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

Diff of /code/trunk/pcre_compile.c

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

revision 964 by ph10, Fri May 4 13:03:39 2012 UTC revision 1576 by ph10, Fri Jul 17 15:15:18 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    
55    
56  /* When PCRE_DEBUG is defined, we need the pcre(16)_printint() function, which  /* When PCRE_DEBUG is defined, we need the pcre(16|32)_printint() function, which
57  is also used by pcretest. PCRE_DEBUG is not defined when building a production  is also used by pcretest. PCRE_DEBUG is not defined when building a production
58  library. We do not need to select pcre16_printint.c specially, because the  library. We do not need to select pcre16_printint.c specially, because the
59  COMPILE_PCREx macro will already be appropriately set. */  COMPILE_PCREx macro will already be appropriately set. */
# Line 68  COMPILE_PCREx macro will already be appr Line 68  COMPILE_PCREx macro will already be appr
68    
69  /* Macro for setting individual bits in class bitmaps. */  /* Macro for setting individual bits in class bitmaps. */
70    
71  #define SETBIT(a,b) a[b/8] |= (1 << (b%8))  #define SETBIT(a,b) a[(b)/8] |= (1 << ((b)&7))
72    
73  /* Maximum length value to check against when making sure that the integer that  /* Maximum length value to check against when making sure that the integer that
74  holds the compiled pattern length does not overflow. We make it a bit less than  holds the compiled pattern length does not overflow. We make it a bit less than
# Line 77  to check them every time. */ Line 77  to check them every time. */
77    
78  #define OFLOW_MAX (INT_MAX - 20)  #define OFLOW_MAX (INT_MAX - 20)
79    
80    /* Definitions to allow mutual recursion */
81    
82    static int
83      add_list_to_class(pcre_uint8 *, pcre_uchar **, int, compile_data *,
84        const pcre_uint32 *, unsigned int);
85    
86    static BOOL
87      compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL, int, int,
88        pcre_uint32 *, pcre_int32 *, pcre_uint32 *, pcre_int32 *, branch_chain *,
89        compile_data *, int *);
90    
91    
92    
93  /*************************************************  /*************************************************
94  *      Code parameters and static tables         *  *      Code parameters and static tables         *
# Line 103  kicks in at the same number of forward r Line 115  kicks in at the same number of forward r
115  #define COMPILE_WORK_SIZE (2048*LINK_SIZE)  #define COMPILE_WORK_SIZE (2048*LINK_SIZE)
116  #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)  #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)
117    
118    /* This value determines the size of the initial vector that is used for
119    remembering named groups during the pre-compile. It is allocated on the stack,
120    but if it is too small, it is expanded using malloc(), in a similar way to the
121    workspace. The value is the number of slots in the list. */
122    
123    #define NAMED_GROUP_LIST_SIZE  20
124    
125  /* The overrun tests check for a slightly smaller size so that they detect the  /* The overrun tests check for a slightly smaller size so that they detect the
126  overrun before it actually does run off the end of the data block. */  overrun before it actually does run off the end of the data block. */
127    
# Line 110  overrun before it actually does run off Line 129  overrun before it actually does run off
129    
130  /* Private flags added to firstchar and reqchar. */  /* Private flags added to firstchar and reqchar. */
131    
132  #define REQ_CASELESS   0x10000000l      /* Indicates caselessness */  #define REQ_CASELESS    (1 << 0)        /* Indicates caselessness */
133  #define REQ_VARY       0x20000000l      /* Reqchar followed non-literal item */  #define REQ_VARY        (1 << 1)        /* Reqchar followed non-literal item */
134    /* Negative values for the firstchar and reqchar flags */
135    #define REQ_UNSET       (-2)
136    #define REQ_NONE        (-1)
137    
138  /* Repeated character flags. */  /* Repeated character flags. */
139    
# Line 152  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 180  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 197  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 238  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 253  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 280  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 307  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 360  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 375  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 440  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 472  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 481  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 490  static const char error_texts[] = Line 546  static const char error_texts[] =
546    "disallowed Unicode code point (>= 0xd800 && <= 0xdfff)\0"    "disallowed Unicode code point (>= 0xd800 && <= 0xdfff)\0"
547    "invalid UTF-16 string\0"    "invalid UTF-16 string\0"
548    /* 75 */    /* 75 */
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"
551      "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 630  static const pcre_uint8 ebcdic_chartab[] Line 699  static const pcre_uint8 ebcdic_chartab[]
699  #endif  #endif
700    
701    
702  /* Definition to allow mutual recursion */  /* 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  static BOOL  
740    compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL, int, int,  /* This table is used to check whether auto-possessification is possible
741      int *, int *, branch_chain *, compile_data *, int *);  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    
# Line 657  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 741  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 a negative value which  positive value for a simple escape such as \n, or 0 for a data character which
985  encodes one of the more complicated things such as \d. A backreference to group  will be placed in chptr. A backreference to group n is returned as negative n.
986  n is returned as -(ESC_REF + n); ESC_REF is the highest ESC_xxx macro. When  When UTF-8 is enabled, a positive value greater than 255 may be returned in
987  UTF-8 is enabled, a positive value greater than 255 may be returned. On entry,  chptr. On entry, ptr is pointing at the \. On exit, it is on the final
988  ptr is pointing at the \. On exit, it is on the final character of the escape  character of the escape sequence.
 sequence.  
989    
990  Arguments:  Arguments:
991    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
992      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
996    isclass        TRUE if inside a character class    isclass        TRUE if inside a character class
997    
998  Returns:         zero or positive => a data character  Returns:         zero => a data character
999                   negative => a special escape sequence                   positive => a special escape sequence
1000                     negative => a back reference
1001                   on error, errorcodeptr is set                   on error, errorcodeptr is set
1002  */  */
1003    
1004  static int  static int
1005  check_escape(const pcre_uchar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,
1006    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. */
1009  BOOL utf = (options & PCRE_UTF8) != 0;  BOOL utf = (options & PCRE_UTF8) != 0;
1010  const pcre_uchar *ptr = *ptrptr + 1;  const pcre_uchar *ptr = *ptrptr + 1;
1011  pcre_int32 c;  pcre_uint32 c;
1012    int escape = 0;
1013  int i;  int i;
1014    
1015  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
# Line 775  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 784  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) c = 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 */
1034  else if (c < 'a' || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}  else if (c < CHAR_a || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}
1035  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1036  #endif  #endif
1037    
1038  /* Escapes that need further processing, or are illegal. */  /* Escapes that need further processing, or are illegal. */
# Line 797  else if ((i = escapes[c - 0x48]) != 0) Line 1040  else if ((i = escapes[c - 0x48]) != 0)
1040  else  else
1041    {    {
1042    const pcre_uchar *oldptr;    const pcre_uchar *oldptr;
1043    BOOL braced, negated;    BOOL braced, negated, overflow;
1044      int s;
1045    
1046    switch (c)    switch (c)
1047      {      {
# Line 822  else Line 1066  else
1066          c = 0;          c = 0;
1067          for (i = 0; i < 4; ++i)          for (i = 0; i < 4; ++i)
1068            {            {
1069            register int cc = *(++ptr);            register pcre_uint32 cc = *(++ptr);
1070  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1071            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1072            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
# Line 831  else Line 1075  else
1075            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1076  #endif  #endif
1077            }            }
1078    
1079    #if defined COMPILE_PCRE8
1080            if (c > (utf ? 0x10ffffU : 0xffU))
1081    #elif defined COMPILE_PCRE16
1082            if (c > (utf ? 0x10ffffU : 0xffffU))
1083    #elif defined COMPILE_PCRE32
1084            if (utf && c > 0x10ffffU)
1085    #endif
1086              {
1087              *errorcodeptr = ERR76;
1088              }
1089            else if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1090          }          }
1091        }        }
1092      else      else
# Line 857  else Line 1113  else
1113      (3) For Oniguruma compatibility we also support \g followed by a name or a      (3) For Oniguruma compatibility we also support \g followed by a name or a
