/[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 1564 by ph10, Tue Jun 9 16:33:27 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 182  static const short int escapes[] = { Line 204  static const short int escapes[] = {
204  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',
205  /*  80 */     0,     7, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,  /*  80 */     0,     7, -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 238  static const verbitem verbs[] = { Line 260  static const verbitem verbs[] = {
260  static const int verbcount = sizeof(verbs)/sizeof(verbitem);  static const int verbcount = sizeof(verbs)/sizeof(verbitem);
261    
262    
263    /* Substitutes for [[:<:]] and [[:>:]], which mean start and end of word in
264    another regex library. */
265    
266    static const pcre_uchar sub_start_of_word[] = {
267      CHAR_BACKSLASH, CHAR_b, CHAR_LEFT_PARENTHESIS, CHAR_QUESTION_MARK,
268      CHAR_EQUALS_SIGN, CHAR_BACKSLASH, CHAR_w, CHAR_RIGHT_PARENTHESIS, '\0' };
269    
270    static const pcre_uchar sub_end_of_word[] = {
271      CHAR_BACKSLASH, CHAR_b, CHAR_LEFT_PARENTHESIS, CHAR_QUESTION_MARK,
272      CHAR_LESS_THAN_SIGN, CHAR_EQUALS_SIGN, CHAR_BACKSLASH, CHAR_w,
273      CHAR_RIGHT_PARENTHESIS, '\0' };
274    
275    
276  /* 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
277  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
278  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
279  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
280  for handling case independence. */  for handling case independence. The indices for graph, print, and punct are
281    needed, so identify them. */
282    
283  static const char posix_names[] =  static const char posix_names[] =
284    STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0    STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
# Line 253  static const char posix_names[] = Line 289  static const char posix_names[] =
289  static const pcre_uint8 posix_name_lengths[] = {  static const pcre_uint8 posix_name_lengths[] = {
290    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 };
291    
292    #define PC_GRAPH  8
293    #define PC_PRINT  9
294    #define PC_PUNCT 10
295    
296    
297  /* 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
298  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
299  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 321  static const int posix_class_maps[] = {
321    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
322  };  };
323    
324  /* 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
325  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. */  
326    
327  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
328  static const pcre_uchar string_PNd[]  = {  static const pcre_uchar string_PNd[]  = {
# Line 307  static const pcre_uchar string_pXwd[] = Line 347  static const pcre_uchar string_pXwd[] =
347  static const pcre_uchar *substitutes[] = {  static const pcre_uchar *substitutes[] = {
348    string_PNd,           /* \D */    string_PNd,           /* \D */
349    string_pNd,           /* \d */    string_pNd,           /* \d */
350    string_PXsp,          /* \S */       /* NOTE: Xsp is Perl space */    string_PXsp,          /* \S */   /* Xsp is Perl space, but from 8.34, Perl */
351    string_pXsp,          /* \s */    string_pXsp,          /* \s */   /* space and POSIX space are the same. */
352    string_PXwd,          /* \W */    string_PXwd,          /* \W */
353    string_pXwd           /* \w */    string_pXwd           /* \w */
354  };  };
355    
356    /* The POSIX class substitutes must be in the order of the POSIX class names,
357    defined above, and there are both positive and negative cases. NULL means no
358    general substitute of a Unicode property escape (\p or \P). However, for some
359    POSIX classes (e.g. graph, print, punct) a special property code is compiled
360    directly. */
361    
362  static const pcre_uchar string_pL[] =   {  static const pcre_uchar string_pL[] =   {
363    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
364    CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };    CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
# Line 360  static const pcre_uchar *posix_substitut Line 406  static const pcre_uchar *posix_substitut
406    NULL,                 /* graph */    NULL,                 /* graph */
407    NULL,                 /* print */    NULL,                 /* print */
408    NULL,                 /* punct */    NULL,                 /* punct */
409    string_pXps,          /* space */    /* NOTE: Xps is POSIX space */    string_pXps,          /* space */   /* Xps is POSIX space, but from 8.34 */
410    string_pXwd,          /* word */    string_pXwd,          /* word  */   /* Perl and POSIX space are the same */
411    NULL,                 /* xdigit */    NULL,                 /* xdigit */
412    /* Negated cases */    /* Negated cases */
413    string_PL,            /* ^alpha */    string_PL,            /* ^alpha */
# Line 375  static const pcre_uchar *posix_substitut Line 421  static const pcre_uchar *posix_substitut
421    NULL,                 /* ^graph */    NULL,                 /* ^graph */
422    NULL,                 /* ^print */    NULL,                 /* ^print */
423    NULL,                 /* ^punct */    NULL,                 /* ^punct */
424    string_PXps,          /* ^space */   /* NOTE: Xps is POSIX space */    string_PXps,          /* ^space */  /* Xps is POSIX space, but from 8.34 */
425    string_PXwd,          /* ^word */    string_PXwd,          /* ^word */   /* Perl and POSIX space are the same */
426    NULL                  /* ^xdigit */    NULL                  /* ^xdigit */
427  };  };
428  #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 486  static const char error_texts[] =
486    "POSIX collating elements are not supported\0"    "POSIX collating elements are not supported\0"
487    "this version of PCRE is compiled without UTF support\0"    "this version of PCRE is compiled without UTF support\0"
488    "spare error\0"  /** DEAD **/    "spare error\0"  /** DEAD **/
489    "character value in \\x{...} sequence is too large\0"    "character value in \\x{} or \\o{} is too large\0"
490    /* 35 */    /* 35 */
491    "invalid condition (?(0)\0"    "invalid condition (?(0)\0"
492    "\\C not allowed in lookbehind assertion\0"    "\\C not allowed in lookbehind assertion\0"
# Line 472  static const char error_texts[] = Line 518  static const char error_texts[] =
518    "a numbered reference must not be zero\0"    "a numbered reference must not be zero\0"
519    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
520    /* 60 */    /* 60 */
521    "(*VERB) not recognized\0"    "(*VERB) not recognized or malformed\0"
522    "number is too big\0"    "number is too big\0"
523    "subpattern name expected\0"    "subpattern name expected\0"
524    "digit expected after (?+\0"    "digit expected after (?+\0"
# Line 490  static const char error_texts[] = Line 536  static const char error_texts[] =
536    "disallowed Unicode code point (>= 0xd800 && <= 0xdfff)\0"    "disallowed Unicode code point (>= 0xd800 && <= 0xdfff)\0"
537    "invalid UTF-16 string\0"    "invalid UTF-16 string\0"
538    /* 75 */    /* 75 */
539    "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"    "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"
540      "character value in \\u.... sequence is too large\0"
541      "invalid UTF-32 string\0"
542      "setting UTF is disabled by the application\0"
543      "non-hex character in \\x{} (closing brace missing?)\0"
544      /* 80 */
545      "non-octal character in \\o{} (closing brace missing?)\0"
546      "missing opening brace after \\o\0"
547      "parentheses are too deeply nested\0"
548      "invalid range in character class\0"
549      "group name must start with a non-digit\0"
550      /* 85 */
551      "parentheses are too deeply nested (stack check)\0"
552      "digits missing in \\x{} or \\o{}\0"
553    ;    ;
554    
555  /* 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 689  static const pcre_uint8 ebcdic_chartab[]
689  #endif  #endif
690    
691    
692  /* Definition to allow mutual recursion */  /* This table is used to check whether auto-possessification is possible
693    between adjacent character-type opcodes. The left-hand (repeated) opcode is
694    used to select the row, and the right-hand opcode is use to select the column.
695    A value of 1 means that auto-possessification is OK. For example, the second
696    value in the first row means that \D+\d can be turned into \D++\d.
697    
698    The Unicode property types (\P and \p) have to be present to fill out the table
699    because of what their opcode values are, but the table values should always be
700    zero because property types are handled separately in the code. The last four
701    columns apply to items that cannot be repeated, so there is no need to have
702    rows for them. Note that OP_DIGIT etc. are generated only when PCRE_UCP is
703    *not* set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
704    
705    #define APTROWS (LAST_AUTOTAB_LEFT_OP - FIRST_AUTOTAB_OP + 1)
706    #define APTCOLS (LAST_AUTOTAB_RIGHT_OP - FIRST_AUTOTAB_OP + 1)
707    
708    static const pcre_uint8 autoposstab[APTROWS][APTCOLS] = {
709    /* \D \d \S \s \W \w  . .+ \C \P \p \R \H \h \V \v \X \Z \z  $ $M */
710      { 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \D */
711      { 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \d */
712      { 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \S */
713      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \s */
714      { 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \W */
715      { 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \w */
716      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .  */
717      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .+ */
718      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \C */
719      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \P */
720      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \p */
721      { 0, 1, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \R */
722      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \H */
723      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \h */
724      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \V */
725      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0 },  /* \v */
