as otherwise there will be an infinite loop. If it is
2447        a recursion, just skip over it. Simple recursions are easily detected. For
2448        mutual recursions we keep a chain on the stack. */
2449    
2450      empty_branch = FALSE;      do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);
2451      scode = cd->start_code + GET(code, 1);      if (code >= scode && code <= endgroup) continue;  /* Simple recursion */
2452      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */      else
2453          {
2454          recurse_check *r = recurses;
2455          for (r = recurses; r != NULL; r = r->prev)
2456            if (r->group == scode) break;
2457          if (r != NULL) continue;   /* Mutual recursion */
2458          }
2459    
2460        /* Completed reference; scan the referenced group, remembering it on the
2461        stack chain to detect mutual recursions. */
2462    
2463      /* Completed backwards reference */      empty_branch = FALSE;
2464        this_recurse.prev = recurses;
2465        this_recurse.group = scode;
2466    
2467      do      do
2468        {        {
2469        if (could_be_empty_branch(scode, endcode, utf, cd))        if (could_be_empty_branch(scode, endcode, utf, cd, &this_recurse))
2470          {          {
2471          empty_branch = TRUE;          empty_branch = TRUE;
2472          break;          break;
# Line 2435  for (code = first_significant_code(code Line 2506  for (code = first_significant_code(code
2506    if (c == OP_BRA  || c == OP_BRAPOS ||    if (c == OP_BRA  || c == OP_BRAPOS ||
2507        c == OP_CBRA || c == OP_CBRAPOS ||        c == OP_CBRA || c == OP_CBRAPOS ||
2508        c == OP_ONCE || c == OP_ONCE_NC ||        c == OP_ONCE || c == OP_ONCE_NC ||
2509        c == OP_COND)        c == OP_COND || c == OP_SCOND)
2510      {      {
2511      BOOL empty_branch;      BOOL empty_branch;
2512      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
# Line 2451  for (code = first_significant_code(code Line 2522  for (code = first_significant_code(code
2522        empty_branch = FALSE;        empty_branch = FALSE;
2523        do        do
2524          {          {
2525          if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd))          if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd,
2526            empty_branch = TRUE;            recurses)) empty_branch = TRUE;
2527          code += GET(code, 1);          code += GET(code, 1);
2528          }          }
2529        while (*code == OP_ALT);        while (*code == OP_ALT);
# Line 2493  for (code = first_significant_code(code Line 2564  for (code = first_significant_code(code
2564        case OP_CRMINSTAR:        case OP_CRMINSTAR:
2565        case OP_CRQUERY:        case OP_CRQUERY:
2566        case OP_CRMINQUERY:        case OP_CRMINQUERY:
2567          case OP_CRPOSSTAR:
2568          case OP_CRPOSQUERY:
2569        break;        break;
2570    
2571        default:                   /* Non-repeat => class must match */        default:                   /* Non-repeat => class must match */
2572        case OP_CRPLUS:            /* These repeats aren't empty */        case OP_CRPLUS:            /* These repeats aren't empty */
2573        case OP_CRMINPLUS:        case OP_CRMINPLUS:
2574          case OP_CRPOSPLUS:
2575        return FALSE;        return FALSE;
2576    
2577        case OP_CRRANGE:        case OP_CRRANGE:
2578        case OP_CRMINRANGE:        case OP_CRMINRANGE:
2579          case OP_CRPOSRANGE:
2580        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */
2581        break;        break;
2582        }        }
# Line 2509  for (code = first_significant_code(code Line 2584  for (code = first_significant_code(code
2584    
2585      /* Opcodes that must match a character */      /* Opcodes that must match a character */
2586    
2587        case OP_ANY:
2588        case OP_ALLANY:
2589        case OP_ANYBYTE:
2590    
2591      case OP_PROP:      case OP_PROP:
2592      case OP_NOTPROP:      case OP_NOTPROP:
2593        case OP_ANYNL:
2594    
2595        case OP_NOT_HSPACE:
2596        case OP_HSPACE:
2597        case OP_NOT_VSPACE:
2598        case OP_VSPACE:
2599      case OP_EXTUNI:      case OP_EXTUNI:
2600    
2601      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2602      case OP_DIGIT:      case OP_DIGIT:
2603      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2604      case OP_WHITESPACE:      case OP_WHITESPACE:
2605      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2606      case OP_WORDCHAR:      case OP_WORDCHAR:
2607      case OP_ANY:  
     case OP_ALLANY:  
     case OP_ANYBYTE:  
2608      case OP_CHAR:      case OP_CHAR:
2609      case OP_CHARI:      case OP_CHARI:
2610      case OP_NOT:      case OP_NOT:
2611      case OP_NOTI:      case OP_NOTI:
2612    
2613      case OP_PLUS:      case OP_PLUS:
2614        case OP_PLUSI:
2615      case OP_MINPLUS:      case OP_MINPLUS:
2616      case OP_POSPLUS:      case OP_MINPLUSI:
2617      case OP_EXACT:  
2618      case OP_NOTPLUS:      case OP_NOTPLUS:
2619        case OP_NOTPLUSI:
2620      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
2621        case OP_NOTMINPLUSI:
2622    
2623        case OP_POSPLUS:
2624        case OP_POSPLUSI:
2625      case OP_NOTPOSPLUS:      case OP_NOTPOSPLUS:
2626        case OP_NOTPOSPLUSI:
2627    
2628        case OP_EXACT:
2629        case OP_EXACTI:
2630      case OP_NOTEXACT:      case OP_NOTEXACT:
2631        case OP_NOTEXACTI:
2632    
2633      case OP_TYPEPLUS:      case OP_TYPEPLUS:
2634      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
2635      case OP_TYPEPOSPLUS:      case OP_TYPEPOSPLUS:
2636      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2637    
2638      return FALSE;      return FALSE;
2639    
2640      /* 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 2556  for (code = first_significant_code(code Line 2654  for (code = first_significant_code(code
2654      case OP_TYPEUPTO:      case OP_TYPEUPTO:
2655      case OP_TYPEMINUPTO:      case OP_TYPEMINUPTO:
2656      case OP_TYPEPOSUPTO:      case OP_TYPEPOSUPTO:
2657      if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)      if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2658        code += 2;        code += 2;
2659      break;      break;
2660    
# Line 2570  for (code = first_significant_code(code Line 2668  for (code = first_significant_code(code
2668      return TRUE;      return TRUE;
2669    
2670      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2671      MINUPTO, and POSUPTO may be followed by a multibyte character */      MINUPTO, and POSUPTO and their caseless and negative versions may be
2672        followed by a multibyte character. */
2673    
2674  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2675      case OP_STAR:      case OP_STAR:
2676      case OP_STARI:      case OP_STARI:
2677        case OP_NOTSTAR:
2678        case OP_NOTSTARI:
2679    
2680      case OP_MINSTAR:      case OP_MINSTAR:
2681      case OP_MINSTARI:      case OP_MINSTARI:
2682        case OP_NOTMINSTAR:
2683        case OP_NOTMINSTARI:
2684    
2685      case OP_POSSTAR:      case OP_POSSTAR:
2686      case OP_POSSTARI:      case OP_POSSTARI:
2687        case OP_NOTPOSSTAR:
2688        case OP_NOTPOSSTARI:
2689    
2690      case OP_QUERY:      case OP_QUERY:
2691      case OP_QUERYI:      case OP_QUERYI:
2692        case OP_NOTQUERY:
2693        case OP_NOTQUERYI:
2694    
2695      case OP_MINQUERY:      case OP_MINQUERY:
2696      case OP_MINQUERYI:      case OP_MINQUERYI:
2697        case OP_NOTMINQUERY:
2698        case OP_NOTMINQUERYI:
2699    
2700      case OP_POSQUERY:      case OP_POSQUERY:
2701      case OP_POSQUERYI:      case OP_POSQUERYI:
2702        case OP_NOTPOSQUERY:
2703        case OP_NOTPOSQUERYI:
2704    
2705      if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);      if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2706      break;      break;
2707    
2708      case OP_UPTO:      case OP_UPTO:
2709      case OP_UPTOI:      case OP_UPTOI:
2710        case OP_NOTUPTO:
2711        case OP_NOTUPTOI:
2712    
2713      case OP_MINUPTO:      case OP_MINUPTO:
2714      case OP_MINUPTOI:      case OP_MINUPTOI:
2715        case OP_NOTMINUPTO:
2716        case OP_NOTMINUPTOI:
2717    
2718      case OP_POSUPTO:      case OP_POSUPTO:
2719      case OP_POSUPTOI:      case OP_POSUPTOI:
2720        case OP_NOTPOSUPTO:
2721        case OP_NOTPOSUPTOI:
2722    
2723      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]);
2724      break;      break;
2725  #endif  #endif
# Line 2604  for (code = first_significant_code(code Line 2730  for (code = first_significant_code(code
2730      case OP_MARK:      case OP_MARK:
2731      case OP_PRUNE_ARG:      case OP_PRUNE_ARG:
2732      case OP_SKIP_ARG:      case OP_SKIP_ARG:
     code += code[1];  
     break;  
   
2733      case OP_THEN_ARG:      case OP_THEN_ARG:
2734      code += code[1];      code += code[1];
2735      break;      break;
# Line 2650  could_be_empty(const pcre_uchar *code, c Line 2773  could_be_empty(const pcre_uchar *code, c
2773  {  {
2774  while (bcptr != NULL && bcptr->current_branch >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2775    {    {
2776    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd))    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd, NULL))
2777      return FALSE;      return FALSE;
2778    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2779    }    }
# Line 2660  return TRUE; Line 2783  return TRUE;
2783    
2784    
2785  /*************************************************  /*************************************************
2786  *           Check for POSIX class syntax         *  *        Base opcode of repeated opcodes         *
2787  *************************************************/  *************************************************/
2788    
2789  /* This function is called when the sequence "[:" or "[." or "[=" is  /* Returns the base opcode for repeated single character type opcodes. If the
2790  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.  