1114      number either in angle brackets or in single quotes. However, these are      number either in angle brackets or in single quotes. However, these are
1115      (possibly recursive) subroutine calls, _not_ backreferences. Just return      (possibly recursive) subroutine calls, _not_ backreferences. Just return
1116      the -ESC_g code (cf \k). */      the ESC_g code (cf \k). */
1117    
1118      case CHAR_g:      case CHAR_g:
1119      if (isclass) break;      if (isclass) break;
1120      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
1121        {        {
1122        c = -ESC_g;        escape = ESC_g;
1123        break;        break;
1124        }        }
1125    
# Line 872  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          c = -ESC_k;          escape = ESC_k;
1136          break;          break;
1137          }          }
1138        braced = TRUE;        braced = TRUE;
# Line 892  else Line 1148  else
1148      else negated = FALSE;      else negated = FALSE;
1149    
1150      /* The integer range is limited by the machine's int representation. */      /* The integer range is limited by the machine's int representation. */
1151      c = 0;      s = 0;
1152        overflow = FALSE;
1153      while (IS_DIGIT(ptr[1]))      while (IS_DIGIT(ptr[1]))
1154        {        {
1155        if (((unsigned int)c) > INT_MAX / 10) /* Integer overflow */        if (s > INT_MAX / 10 - 1) /* Integer overflow */
1156          {          {
1157          c = -1;          overflow = TRUE;
1158          break;          break;
1159          }          }
1160        c = c * 10 + *(++ptr) - CHAR_0;        s = s * 10 + (int)(*(++ptr) - CHAR_0);
1161        }        }
1162      if (((unsigned int)c) > INT_MAX) /* Integer overflow */      if (overflow) /* Integer overflow */
1163        {        {
1164        while (IS_DIGIT(ptr[1]))        while (IS_DIGIT(ptr[1]))
1165          ptr++;          ptr++;
# Line 916  else Line 1173  else
1173        break;        break;
1174        }        }
1175    
1176      if (c == 0)      if (s == 0)
1177        {        {
1178        *errorcodeptr = ERR58;        *errorcodeptr = ERR58;
1179        break;        break;
# Line 924  else Line 1181  else
1181    
1182      if (negated)      if (negated)
1183        {        {
1184        if (c > bracount)        if (s > bracount)
1185          {          {
1186          *errorcodeptr = ERR15;          *errorcodeptr = ERR15;
1187          break;          break;
1188          }          }
1189        c = bracount - (c - 1);        s = bracount - (s - 1);
1190        }        }
1191    
1192      c = -(ESC_REF + c);      escape = -s;
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 954  else Line 1215  else
1215        {        {
1216        oldptr = ptr;        oldptr = ptr;
1217        /* The integer range is limited by the machine's int representation. */        /* The integer range is limited by the machine's int representation. */
1218        c -= CHAR_0;        s = (int)(c -CHAR_0);
1219          overflow = FALSE;
1220        while (IS_DIGIT(ptr[1]))        while (IS_DIGIT(ptr[1]))
1221          {          {
1222          if (((unsigned int)c) > INT_MAX / 10) /* Integer overflow */          if (s > INT_MAX / 10 - 1) /* Integer overflow */
1223            {            {
1224            c = -1;            overflow = TRUE;
1225            break;            break;
1226            }            }
1227          c = c * 10 + *(++ptr) - CHAR_0;          s = s * 10 + (int)(*(++ptr) - CHAR_0);
1228          }          }
1229        if (((unsigned int)c) > INT_MAX) /* Integer overflow */        if (overflow) /* Integer overflow */
1230          {          {
1231          while (IS_DIGIT(ptr[1]))          while (IS_DIGIT(ptr[1]))
1232            ptr++;            ptr++;
1233          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
1234          break;          break;
1235          }          }
1236        if (c < 10 || c <= bracount)        if (s < 8 || s <= bracount)  /* Check for back reference */
1237          {          {
1238          c = -(ESC_REF + c);          escape = -s;
1239          break;          break;
1240          }          }
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 1005  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          {          {
1316          c = 0;          c = 0;
1317          for (i = 0; i < 2; ++i)          for (i = 0; i < 2; ++i)
1318            {            {
1319            register int cc = *(++ptr);            register pcre_uint32 cc = *(++ptr);
1320  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1321            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1322            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
# Line 1030  else Line 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        while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0)        {
1340          if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1341          {          {
1342          register int 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
1356              if (c >= 0x10000000l) { overflow = TRUE; break; }
1357    #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  #ifdef COMPILE_PCRE8  #if defined COMPILE_PCRE8
1368          if (c > (utf ? 0x10ffff : 0xff)) { c = -1; break; }            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1369  #else  #elif defined COMPILE_PCRE16
1370  #ifdef COMPILE_PCRE16            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1371          if (c > (utf ? 0x10ffff : 0xffff)) { c = -1; break; }  #elif defined COMPILE_PCRE32
1372  #endif            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1373  #endif  #endif
1374          }            }
1375    
1376        if (c < 0)          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        int 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 1101  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 1116  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 1141  else Line 1469  else
1469  newline". PCRE does not support \N{name}. However, it does support  newline". PCRE does not support \N{name}. However, it does support
1470  quantification such as \N{2,3}. */  quantification such as \N{2,3}. */
1471    
1472  if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&  if (escape == ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1473       !is_counted_repeat(ptr+2))       !is_counted_repeat(ptr+2))
1474    *errorcodeptr = ERR37;    *errorcodeptr = ERR37;
1475    
1476  /* If PCRE_UCP is set, we change the values for \d etc. */  /* If PCRE_UCP is set, we change the values for \d etc. */
1477    
1478  if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)  if ((options & PCRE_UCP) != 0 && escape >= ESC_D && escape <= ESC_w)
1479    c -= (ESC_DU - ESC_D);    escape += (ESC_DU - ESC_D);
1480    
1481  /* Set the pointer to the final character before returning. */  /* Set the pointer to the final character before returning. */
1482    
1483  *ptrptr = ptr;  *ptrptr = ptr;
1484  return c;  *chptr = c;
1485    return escape;
1486  }  }
1487    
1488    
# Line 1171  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 1202  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 1227  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 1276  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 1308  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 1319  return p; Line 1649  return p;
1649    
1650    
1651  /*************************************************  /*************************************************
1652  *  Subroutine for finding forward reference      *  *      Find first significant op code            *
1653  *************************************************/  *************************************************/
1654    
1655  /* This recursive function is called only from find_parens() below. The  /* This is called by several functions that scan a compiled expression looking
1656  top-level call starts at the beginning of the pattern. All other calls must  for a fixed first character, or an anchoring op code etc. It skips over things
1657  start at a parenthesis. It scans along a pattern's text looking for capturing  that do not influence this. For some calls, it makes sense to skip negative
1658  subpatterns, and counting them. If it finds a named pattern that matches the  forward and all backward assertions, and also the \b assertion; for others it
1659  name it is given, it returns its number. Alternatively, if the name is NULL, it  does not.
 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.)  
1660    
1661  Arguments:  Arguments:
1662    ptrptr       address of the current character pointer (updated)    code         pointer to the start of the group
1663    cd           compile background data    skipassert   TRUE if certain assertions are to be skipped
   name         name to seek, or NULL if seeking a numbered subpattern  
   lorn         name length, or subpattern number if name is NULL  
   xmode        TRUE if we are in /x mode  
   utf          TRUE if we are in UTF-8 / UTF-16 mode  
   count        pointer to the current capturing subpattern number (updated)  
1664    
1665  Returns:       the number of the named subpattern, or -1 if not found  Returns:       pointer to the first significant opcode
1666  */  */
1667    
1668  static int  static const pcre_uchar*
1669  find_parens_sub(pcre_uchar **ptrptr, compile_data *cd, const pcre_uchar *name, int lorn,  first_significant_code(const pcre_uchar *code, BOOL skipassert)
   BOOL xmode, BOOL utf, int *count)  
1670  {  {
1671  pcre_uchar *ptr = *ptrptr;  for (;;)
 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)  
1672    {    {
1673    /* Handle specials such as (*SKIP) or (*UTF8) etc. */    switch ((int)*code)
   