726      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }   /* \X */
727    };
728    
729    
730  static BOOL  /* This table is used to check whether auto-possessification is possible
731    compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL, int, int,  between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP). The
732      int *, int *, branch_chain *, compile_data *, int *);  left-hand (repeated) opcode is used to select the row, and the right-hand
733    opcode is used to select the column. The values are as follows:
734    
735      0   Always return FALSE (never auto-possessify)
736      1   Character groups are distinct (possessify if both are OP_PROP)
737      2   Check character categories in the same group (general or particular)
738      3   TRUE if the two opcodes are not the same (PROP vs NOTPROP)
739    
740      4   Check left general category vs right particular category
741      5   Check right general category vs left particular category
742    
743      6   Left alphanum vs right general category
744      7   Left space vs right general category
745      8   Left word vs right general category
746    
747      9   Right alphanum vs left general category
748     10   Right space vs left general category
749     11   Right word vs left general category
750    
751     12   Left alphanum vs right particular category
752     13   Left space vs right particular category
753     14   Left word vs right particular category
754    
755     15   Right alphanum vs left particular category
756     16   Right space vs left particular category
757     17   Right word vs left particular category
758    */
759    
760    static const pcre_uint8 propposstab[PT_TABSIZE][PT_TABSIZE] = {
761    /* ANY LAMP GC  PC  SC ALNUM SPACE PXSPACE WORD CLIST UCNC */
762      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_ANY */
763      { 0,  3,  0,  0,  0,    3,    1,      1,   0,    0,   0 },  /* PT_LAMP */
764      { 0,  0,  2,  4,  0,    9,   10,     10,  11,    0,   0 },  /* PT_GC */
765      { 0,  0,  5,  2,  0,   15,   16,     16,  17,    0,   0 },  /* PT_PC */
766      { 0,  0,  0,  0,  2,    0,    0,      0,   0,    0,   0 },  /* PT_SC */
767      { 0,  3,  6, 12,  0,    3,    1,      1,   0,    0,   0 },  /* PT_ALNUM */
768      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_SPACE */
769      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_PXSPACE */
770      { 0,  0,  8, 14,  0,    0,    1,      1,   3,    0,   0 },  /* PT_WORD */
771      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_CLIST */
772      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   3 }   /* PT_UCNC */
773    };
774    
775    /* This table is used to check whether auto-possessification is possible
776    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP) when one
777    specifies a general category and the other specifies a particular category. The
778    row is selected by the general category and the column by the particular
779    category. The value is 1 if the particular category is not part of the general
780    category. */
781    
782    static const pcre_uint8 catposstab[7][30] = {
783    /* 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 */
784      { 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 */
785      { 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 */
786      { 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 */
787      { 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 */
788      { 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 */
789      { 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 */
790      { 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 */
791    };
792    
793    /* This table is used when checking ALNUM, (PX)SPACE, SPACE, and WORD against
794    a general or particular category. The properties in each row are those
795    that apply to the character set in question. Duplication means that a little
796    unnecessary work is done when checking, but this keeps things much simpler
797    because they can all use the same code. For more details see the comment where
798    this table is used.
799    
800    Note: SPACE and PXSPACE used to be different because Perl excluded VT from
801    "space", but from Perl 5.18 it's included, so both categories are treated the
802    same here. */
803    
804    static const pcre_uint8 posspropstab[3][4] = {
805      { ucp_L, ucp_N, ucp_N, ucp_Nl },  /* ALNUM, 3rd and 4th values redundant */
806      { ucp_Z, ucp_Z, ucp_C, ucp_Cc },  /* SPACE and PXSPACE, 2nd value redundant */
807      { ucp_L, ucp_N, ucp_P, ucp_Po }   /* WORD */
808    };
809    
810    /* This table is used when converting repeating opcodes into possessified
811    versions as a result of an explicit possessive quantifier such as ++. A zero
812    value means there is no possessified version - in those cases the item in
813    question must be wrapped in ONCE brackets. The table is truncated at OP_CALLOUT
814    because all relevant opcodes are less than that. */
815    
816    static const pcre_uint8 opcode_possessify[] = {
817      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 0 - 15  */
818      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 16 - 31 */
819    
820      0,                       /* NOTI */
821      OP_POSSTAR, 0,           /* STAR, MINSTAR */
822      OP_POSPLUS, 0,           /* PLUS, MINPLUS */
823      OP_POSQUERY, 0,          /* QUERY, MINQUERY */
824      OP_POSUPTO, 0,           /* UPTO, MINUPTO */
825      0,                       /* EXACT */
826      0, 0, 0, 0,              /* POS{STAR,PLUS,QUERY,UPTO} */
827    
828      OP_POSSTARI, 0,          /* STARI, MINSTARI */
829      OP_POSPLUSI, 0,          /* PLUSI, MINPLUSI */
830      OP_POSQUERYI, 0,         /* QUERYI, MINQUERYI */
831      OP_POSUPTOI, 0,          /* UPTOI, MINUPTOI */
832      0,                       /* EXACTI */
833      0, 0, 0, 0,              /* POS{STARI,PLUSI,QUERYI,UPTOI} */
834    
835      OP_NOTPOSSTAR, 0,        /* NOTSTAR, NOTMINSTAR */
836      OP_NOTPOSPLUS, 0,        /* NOTPLUS, NOTMINPLUS */
837      OP_NOTPOSQUERY, 0,       /* NOTQUERY, NOTMINQUERY */
838      OP_NOTPOSUPTO, 0,        /* NOTUPTO, NOTMINUPTO */
839      0,                       /* NOTEXACT */
840      0, 0, 0, 0,              /* NOTPOS{STAR,PLUS,QUERY,UPTO} */
841    
842      OP_NOTPOSSTARI, 0,       /* NOTSTARI, NOTMINSTARI */
843      OP_NOTPOSPLUSI, 0,       /* NOTPLUSI, NOTMINPLUSI */
844      OP_NOTPOSQUERYI, 0,      /* NOTQUERYI, NOTMINQUERYI */
845      OP_NOTPOSUPTOI, 0,       /* NOTUPTOI, NOTMINUPTOI */
846      0,                       /* NOTEXACTI */
847      0, 0, 0, 0,              /* NOTPOS{STARI,PLUSI,QUERYI,UPTOI} */
848    
849      OP_TYPEPOSSTAR, 0,       /* TYPESTAR, TYPEMINSTAR */
850      OP_TYPEPOSPLUS, 0,       /* TYPEPLUS, TYPEMINPLUS */
851      OP_TYPEPOSQUERY, 0,      /* TYPEQUERY, TYPEMINQUERY */
852      OP_TYPEPOSUPTO, 0,       /* TYPEUPTO, TYPEMINUPTO */
853      0,                       /* TYPEEXACT */
854      0, 0, 0, 0,              /* TYPEPOS{STAR,PLUS,QUERY,UPTO} */
855    
856      OP_CRPOSSTAR, 0,         /* CRSTAR, CRMINSTAR */
857      OP_CRPOSPLUS, 0,         /* CRPLUS, CRMINPLUS */
858      OP_CRPOSQUERY, 0,        /* CRQUERY, CRMINQUERY */
859      OP_CRPOSRANGE, 0,        /* CRRANGE, CRMINRANGE */
860      0, 0, 0, 0,              /* CRPOS{STAR,PLUS,QUERY,RANGE} */
861    
862      0, 0, 0,                 /* CLASS, NCLASS, XCLASS */
863      0, 0,                    /* REF, REFI */
864      0, 0,                    /* DNREF, DNREFI */
865      0, 0                     /* RECURSE, CALLOUT */
866    };
867    
868    
869    
# Line 657  find_error_text(int n) Line 886  find_error_text(int n)
886  const char *s = error_texts;  const char *s = error_texts;
887  for (; n > 0; n--)  for (; n > 0; n--)
888    {    {
889    while (*s++ != 0) {};    while (*s++ != CHAR_NULL) {};
890    if (*s == 0) return "Error text not found (please report)";    if (*s == CHAR_NULL) return "Error text not found (please report)";
891    }    }
892  return s;  return s;
893  }  }
894    
895    
896    
897  /*************************************************  /*************************************************
898  *           Expand the workspace                 *  *           Expand the workspace                 *
899  *************************************************/  *************************************************/
# Line 741  return (*p == CHAR_RIGHT_CURLY_BRACKET); Line 971  return (*p == CHAR_RIGHT_CURLY_BRACKET);
971  *************************************************/  *************************************************/
972    
973  /* 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
974  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
975  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.
976  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
977  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
978  ptr is pointing at the \. On exit, it is on the final character of the escape  character of the escape sequence.
 sequence.  
979    
980  Arguments:  Arguments:
981    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
982      chptr          points to a returned data character
983    errorcodeptr   points to the errorcode variable    errorcodeptr   points to the errorcode variable
984    bracount       number of previous extracting brackets    bracount       number of previous extracting brackets
985    options        the options bits    options        the options bits
986    isclass        TRUE if inside a character class    isclass        TRUE if inside a character class
987    
988  Returns:         zero or positive => a data character  Returns:         zero => a data character
989                   negative => a special escape sequence                   positive => a special escape sequence
990                     negative => a back reference
991                   on error, errorcodeptr is set                   on error, errorcodeptr is set
992  */  */
993    
994  static int  static int
995  check_escape(const pcre_uchar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,
996    int options, BOOL isclass)    int bracount, int options, BOOL isclass)
997  {  {
998  /* PCRE_UTF16 has the same value as PCRE_UTF8. */  /* PCRE_UTF16 has the same value as PCRE_UTF8. */
999  BOOL utf = (options & PCRE_UTF8) != 0;  BOOL utf = (options & PCRE_UTF8) != 0;
1000  const pcre_uchar *ptr = *ptrptr + 1;  const pcre_uchar *ptr = *ptrptr + 1;