2791    
2792  Originally, this function only recognized a sequence of letters between the  Arguments:  c opcode
2793  terminators, but it seems that Perl recognizes any sequence of characters,  Returns:    base opcode for the type
2794  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:].  
2795    
2796  The problem in trying to be exactly like Perl is in the handling of escapes. We  static pcre_uchar
2797  have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX  get_repeat_base(pcre_uchar c)
2798  class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code  {
2799  below handles the special case of \], but does not try to do any other escape  return (c > OP_TYPEPOSUPTO)? c :
2800  processing. This makes it different from Perl for cases such as [:l\ower:]         (c >= OP_TYPESTAR)?   OP_TYPESTAR :
2801  where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize         (c >= OP_NOTSTARI)?   OP_NOTSTARI :
2802  "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,         (c >= OP_NOTSTAR)?    OP_NOTSTAR :
2803  I think.         (c >= OP_STARI)?      OP_STARI :
2804                                 OP_STAR;
2805    }
2806    
 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.  
2807    
2808  In Perl, unescaped square brackets may also appear as part of class names. For  
2809  example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for  #ifdef SUPPORT_UCP
2810  [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not  /*************************************************
2811  seem right at all. PCRE does not allow closing square brackets in POSIX class  *        Check a character and a property        *
2812  names.  *************************************************/
2813    
2814    /* This function is called by check_auto_possessive() when a property item
2815    is adjacent to a fixed character.
2816    
2817  Arguments:  Arguments:
2818    ptr      pointer to the initial [    c            the character
2819    endptr   where to return the end pointer    ptype        the property type
2820      pdata        the data for the type
2821      negated      TRUE if it's a negated property (\P or \p{^)
2822    
2823  Returns:   TRUE or FALSE  Returns:       TRUE if auto-possessifying is OK
2824  */  */
2825    
2826  static BOOL  static BOOL
2827  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)  check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata,
2828      BOOL negated)
2829  {  {
2830  int terminator;          /* Don't combine these lines; the Solaris cc */  const pcre_uint32 *p;
2831  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  const ucd_record *prop = GET_UCD(c);
2832  for (++ptr; *ptr != 0; ptr++)  
2833    switch(ptype)
2834    {    {
2835    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)    case PT_LAMP:
2836      ptr++;    return (prop->chartype == ucp_Lu ||
2837    else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;            prop->chartype == ucp_Ll ||
2838    else            prop->chartype == ucp_Lt) == negated;
2839    
2840      case PT_GC:
2841      return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2842    
2843      case PT_PC:
2844      return (pdata == prop->chartype) == negated;
2845    
2846      case PT_SC:
2847      return (pdata == prop->script) == negated;
2848    
2849      /* These are specials */
2850    
2851      case PT_ALNUM:
2852      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2853              PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2854    
2855      /* Perl space used to exclude VT, but from Perl 5.18 it is included, which
2856      means that Perl space and POSIX space are now identical. PCRE was changed
2857      at release 8.34. */
2858    
2859      case PT_SPACE:    /* Perl space */
2860      case PT_PXSPACE:  /* POSIX space */
2861      switch(c)
2862      {      {
2863      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)      HSPACE_CASES:
2864        {      VSPACE_CASES:
2865        *endptr = ptr;      return negated;
2866        return TRUE;  
2867        }      default:
2868      if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z) == negated;
2869           (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||      }
2870            ptr[1] == CHAR_EQUALS_SIGN) &&    break;  /* Control never reaches here */
2871          check_posix_syntax(ptr, endptr))  
2872        return FALSE;    case PT_WORD:
2873      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2874              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2875              c == CHAR_UNDERSCORE) == negated;
2876    
2877      case PT_CLIST:
2878      p = PRIV(ucd_caseless_sets) + prop->caseset;
2879      for (;;)
2880        {
2881        if (c < *p) return !negated;
2882        if (c == *p++) return negated;
2883      }      }
2884      break;  /* Control never reaches here */
2885    }    }
2886    
2887  return FALSE;  return FALSE;
2888  }  }
2889    #endif  /* SUPPORT_UCP */
2890    
2891    
2892    
2893  /*************************************************  /*************************************************
2894  *          Check POSIX class name                *  *        Fill the character property list        *
2895  *************************************************/  *************************************************/
2896    
2897  /* This function is called to check the name given in a POSIX-style class entry  /* Checks whether the code points to an opcode that can take part in auto-
2898  such as [:alnum:].  possessification, and if so, fills a list with its properties.
2899    
2900  Arguments:  Arguments:
2901    ptr        points to the first letter    code        points to start of expression
2902    len        the length of the name    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2903      fcc         points to case-flipping table
2904      list        points to output list
2905                  list[0] will be filled with the opcode
2906                  list[1] will be non-zero if this opcode
2907                    can match an empty character string
2908                  list[2..7] depends on the opcode
2909    
2910  Returns:     a value representing the name, or -1 if unknown  Returns:      points to the start of the next opcode if *code is accepted
2911                  NULL if *code is not accepted
2912  */  */
2913    
2914  static int  static const pcre_uchar *
2915  check_posix_name(const pcre_uchar *ptr, int len)  get_chr_property_list(const pcre_uchar *code, BOOL utf,
2916      const pcre_uint8 *fcc, pcre_uint32 *list)
2917  {  {
2918  const char *pn = posix_names;  pcre_uchar c = *code;
2919  register int yield = 0;  pcre_uchar base;
2920  while (posix_name_lengths[yield] != 0)  const pcre_uchar *end;
2921    pcre_uint32 chr;
2922    
2923    #ifdef SUPPORT_UCP
2924    pcre_uint32 *clist_dest;
2925    const pcre_uint32 *clist_src;
2926    #else
2927    utf = utf;  /* Suppress "unused parameter" compiler warning */
2928    #endif
2929    
2930    list[0] = c;
2931    list[1] = FALSE;
2932    code++;
2933    
2934    if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
2935    {    {
2936    if (len == posix_name_lengths[yield] &&    base = get_repeat_base(c);
2937      STRNCMP_UC_C8(ptr, pn, len) == 0) return yield;    c -= (base - OP_STAR);
   pn += posix_name_lengths[yield] + 1;  
   yield++;  
   }  
 return -1;  
 }  
2938    
2939      if (c == OP_UPTO || c == OP_MINUPTO || c == OP_EXACT || c == OP_POSUPTO)
2940        code += IMM2_SIZE;
2941    
2942  /*************************************************    list[1] = (c != OP_PLUS && c != OP_MINPLUS && c != OP_EXACT && c != OP_POSPLUS);
 *    Adjust OP_RECURSE items in repeated group   *  
 *************************************************/  
2943    
2944  /* OP_RECURSE items contain an offset from the start of the regex to the group    switch(base)
2945  that is referenced. This means that groups can be replicated for fixed      {
2946  repetition simply by copying (because the recursion is allowed to refer to      case OP_STAR:
2947  earlier groups that are outside the current group). However, when a group is      list[0] = OP_CHAR;
2948  optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is      break;
 inserted before it, after it has been compiled. This means that any OP_RECURSE  
 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.  
2949    
2950  This function has been extended with the possibility of forward references for      case OP_STARI:
2951  recursions and subroutine calls. It must also check the list of such references      list[0] = OP_CHARI;
2952  for the group we are dealing with. If it finds that one of the recursions in      break;
 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).  
2953    
2954  Arguments:      case OP_NOTSTAR:
2955    group      points to the start of the group      list[0] = OP_NOT;
2956    adjust     the amount by which the group is to be moved      break;
   utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode  
   cd         contains pointers to tables etc.  