   if (ptr[1] == CHAR_ASTERISK) ptr += 2;  
   
   /* Handle a normal, unnamed capturing parenthesis. */  
   
   else if (ptr[1] != CHAR_QUESTION_MARK)  
1674      {      {
1675      *count += 1;      case OP_ASSERT_NOT:
1676      if (name == NULL && *count == lorn) return *count;      case OP_ASSERTBACK:
1677      ptr++;      case OP_ASSERTBACK_NOT:
1678      }      if (!skipassert) return code;
1679        do code += GET(code, 1); while (*code == OP_ALT);
1680        code += PRIV(OP_lengths)[*code];
1681        break;
1682    
1683        case OP_WORD_BOUNDARY:
1684        case OP_NOT_WORD_BOUNDARY:
1685        if (!skipassert) return code;
1686        /* Fall through */
1687    
1688    /* All cases now have (? at the start. Remember when we are in a group      case OP_CALLOUT:
1689    where the parenthesis numbers are duplicated. */      case OP_CREF:
1690        case OP_DNCREF:
1691        case OP_RREF:
1692        case OP_DNRREF:
1693        case OP_DEF:
1694        code += PRIV(OP_lengths)[*code];
1695        break;
1696    
1697    else if (ptr[2] == CHAR_VERTICAL_LINE)      default:
1698      {      return code;
     ptr += 3;  
     dup_parens = TRUE;  
1699      }      }
1700      }
1701    /* Control never reaches here */
1702    }
1703    
   /* Handle comments; all characters are allowed until a ket is reached. */  
1704    
   else if (ptr[2] == CHAR_NUMBER_SIGN)  
     {  
     for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;  
     goto FAIL_EXIT;  
     }  
1705    
1706    /* Handle a condition. If it is an assertion, just carry on so that it  /*************************************************
1707    is processed as normal. If not, skip to the closing parenthesis of the  *        Find the fixed length of a branch       *
1708    condition (there can't be any nested parens). */  *************************************************/
1709    
1710    else if (ptr[2] == CHAR_LEFT_PARENTHESIS)  /* Scan a branch and compute the fixed length of subject that will match it,
1711      {  if the length is fixed. This is needed for dealing with backward assertions.
1712      ptr += 2;  In UTF8 mode, the result is in characters rather than bytes. The branch is
1713      if (ptr[1] != CHAR_QUESTION_MARK)  temporarily terminated with OP_END when this function is called.
       {  
       while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;  
       if (*ptr != 0) ptr++;  
       }  
     }  
1714    
1715    /* Start with (? but not a condition. */  This function is called when a backward assertion is encountered, so that if it
1716    fails, the error message can point to the correct place in the pattern.
1717    However, we cannot do this when the assertion contains subroutine calls,
1718    because they can be forward references. We solve this by remembering this case
1719    and doing the check at the end; a flag specifies which mode we are running in.
1720    
1721    else  Arguments:
1722      {    code     points to the start of the pattern (the bracket)
1723      ptr += 2;    utf      TRUE in UTF-8 / UTF-16 / UTF-32 mode
1724      if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */    atend    TRUE if called when the pattern is complete
1725      cd       the "compile data" structure
1726      recurses    chain of recurse_check to catch mutual recursion
1727    
1728      /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */  Returns:   the fixed length,
1729                 or -1 if there is no fixed length,
1730                 or -2 if \C was encountered (in UTF-8 mode only)
1731                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1732                 or -4 if an unknown opcode was encountered (internal error)
1733    */
1734    
1735      if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&  static int
1736          ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)  find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd,
1737        {    recurse_check *recurses)
1738        int term;  {
1739        const pcre_uchar *thisname;  int length = -1;
1740        *count += 1;  recurse_check this_recurse;
1741        if (name == NULL && *count == lorn) return *count;  register int branchlength = 0;
1742        term = *ptr++;  register pcre_uchar *cc = code + 1 + LINK_SIZE;
       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++;  
       }  
     }  
   }  
1743    
1744  /* Past any initial parenthesis handling, scan for parentheses or vertical  /* Scan along the opcodes for this branch. If we get to the end of the
1745  bars. Stop if we get to cd->end_pattern. Note that this is important for the  branch, check the length against that of the other branches. */
 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. */  
1746    
1747  for (; ptr < cd->end_pattern; ptr++)  for (;;)
1748    {    {
1749    /* Skip over backslashed characters and also entire \Q...\E */    int d;
1750      pcre_uchar *ce, *cs;
1751      register pcre_uchar op = *cc;
1752    
1753    if (*ptr == CHAR_BACKSLASH)    switch (op)
1754      {      {
1755      if (*(++ptr) == 0) goto FAIL_EXIT;      /* We only need to continue for OP_CBRA (normal capturing bracket) and
1756      if (*ptr == CHAR_Q) for (;;)      OP_BRA (normal non-capturing bracket) because the other variants of these
1757        {      opcodes are all concerned with unlimited repeated groups, which of course
1758        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};      are not of fixed length. */
       if (*ptr == 0) goto FAIL_EXIT;  
       if (*(++ptr) == CHAR_E) break;  
       }  
     continue;  
     }  
1759    
1760    /* Skip over character classes; this logic must be similar to the way they      case OP_CBRA:
1761    are handled for real. If the first character is '^', skip it. Also, if the      case OP_BRA:
1762    first few characters (either before or after ^) are \Q\E or \E we skip them      case OP_ONCE:
1763    too. This makes for compatibility with Perl. Note the use of STR macros to      case OP_ONCE_NC:
1764    encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */      case OP_COND:
1765        d = find_fixedlength(cc + ((op == OP_CBRA)? IMM2_SIZE : 0), utf, atend, cd,
1766          recurses);
1767        if (d < 0) return d;
1768        branchlength += d;
1769        do cc += GET(cc, 1); while (*cc == OP_ALT);
1770        cc += 1 + LINK_SIZE;
1771        break;
1772    
1773    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)      /* Reached end of a branch; if it's a ket it is the end of a nested call.
1774      {      If it's ALT it is an alternation in a nested call. An ACCEPT is effectively
1775      BOOL negate_class = FALSE;      an ALT. If it is END it's the end of the outer call. All can be handled by
1776      for (;;)      the same code. Note that we must not include the OP_KETRxxx opcodes here,
1777        {      because they all imply an unlimited repeat. */
       if (ptr[1] == CHAR_BACKSLASH)  
         {  
         if (ptr[2] == CHAR_E)  
           ptr+= 2;  
         else if (STRNCMP_UC_C8(ptr + 2,  
                  STR_Q STR_BACKSLASH STR_E, 3) == 0)  
           ptr += 4;  
         else  
           break;  
         }  
       else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)  
         {  
         negate_class = TRUE;  
         ptr++;  
         }  
       else break;  
       }  
   