1001  pcre_int32 c;  pcre_uint32 c;
1002    int escape = 0;
1003  int i;  int i;
1004    
1005  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
# Line 775  ptr--;                            /* Set Line 1007  ptr--;                            /* Set
1007    
1008  /* 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. */
1009    
1010  if (c == 0) *errorcodeptr = ERR1;  if (c == CHAR_NULL) *errorcodeptr = ERR1;
1011    
1012  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
1013  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 1016  Otherwise further processing may be requ
1016  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1017  /* Not alphanumeric */  /* Not alphanumeric */
1018  else if (c < CHAR_0 || c > CHAR_z) {}  else if (c < CHAR_0 || c > CHAR_z) {}
1019  else if ((i = escapes[c - CHAR_0]) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0)
1020      { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1021    
1022  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1023  /* Not alphanumeric */  /* Not alphanumeric */
1024  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)) {}
1025  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; }
1026  #endif  #endif
1027    
1028  /* 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 1030  else if ((i = escapes[c - 0x48]) != 0)
1030  else  else
1031    {    {
1032    const pcre_uchar *oldptr;    const pcre_uchar *oldptr;
1033    BOOL braced, negated;    BOOL braced, negated, overflow;
1034      int s;
1035    
1036    switch (c)    switch (c)
1037      {      {
# Line 822  else Line 1056  else
1056          c = 0;          c = 0;
1057          for (i = 0; i < 4; ++i)          for (i = 0; i < 4; ++i)
1058            {            {
1059            register int cc = *(++ptr);            register pcre_uint32 cc = *(++ptr);
1060  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1061            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1062            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 1065  else
1065            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1066  #endif  #endif
1067            }            }
1068    
1069    #if defined COMPILE_PCRE8
1070            if (c > (utf ? 0x10ffffU : 0xffU))
1071    #elif defined COMPILE_PCRE16
1072            if (c > (utf ? 0x10ffffU : 0xffffU))
1073    #elif defined COMPILE_PCRE32
1074            if (utf && c > 0x10ffffU)
1075    #endif
1076              {
1077              *errorcodeptr = ERR76;
1078              }
1079            else if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1080          }          }
1081        }        }
1082      else      else
# Line 857  else Line 1103  else
1103      (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
1104      number either in angle brackets or in single quotes. However, these are      number either in angle brackets or in single quotes. However, these are
1105      (possibly recursive) subroutine calls, _not_ backreferences. Just return      (possibly recursive) subroutine calls, _not_ backreferences. Just return
1106      the -ESC_g code (cf \k). */      the ESC_g code (cf \k). */
1107    
1108      case CHAR_g:      case CHAR_g:
1109      if (isclass) break;      if (isclass) break;
1110      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
1111        {        {
1112        c = -ESC_g;        escape = ESC_g;
1113        break;        break;
1114        }        }
1115    
# Line 872  else Line 1118  else
1118      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1119        {        {
1120        const pcre_uchar *p;        const pcre_uchar *p;
1121        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++)
1122          if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;          if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
1123        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)        if (*p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET)
1124          {          {
1125          c = -ESC_k;          escape = ESC_k;
1126          break;          break;
1127          }          }
1128        braced = TRUE;        braced = TRUE;
# Line 892  else Line 1138  else
1138      else negated = FALSE;      else negated = FALSE;
1139    
1140      /* The integer range is limited by the machine's int representation. */      /* The integer range is limited by the machine's int representation. */
1141      c = 0;      s = 0;
1142        overflow = FALSE;
1143      while (IS_DIGIT(ptr[1]))      while (IS_DIGIT(ptr[1]))
1144        {        {
1145        if (((unsigned int)c) > INT_MAX / 10) /* Integer overflow */        if (s > INT_MAX / 10 - 1) /* Integer overflow */
1146          {          {
1147          c = -1;          overflow = TRUE;
1148          break;          break;
1149          }          }
1150        c = c * 10 + *(++ptr) - CHAR_0;        s = s * 10 + (int)(*(++ptr) - CHAR_0);
1151        }        }
1152      if (((unsigned int)c) > INT_MAX) /* Integer overflow */      if (overflow) /* Integer overflow */
1153        {        {
1154        while (IS_DIGIT(ptr[1]))        while (IS_DIGIT(ptr[1]))
1155          ptr++;          ptr++;
# Line 916  else Line 1163  else
1163        break;        break;
1164        }        }
1165    
1166      if (c == 0)      if (s == 0)
1167        {        {
1168        *errorcodeptr = ERR58;        *errorcodeptr = ERR58;
1169        break;        break;
# Line 924  else Line 1171  else
1171    
1172      if (negated)      if (negated)
1173        {        {
1174        if (c > bracount)        if (s > bracount)
1175          {          {
1176          *errorcodeptr = ERR15;          *errorcodeptr = ERR15;
1177          break;          break;
1178          }          }
1179        c = bracount - (c - 1);        s = bracount - (s - 1);
1180        }        }
1181    
1182      c = -(ESC_REF + c);      escape = -s;
1183      break;      break;
1184    
1185      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
1186      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
1187      the way Perl works seems to be as follows:      over the years. Nowadays \g{} for backreferences and \o{} for octal are
1188        recommended to avoid the ambiguities in the old syntax.
1189    
1190      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
1191      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
1192      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
1193      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
1194      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
1195      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
1196      character class, \ followed by a digit is always an octal number. */      taken. \8 and \9 are treated as the literal characters 8 and 9.
1197    
1198        Inside a character class, \ followed by a digit is always either a literal
1199        8 or 9 or an octal number. */
1200    
1201      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:
1202      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 1205  else
1205        {        {
1206        oldptr = ptr;        oldptr = ptr;
1207        /* The integer range is limited by the machine's int representation. */        /* The integer range is limited by the machine's int representation. */
1208        c -= CHAR_0;        s = (int)(c -CHAR_0);
1209          overflow = FALSE;
1210        while (IS_DIGIT(ptr[1]))        while (IS_DIGIT(ptr[1]))
1211          {          {
1212          if (((unsigned int)c) > INT_MAX / 10) /* Integer overflow */          if (s > INT_MAX / 10 - 1) /* Integer overflow */
1213            {            {
1214            c = -1;            overflow = TRUE;
1215            break;            break;
1216            }            }
1217          c = c * 10 + *(++ptr) - CHAR_0;          s = s * 10 + (int)(*(++ptr) - CHAR_0);
1218          }          }
1219        if (((unsigned int)c) > INT_MAX) /* Integer overflow */        if (overflow) /* Integer overflow */
1220          {          {
1221          while (IS_DIGIT(ptr[1]))          while (IS_DIGIT(ptr[1]))
1222            ptr++;            ptr++;
1223          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
1224          break;          break;
1225          }          }
1226        if (c < 10 || c <= bracount)        if (s < 8 || s <= bracount)  /* Check for back reference */
1227          {          {
1228          c = -(ESC_REF + c);          escape = -s;
1229          break;          break;
1230          }          }
1231        ptr = oldptr;      /* Put the pointer back and fall through */        ptr = oldptr;      /* Put the pointer back and fall through */
1232        }        }
1233    
1234      /* 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
1235      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
1236      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
1237        changed so as not to insert the binary zero. */
1238    
1239      if ((c = *ptr) >= CHAR_8)      if ((c = *ptr) >= CHAR_8) break;
1240        {  
1241        ptr--;      /* Fall through with a digit less than 8 */
       c = 0;  
       break;  
       }  
1242    
1243      /* \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
1244      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 1255  else
1255  #endif  #endif
1256      break;      break;
1257    
1258      /* \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
1259      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}. */
1260      If not, { is treated as a data character. */  
1261        case CHAR_o:
1262        if (ptr[1] != CHAR_LEFT_CURLY_BRACKET) *errorcodeptr = ERR81; else
1263        if (ptr[2] == CHAR_RIGHT_CURLY_BRACKET) *errorcodeptr = ERR86; else
1264          {
1265          ptr += 2;
1266          c = 0;
1267          overflow = FALSE;
1268          while (*ptr >= CHAR_0 && *ptr <= CHAR_7)
1269            {
1270            register pcre_uint32 cc = *ptr++;
1271            if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1272    #ifdef COMPILE_PCRE32
1273            if (c >= 0x20000000l) { overflow = TRUE; break; }
1274    #endif
1275            c = (c << 3) + cc - CHAR_0 ;
1276    #if defined COMPILE_PCRE8
1277            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1278    #elif defined COMPILE_PCRE16
1279            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1280    #elif defined COMPILE_PCRE32