   save_hwm   the hwm forward reference pointer at the start of the group  
2957    
2958  Returns:     nothing      case OP_NOTSTARI:
2959  */      list[0] = OP_NOTI;
2960        break;
2961    
2962  static void      case OP_TYPESTAR:
2963  adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,      list[0] = *code;
2964    pcre_uchar *save_hwm)      code++;
2965  {      break;
2966  pcre_uchar *ptr = group;      }
2967      c = list[0];
2968      }
2969    
2970  while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)  switch(c)
2971    {    {
2972    int offset;    case OP_NOT_DIGIT:
2973    pcre_uchar *hc;    case OP_DIGIT:
2974      case OP_NOT_WHITESPACE:
2975      case OP_WHITESPACE:
2976      case OP_NOT_WORDCHAR:
2977      case OP_WORDCHAR:
2978      case OP_ANY:
2979      case OP_ALLANY:
2980      case OP_ANYNL:
2981      case OP_NOT_HSPACE:
2982      case OP_HSPACE:
2983      case OP_NOT_VSPACE:
2984      case OP_VSPACE:
2985      case OP_EXTUNI:
2986      case OP_EODN:
2987      case OP_EOD:
2988      case OP_DOLL:
2989      case OP_DOLLM:
2990      return code;
2991    
2992    /* See if this recursion is on the forward reference list. If so, adjust the    case OP_CHAR:
2993    reference. */    case OP_NOT:
2994      GETCHARINCTEST(chr, code);
2995      list[2] = chr;
2996      list[3] = NOTACHAR;
2997      return code;
2998    
2999    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)    case OP_CHARI:
3000      {    case OP_NOTI:
3001      offset = GET(hc, 0);    list[0] = (c == OP_CHARI) ? OP_CHAR : OP_NOT;
3002      if (cd->start_code + offset == ptr + 1)    GETCHARINCTEST(chr, code);
3003        {    list[2] = chr;
       PUT(hc, 0, offset + adjust);  
       break;  
       }  
     }  
3004    
3005    /* Otherwise, adjust the recursion offset if it's after the start of this  #ifdef SUPPORT_UCP
3006    group. */    if (chr < 128 || (chr < 256 && !utf))
3007        list[3] = fcc[chr];
3008      else
3009        list[3] = UCD_OTHERCASE(chr);
3010    #elif defined SUPPORT_UTF || !defined COMPILE_PCRE8
3011      list[3] = (chr < 256) ? fcc[chr] : chr;
3012    #else
3013      list[3] = fcc[chr];
3014    #endif
3015    
3016    if (hc >= cd->hwm)    /* The othercase might be the same value. */
3017    
3018      if (chr == list[3])
3019        list[3] = NOTACHAR;
3020      else
3021        list[4] = NOTACHAR;
3022      return code;
3023    
3024    #ifdef SUPPORT_UCP
3025      case OP_PROP:
3026      case OP_NOTPROP:
3027      if (code[0] != PT_CLIST)
3028      {      {
3029      offset = GET(ptr, 1);      list[2] = code[0];
3030      if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);      list[3] = code[1];
3031        return code + 2;
3032      }      }
3033    
3034    ptr += 1 + LINK_SIZE;    /* Convert only if we have enough space. */
3035    }  
3036  }    clist_src = PRIV(ucd_caseless_sets) + code[1];
3037      clist_dest = list + 2;
3038      code += 2;
3039    
3040      do {
3041         if (clist_dest >= list + 8)
3042           {
3043           /* Early return if there is not enough space. This should never
3044           happen, since all clists are shorter than 5 character now. */
3045           list[2] = code[0];
3046           list[3] = code[1];
3047           return code;
3048           }
3049         *clist_dest++ = *clist_src;
3050         }
3051      while(*clist_src++ != NOTACHAR);
3052    
3053      /* All characters are stored. The terminating NOTACHAR
3054      is copied form the clist itself. */
3055    
3056  /*************************************************    list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
3057  *        Insert an automatic callout point       *    return code;
3058  *************************************************/  #endif
3059    
3060  /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert    case OP_NCLASS:
3061  callout points before each pattern item.    case OP_CLASS:
3062    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3063      case OP_XCLASS:
3064      if (c == OP_XCLASS)
3065        end = code + GET(code, 0) - 1;
3066      else
3067    #endif
3068        end = code + 32 / sizeof(pcre_uchar);
3069    
3070  Arguments:    switch(*end)
3071    code           current code pointer      {
3072    ptr            current pattern pointer      case OP_CRSTAR:
3073    cd             pointers to tables etc      case OP_CRMINSTAR:
3074        case OP_CRQUERY:
3075        case OP_CRMINQUERY:
3076        case OP_CRPOSSTAR:
3077        case OP_CRPOSQUERY:
3078        list[1] = TRUE;
3079        end++;
3080        break;
3081    
3082  Returns:         new code pointer      case OP_CRPLUS:
3083  */      case OP_CRMINPLUS:
3084        case OP_CRPOSPLUS:
3085        end++;
3086        break;
3087    
3088  static pcre_uchar *      case OP_CRRANGE:
3089  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)      case OP_CRMINRANGE:
3090  {      case OP_CRPOSRANGE:
3091  *code++ = OP_CALLOUT;      list[1] = (GET2(end, 1) == 0);
3092  *code++ = 255;      end += 1 + 2 * IMM2_SIZE;
3093  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */      break;
3094  PUT(code, LINK_SIZE, 0);                       /* Default length */      }
3095  return code + 2 * LINK_SIZE;    list[2] = (pcre_uint32)(end - code);
3096      return end;
3097      }
3098    return NULL;    /* Opcode not accepted */
3099  }  }
3100    
3101    
3102    
3103  /*************************************************  /*************************************************
3104  *         Complete a callout item                *  *    Scan further character sets for match       *
3105  *************************************************/  *************************************************/
3106    
3107  /* A callout item contains the length of the next item in the pattern, which  /* Checks whether the base and the current opcode have a common character, in
3108  we can't fill in till after we have reached the relevant point. This is used  which case the base cannot be possessified.
 for both automatic and manual callouts.  
3109    
3110  Arguments:  Arguments:
3111    previous_callout   points to previous callout item    code        points to the byte code
3112    ptr                current pattern pointer    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3113    cd                 pointers to tables etc    cd          static compile data
3114      base_list   the data list of the base opcode
3115    
3116  Returns:             nothing  Returns:      TRUE if the auto-possessification is possible
3117  */  */
3118    
3119  static void  static BOOL
3120  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)  compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
3121      const pcre_uint32 *base_list, const pcre_uchar *base_end, int *rec_limit)
3122  {  {
3123  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));  pcre_uchar c;
3124  PUT(previous_callout, 2 + LINK_SIZE, length);  pcre_uint32 list[8];
3125  }  const pcre_uint32 *chr_ptr;
3126    const pcre_uint32 *ochr_ptr;
3127    const pcre_uint32 *list_ptr;
3128    const pcre_uchar *next_code;
3129    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3130    const pcre_uchar *xclass_flags;
3131    #endif
3132    const pcre_uint8 *class_bitset;
3133    const pcre_uint8 *set1, *set2, *set_end;
3134    pcre_uint32 chr;
3135    BOOL accepted, invert_bits;
3136    BOOL entered_a_group = FALSE;
3137    
3138    if (*rec_limit == 0) return FALSE;
3139    --(*rec_limit);
3140    
3141    /* Note: the base_list[1] contains whether the current opcode has greedy
3142    (represented by a non-zero value) quantifier. This is a different from
3143    other character type lists, which stores here that the character iterator
3144    matches to an empty string (also represented by a non-zero value). */
3145    
3146    for(;;)
3147      {
3148      /* All operations move the code pointer forward.
3149      Therefore infinite recursions are not possible. */
3150    
3151      c = *code;
3152    
3153  #ifdef SUPPORT_UCP    /* Skip over callouts */
 /*************************************************  
 *           Get othercase range                  *  
 *************************************************/  
3154    
3155  /* This function is passed the start and end of a class range, in UTF-8 mode    if (c == OP_CALLOUT)
3156  with UCP support. It searches up the characters, looking for ranges of      {
3157  characters in the "other" case. Each call returns the next one, updating the      code += PRIV(OP_lengths)[c];
3158  start address. A character with multiple other cases is returned on its own      continue;
3159  with a special return value.      }
3160    
3161  Arguments:    if (c == OP_ALT)
3162    cptr        points to starting character value; updated      {
3163    d           end value      do code += GET(code, 1); while (*code == OP_ALT);
3164    ocptr       where to put start of othercase range      c = *code;
3165    odptr       where to put end of othercase range      }
3166    
3167  Yield:        -1 when no more    switch(c)
3168                 0 when a range is returned      {
3169                >0 the CASESET offset for char with multiple other cases      case OP_END:
3170                  in this case, ocptr contains the original      case OP_KETRPOS:
3171  */      /* TRUE only in greedy case. The non-greedy case could be replaced by
3172        an OP_EXACT, but it is probably not worth it. (And note that OP_EXACT
3173        uses more memory, which we cannot get at this stage.) */
3174    
3175  static int      return base_list[1] != 0;
 get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,  
   unsigned int *odptr)  
 {  
 unsigned int c, othercase, next;  
 int co;  
3176    
3177  /* Find the first character that has an other case. If it has multiple other      case OP_KET:
3178  cases, return its case offset value. */      /* If the bracket is capturing, and referenced by an OP_RECURSE, or
3179        it is an atomic sub-pattern (assert, once, etc.) the non-greedy case
3180        cannot be converted to a possessive form. */
3181    
3182        if (base_list[1] == 0) return FALSE;
3183    
3184        switch(*(code - GET(code, 1)))
3185          {
3186          case OP_ASSERT:
3187          case OP_ASSERT_NOT:
3188          case OP_ASSERTBACK:
3189          case OP_ASSERTBACK_NOT:
3190          case OP_ONCE:
3191          case OP_ONCE_NC:
3192          /* Atomic sub-patterns and assertions can always auto-possessify their
3193          last iterator. However, if the group was entered as a result of checking
3194          a previous iterator, this is not possible. */
3195    
3196  for (c = *cptr; c <= d; c++)        return !entered_a_group;
3197    {        }
   if ((co = UCD_CASESET(c)) != 0)  
     {  
     *ocptr = c++;   /* Character that has the set */  
     *cptr = c;      /* Rest of input range */  
     return co;  
     }  
   if ((othercase = UCD_OTHERCASE(c)) != c) break;  
   }  
3198    
3199  if (c > d) return -1;  /* Reached end of range */      code += PRIV(OP_lengths)[c];
3200        continue;
3201    
3202  *ocptr = othercase;      case OP_ONCE:
3203  next = othercase + 1;      case OP_ONCE_NC:
3204        case OP_BRA:
3205        case OP_CBRA:
3206        next_code = code + GET(code, 1);
3207        code += PRIV(OP_lengths)[c];
3208    
3209  for (++c; c <= d; c++)      while (*next_code == OP_ALT)
3210    {        {
3211    if (UCD_OTHERCASE(c) != next) break;        if (!compare_opcodes(code, utf, cd, base_list, base_end, rec_limit))
3212    next++;          return FALSE;
3213    }        code = next_code + 1 + LINK_SIZE;
3214          next_code += GET(next_code, 1);
3215          }
3216    
3217  *odptr = next - 1;     /* End of othercase range */      entered_a_group = TRUE;
3218  *cptr = c;             /* Rest of input range */      continue;
 return 0;  
 }  
3219    
3220        case OP_BRAZERO:
3221        case OP_BRAMINZERO:
3222    
3223        next_code = code + 1;
3224        if (*next_code != OP_BRA && *next_code != OP_CBRA
3225            && *next_code != OP_ONCE && *next_code != OP_ONCE_NC) return FALSE;
3226    
3227        do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
3228    