     /* If the next character is ']', it is a data character that must be  
     skipped, except in JavaScript compatibility mode. */  
   
     if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&  
         (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)  
       ptr++;  
   
     while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)  
       {  
       if (*ptr == 0) return -1;  
       if (*ptr == CHAR_BACKSLASH)  
         {  
         if (*(++ptr) == 0) goto FAIL_EXIT;  
         if (*ptr == CHAR_Q) for (;;)  
           {  
           while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};  
           if (*ptr == 0) goto FAIL_EXIT;  
           if (*(++ptr) == CHAR_E) break;  
           }  
         continue;  
         }  
       }  
     continue;  
     }  
   
   /* Skip comments in /x mode */  
   
   if (xmode && *ptr == CHAR_NUMBER_SIGN)  
     {  
     ptr++;  
     while (*ptr != 0)  
       {  
       if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }  
       ptr++;  
 #ifdef SUPPORT_UTF  
       if (utf) FORWARDCHAR(ptr);  
 #endif  
       }  
     if (*ptr == 0) goto FAIL_EXIT;  
     continue;  
     }  
   
   /* Check for the special metacharacters */  
   
   if (*ptr == CHAR_LEFT_PARENTHESIS)  
     {  
     int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, count);  
     if (rc > 0) return rc;  
     if (*ptr == 0) goto FAIL_EXIT;  
     }  
   
   else if (*ptr == CHAR_RIGHT_PARENTHESIS)  
     {  
     if (dup_parens && *count < hwm_count) *count = hwm_count;  
     goto FAIL_EXIT;  
     }  
   
   else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)  
     {  
     if (*count > hwm_count) hwm_count = *count;  
     *count = start_count;  
     }  
   }  
   
 FAIL_EXIT:  
 *ptrptr = ptr;  
 return -1;  
 }  
   
   
   
   
 /*************************************************  
 *       Find forward referenced subpattern       *  
 *************************************************/  
   
 /* This function scans along a pattern's text looking for capturing  
 subpatterns, and counting them. If it finds a named pattern that matches the  
 name it is given, it returns its number. Alternatively, if the name is NULL, it  
 returns when it reaches a given numbered subpattern. This is used for forward  
 references to subpatterns. We used to be able to start this scan from the  
 current compiling point, using the current count value from cd->bracount, and  
 do it all in a single loop, but the addition of the possibility of duplicate  
 subpattern numbers means that we have to scan from the very start, in order to  
 take account of such duplicates, and to use a recursive function to keep track  
 of the different types of group.  
   
 Arguments:  
   cd           compile background data  
   name         name to seek, or NULL if seeking a numbered subpattern  
   lorn         name length, or subpattern number if name is NULL  
   xmode        TRUE if we are in /x mode  
   utf          TRUE if we are in UTF-8 / UTF-16 mode  
   
 Returns:       the number of the found subpattern, or -1 if not found  
 */  
   
 static int  
 find_parens(compile_data *cd, const pcre_uchar *name, int lorn, BOOL xmode,  
   BOOL utf)  
 {  
 pcre_uchar *ptr = (pcre_uchar *)cd->start_pattern;  
 int count = 0;  
 int rc;  
   
 /* If the pattern does not start with an opening parenthesis, the first call  
 to find_parens_sub() will scan right to the end (if necessary). However, if it  
 does start with a parenthesis, find_parens_sub() will return when it hits the  
 matching closing parens. That is why we have to have a loop. */  
   
 for (;;)  
   {  
   rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, &count);  
   if (rc > 0 || *ptr++ == 0) break;  
   }  
   
 return rc;  
 }  
   
   
   
   
 /*************************************************  
 *      Find first significant op code            *  
 *************************************************/  
   
 /* This is called by several functions that scan a compiled expression looking  
 for a fixed first character, or an anchoring op code etc. It skips over things  
 that do not influence this. For some calls, it makes sense to skip negative  
 forward and all backward assertions, and also the \b assertion; for others it  
 does not.  
   
 Arguments:  
   code         pointer to the start of the group  
   skipassert   TRUE if certain assertions are to be skipped  
   
 Returns:       pointer to the first significant opcode  
 */  
   
 static const pcre_uchar*  
 first_significant_code(const pcre_uchar *code, BOOL skipassert)  
 {  
 for (;;)  
   {  
   switch ((int)*code)  
     {  
     case OP_ASSERT_NOT:  
     case OP_ASSERTBACK:  
     case OP_ASSERTBACK_NOT:  
     if (!skipassert) return code;  
     do code += GET(code, 1); while (*code == OP_ALT);  
     code += PRIV(OP_lengths)[*code];  
     break;  
   
     case OP_WORD_BOUNDARY:  
     case OP_NOT_WORD_BOUNDARY:  
     if (!skipassert) return code;  
     /* Fall through */  
   
     case OP_CALLOUT:  
     case OP_CREF:  
     case OP_NCREF:  
     case OP_RREF:  
     case OP_NRREF:  
     case OP_DEF:  
     code += PRIV(OP_lengths)[*code];  
     break;  
   
     default:  
     return code;  
     }  
   }  
 /* Control never reaches here */  
 }  
   
   
   
   
 /*************************************************  
 *        Find the fixed length of a branch       *  
 *************************************************/  
   
 /* Scan a branch and compute the fixed length of subject that will match it,  
 if the length is fixed. This is needed for dealing with backward assertions.  
 In UTF8 mode, the result is in characters rather than bytes. The branch is  
 temporarily terminated with OP_END when this function is called.  
   
 This function is called when a backward assertion is encountered, so that if it  
 fails, the error message can point to the correct place in the pattern.  
 However, we cannot do this when the assertion contains subroutine calls,  
 because they can be forward references. We solve this by remembering this case  
 and doing the check at the end; a flag specifies which mode we are running in.  
   