1281            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1282    #endif
1283            }
1284          if (overflow)
1285            {
1286            while (*ptr >= CHAR_0 && *ptr <= CHAR_7) ptr++;
1287            *errorcodeptr = ERR34;
1288            }
1289          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1290            {
1291            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1292            }
1293          else *errorcodeptr = ERR80;
1294          }
1295        break;
1296    
1297        /* \x is complicated. In JavaScript, \x must be followed by two hexadecimal
1298        numbers. Otherwise it is a lowercase x letter. */
1299    
1300      case CHAR_x:      case CHAR_x:
1301      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1302        {        {
       /* In JavaScript, \x must be followed by two hexadecimal numbers.  
       Otherwise it is a lowercase x letter. */  
1303        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1304          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1305          {          {
1306          c = 0;          c = 0;
1307          for (i = 0; i < 2; ++i)          for (i = 0; i < 2; ++i)
1308            {            {
1309            register int cc = *(++ptr);            register pcre_uint32 cc = *(++ptr);
1310  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1311            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1312            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 1316  else
1316  #endif  #endif
1317            }            }
1318          }          }
1319        break;        }    /* End JavaScript handling */
       }  
1320    
1321      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      /* Handle \x in Perl's style. \x{ddd} is a character number which can be
1322        {      greater than 0xff in utf or non-8bit mode, but only if the ddd are hex
1323        const pcre_uchar *pt = ptr + 2;      digits. If not, { used to be treated as a data character. However, Perl
1324        seems to read hex digits up to the first non-such, and ignore the rest, so
1325        that, for example \x{zz} matches a binary zero. This seems crazy, so PCRE
1326        now gives an error. */
1327    
1328        c = 0;      else
1329        while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0)        {
1330          if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1331          {          {
1332          register int cc = *pt++;          ptr += 2;
1333          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */          if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1334              {
1335              *errorcodeptr = ERR86;
1336              break;
1337              }
1338            c = 0;
1339            overflow = FALSE;
1340            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0)
1341              {
1342              register pcre_uint32 cc = *ptr++;
1343              if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1344    
1345    #ifdef COMPILE_PCRE32
1346              if (c >= 0x10000000l) { overflow = TRUE; break; }
1347    #endif
1348    
1349  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1350          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1351          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1352  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1353          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 */
1354          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1355  #endif  #endif
1356    
1357  #ifdef COMPILE_PCRE8  #if defined COMPILE_PCRE8
1358          if (c > (utf ? 0x10ffff : 0xff)) { c = -1; break; }            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1359  #else  #elif defined COMPILE_PCRE16
1360  #ifdef COMPILE_PCRE16            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1361          if (c > (utf ? 0x10ffff : 0xffff)) { c = -1; break; }  #elif defined COMPILE_PCRE32
1362  #endif            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1363  #endif  #endif
1364          }            }
1365    
1366        if (c < 0)          if (overflow)
1367          {            {
1368          while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0) pt++;            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0) ptr++;
1369          *errorcodeptr = ERR34;            *errorcodeptr = ERR34;
1370          }            }
1371    
1372        if (*pt == CHAR_RIGHT_CURLY_BRACKET)          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1373          {            {
1374          if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1375          ptr = pt;            }
         break;  
         }  
1376    
1377        /* 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.
1378        recognize this construct; fall through to the normal \x handling. */          We used just to recognize this construct and fall through to the normal
1379        }          \x handling, but nowadays Perl gives an error, which seems much more
1380            sensible, so we do too. */
1381    
1382      /* Read just a single-byte hex-defined char */          else *errorcodeptr = ERR79;
1383            }   /* End of \x{} processing */
1384    
1385      c = 0;        /* Read a single-byte hex-defined char (up to two hex digits after \x) */
1386      while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)  
1387        {        else
1388        int cc;                                  /* Some compilers don't like */          {
1389        cc = *(++ptr);                           /* ++ in initializers */          c = 0;
1390            while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1391              {
1392              pcre_uint32 cc;                          /* Some compilers don't like */
1393              cc = *(++ptr);                           /* ++ in initializers */
1394  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1395        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1396        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1397  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1398        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */            if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1399        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1400  #endif  #endif
1401        }            }
1402            }     /* End of \xdd handling */
1403          }       /* End of Perl-style \x handling */
1404      break;      break;
1405    
1406      /* 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 1410  else
1410    
1411      case CHAR_c:      case CHAR_c:
1412      c = *(++ptr);      c = *(++ptr);
1413      if (c == 0)      if (c == CHAR_NULL)
1414        {        {
1415        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
1416        break;        break;
# Line 1141  else Line 1450  else
1450  newline". PCRE does not support \N{name}. However, it does support  newline". PCRE does not support \N{name}. However, it does support
1451  quantification such as \N{2,3}. */  quantification such as \N{2,3}. */
1452    
1453  if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&  if (escape == ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1454       !is_counted_repeat(ptr+2))       !is_counted_repeat(ptr+2))
1455    *errorcodeptr = ERR37;    *errorcodeptr = ERR37;
1456    
1457  /* If PCRE_UCP is set, we change the values for \d etc. */  /* If PCRE_UCP is set, we change the values for \d etc. */
1458    
1459  if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)  if ((options & PCRE_UCP) != 0 && escape >= ESC_D && escape <= ESC_w)
1460    c -= (ESC_DU - ESC_D);    escape += (ESC_DU - ESC_D);
1461    
1462  /* Set the pointer to the final character before returning. */  /* Set the pointer to the final character before returning. */
1463    
1464  *ptrptr = ptr;  *ptrptr = ptr;
1465  return c;  *chptr = c;
1466    return escape;
1467  }  }
1468    
1469    
# Line 1171  escape sequence. Line 1481  escape sequence.
1481  Argument:  Argument:
1482    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
1483    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
1484    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
1485      pdataptr       points to an unsigned int that is set to the detailed property value
1486    errorcodeptr   points to the error code variable    errorcodeptr   points to the error code variable
1487    
1488  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
1489  */  */
1490    
1491  static int  static BOOL
1492  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, unsigned int *ptypeptr,
1493      unsigned int *pdataptr, int *errorcodeptr)
1494  {  {
1495  int c, i, bot, top;  pcre_uchar c;
1496    int i, bot, top;
1497  const pcre_uchar *ptr = *ptrptr;  const pcre_uchar *ptr = *ptrptr;
1498  pcre_uchar name[32];  pcre_uchar name[32];
1499    
1500  c = *(++ptr);  c = *(++ptr);
1501  if (c == 0) goto ERROR_RETURN;  if (c == CHAR_NULL) goto ERROR_RETURN;
1502    
1503  *negptr = FALSE;  *negptr = FALSE;
1504    
# Line 1202  if (c == CHAR_LEFT_CURLY_BRACKET) Line 1515  if (c == CHAR_LEFT_CURLY_BRACKET)
1515    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1516      {      {
1517      c = *(++ptr);      c = *(++ptr);
1518      if (c == 0) goto ERROR_RETURN;      if (c == CHAR_NULL) goto ERROR_RETURN;
1519      if (c == CHAR_RIGHT_CURLY_BRACKET) break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1520      name[i] = c;      name[i] = c;
1521      }      }
# Line 1227  top = PRIV(utt_size); Line 1540  top = PRIV(utt_size);
1540    
1541  while (bot < top)  while (bot < top)
1542    {    {
1543      int r;
1544    i = (bot + top) >> 1;    i = (bot + top) >> 1;
1545    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);
1546    if (c == 0)    if (r == 0)
1547      {      {
1548      *dptr = PRIV(utt)[i].value;      *ptypeptr = PRIV(utt)[i].type;
1549      return PRIV(utt)[i].type;      *pdataptr = PRIV(utt)[i].value;
1550        return TRUE;
1551      }      }
1552    if (c > 0) bot = i + 1; else top = i;    if (r > 0) bot = i + 1; else top = i;
1553    }    }
1554    
1555  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
1556  *ptrptr = ptr;  *ptrptr = ptr;
1557  return -1;  return FALSE;
1558    
1559  ERROR_RETURN:  ERROR_RETURN:
1560  *errorcodeptr = ERR46;  *errorcodeptr = ERR46;
1561  *ptrptr = ptr;  *ptrptr = ptr;
1562  return -1;  return FALSE;
1563  }  }
1564  #endif  #endif
1565    
1566    
1567    
   