3229        /* The bracket content will be checked by the
3230        OP_BRA/OP_CBRA case above. */
3231        next_code += 1 + LINK_SIZE;
3232        if (!compare_opcodes(next_code, utf, cd, base_list, base_end, rec_limit))
3233          return FALSE;
3234    
3235  /*************************************************      code += PRIV(OP_lengths)[c];
3236  *        Check a character and a property        *      continue;
 *************************************************/  
3237    
3238  /* This function is called by check_auto_possessive() when a property item      default:
3239  is adjacent to a fixed character.      break;
3240        }
3241    
3242  Arguments:    /* Check for a supported opcode, and load its properties. */
   c            the character  
   ptype        the property type  
   pdata        the data for the type  
   negated      TRUE if it's a negated property (\P or \p{^)  
3243    
3244  Returns:       TRUE if auto-possessifying is OK    code = get_chr_property_list(code, utf, cd->fcc, list);
3245  */    if (code == NULL) return FALSE;    /* Unsupported */
3246    
3247  static BOOL    /* If either opcode is a small character list, set pointers for comparing
3248  check_char_prop(pcre_uint32 c, int ptype, int pdata, BOOL negated)    characters from that list with another list, or with a property. */
 {  
 #ifdef SUPPORT_UCP  
 const pcre_uint32 *p;  
 #endif  
3249    
3250  const ucd_record *prop = GET_UCD(c);    if (base_list[0] == OP_CHAR)
3251        {
3252        chr_ptr = base_list + 2;
3253        list_ptr = list;
3254        }
3255      else if (list[0] == OP_CHAR)
3256        {
3257        chr_ptr = list + 2;
3258        list_ptr = base_list;
3259        }
3260    
3261  switch(ptype)    /* Character bitsets can also be compared to certain opcodes. */
   {  
   case PT_LAMP:  
   return (prop->chartype == ucp_Lu ||  
           prop->chartype == ucp_Ll ||  
           prop->chartype == ucp_Lt) == negated;  
3262    
3263    case PT_GC:    else if (base_list[0] == OP_CLASS || list[0] == OP_CLASS
3264    return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;  #ifdef COMPILE_PCRE8
3265          /* In 8 bit, non-UTF mode, OP_CLASS and OP_NCLASS are the same. */
3266          || (!utf && (base_list[0] == OP_NCLASS || list[0] == OP_NCLASS))
3267    #endif
3268          )
3269        {
3270    #ifdef COMPILE_PCRE8
3271        if (base_list[0] == OP_CLASS || (!utf && base_list[0] == OP_NCLASS))
3272    #else
3273        if (base_list[0] == OP_CLASS)
3274    #endif
3275          {
3276          set1 = (pcre_uint8 *)(base_end - base_list[2]);
3277          list_ptr = list;
3278          }
3279        else
3280          {
3281          set1 = (pcre_uint8 *)(code - list[2]);
3282          list_ptr = base_list;
3283          }
3284    
3285    case PT_PC:      invert_bits = FALSE;
3286    return (pdata == prop->chartype) == negated;      switch(list_ptr[0])
3287          {
3288          case OP_CLASS:
3289          case OP_NCLASS:
3290          set2 = (pcre_uint8 *)
3291            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3292          break;
3293    
3294    case PT_SC:  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3295    return (pdata == prop->script) == negated;        case OP_XCLASS:
3296          xclass_flags = (list_ptr == list ? code : base_end) - list_ptr[2] + LINK_SIZE;
3297          if ((*xclass_flags & XCL_HASPROP) != 0) return FALSE;
3298          if ((*xclass_flags & XCL_MAP) == 0)
3299            {
3300            /* No bits are set for characters < 256. */
3301            if (list[1] == 0) return TRUE;
3302            /* Might be an empty repeat. */
3303            continue;
3304            }
3305          set2 = (pcre_uint8 *)(xclass_flags + 1);
3306          break;
3307    #endif
3308    
3309    /* These are specials */        case OP_NOT_DIGIT:
3310          invert_bits = TRUE;
3311          /* Fall through */
3312          case OP_DIGIT:
3313          set2 = (pcre_uint8 *)(cd->cbits + cbit_digit);
3314          break;
3315    
3316    case PT_ALNUM:        case OP_NOT_WHITESPACE:
3317    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||        invert_bits = TRUE;
3318            PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;        /* Fall through */
3319          case OP_WHITESPACE:
3320          set2 = (pcre_uint8 *)(cd->cbits + cbit_space);
3321          break;
3322    
3323    case PT_SPACE:    /* Perl space */        case OP_NOT_WORDCHAR:
3324    return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||        invert_bits = TRUE;
3325            c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)        /* Fall through */
3326            == negated;        case OP_WORDCHAR:
3327          set2 = (pcre_uint8 *)(cd->cbits + cbit_word);
3328          break;
3329    
3330    case PT_PXSPACE:  /* POSIX space */        default:
3331    return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||        return FALSE;
3332            c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||        }
           c == CHAR_FF || c == CHAR_CR)  
           == negated;  
3333    
3334    case PT_WORD:      /* Because the sets are unaligned, we need
3335    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||      to perform byte comparison here. */
3336            PRIV(ucp_gentype)[prop->chartype] == ucp_N ||      set_end = set1 + 32;
3337            c == CHAR_UNDERSCORE) == negated;      if (invert_bits)
3338          {
3339  #ifdef SUPPORT_UCP        do
3340    case PT_CLIST:          {
3341    p = PRIV(ucd_caseless_sets) + prop->caseset;          if ((*set1++ & ~(*set2++)) != 0) return FALSE;
3342    for (;;)          }
3343      {        while (set1 < set_end);
3344      if ((unsigned int)c < *p) return !negated;        }
3345      if ((unsigned int)c == *p++) return negated;      else
3346          {
3347          do
3348            {
3349            if ((*set1++ & *set2++) != 0) return FALSE;
3350            }
3351          while (set1 < set_end);
3352          }
3353    
3354        if (list[1] == 0) return TRUE;
3355        /* Might be an empty repeat. */
3356        continue;
3357      }      }
3358    break;  /* Control never reaches here */  
3359      /* Some property combinations also acceptable. Unicode property opcodes are
3360      processed specially; the rest can be handled with a lookup table. */
3361    
3362      else
3363        {
3364        pcre_uint32 leftop, rightop;
3365    
3366        leftop = base_list[0];
3367        rightop = list[0];
3368    
3369    #ifdef SUPPORT_UCP
3370        accepted = FALSE; /* Always set in non-unicode case. */
3371        if (leftop == OP_PROP || leftop == OP_NOTPROP)
3372          {
3373          if (rightop == OP_EOD)
3374            accepted = TRUE;
3375          else if (rightop == OP_PROP || rightop == OP_NOTPROP)
3376            {
3377            int n;
3378            const pcre_uint8 *p;
3379            BOOL same = leftop == rightop;
3380            BOOL lisprop = leftop == OP_PROP;
3381            BOOL risprop = rightop == OP_PROP;
3382            BOOL bothprop = lisprop && risprop;
3383    
3384            /* There's a table that specifies how each combination is to be
3385            processed:
3386              0   Always return FALSE (never auto-possessify)
3387              1   Character groups are distinct (possessify if both are OP_PROP)
3388              2   Check character categories in the same group (general or particular)
3389              3   Return TRUE if the two opcodes are not the same
3390              ... see comments below
3391            */
3392    
3393            n = propposstab[base_list[2]][list[2]];
3394            switch(n)
3395              {
3396              case 0: break;
3397              case 1: accepted = bothprop; break;
3398              case 2: accepted = (base_list[3] == list[3]) != same; break;
3399              case 3: accepted = !same; break;
3400    
3401              case 4:  /* Left general category, right particular category */
3402              accepted = risprop && catposstab[base_list[3]][list[3]] == same;
3403              break;
3404    
3405              case 5:  /* Right general category, left particular category */
3406              accepted = lisprop && catposstab[list[3]][base_list[3]] == same;
3407              break;
3408    
3409              /* This code is logically tricky. Think hard before fiddling with it.
3410              The posspropstab table has four entries per row. Each row relates to
3411              one of PCRE's special properties such as ALNUM or SPACE or WORD.
3412              Only WORD actually needs all four entries, but using repeats for the
3413              others means they can all use the same code below.
3414    
3415              The first two entries in each row are Unicode general categories, and
3416              apply always, because all the characters they include are part of the
3417              PCRE character set. The third and fourth entries are a general and a
3418              particular category, respectively, that include one or more relevant
3419              characters. One or the other is used, depending on whether the check
3420              is for a general or a particular category. However, in both cases the
3421              category contains more characters than the specials that are defined
3422              for the property being tested against. Therefore, it cannot be used
3423              in a NOTPROP case.
3424    
3425              Example: the row for WORD contains ucp_L, ucp_N, ucp_P, ucp_Po.
3426              Underscore is covered by ucp_P or ucp_Po. */
3427    
3428              case 6:  /* Left alphanum vs right general category */
3429              case 7:  /* Left space vs right general category */
3430              case 8:  /* Left word vs right general category */
3431              p = posspropstab[n-6];
3432              accepted = risprop && lisprop ==
3433                (list[3] != p[0] &&
3434                 list[3] != p[1] &&
3435                (list[3] != p[2] || !lisprop));
3436              break;
3437    
3438              case 9:   /* Right alphanum vs left general category */
3439              case 10:  /* Right space vs left general category */
3440              case 11:  /* Right word vs left general category */
3441              p = posspropstab[n-9];
3442              accepted = lisprop && risprop ==
3443                (base_list[3] != p[0] &&
3444                 base_list[3] != p[1] &&
3445                (base_list[3] != p[2] || !risprop));
3446              break;
3447    
3448              case 12:  /* Left alphanum vs right particular category */
3449              case 13:  /* Left space vs right particular category */
3450              case 14:  /* Left word vs right particular category */
3451              p = posspropstab[n-12];
3452              accepted = risprop && lisprop ==
3453                (catposstab[p[0]][list[3]] &&
3454                 catposstab[p[1]][list[3]] &&
3455                (list[3] != p[3] || !lisprop));
3456              break;
3457    
3458              case 15:  /* Right alphanum vs left particular category */
3459              case 16:  /* Right space vs left particular category */
3460              case 17:  /* Right word vs left particular category */
3461              p = posspropstab[n-15];
3462              accepted = lisprop && risprop ==
3463                (catposstab[p[0]][base_list[3]] &&
3464                 catposstab[p[1]][base_list[3]] &&
3465                (base_list[3] != p[3] || !risprop));
3466              break;
3467              }
3468            }
3469          }
3470    
3471        else
3472    #endif  /* SUPPORT_UCP */
3473    
3474        accepted = leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&
3475               rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&
3476               autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];
3477    
3478        if (!accepted) return FALSE;
3479    
3480        if (list[1] == 0) return TRUE;
3481        /* Might be an empty repeat. */
3482        continue;
3483        }
3484    
3485      /* Control reaches here only if one of the items is a small character list.