 Arguments:  
   code     points to the start of the pattern (the bracket)  
   utf      TRUE in UTF-8 / UTF-16 mode  
   atend    TRUE if called when the pattern is complete  
   cd       the "compile data" structure  
   
 Returns:   the fixed length,  
              or -1 if there is no fixed length,  
              or -2 if \C was encountered (in UTF-8 mode only)  
              or -3 if an OP_RECURSE item was encountered and atend is FALSE  
              or -4 if an unknown opcode was encountered (internal error)  
 */  
   
 static int  
 find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd)  
 {  
 int length = -1;  
   
 register int branchlength = 0;  
 register pcre_uchar *cc = code + 1 + LINK_SIZE;  
   
 /* Scan along the opcodes for this branch. If we get to the end of the  
 branch, check the length against that of the other branches. */  
   
 for (;;)  
   {  
   int d;  
   pcre_uchar *ce, *cs;  
   register int op = *cc;  
   
   switch (op)  
     {  
     /* We only need to continue for OP_CBRA (normal capturing bracket) and  
     OP_BRA (normal non-capturing bracket) because the other variants of these  
     opcodes are all concerned with unlimited repeated groups, which of course  
     are not of fixed length. */  
   
     case OP_CBRA:  
     case OP_BRA:  
     case OP_ONCE:  
     case OP_ONCE_NC:  
     case OP_COND:  
     d = find_fixedlength(cc + ((op == OP_CBRA)? IMM2_SIZE : 0), utf, atend, cd);  
     if (d < 0) return d;  
     branchlength += d;  
     do cc += GET(cc, 1); while (*cc == OP_ALT);  
     cc += 1 + LINK_SIZE;  
     break;  
   
     /* Reached end of a branch; if it's a ket it is the end of a nested call.  
     If it's ALT it is an alternation in a nested call. An ACCEPT is effectively  
     an ALT. If it is END it's the end of the outer call. All can be handled by  
     the same code. Note that we must not include the OP_KETRxxx opcodes here,  
     because they all imply an unlimited repeat. */  
1778    
1779      case OP_ALT:      case OP_ALT:
1780      case OP_KET:      case OP_KET:
# Line 1760  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 1773  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 1792  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 1832  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  #ifdef SUPPORT_UTF  #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]);
# Line 1841  for (;;) Line 1886  for (;;)
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) cc += 2;      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)
1890          cc += 2;
1891      cc += 1 + IMM2_SIZE + 1;      cc += 1 + IMM2_SIZE + 1;
1892      break;      break;
1893    
# Line 1876  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  
     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 1966  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 2002  for (;;) Line 2058  for (;;)
2058    
2059    
2060    
   