1568  /*************************************************  /*************************************************
1569  *         Read repeat counts                     *  *         Read repeat counts                     *
1570  *************************************************/  *************************************************/
# Line 1276  read_repeat_counts(const pcre_uchar *p, Line 1590  read_repeat_counts(const pcre_uchar *p,
1590  int min = 0;  int min = 0;
1591  int max = -1;  int max = -1;
1592    
1593  /* 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)  
1594    {    {
1595    *errorcodeptr = ERR5;    min = min * 10 + (int)(*p++ - CHAR_0);
1596    return p;    if (min > 65535)
1597        {
1598        *errorcodeptr = ERR5;
1599        return p;
1600        }
1601    }    }
1602    
 /* Read the maximum value if there is one, and again do a paranoid on its size.  
 Also, max must not be less than min. */  
   
1603  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1604    {    {
1605    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1606      {      {
1607      max = 0;      max = 0;
1608      while(IS_DIGIT(*p)) max = max * 10 + *p++ - CHAR_0;      while(IS_DIGIT(*p))
     if (max < 0 || max > 65535)  
1609        {        {
1610        *errorcodeptr = ERR5;        max = max * 10 + (int)(*p++ - CHAR_0);
1611        return p;        if (max > 65535)
1612            {
1613            *errorcodeptr = ERR5;
1614            return p;
1615            }
1616        }        }
1617      if (max < min)      if (max < min)
1618        {        {
# Line 1308  if (*p == CHAR_RIGHT_CURLY_BRACKET) max Line 1622  if (*p == CHAR_RIGHT_CURLY_BRACKET) max
1622      }      }
1623    }    }
1624    
 /* Fill in the required variables, and pass back the pointer to the terminating  
 '}'. */  
   