3486      All characters are checked against the other side. */
3487    
3488      do
3489        {
3490        chr = *chr_ptr;
3491    
3492        switch(list_ptr[0])
3493          {
3494          case OP_CHAR:
3495          ochr_ptr = list_ptr + 2;
3496          do
3497            {
3498            if (chr == *ochr_ptr) return FALSE;
3499            ochr_ptr++;
3500            }
3501          while(*ochr_ptr != NOTACHAR);
3502          break;
3503    
3504          case OP_NOT:
3505          ochr_ptr = list_ptr + 2;
3506          do
3507            {
3508            if (chr == *ochr_ptr)
3509              break;
3510            ochr_ptr++;
3511            }
3512          while(*ochr_ptr != NOTACHAR);
3513          if (*ochr_ptr == NOTACHAR) return FALSE;   /* Not found */
3514          break;
3515    
3516          /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not*
3517          set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3518    
3519          case OP_DIGIT:
3520          if (chr < 256 && (cd->ctypes[chr] & ctype_digit) != 0) return FALSE;
3521          break;
3522    
3523          case OP_NOT_DIGIT:
3524          if (chr > 255 || (cd->ctypes[chr] & ctype_digit) == 0) return FALSE;
3525          break;
3526    
3527          case OP_WHITESPACE:
3528          if (chr < 256 && (cd->ctypes[chr] & ctype_space) != 0) return FALSE;
3529          break;
3530    
3531          case OP_NOT_WHITESPACE:
3532          if (chr > 255 || (cd->ctypes[chr] & ctype_space) == 0) return FALSE;
3533          break;
3534    
3535          case OP_WORDCHAR:
3536          if (chr < 255 && (cd->ctypes[chr] & ctype_word) != 0) return FALSE;
3537          break;
3538    
3539          case OP_NOT_WORDCHAR:
3540          if (chr > 255 || (cd->ctypes[chr] & ctype_word) == 0) return FALSE;
3541          break;
3542    
3543          case OP_HSPACE:
3544          switch(chr)
3545            {
3546            HSPACE_CASES: return FALSE;
3547            default: break;
3548            }
3549          break;
3550    
3551          case OP_NOT_HSPACE:
3552          switch(chr)
3553            {
3554            HSPACE_CASES: break;
3555            default: return FALSE;
3556            }
3557          break;
3558    
3559          case OP_ANYNL:
3560          case OP_VSPACE:
3561          switch(chr)
3562            {
3563            VSPACE_CASES: return FALSE;
3564            default: break;
3565            }
3566          break;
3567    
3568          case OP_NOT_VSPACE:
3569          switch(chr)
3570            {
3571            VSPACE_CASES: break;
3572            default: return FALSE;
3573            }
3574          break;
3575    
3576          case OP_DOLL:
3577          case OP_EODN:
3578          switch (chr)
3579            {
3580            case CHAR_CR:
3581            case CHAR_LF:
3582            case CHAR_VT:
3583            case CHAR_FF:
3584            case CHAR_NEL:
3585    #ifndef EBCDIC
3586            case 0x2028:
3587            case 0x2029:
3588    #endif  /* Not EBCDIC */
3589            return FALSE;
3590            }
3591          break;
3592    
3593          case OP_EOD:    /* Can always possessify before \z */
3594          break;
3595    
3596    #ifdef SUPPORT_UCP
3597          case OP_PROP:
3598          case OP_NOTPROP:
3599          if (!check_char_prop(chr, list_ptr[2], list_ptr[3],
3600                list_ptr[0] == OP_NOTPROP))
3601            return FALSE;
3602          break;
3603    #endif
3604    
3605          case OP_NCLASS:
3606          if (chr > 255) return FALSE;
3607          /* Fall through */
3608    
3609          case OP_CLASS:
3610          if (chr > 255) break;
3611          class_bitset = (pcre_uint8 *)
3612            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3613          if ((class_bitset[chr >> 3] & (1 << (chr & 7))) != 0) return FALSE;
3614          break;
3615    
3616    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3617          case OP_XCLASS:
3618          if (PRIV(xclass)(chr, (list_ptr == list ? code : base_end) -
3619              list_ptr[2] + LINK_SIZE, utf)) return FALSE;
3620          break;
3621  #endif  #endif
3622    
3623          default:
3624          return FALSE;
3625          }
3626    
3627        chr_ptr++;
3628        }
3629      while(*chr_ptr != NOTACHAR);
3630    
3631      /* At least one character must be matched from this opcode. */
3632    
3633      if (list[1] == 0) return TRUE;
3634    }    }
3635    
3636  return FALSE;  /* Control never reaches here. There used to be a fail-save return FALSE; here,
3637    but some compilers complain about an unreachable statement. */
3638    
3639  }  }
 #endif  /* SUPPORT_UCP */  
3640    
3641    
3642    
3643  /*************************************************  /*************************************************
3644  *     Check if auto-possessifying is possible    *  *    Scan compiled regex for auto-possession     *
3645  *************************************************/  *************************************************/
3646    
3647  /* This function is called for unlimited repeats of certain items, to see  /* Replaces single character iterations with their possessive alternatives
3648  whether the next thing could possibly match the repeated item. If not, it makes  if appropriate. This function modifies the compiled opcode!
 sense to automatically possessify the repeated item.  
3649    
3650  Arguments:  Arguments:
3651    previous      pointer to the repeated opcode    code        points to start of the byte code
3652    utf           TRUE in UTF-8 / UTF-16 / UTF-32 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3653    ptr           next character in pattern    cd          static compile data
   options       options bits  
   cd            contains pointers to tables etc.  
3654    
3655  Returns:        TRUE if possessifying is wanted  Returns:      nothing
3656  */  */
3657    
3658  static BOOL  static void
3659  check_auto_possessive(const pcre_uchar *previous, BOOL utf,  auto_possessify(pcre_uchar *code, BOOL utf, const compile_data *cd)
   const pcre_uchar *ptr, int options, compile_data *cd)  
3660  {  {
3661  pcre_uint32 c = NOTACHAR;  register pcre_uchar c;
3662  pcre_uint32 next;  const pcre_uchar *end;
3663  int escape;  pcre_uchar *repeat_opcode;
3664  int op_code = *previous++;  pcre_uint32 list[8];
3665    int rec_limit;
3666    
3667  /* Skip whitespace and comments in extended mode */  for (;;)
   
 if ((options & PCRE_EXTENDED) != 0)  
3668    {    {
3669    for (;;)    c = *code;
3670    
3671      /* When a pattern with bad UTF-8 encoding is compiled with NO_UTF_CHECK,
3672      it may compile without complaining, but may get into a loop here if the code
3673      pointer points to a bad value. This is, of course a documentated possibility,
3674      when NO_UTF_CHECK is set, so it isn't a bug, but we can detect this case and
3675      just give up on this optimization. */
3676    
3677      if (c >= OP_TABLE_LENGTH) return;
3678    
3679      if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
3680      {      {
3681      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      c -= get_repeat_base(c) - OP_STAR;
3682      if (*ptr == CHAR_NUMBER_SIGN)      end = (c <= OP_MINUPTO) ?