2061  /*************************************************  /*************************************************
2062  *    Scan compiled regex for specific bracket    *  *    Scan compiled regex for specific bracket    *
2063  *************************************************/  *************************************************/
# Line 2015  length. Line 2070  length.
2070    
2071  Arguments:  Arguments:
2072    code        points to start of expression    code        points to start of expression
2073    utf         TRUE in UTF-8 / UTF-16 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2074    number      the required bracket number or negative to find a lookbehind    number      the required bracket number or negative to find a lookbehind
2075    
2076  Returns:      pointer to the opcode for the bracket, or NULL if not found  Returns:      pointer to the opcode for the bracket, or NULL if not found
# Line 2026  PRIV(find_bracket)(const pcre_uchar *cod Line 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 2049  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 2079  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        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2138          || code[1 + IMM2_SIZE] == OP_NOTPROP) 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 2102  for (;;) Line 2154  for (;;)
2154    a multi-byte character. The length in the table is a minimum, so we have to    a multi-byte character. The length in the table is a minimum, so we have to
2155    arrange to skip the extra bytes. */    arrange to skip the extra bytes. */
2156    
2157  #ifdef SUPPORT_UTF  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2158      if (utf) switch(c)      if (utf) switch(c)
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 2154  instance of OP_RECURSE. Line 2234  instance of OP_RECURSE.
2234    
2235  Arguments:  Arguments:
2236    code        points to start of expression    code        points to start of expression
2237    utf         TRUE in UTF-8 / UTF-16 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2238    
2239  Returns:      pointer to the opcode for OP_RECURSE, or NULL if not found  Returns:      pointer to the opcode for OP_RECURSE, or NULL if not found
2240  */  */
# Line 2164  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 2199  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        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2283          || code[1 + IMM2_SIZE] == OP_NOTPROP) 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 2222  for (;;) Line 2299  for (;;)
2299      by a multi-byte character. The length in the table is a minimum, so we have      by a multi-byte character. The length in the table is a minimum, so we have
2300      to arrange to skip the extra bytes. */      to arrange to skip the extra bytes. */
2301    
2302  #ifdef SUPPORT_UTF  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2303      if (utf) switch(c)      if (utf) switch(c)
2304        {        {
2305        case OP_CHAR:        case OP_CHAR:
# Line 2308  bracket whose current branch will alread Line 2385  bracket whose current branch will alread
2385  Arguments:  Arguments:
2386    code        points to start of search    code        points to start of search
2387    endcode     points to where to stop    endcode     points to where to stop
2388    utf         TRUE if in UTF-8 / UTF-16 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 2346  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 backwards reference */      /* Completed reference; scan the referenced group, remembering it on the
2461        stack chain to detect mutual recursions. */
2462    
2463        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 2404  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 2420  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 2462  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 2478  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 2525  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      if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2658        || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;        code += 2;
2659      break;      break;
2660    
2661      /* End of branch */      /* End of branch */
# Line 2539  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  #ifdef SUPPORT_UTF  #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
2726    
2727      /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument      /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2728      string. */      string. */
2729    
2730        case OP_MARK:
2731        case OP_PRUNE_ARG:
2732        case OP_SKIP_ARG:
2733        case OP_THEN_ARG:
2734        code += code[1];
2735        break;
2736    
2737        /* None of the remaining opcodes are required to match a character. */
2738    
2739        default:
2740        break;
2741        }
2742      }
2743    
2744    return TRUE;
2745    }
2746    
2747    
2748    
2749    /*************************************************
2750    *    Scan compiled regex for non-emptiness       *
2751    *************************************************/
2752    
2753    /* This function is called to check for left recursive calls. We want to check
2754    the current branch of the current pattern to see if it could match the empty
2755    string. If it could, we must look outwards for branches at other levels,
2756    stopping when we pass beyond the bracket which is the subject of the recursion.
2757    This function is called only during the real compile, not during the
2758    pre-compile.
2759    
2760    Arguments:
2761      code        points to start of the recursion
2762      endcode     points to where to stop (current RECURSE item)
2763      bcptr       points to the chain of current (unclosed) branch starts
2764      utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2765      cd          pointers to tables etc
2766    
2767    Returns:      TRUE if what is matched could be empty
2768    */
2769    
2770    static BOOL
2771    could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
2772      branch_chain *bcptr, BOOL utf, compile_data *cd)
2773    {
2774    while (bcptr != NULL && bcptr->current_branch >= code)
2775      {
2776      if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd, NULL))
2777        return FALSE;
2778      bcptr = bcptr->outer;
2779      }
2780    return TRUE;
2781    }
2782    
2783    
2784    
2785    /*************************************************
2786    *        Base opcode of repeated opcodes         *
2787    *************************************************/
2788    
2789    /* Returns the base opcode for repeated single character type opcodes. If the
2790    opcode is not a repeated character type, it returns with the original value.
2791    
2792    Arguments:  c opcode
2793    Returns:    base opcode for the type
2794    */
2795    
2796    static pcre_uchar
2797    get_repeat_base(pcre_uchar c)
2798    {
2799    return (c > OP_TYPEPOSUPTO)? c :
2800           (c >= OP_TYPESTAR)?   OP_TYPESTAR :
2801           (c >= OP_NOTSTARI)?   OP_NOTSTARI :
2802           (c >= OP_NOTSTAR)?    OP_NOTSTAR :
2803           (c >= OP_STARI)?      OP_STARI :
2804                                 OP_STAR;
2805    }
2806    
2807    
2808    
2809    #ifdef SUPPORT_UCP
2810    /*************************************************
2811    *        Check a character and a property        *
2812    *************************************************/
2813    
2814    /* This function is called by check_auto_possessive() when a property item
2815    is adjacent to a fixed character.
2816    
2817    Arguments:
2818      c            the character
2819      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 if auto-possessifying is OK
2824    */
2825    
2826    static BOOL
2827    check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata,
2828      BOOL negated)
2829    {
2830    const pcre_uint32 *p;
2831    const ucd_record *prop = GET_UCD(c);
2832    
2833    switch(ptype)
2834      {
2835      case PT_LAMP:
2836      return (prop->chartype == ucp_Lu ||
2837              prop->chartype == ucp_Ll ||
2838              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        HSPACE_CASES:
2864        VSPACE_CASES:
2865        return negated;
2866    
2867        default:
2868        return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z) == negated;
2869        }
2870      break;  /* Control never reaches here */
2871    
2872      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;
2888    }
2889    #endif  /* SUPPORT_UCP */
2890    
2891    
2892    
2893    /*************************************************
2894    *        Fill the character property list        *
2895    *************************************************/
2896    
2897    /* Checks whether the code points to an opcode that can take part in auto-
2898    possessification, and if so, fills a list with its properties.
2899    
2900    Arguments:
2901      code        points to start of expression
2902      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:      points to the start of the next opcode if *code is accepted
2911                  NULL if *code is not accepted
2912    */
2913    
2914    static const pcre_uchar *
2915    get_chr_property_list(const pcre_uchar *code, BOOL utf,
2916      const pcre_uint8 *fcc, pcre_uint32 *list)
2917    {
2918    pcre_uchar c = *code;
2919    pcre_uchar base;
2920    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      base = get_repeat_base(c);
2937      c -= (base - OP_STAR);
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);
2943    
2944      switch(base)
2945        {
2946        case OP_STAR:
2947        list[0] = OP_CHAR;
2948        break;
2949    
2950        case OP_STARI:
2951        list[0] = OP_CHARI;
2952        break;
2953    
2954        case OP_NOTSTAR:
2955        list[0] = OP_NOT;
2956        break;
2957    
2958        case OP_NOTSTARI:
2959        list[0] = OP_NOTI;
2960        break;
2961    
2962        case OP_TYPESTAR:
2963        list[0] = *code;
2964        code++;
2965        break;
2966        }
2967      c = list[0];
2968      }
2969    
2970    switch(c)
2971      {
2972      case OP_NOT_DIGIT:
2973      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      case OP_CHAR:
2993      case OP_NOT:
2994      GETCHARINCTEST(chr, code);
2995      list[2] = chr;
2996      list[3] = NOTACHAR;
2997      return code;
2998    
2999      case OP_CHARI:
3000      case OP_NOTI:
3001      list[0] = (c == OP_CHARI) ? OP_CHAR : OP_NOT;
3002      GETCHARINCTEST(chr, code);
3003      list[2] = chr;
3004    
3005    #ifdef SUPPORT_UCP
3006      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      /* 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        list[2] = code[0];
3030        list[3] = code[1];
3031        return code + 2;
3032        }
3033    
3034      /* 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      return code;
3058    #endif
3059    
3060      case OP_NCLASS:
3061      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      switch(*end)
3071        {
3072        case OP_CRSTAR:
3073        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        case OP_CRPLUS:
3083        case OP_CRMINPLUS:
3084        case OP_CRPOSPLUS:
3085        end++;
3086        break;
3087    
3088        case OP_CRRANGE:
3089        case OP_CRMINRANGE:
3090        case OP_CRPOSRANGE:
3091        list[1] = (GET2(end, 1) == 0);
3092        end += 1 + 2 * IMM2_SIZE;
3093        break;
3094        }
3095      list[2] = (pcre_uint32)(end - code);
3096      return end;
3097      }
3098    return NULL;    /* Opcode not accepted */
3099    }
3100    
3101    
3102    
3103    /*************************************************
3104    *    Scan further character sets for match       *
3105    *************************************************/
3106    
3107    /* Checks whether the base and the current opcode have a common character, in
3108    which case the base cannot be possessified.
3109    
3110    Arguments:
3111      code        points to the byte code
3112      utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3113      cd          static compile data
3114      base_list   the data list of the base opcode
3115    
3116    Returns:      TRUE if the auto-possessification is possible
3117    */
3118    
3119    static BOOL
3120    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    pcre_uchar c;
3124    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      /* Skip over callouts */
3154    
3155      if (c == OP_CALLOUT)
3156        {
3157        code += PRIV(OP_lengths)[c];
3158        continue;
3159        }
3160    
3161      if (c == OP_ALT)
3162        {
3163        do code += GET(code, 1); while (*code == OP_ALT);
3164        c = *code;
3165        }
3166    
3167      switch(c)
3168        {
3169        case OP_END:
3170        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        return base_list[1] != 0;
3176    
3177        case OP_KET:
3178        /* 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          return !entered_a_group;
3197          }
3198    
3199        code += PRIV(OP_lengths)[c];
3200        continue;
3201    
3202        case OP_ONCE:
3203        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        while (*next_code == OP_ALT)
3210          {
3211          if (!compare_opcodes(code, utf, cd, base_list, base_end, rec_limit))
3212            return FALSE;
3213          code = next_code + 1 + LINK_SIZE;
3214          next_code += GET(next_code, 1);
3215          }
3216    
3217        entered_a_group = TRUE;
3218        continue;
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        continue;
3237    
3238        default:
3239        break;
3240        }
3241    
3242      /* Check for a supported opcode, and load its properties. */
3243    
3244      code = get_chr_property_list(code, utf, cd->fcc, list);
3245      if (code == NULL) return FALSE;    /* Unsupported */
3246    
3247      /* If either opcode is a small character list, set pointers for comparing
3248      characters from that list with another list, or with a property. */
3249    
3250      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      /* Character bitsets can also be compared to certain opcodes. */
3262    
3263      else if (base_list[0] == OP_CLASS || list[0] == OP_CLASS
3264    #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        invert_bits = FALSE;
3286        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    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3295          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          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 OP_NOT_WHITESPACE:
3317          invert_bits = TRUE;
3318          /* Fall through */
3319          case OP_WHITESPACE:
3320          set2 = (pcre_uint8 *)(cd->cbits + cbit_space);
3321          break;
3322    
3323          case OP_NOT_WORDCHAR:
3324          invert_bits = TRUE;
3325          /* Fall through */
3326          case OP_WORDCHAR:
3327          set2 = (pcre_uint8 *)(cd->cbits + cbit_word);
3328          break;
3329    
3330          default:
3331          return FALSE;
3332          }
3333    
3334        /* Because the sets are unaligned, we need
3335        to perform byte comparison here. */
3336        set_end = set1 + 32;
3337        if (invert_bits)
3338          {
3339          do
3340            {
3341            if ((*set1++ & ~(*set2++)) != 0) return FALSE;
3342            }
3343          while (set1 < set_end);
3344          }
3345        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    
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
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    /* 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    }
3640    
3641    
3642    
3643    /*************************************************
3644    *    Scan compiled regex for auto-possession     *
3645    *************************************************/
3646    
3647    /* Replaces single character iterations with their possessive alternatives
3648    if appropriate. This function modifies the compiled opcode!
3649    
3650    Arguments:
3651      code        points to start of the byte code
3652      utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3653      cd          static compile data
3654    
3655    Returns:      nothing
3656    */
3657    
3658    static void
3659    auto_possessify(pcre_uchar *code, BOOL utf, const compile_data *cd)
3660    {
3661    register pcre_uchar c;
3662    const pcre_uchar *end;
3663    pcre_uchar *repeat_opcode;
3664    pcre_uint32 list[8];
3665    int rec_limit;
3666    
3667    for (;;)
3668      {
3669      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        c -= get_repeat_base(c) - OP_STAR;
3682        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          switch(c)
3690            {
3691            case OP_STAR:
3692            *code += OP_POSSTAR - OP_STAR;
3693            break;
3694    
3695            case OP_MINSTAR:
3696            *code += OP_POSSTAR - OP_MINSTAR;
3697            break;
3698    
3699            case OP_PLUS:
3700            *code += OP_POSPLUS - OP_PLUS;
3701            break;
3702    
3703            case OP_MINPLUS:
3704            *code += OP_POSPLUS - OP_MINPLUS;
3705            break;
3706    
3707            case OP_QUERY:
3708            *code += OP_POSQUERY - OP_QUERY;
3709            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    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3729        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          /* end must not be NULL. */
3739          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            switch (c)
3747              {
3748              case OP_CRSTAR:
3749              case OP_CRMINSTAR:
3750              *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        c = *code;
3771        }
3772    
3773      switch(c)
3774        {
3775        case OP_END:
3776        return;
3777    
3778        case OP_TYPESTAR:
3779        case OP_TYPEMINSTAR:
3780        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        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 defined SUPPORT_UTF || !defined COMPILE_PCRE8
3799        case OP_XCLASS:
3800        code += GET(code, 1);
3801        break;
3802    #endif
3803    
3804      case OP_MARK:      case OP_MARK:
3805      case OP_PRUNE_ARG:      case OP_PRUNE_ARG:
3806      case OP_SKIP_ARG:      case OP_SKIP_ARG:
     code += code[1];  
     break;  
   