1625  *minp = min;  *minp = min;
1626  *maxp = max;  *maxp = max;
1627  return p;  return p;
# Line 1319  return p; Line 1630  return p;
1630    
1631    
1632  /*************************************************  /*************************************************
 *  Subroutine for finding forward reference      *  
 *************************************************/  
   
 /* This recursive function is called only from find_parens() below. The  
 top-level call starts at the beginning of the pattern. All other calls must  
 start at a parenthesis. It scans along a pattern's text looking for capturing  
 subpatterns, and counting them. If it finds a named pattern that matches the  
 name it is given, it returns its number. Alternatively, if the name is NULL, it  
 returns when it reaches a given numbered subpattern. Recursion is used to keep  
 track of subpatterns that reset the capturing group numbers - the (?| feature.  
   
 This function was originally called only from the second pass, in which we know  
 that if (?< or (?' or (?P< is encountered, the name will be correctly  
 terminated because that is checked in the first pass. There is now one call to  
 this function in the first pass, to check for a recursive back reference by  
 name (so that we can make the whole group atomic). In this case, we need check  
 only up to the current position in the pattern, and that is still OK because  
 and previous occurrences will have been checked. To make this work, the test  
 for "end of pattern" is a check against cd->end_pattern in the main loop,  
 instead of looking for a binary zero. This means that the special first-pass  
 call can adjust cd->end_pattern temporarily. (Checks for binary zero while  
 processing items within the loop are OK, because afterwards the main loop will  
 terminate.)  
   