3683          get_chr_property_list(code, utf, cd->fcc, list) : NULL;
3684        list[1] = c == OP_STAR || c == OP_PLUS || c == OP_QUERY || c == OP_UPTO;
3685    
3686        rec_limit = 1000;
3687        if (end != NULL && compare_opcodes(end, utf, cd, list, end, &rec_limit))
3688        {        {
3689        ptr++;        switch(c)
       while (*ptr != 0)  
3690          {          {
3691          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          case OP_STAR:
3692          ptr++;          *code += OP_POSSTAR - OP_STAR;
3693  #ifdef SUPPORT_UTF          break;
         if (utf) FORWARDCHAR(ptr);  
 #endif  
         }  
       }  
     else break;  
     }  
   }  
3694    
3695  /* If the next item is one that we can handle, get its value. A non-negative          case OP_MINSTAR:
3696  value is a character, a negative value is an escape value. */          *code += OP_POSSTAR - OP_MINSTAR;
3697            break;
3698    
3699  if (*ptr == CHAR_BACKSLASH)          case OP_PLUS:
3700    {          *code += OP_POSPLUS - OP_PLUS;
3701    int temperrorcode = 0;          break;
   escape = check_escape(&ptr, &next, &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)  
   {  
   escape = 0;  
 #ifdef SUPPORT_UTF  
   if (utf) { GETCHARINC(next, ptr); } else  
 #endif  
   next = *ptr++;  
   }  
 else return FALSE;  
3702    
3703  /* Skip whitespace and comments in extended mode */          case OP_MINPLUS:
3704            *code += OP_POSPLUS - OP_MINPLUS;
3705            break;
3706    
3707  if ((options & PCRE_EXTENDED) != 0)          case OP_QUERY:
3708    {          *code += OP_POSQUERY - OP_QUERY;
3709    for (;;)          break;
3710    
3711            case OP_MINQUERY:
3712            *code += OP_POSQUERY - OP_MINQUERY;
3713            break;
3714    
3715            case OP_UPTO:
3716            *code += OP_POSUPTO - OP_UPTO;
3717            break;
3718    
3719            case OP_MINUPTO:
3720            *code += OP_POSUPTO - OP_MINUPTO;
3721            break;
3722            }
3723          }
3724        c = *code;
3725        }
3726      else if (c == OP_CLASS || c == OP_NCLASS || c == OP_XCLASS)
3727      {      {
3728      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3729      if (*ptr == CHAR_NUMBER_SIGN)      if (c == OP_XCLASS)
3730          repeat_opcode = code + GET(code, 1);
3731        else
3732    #endif
3733          repeat_opcode = code + 1 + (32 / sizeof(pcre_uchar));
3734    
3735        c = *repeat_opcode;
3736        if (c >= OP_CRSTAR && c <= OP_CRMINRANGE)
3737        {        {
3738        ptr++;        /* end must not be NULL. */
3739        while (*ptr != 0)        end = get_chr_property_list(code, utf, cd->fcc, list);
3740    
3741          list[1] = (c & 1) == 0;
3742    
3743          rec_limit = 1000;
3744          if (compare_opcodes(end, utf, cd, list, end, &rec_limit))
3745          {          {
3746          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          switch (c)
3747          ptr++;            {
3748  #ifdef SUPPORT_UTF            case OP_CRSTAR:
3749          if (utf) FORWARDCHAR(ptr);            case OP_CRMINSTAR:
3750  #endif            *repeat_opcode = OP_CRPOSSTAR;
3751              break;
3752    
3753              case OP_CRPLUS:
3754              case OP_CRMINPLUS:
3755              *repeat_opcode = OP_CRPOSPLUS;
3756              break;
3757    
3758              case OP_CRQUERY:
3759              case OP_CRMINQUERY:
3760              *repeat_opcode = OP_CRPOSQUERY;
3761              break;
3762    
3763              case OP_CRRANGE:
3764              case OP_CRMINRANGE:
3765              *repeat_opcode = OP_CRPOSRANGE;
3766              break;
3767              }
3768          }          }
3769        }        }
3770      else break;      c = *code;
3771      }      }
   }  
3772    
3773  /* If the next thing is itself optional, we have to give up. */    switch(c)
3774        {
3775        case OP_END:
3776        return;
3777    
3778  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||      case OP_TYPESTAR:
3779    STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)      case OP_TYPEMINSTAR:
3780      return FALSE;      case OP_TYPEPLUS:
3781        case OP_TYPEMINPLUS:
3782        case OP_TYPEQUERY:
3783        case OP_TYPEMINQUERY:
3784        case OP_TYPEPOSSTAR:
3785        case OP_TYPEPOSPLUS:
3786        case OP_TYPEPOSQUERY:
3787        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
3788        break;
3789    
3790  /* If the previous item is a character, get its value. */      case OP_TYPEUPTO:
3791        case OP_TYPEMINUPTO:
3792        case OP_TYPEEXACT:
3793        case OP_TYPEPOSUPTO:
3794        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
3795          code += 2;
3796        break;
3797    
3798  if (op_code == OP_CHAR || op_code == OP_CHARI ||  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3799      op_code == OP_NOT || op_code == OP_NOTI)      case OP_XCLASS:
3800    //if (escape == 0) switch(op_code)      code += GET(code, 1);
3801    {      break;
 #ifdef SUPPORT_UTF  
   GETCHARTEST(c, previous);  
 #else  
   c = *previous;  
3802  #endif  #endif
   }  
3803    
3804  /* Now compare the next item with the previous opcode. First, handle cases when      case OP_MARK:
3805  the next item is a character. */      case OP_PRUNE_ARG:
3806        case OP_SKIP_ARG:
3807        case OP_THEN_ARG:
3808        code += code[1];
3809        break;
3810        }
3811    
3812  if (escape == 0)    /* Add in the fixed length from the table */
3813    {  
3814    /* For a caseless UTF match, the next character may have more than one other    code += PRIV(OP_lengths)[c];
3815    case, which maps to the special PT_CLIST property. Check this first. */  
3816      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
3817  #ifdef SUPPORT_UCP    a multi-byte character. The length in the table is a minimum, so we have to
3818    if (utf && c != NOTACHAR && (options & PCRE_CASELESS) != 0)    arrange to skip the extra bytes. */
3819    
3820    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3821      if (utf) switch(c)
3822      {      {
3823      int ocs = UCD_CASESET(next);      case OP_CHAR:
3824      if (ocs > 0) return check_char_prop(c, PT_CLIST, ocs, op_code >= OP_NOT);      case OP_CHARI:
3825        case OP_NOT:
3826        case OP_NOTI:
3827        case OP_STAR:
3828        case OP_MINSTAR:
3829        case OP_PLUS:
3830        case OP_MINPLUS:
3831        case OP_QUERY:
3832        case OP_MINQUERY:
3833        case OP_UPTO:
3834        case OP_MINUPTO:
3835        case OP_EXACT:
3836        case OP_POSSTAR:
3837        case OP_POSPLUS:
3838        case OP_POSQUERY:
3839        case OP_POSUPTO:
3840        case OP_STARI:
3841        case OP_MINSTARI:
3842        case OP_PLUSI:
3843        case OP_MINPLUSI:
3844        case OP_QUERYI:
3845        case OP_MINQUERYI:
3846        case OP_UPTOI:
3847        case OP_MINUPTOI:
3848        case OP_EXACTI:
3849        case OP_POSSTARI:
3850        case OP_POSPLUSI:
3851        case OP_POSQUERYI:
3852        case OP_POSUPTOI:
3853        case OP_NOTSTAR:
3854        case OP_NOTMINSTAR:
3855        case OP_NOTPLUS:
3856        case OP_NOTMINPLUS:
3857        case OP_NOTQUERY:
3858        case OP_NOTMINQUERY:
3859        case OP_NOTUPTO:
3860        case OP_NOTMINUPTO:
3861        case OP_NOTEXACT:
3862        case OP_NOTPOSSTAR:
3863        case OP_NOTPOSPLUS:
3864        case OP_NOTPOSQUERY:
3865        case OP_NOTPOSUPTO:
3866        case OP_NOTSTARI:
3867        case OP_NOTMINSTARI:
3868        case OP_NOTPLUSI:
3869        case OP_NOTMINPLUSI:
3870        case OP_NOTQUERYI:
3871        case OP_NOTMINQUERYI:
3872        case OP_NOTUPTOI:
3873        case OP_NOTMINUPTOI:
3874        case OP_NOTEXACTI:
3875        case OP_NOTPOSSTARI:
3876        case OP_NOTPOSPLUSI:
3877        case OP_NOTPOSQUERYI:
3878        case OP_NOTPOSUPTOI:
3879        if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
3880        break;
3881      }      }
3882    #else
3883      (void)(utf);  /* Keep compiler happy by referencing function argument */
3884  #endif  #endif
3885      }
3886    }
3887    
3888    
3889    
3890    /*************************************************
3891    *           Check for POSIX class syntax         *
3892    *************************************************/
3893    
3894    switch(op_code)  /* This function is called when the sequence "[:" or "[." or "[=" is
3895    encountered in a character class. It checks whether this is followed by a
3896    sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
3897    reach an unescaped ']' without the special preceding character, return FALSE.
3898    
3899    Originally, this function only recognized a sequence of letters between the
3900    terminators, but it seems that Perl recognizes any sequence of characters,
3901    though of course unknown POSIX names are subsequently rejected. Perl gives an
3902    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
3903    didn't consider this to be a POSIX class. Likewise for [:1234:].
3904    
3905    The problem in trying to be exactly like Perl is in the handling of escapes. We
3906    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
3907    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
3908    below handles the special cases \\ and \], but does not try to do any other
3909    escape processing. This makes it different from Perl for cases such as
3910    [:l\ower:] where Perl recognizes it as the POSIX class "lower" but PCRE does
3911    not recognize "l\ower". This is a lesser evil than not diagnosing bad classes
3912    when Perl does, I think.
3913    
3914    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
3915    It seems that the appearance of a nested POSIX class supersedes an apparent
3916    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
3917    a digit.
3918    
3919    In Perl, unescaped square brackets may also appear as part of class names. For
3920    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
3921    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
3922    seem right at all. PCRE does not allow closing square brackets in POSIX class
3923    names.
3924    
3925    Arguments:
3926      ptr      pointer to the initial [
3927      endptr   where to return the end pointer
3928    
3929    Returns:   TRUE or FALSE
3930    */
3931    
3932    static BOOL
3933    check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
3934    {
3935    pcre_uchar terminator;          /* Don't combine these lines; the Solaris cc */
3936    terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
3937    for (++ptr; *ptr != CHAR_NULL; ptr++)
3938      {
3939      if (*ptr == CHAR_BACKSLASH &&
3940          (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET ||
3941           ptr[1] == CHAR_BACKSLASH))
3942        ptr++;
3943      else if ((*ptr == CHAR_LEFT_SQUARE_BRACKET && ptr[1] == terminator) ||
3944                *ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
3945      else if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3946      {      {
3947      case OP_CHAR:      *endptr = ptr;
3948      return c != next;      return TRUE;
3949        }
3950      }
3951    return FALSE;
3952    }
3953    
     /* 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. We know that next can have only one other case,  
     because multi-other-case characters are dealt with above. */  
3954    
     case OP_CHARI:  
     if (c == next) return FALSE;  
 #ifdef SUPPORT_UTF  
     if (utf)  
       {  
       pcre_uint32 othercase;  
       if (next < 128) othercase = cd->fcc[next]; else  
 #ifdef SUPPORT_UCP  
       othercase = UCD_OTHERCASE(next);  
 #else  
       othercase = NOTACHAR;  
 #endif  
       return c != othercase;  
       }  
     else  
 #endif  /* SUPPORT_UTF */  
     return (c != TABLE_GET(next, cd->fcc, next));  /* Not UTF */  
   
     case OP_NOT:  
     return c == next;  
   
     case OP_NOTI:  
     if (c == next) return TRUE;  
 #ifdef SUPPORT_UTF  
     if (utf)  
       {  
       pcre_uint32 othercase;  
       if (next < 128) othercase = cd->fcc[next]; else  
 #ifdef SUPPORT_UCP  
       othercase = UCD_OTHERCASE(next);  
 #else  
       othercase = NOTACHAR;  
 #endif  
       return c == othercase;  
       }  
     else  
 #endif  /* SUPPORT_UTF */  
     return (c == TABLE_GET(next, cd->fcc, next));  /* Not UTF */  
3955    
     /* 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. */  
3956    
3957      case OP_DIGIT:  /*************************************************
3958      return next > 255 || (cd->ctypes[next] & ctype_digit) == 0;  *          Check POSIX class name                *
3959    *************************************************/
3960    
3961      case OP_NOT_DIGIT:  /* This function is called to check the name given in a POSIX-style class entry
3962      return next <= 255 && (cd->ctypes[next] & ctype_digit) != 0;  such as [:alnum:].