3807      case OP_THEN_ARG:      case OP_THEN_ARG:
3808      code += code[1];      code += code[1];
3809      break;      break;
   
     /* None of the remaining opcodes are required to match a character. */  
   
     default:  
     break;  
3810      }      }
   }  
   
 return TRUE;  
 }  
   
   
   
 /*************************************************  
 *    Scan compiled regex for non-emptiness       *  
 *************************************************/  
3811    
3812  /* This function is called to check for left recursive calls. We want to check    /* Add in the fixed length from the table */
 the current branch of the current pattern to see if it could match the empty  
 string. If it could, we must look outwards for branches at other levels,  
 stopping when we pass beyond the bracket which is the subject of the recursion.  
 This function is called only during the real compile, not during the  
 pre-compile.  
3813    
3814  Arguments:    code += PRIV(OP_lengths)[c];
   code        points to start of the recursion  
   endcode     points to where to stop (current RECURSE item)  
   bcptr       points to the chain of current (unclosed) branch starts  
   utf         TRUE if in UTF-8 / UTF-16 mode  
   cd          pointers to tables etc  
3815    
3816  Returns:      TRUE if what is matched could be empty    /* In UTF-8 mode, opcodes that are followed by a character may be followed by
3817  */    a multi-byte character. The length in the table is a minimum, so we have to
3818      arrange to skip the extra bytes. */
3819    
3820  static BOOL  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3821  could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,    if (utf) switch(c)
3822    branch_chain *bcptr, BOOL utf, compile_data *cd)      {
3823  {      case OP_CHAR:
3824  while (bcptr != NULL && bcptr->current_branch >= code)      case OP_CHARI:
3825    {      case OP_NOT:
3826    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd))      case OP_NOTI:
3827      return FALSE;      case OP_STAR:
3828    bcptr = bcptr->outer;      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
3885    }    }
 return TRUE;  
3886  }  }
3887    
3888    
# Line 2649  class, but [abc[:x\]pqr:]] is (so that a Line 3908  class, but [abc[:x\]pqr:]] is (so that a
3908  below handles the special case of \], but does not try to do any other escape  below handles the special case of \], but does not try to do any other escape
3909  processing. This makes it different from Perl for cases such as [:l\ower:]  processing. This makes it different from Perl for cases such as [:l\ower:]
3910  where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize  where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
3911  "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,  "l\ower". This is a lesser evil than not diagnosing bad classes when Perl does,
3912  I think.  I think.
3913    
3914  A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.  A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
# Line 2673  Returns:   TRUE or FALSE Line 3932  Returns:   TRUE or FALSE
3932  static BOOL  static BOOL
3933  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
3934  {  {
3935  int terminator;          /* Don't combine these lines; the Solaris cc */  pcre_uchar terminator;          /* Don't combine these lines; the Solaris cc */
3936  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
3937  for (++ptr; *ptr != 0; ptr++)  for (++ptr; *ptr != CHAR_NULL; ptr++)
3938    {    {
3939    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3940      ptr++;      ptr++;
# Line 2722  register int yield = 0; Line 3981  register int yield = 0;
3981  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
3982    {    {
3983    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
3984      STRNCMP_UC_C8(ptr, pn, len) == 0) return yield;      STRNCMP_UC_C8(ptr, pn, (unsigned int)len) == 0) return yield;
3985    pn += posix_name_lengths[yield] + 1;    pn += posix_name_lengths[yield] + 1;
3986    yield++;    yield++;
3987    }    }
# Line 2745  have their offsets adjusted. That one of Line 4004  have their offsets adjusted. That one of
4004  is called, the partially compiled regex must be temporarily terminated with  is called, the partially compiled regex must be temporarily terminated with
4005  OP_END.  OP_END.
4006    
4007  This function has been extended with the possibility of forward references for  This function has been extended to cope with forward references for recursions
4008  recursions and subroutine calls. It must also check the list of such references  and subroutine calls. It must check the list of such references for the
4009  for the group we are dealing with. If it finds that one of the recursions in  group we are dealing with. If it finds that one of the recursions in the
4010  the current group is on this list, it adjusts the offset in the list, not the  current group is on this list, it does not adjust the value in the reference
4011  value in the reference (which is a group number).  (which is a group number). After the group has been scanned, all the offsets in
4012    the forward reference list for the group are adjusted.
4013    
4014  Arguments:  Arguments:
4015    group      points to the start of the group    group      points to the start of the group
4016    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
4017    utf        TRUE in UTF-8 / UTF-16 mode    utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode
4018    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
4019    save_hwm   the hwm forward reference pointer at the start of the group    save_hwm_offset   the hwm forward reference offset at the start of the group
4020    
4021  Returns:     nothing  Returns:     nothing
4022  */  */
4023    
4024  static void  static void
4025  adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,  adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
4026    pcre_uchar *save_hwm)    size_t save_hwm_offset)
4027  {  {
4028    int offset;
4029    pcre_uchar *hc;
4030  pcre_uchar *ptr = group;  pcre_uchar *ptr = group;
4031    
4032  while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)  while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
4033    {    {
4034    int offset;    for (hc = (pcre_uchar *)cd->start_workspace + save_hwm_offset; hc < cd->hwm;
4035    pcre_uchar *hc;         hc += LINK_SIZE)
   