 Arguments:  
   ptrptr       address of the current character pointer (updated)  
   cd           compile background data  
   name         name to seek, or NULL if seeking a numbered subpattern  
   lorn         name length, or subpattern number if name is NULL  
   xmode        TRUE if we are in /x mode  
   utf          TRUE if we are in UTF-8 / UTF-16 mode  
   count        pointer to the current capturing subpattern number (updated)  
   
 Returns:       the number of the named subpattern, or -1 if not found  
 */  
   
 static int  
 find_parens_sub(pcre_uchar **ptrptr, compile_data *cd, const pcre_uchar *name, int lorn,  
   BOOL xmode, BOOL utf, int *count)  
 {  
 pcre_uchar *ptr = *ptrptr;  
 int start_count = *count;  
 int hwm_count = start_count;  
 BOOL dup_parens = FALSE;  
   
 /* If the first character is a parenthesis, check on the type of group we are  
 dealing with. The very first call may not start with a parenthesis. */  
   
 if (ptr[0] == CHAR_LEFT_PARENTHESIS)  
   {  
   /* Handle specials such as (*SKIP) or (*UTF8) etc. */  
   
   if (ptr[1] == CHAR_ASTERISK) ptr += 2;  
   
   /* Handle a normal, unnamed capturing parenthesis. */  
   
   else if (ptr[1] != CHAR_QUESTION_MARK)  
     {  
     *count += 1;  
     if (name == NULL && *count == lorn) return *count;  
     ptr++;  
     }  
   
   /* All cases now have (? at the start. Remember when we are in a group  
   where the parenthesis numbers are duplicated. */  
   
   else if (ptr[2] == CHAR_VERTICAL_LINE)  
     {  
     ptr += 3;  
     dup_parens = TRUE;  
     }  
   
   /* Handle comments; all characters are allowed until a ket is reached. */  
   
   else if (ptr[2] == CHAR_NUMBER_SIGN)  
     {  
     for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;  
     goto FAIL_EXIT;  
     }  
   
   /* Handle a condition. If it is an assertion, just carry on so that it  
   is processed as normal. If not, skip to the closing parenthesis of the  
   condition (there can't be any nested parens). */  
   
   else if (ptr[2] == CHAR_LEFT_PARENTHESIS)  
     {  
     ptr += 2;  
     if (ptr[1] != CHAR_QUESTION_MARK)  
       {  
       while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;  
       if (*ptr != 0) ptr++;  
       }  
     }  
   
   /* Start with (? but not a condition. */  
   
   else  
     {  
     ptr += 2;  
     if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */  
   
     /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */  
   
     if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&  
         ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)  
       {  
       int term;  
       const pcre_uchar *thisname;  
       *count += 1;  
       if (name == NULL && *count == lorn) return *count;  
       term = *ptr++;  
       if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;  
       thisname = ptr;  
       while (*ptr != term) ptr++;  
       if (name != NULL && lorn == ptr - thisname &&  
           STRNCMP_UC_UC(name, thisname, lorn) == 0)  
         return *count;  
       term++;  
       }  
     }  
   }  
   