3963    
3964      case OP_WHITESPACE:  Arguments:
3965      return next > 255 || (cd->ctypes[next] & ctype_space) == 0;    ptr        points to the first letter
3966      len        the length of the name
3967    
3968      case OP_NOT_WHITESPACE:  Returns:     a value representing the name, or -1 if unknown
3969      return next <= 255 && (cd->ctypes[next] & ctype_space) != 0;  */
3970    
3971      case OP_WORDCHAR:  static int
3972      return next > 255 || (cd->ctypes[next] & ctype_word) == 0;  check_posix_name(const pcre_uchar *ptr, int len)
3973    {
3974    const char *pn = posix_names;
3975    register int yield = 0;
3976    while (posix_name_lengths[yield] != 0)
3977      {
3978      if (len == posix_name_lengths[yield] &&
3979        STRNCMP_UC_C8(ptr, pn, (unsigned int)len) == 0) return yield;
3980      pn += posix_name_lengths[yield] + 1;
3981      yield++;
3982      }
3983    return -1;
3984    }
3985    
     case OP_NOT_WORDCHAR:  
     return next <= 255 && (cd->ctypes[next] & ctype_word) != 0;  
3986    
3987      case OP_HSPACE:  /*************************************************
3988      case OP_NOT_HSPACE:  *    Adjust OP_RECURSE items in repeated group   *
3989      switch(next)  *************************************************/
       {  
       HSPACE_CASES:  
       return op_code == OP_NOT_HSPACE;  
3990    
3991        default:  /* OP_RECURSE items contain an offset from the start of the regex to the group
3992        return op_code != OP_NOT_HSPACE;  that is referenced. This means that groups can be replicated for fixed
3993        }  repetition simply by copying (because the recursion is allowed to refer to
3994    earlier groups that are outside the current group). However, when a group is
3995    optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is
3996    inserted before it, after it has been compiled. This means that any OP_RECURSE
3997    items within it that refer to the group itself or any contained groups have to
3998    have their offsets adjusted. That one of the jobs of this function. Before it
3999    is called, the partially compiled regex must be temporarily terminated with
4000    OP_END.
4001    
4002      case OP_ANYNL:  This function has been extended to cope with forward references for recursions
4003      case OP_VSPACE:  and subroutine calls. It must check the list of such references for the
4004      case OP_NOT_VSPACE:  group we are dealing with. If it finds that one of the recursions in the
4005      switch(next)  current group is on this list, it does not adjust the value in the reference
4006        {  (which is a group number). After the group has been scanned, all the offsets in
4007        VSPACE_CASES:  the forward reference list for the group are adjusted.
       return op_code == OP_NOT_VSPACE;  
4008    
4009        default:  Arguments:
4010        return op_code != OP_NOT_VSPACE;    group      points to the start of the group
4011        }    adjust     the amount by which the group is to be moved
4012      utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode
4013      cd         contains pointers to tables etc.
4014      save_hwm_offset   the hwm forward reference offset at the start of the group
4015    
4016  #ifdef SUPPORT_UCP  Returns:     nothing
4017      case OP_PROP:  */
     return check_char_prop(next, previous[0], previous[1], FALSE);  
4018    
4019      case OP_NOTPROP:  static void
4020      return check_char_prop(next, previous[0], previous[1], TRUE);  adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
4021  #endif    size_t save_hwm_offset)
4022    {
4023    int offset;
4024    pcre_uchar *hc;
4025    pcre_uchar *ptr = group;
4026    
4027      default:  while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
4028      return FALSE;    {
4029      for (hc = (pcre_uchar *)cd->start_workspace + save_hwm_offset; hc < cd->hwm;
4030           hc += LINK_SIZE)
4031        {
4032        offset = (int)GET(hc, 0);
4033        if (cd->start_code + offset == ptr + 1) break;
4034      }      }
   }  
4035    
4036  /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP    /* If we have not found this recursion on the forward reference list, adjust
4037  is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are    the recursion's offset if it's after the start of this group. */
 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. */  
4038    
4039  switch(op_code)    if (hc >= cd->hwm)
   {  
   case OP_CHAR:  
   case OP_CHARI:  
   switch(escape)  
4040      {      {
4041      case ESC_d:      offset = (int)GET(ptr, 1);
4042      return c > 255 || (cd->ctypes[c] & ctype_digit) == 0;      if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
4043        }
4044    
4045      case ESC_D:    ptr += 1 + LINK_SIZE;
4046      return c <= 255 && (cd->ctypes[c] & ctype_digit) != 0;    }
4047    
4048      case ESC_s:  /* Now adjust all forward reference offsets for the group. */
     return c > 255 || (cd->ctypes[c] & ctype_space) == 0;  
4049    
4050      case ESC_S:  for (hc = (pcre_uchar *)cd->start_workspace + save_hwm_offset; hc < cd->hwm;
4051      return c <= 255 && (cd->ctypes[c] & ctype_space) != 0;       hc += LINK_SIZE)
4052      {
4053      offset = (int)GET(hc, 0);
4054      PUT(hc, 0, offset + adjust);
4055      }
4056    }
4057    
     case ESC_w:  
     return c > 255 || (cd->ctypes[c] & ctype_word) == 0;  
4058    
     case ESC_W:  
     return c <= 255 && (cd->ctypes[c] & ctype_word) != 0;  
4059    
4060      case ESC_h:  /*************************************************
4061      case ESC_H:  *        Insert an automatic callout point       *
4062      switch(c)  *************************************************/
       {  
       HSPACE_CASES:  
       return escape != ESC_h;  
   
       default:  
       return escape == ESC_h;  
       }  
4063    
4064      case ESC_v:  /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert
4065      case ESC_V:  callout points before each pattern item.
     switch(c)  
       {  
       VSPACE_CASES:  
       return escape != ESC_v;  
4066    
4067        default:  Arguments:
4068        return escape == ESC_v;    code           current code pointer
4069        }    ptr            current pattern pointer
4070      cd             pointers to tables etc
4071    
4072      /* When PCRE_UCP is set, these values get generated for \d etc. Find  Returns:         new code pointer
4073      their substitutions and process them. The result will always be either  */
     ESC_p or ESC_P. Then fall through to process those values. */  
4074    
4075  #ifdef SUPPORT_UCP  static pcre_uchar *
4076      case ESC_du:  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
4077      case ESC_DU:  {
4078      case ESC_wu:  *code++ = OP_CALLOUT;
4079      case ESC_WU:  *code++ = 255;
4080      case ESC_su:  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
4081      case ESC_SU:  PUT(code, LINK_SIZE, 0);                       /* Default length */
4082        {  return code + 2 * LINK_SIZE;
4083        int temperrorcode = 0;  }
       ptr = substitutes[escape - ESC_DU];  
       escape = check_escape(&ptr, &next, &temperrorcode, 0, options, FALSE);  
       if (temperrorcode != 0) return FALSE;  
       ptr++;    /* For compatibility */  
       }  
     /* Fall through */  
4084    
     case ESC_p:  
     case ESC_P:  
       {  
       int ptype, pdata, errorcodeptr;  
       BOOL negated;  
4085    
       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 */  
4086    
4087        /* If the property item is optional, we have to give up. (When generated  /*************************************************
4088        from \d etc by PCRE_UCP, this test will have been applied much earlier,  *         Complete a callout item                *
4089        to the original \d etc. At this point, ptr will point to a zero byte. */  *************************************************/
4090    
4091        if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||  /* A callout item contains the length of the next item in the pattern, which
4092          STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)  we can't fill in till after we have reached the relevant point. This is used
4093            return FALSE;  for both automatic and manual callouts.
4094    
4095        /* Do the property check. */  Arguments:
4096      previous_callout   points to previous callout item
4097      ptr                current pattern pointer
4098      cd                 pointers to tables etc
4099    
4100        return check_char_prop(c, ptype, pdata, (escape == ESC_P) != negated);  Returns:             nothing
4101        }  */
 #endif  
4102    
4103      default:  static void
4104      return FALSE;  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
4105      }  {
4106    int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
4107    PUT(previous_callout, 2 + LINK_SIZE, length);
4108    }
4109    
   /* In principle, support for Unicode properties should be integrated here as  
   well. It means re-organizing the above code so as to get hold of the property  
   values before switching on the op-code. However, I wonder how many patterns  
   combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,  
   these op-codes are never generated.) */  
4110    
   case OP_DIGIT:  
   return escape == ESC_D || escape == ESC_s || escape == ESC_W ||  
          escape == ESC_h || escape == ESC_v || escape == ESC_R;  
4111    
4112    case OP_NOT_DIGIT:  #ifdef SUPPORT_UCP
4113    return escape == ESC_d;  /*************************************************
4114    *           Get othercase range                  *
4115    *************************************************/
4116    
4117    case OP_WHITESPACE:  /* This function is passed the start and end of a class range, in UTF-8 mode
4118    return escape == ESC_S || escape == ESC_d || escape == ESC_w;  with UCP support. It searches up the characters, looking for ranges of
4119    characters in the "other" case. Each call returns the next one, updating the
4120    start address. A character with multiple other cases is returned on its own
4121    with a special return value.