   /* See if this recursion is on the forward reference list. If so, adjust the  
   reference. */  
   
   for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)  
4036      {      {
4037      offset = GET(hc, 0);      offset = (int)GET(hc, 0);
4038      if (cd->start_code + offset == ptr + 1)      if (cd->start_code + offset == ptr + 1) break;
       {  
       PUT(hc, 0, offset + adjust);  
       break;  
       }  
4039      }      }
4040    
4041    /* Otherwise, adjust the recursion offset if it's after the start of this    /* If we have not found this recursion on the forward reference list, adjust
4042    group. */    the recursion's offset if it's after the start of this group. */
4043    
4044    if (hc >= cd->hwm)    if (hc >= cd->hwm)
4045      {      {
4046      offset = GET(ptr, 1);      offset = (int)GET(ptr, 1);
4047      if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);      if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
4048      }      }
4049    
4050    ptr += 1 + LINK_SIZE;    ptr += 1 + LINK_SIZE;
4051    }    }
4052    
4053    /* Now adjust all forward reference offsets for the group. */
4054    
4055    for (hc = (pcre_uchar *)cd->start_workspace + save_hwm_offset; hc < cd->hwm;
4056         hc += LINK_SIZE)
4057      {
4058      offset = (int)GET(hc, 0);
4059      PUT(hc, 0, offset + adjust);
4060      }
4061  }  }
4062    
4063    
# Line 2858  PUT(previous_callout, 2 + LINK_SIZE, len Line 4120  PUT(previous_callout, 2 + LINK_SIZE, len
4120  *************************************************/  *************************************************/
4121    
4122  /* This function is passed the start and end of a class range, in UTF-8 mode  /* This function is passed the start and end of a class range, in UTF-8 mode
4123  with UCP support. It searches up the characters, looking for internal ranges of  with UCP support. It searches up the characters, looking for ranges of
4124  characters in the "other" case. Each call returns the next one, updating the  characters in the "other" case. Each call returns the next one, updating the
4125  start address.  start address. A character with multiple other cases is returned on its own
4126    with a special return value.
4127    
4128  Arguments:  Arguments:
4129    cptr        points to starting character value; updated    cptr        points to starting character value; updated
# Line 2868  Arguments: Line 4131  Arguments:
4131    ocptr       where to put start of othercase range    ocptr       where to put start of othercase range
4132    odptr       where to put end of othercase range    odptr       where to put end of othercase range
4133    
4134  Yield:        TRUE when range returned; FALSE when no more  Yield:        -1 when no more
4135                   0 when a range is returned
4136                  >0 the CASESET offset for char with multiple other cases
4137                    in this case, ocptr contains the original
4138  */  */
4139    
4140  static BOOL  static int
4141  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,  get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,
4142    unsigned int *odptr)    pcre_uint32 *odptr)
4143  {  {
4144  unsigned int c, othercase, next;  pcre_uint32 c, othercase, next;
4145    unsigned int co;
4146    
4147    /* Find the first character that has an other case. If it has multiple other
4148    cases, return its case offset value. */
4149    
4150  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
4151    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }    {
4152      if ((co = UCD_CASESET(c)) != 0)
4153        {
4154        *ocptr = c++;   /* Character that has the set */
4155        *cptr = c;      /* Rest of input range */
4156        return (int)co;
4157        }
4158      if ((othercase = UCD_OTHERCASE(c)) != c) break;
4159      }
4160    
4161  if (c > d) return FALSE;  if (c > d) return -1;  /* Reached end of range */
4162    
4163    /* Found a character that has a single other case. Search for the end of the
4164    range, which is either the end of the input range, or a character that has zero
4165    or more than one other cases. */
4166    
4167  *ocptr = othercase;  *ocptr = othercase;
4168  next = othercase + 1;  next = othercase + 1;
4169    
4170  for (++c; c <= d; c++)  for (++c; c <= d; c++)
4171    {    {
4172    if (UCD_OTHERCASE(c) != next) break;    if ((co = UCD_CASESET(c)) != 0 || UCD_OTHERCASE(c) != next) break;
4173    next++;    next++;
4174    }    }
4175    
4176  *odptr = next - 1;  *odptr = next - 1;     /* End of othercase range */
4177  *cptr = c;  *cptr = c;             /* Rest of input range */
4178    return 0;
 return TRUE;  
 }  
   
   
   
 /*************************************************  
 *        Check a character and a property        *  
 *************************************************/  
   
 /* This function is called by check_auto_possessive() when a property item  
 is adjacent to a fixed character.  
   
 Arguments:  
   c            the character  
   ptype        the property type  
   pdata        the data for the type  
   negated      TRUE if it's a negated property (\P or \p{^)  
   
 Returns:       TRUE if auto-possessifying is OK  
 */  
   
 static BOOL  
 check_char_prop(int c, int ptype, int pdata, BOOL negated)  
 {  
 const ucd_record *prop = GET_UCD(c);  
 switch(ptype)  
   {  
   case PT_LAMP:  
   return (prop->chartype == ucp_Lu ||  
           prop->chartype == ucp_Ll ||  
           prop->chartype == ucp_Lt) == negated;  
   
   case PT_GC:  
   return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;  
   
   case PT_PC:  
   return (pdata == prop->chartype) == negated;  
   
   case PT_SC:  
   return (pdata == prop->script) == negated;  
   
   /* These are specials */  
   
   case PT_ALNUM:  
   return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||  
           PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;  
   
   case PT_SPACE:    /* Perl space */  
   return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||  
           c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)  
           == negated;  
   
   case PT_PXSPACE:  /* POSIX space */  
   return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||  
           c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||  
           c == CHAR_FF || c == CHAR_CR)  
           == negated;  
   
   case PT_WORD:  
   return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||  
           PRIV(ucp_gentype)[prop->chartype] == ucp_N ||  
           c == CHAR_UNDERSCORE) == negated;  
   }  
 return FALSE;  
4179  }  }
4180  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
4181    
4182    
4183    
4184  /*************************************************  /*************************************************
4185  *     Check if auto-possessifying is possible    *  *        Add a character or range to a class     *
4186  *************************************************/  *************************************************/
4187    
4188  /* This function is called for unlimited repeats of certain items, to see  /* This function packages up the logic of adding a character or range of
4189  whether the next thing could possibly match the repeated item. If not, it makes  characters to a class. The character values in the arguments will be within the
4190  sense to automatically possessify the repeated item.  valid values for the current mode (8-bit, 16-bit, UTF, etc). This function is
4191    mutually recursive with the function immediately below.
4192    
4193  Arguments:  Arguments:
4194    previous      pointer to the repeated opcode    classbits     the bit map for characters < 256
4195    utf           TRUE in UTF-8 / UTF-16 mode    uchardptr     points to the pointer for extra data
4196    ptr           next character in pattern    options       the options word
   options       options bits  
4197    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
4198      start         start of range character
4199      end           end of range character
4200    
4201  Returns:        TRUE if possessifying is wanted  Returns:        the number of < 256 characters added
4202 &nb