4122    
4123    case OP_NOT_WHITESPACE:  Arguments:
4124    return escape == ESC_s || escape == ESC_h || escape == ESC_v || escape == ESC_R;    cptr        points to starting character value; updated
4125      d           end value
4126      ocptr       where to put start of othercase range
4127      odptr       where to put end of othercase range
4128    
4129    case OP_HSPACE:  Yield:        -1 when no more
4130    return escape == ESC_S || escape == ESC_H || escape == ESC_d ||                 0 when a range is returned
4131           escape == ESC_w || escape == ESC_v || escape == ESC_R;                >0 the CASESET offset for char with multiple other cases
4132                    in this case, ocptr contains the original
4133    */
4134    
4135    case OP_NOT_HSPACE:  static int
4136    return escape == ESC_h;  get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,
4137      pcre_uint32 *odptr)
4138    {
4139    pcre_uint32 c, othercase, next;
4140    unsigned int co;
4141    
4142    /* Can't have \S in here because VT matches \S (Perl anomaly) */  /* Find the first character that has an other case. If it has multiple other
4143    case OP_ANYNL:  cases, return its case offset value. */
   case OP_VSPACE:  
   return escape == ESC_V || escape == ESC_d || escape == ESC_w;  
4144    
4145    case OP_NOT_VSPACE:  for (c = *cptr; c <= d; c++)
4146    return escape == ESC_v || escape == ESC_R;    {
4147      if ((co = UCD_CASESET(c)) != 0)
4148        {
4149        *ocptr = c++;   /* Character that has the set */
4150        *cptr = c;      /* Rest of input range */
4151        return (int)co;
4152        }
4153      if ((othercase = UCD_OTHERCASE(c)) != c) break;
4154      }
4155    
4156    case OP_WORDCHAR:  if (c > d) return -1;  /* Reached end of range */
   return escape == ESC_W || escape == ESC_s || escape == ESC_h ||  
          escape == ESC_v || escape == ESC_R;  
4157    
4158    case OP_NOT_WORDCHAR:  /* Found a character that has a single other case. Search for the end of the
4159    return escape == ESC_w || escape == ESC_d;  range, which is either the end of the input range, or a character that has zero
4160    or more than one other cases. */
4161    
4162    default:  *ocptr = othercase;
4163    return FALSE;  next = othercase + 1;
4164    
4165    for (++c; c <= d; c++)
4166      {
4167      if ((co = UCD_CASESET(c)) != 0 || UCD_OTHERCASE(c) != next) break;
4168      next++;
4169    }    }
4170    
4171  /* Control does not reach here */  *odptr = next - 1;     /* End of othercase range */
4172    *cptr = c;             /* Rest of input range */
4173    return 0;
4174  }  }
4175    #endif  /* SUPPORT_UCP */
4176    
4177    
4178    
# Line 3416  switch(op_code) Line 4181  switch(op_code)
4181  *************************************************/  *************************************************/
4182    
4183  /* This function packages up the logic of adding a character or range of  /* This function packages up the logic of adding a character or range of
4184  characters to a class. The character values in the arguments will be within the  characters to a class. The character values in the arguments will be within the
4185  valid values for the current mode (8-bit, 16-bit, UTF, etc). This function is  valid values for the current mode (8-bit, 16-bit, UTF, etc). This function is
4186  mutually recursive with the function immediately below.  mutually recursive with the function immediately below.
4187    
4188  Arguments:  Arguments:
4189    classbits     the bit map for characters < 256    classbits     the bit map for characters < 256
4190    uchardptr     points to the pointer for extra data    uchardptr     points to the pointer for extra data
4191    options       the options word    options       the options word
4192    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
4193    start         start of range character    start         start of range character
4194    end           end of range character    end           end of range character
4195    
4196  Returns:        the number of < 256 characters added  Returns:        the number of < 256 characters added
4197                  the pointer to extra data is updated                  the pointer to extra data is updated
4198  */  */
# Line 3437  add_to_class(pcre_uint8 *classbits, pcre Line 4202  add_to_class(pcre_uint8 *classbits, pcre
4202    compile_data *cd, pcre_uint32 start, pcre_uint32 end)    compile_data *cd, pcre_uint32 start, pcre_uint32 end)
4203  {  {
4204  pcre_uint32 c;  pcre_uint32 c;
4205    pcre_uint32 classbits_end = (end <= 0xff ? end : 0xff);
4206  int n8 = 0;  int n8 = 0;
4207    
4208  /* If caseless matching is required, scan the range and process alternate  /* If caseless matching is required, scan the range and process alternate
4209  cases. In Unicode, there are 8-bit characters that have alternate cases that  cases. In Unicode, there are 8-bit characters that have alternate cases that
4210  are greater than 255 and vice-versa. Sometimes we can just extend the original  are greater than 255 and vice-versa. Sometimes we can just extend the original
4211  range. */  range. */
4212    
4213  if ((options & PCRE_CASELESS) != 0)  if ((options & PCRE_CASELESS) != 0)
4214    {    {
4215  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4216    if ((options & PCRE_UTF8) != 0)    if ((options & PCRE_UTF8) != 0)
4217      {      {
4218      int rc;      int rc;
4219      pcre_uint32 oc, od;      pcre_uint32 oc, od;
4220    
4221      options &= ~PCRE_CASELESS;   /* Remove for recursive calls */      options &= ~PCRE_CASELESS;   /* Remove for recursive calls */
4222      c = start;      c = start;
4223    
4224      while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)      while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)
4225        {        {
4226        /* Handle a single character that has more than one other case. */        /* Handle a single character that has more than one other case. */
4227    
4228        if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,        if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,
4229          PRIV(ucd_caseless_sets) + rc, oc);          PRIV(ucd_caseless_sets) + rc, oc);
4230    
4231        /* Do nothing if the other case range is within the original range. */        /* Do nothing if the other case range is within the original range. */
4232    
4233        else if (oc >= start && od <= end) continue;        else if (oc >= start && od <= end) continue;
4234    
4235        /* Extend the original range if there is overlap, noting that if oc < c, we        /* Extend the original range if there is overlap, noting that if oc < c, we
4236        can't have od > end because a subrange is always shorter than the basic        can't have od > end because a subrange is always shorter than the basic
4237        range. Otherwise, use a recursive call to add the additional range. */        range. Otherwise, use a recursive call to add the additional range. */
4238    
4239        else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */        else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */
4240        else if (od > end && oc <= end + 1) end = od;       /* Extend upwards */        else if (od > end && oc <= end + 1)
4241            {
4242            end = od;       /* Extend upwards */
4243            if (end > classbits_end) classbits_end = (end <= 0xff ? end : 0xff);
4244            }
4245        else n8 += add_to_class(classbits, uchardptr, options, cd, oc, od);        else n8 += add_to_class(classbits, uchardptr, options, cd, oc, od);
4246        }        }
4247      }      }
# Line 3479  if ((options & PCRE_CASELESS) != 0) Line 4249  if ((options & PCRE_CASELESS) != 0)
4249  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
4250    
4251    /* Not UTF-mode, or no UCP */    /* Not UTF-mode, or no UCP */
4252    
4253    for (c = start; c <= end && c < 256; c++)    for (c = start; c <= classbits_end; c++)
4254      {      {
4255      SETBIT(classbits, cd->fcc[c]);      SETBIT(classbits, cd->fcc[c]);
4256      n8++;      n8++;
4257      }      }
4258    }    }
4259    
4260  /* Now handle the original range. Adjust the final value according to the bit  /* Now handle the original range. Adjust the final value according to the bit
4261  length - this means that the same lists of (e.g.) horizontal spaces can be used  length - this means that the same lists of (e.g.) horizontal spaces can be used
4262  in all cases. */  in all cases. */
# Line 3505  in all cases. */ Line 4275  in all cases. */
4275    
4276  #endif /* COMPILE_PCRE[8|16] */  #endif /* COMPILE_PCRE[8|16] */
4277    
4278  /* If all characters are less than 256, use the bit map. Otherwise use extra  /* Use the bitmap for characters < 256. Otherwise use extra data.*/
 data. */  
4279    
4280  if (end < 0x100)  for (c = start; c <= classbits_end; c++)
4281    {    {
4282    for (c = start; c <= end; c++)    /* Regardless of start, c will always be <= 255. */
4283      {    SETBIT(classbits, c);
4284      n8++;    n8++;
     SETBIT(classbits, c);  
     }  
4285    }    }
4286    
4287  else  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4288    {  if (start <= 0xff) start = 0xff + 1;
4289    
4290    if (end >= start)
4291      {
4292    pcre_uchar *uchardata = *uchardptr;    pcre_uchar *uchardata = *uchardptr;
   
4293  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4294    if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */    if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */
4295      {      {
4296      if (start < end)      if (start < end)
4297        {        {
4298        *uchardata++ = XCL_RANGE;        *uchardata++ = XCL_RANGE;
4299        uchardata += PRIV(ord2utf)(start, uchardata);        uchardata += PRIV(ord2utf)(start, uchardata);
4300        uchardata += PRIV(ord2utf)(end, uchardata);        uchardata += PRIV(ord2utf)(end, uchardata);
4301        }        }
4302      else if (start == end)      else if (start == end)
4303        {        {
4304        *uchardata++ = XCL_SINGLE;        *uchardata++ = XCL_SINGLE;
4305        uchardata += PRIV(ord2utf)(start, uchardata);        uchardata += PRIV(ord2utf)(start, uchardata);
4306        }        }
4307      }      }
4308    else    else
4309  #endif  /* SUPPORT_UTF */  #endif  /* SUPPORT_UTF */
4310    
4311    /* Without UTF support, character values are constrained by the bit length,    /* Without UTF support, character values are constrained by the bit length,
4312    and can only be > 256 for 16-bit and 32-bit libraries. */    and can only be > 256 for 16-bit and 32-bit libraries. */
4313    
4314  #ifdef COMPILE_PCRE8  #ifdef COMPILE_PCRE8
4315      {}      {}
4316  #else  #else
4317    if (start < end)    if (start < end)
4318      {      {
4319      *uchardata++ = XCL_RANGE;      *uchardata++ = XCL_RANGE;
# Line 3555  else Line 4324