 /* Past any initial parenthesis handling, scan for parentheses or vertical  
 bars. Stop if we get to cd->end_pattern. Note that this is important for the  
 first-pass call when this value is temporarily adjusted to stop at the current  
 position. So DO NOT change this to a test for binary zero. */  
   
 for (; ptr < cd->end_pattern; ptr++)  
   {  
   /* Skip over backslashed characters and also entire \Q...\E */  
   
   if (*ptr == CHAR_BACKSLASH)  
     {  
     if (*(++ptr) == 0) goto FAIL_EXIT;  
     if (*ptr == CHAR_Q) for (;;)  
       {  
       while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};  
       if (*ptr == 0) goto FAIL_EXIT;  
       if (*(++ptr) == CHAR_E) break;  
       }  
     continue;  
     }  
   
   /* Skip over character classes; this logic must be similar to the way they  
   are handled for real. If the first character is '^', skip it. Also, if the  
   first few characters (either before or after ^) are \Q\E or \E we skip them  
   too. This makes for compatibility with Perl. Note the use of STR macros to  
   encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */  
   
   if (*ptr == CHAR_LEFT_SQUARE_BRACKET)  
     {  
     BOOL negate_class = FALSE;  
     for (;;)  
       {  
       if (ptr[1] == CHAR_BACKSLASH)  
         {  
         if (ptr[2] == CHAR_E)  
           ptr+= 2;  
         else if (STRNCMP_UC_C8(ptr + 2,  
                  STR_Q STR_BACKSLASH STR_E, 3) == 0)  
           ptr += 4;  
         else  
           break;  
         }  
       else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)  
         {  
         negate_class = TRUE;  
         ptr++;  
         }  
       else break;  
       }  
   
     /* If the next character is ']', it is a data character that must be  
     skipped, except in JavaScript compatibility mode. */  
   
     if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&  
         (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)  
       ptr++;  
   
     while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)  
       {  
       if (*ptr == 0) return -1;  
       if (*ptr == CHAR_BACKSLASH)  
         {  
         if (*(++ptr) == 0) goto FAIL_EXIT;  
         if (*ptr == CHAR_Q) for (;;)  
           {  
           while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};  
           if (*ptr == 0) goto FAIL_EXIT;  
           if (*(++ptr) == CHAR_E) break;  
           }  
         continue;  
         }  
       }  
     continue;  
     }  
   
   /* Skip comments in /x mode */  
   
   if (xmode && *ptr == CHAR_NUMBER_SIGN)  
     {  
     ptr++;  
     while (*ptr != 0)  
       {  
       if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }  
       ptr++;  
 #ifdef SUPPORT_UTF  
       if (utf) FORWARDCHAR(ptr);  
 #endif  
       }  
     if (*ptr == 0) goto FAIL_EXIT;  
     continue;  
     }  
   
   /* Check for the special metacharacters */  
   
   if (*ptr == CHAR_LEFT_PARENTHESIS)  
     {  
     int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, count);  
     if (rc > 0) return rc;  
     if (*ptr == 0) goto FAIL_EXIT;  
     }  
   
   else if (*ptr == CHAR_RIGHT_PARENTHESIS)  
     {  
     if (dup_parens && *count < hwm_count) *count = hwm_count;  
     goto FAIL_EXIT;  
     }  
   
   else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)  
     {  
     if (*count > hwm_count) hwm_count = *count;  
     *count = start_count;  
     }  
   }  
   
 FAIL_EXIT:  
 *ptrptr = ptr;  
 return -1;  
 }  
   
   
   
   
 /*************************************************  
 *       Find forward referenced subpattern       *  
 *************************************************/  
   
 /* This function scans along a pattern's text looking for capturing  
 subpatterns, and counting them. If it finds a named pattern that matches the  
 name it is given, it returns its number. Alternatively, if the name is NULL, it  
 returns when it reaches a given numbered subpattern. This is used for forward  
 references to subpatterns. We used to be able to start this scan from the  
 current compiling point, using the current count value from cd->bracount, and  
 do it all in a single loop, but the addition of the possibility of duplicate  
 subpattern numbers means that we have to scan from the very start, in order to  
 take account of such duplicates, and to use a recursive function to keep track  
 of the different types of group.  
   
 Arguments:  
   cd           compile background data  
   name         name to seek, or NULL if seeking a numbered subpattern  
   lorn         name length, or subpattern number if name is NULL  
   xmode        TRUE if we are in /x mode  
   utf          TRUE if we are in UTF-8 / UTF-16 mode  
   
 Returns:       the number of the found subpattern, or -1 if not found  
 */  
   
 static int  
 find_parens(compile_data *cd, const pcre_uchar *name, int lorn, BOOL xmode,  
   BOOL utf)  
 {  
 pcre_uchar *ptr = (pcre_uchar *)cd->start_pattern;  
 int count = 0;  
 int rc;  
   
 /* If the pattern does not start with an opening parenthesis, the first call  
 to find_parens_sub() will scan right to the end (if necessary). However, if it  
 does start with a parenthesis, find_parens_sub() will return when it hits the  
 matching closing parens. That is why we have to have a loop. */  
   
 for (;;)  
   {  
   rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, &count);  
   if (rc > 0 || *ptr++ == 0) break;  
   }  
   
 return rc;  
 }  
   
   
   
   
 /*************************************************  
1633  *      Find first significant op code            *  *      Find first significant op code            *
1634  *************************************************/  *************************************************/
1635    
# Line 1652  for (;;) Line 1668  for (;;)
1668    
1669      case OP_CALLOUT:      case OP_CALLOUT:
1670      case OP_CREF:      case OP_CREF:
1671      case OP_NCREF:      case OP_DNCREF:
1672      case OP_RREF:      case OP_RREF:
1673      case OP_NRREF:      case OP_DNRREF:
1674      case OP_DEF:      case OP_DEF:
1675      code += PRIV(OP_lengths)[*code];      code += PRIV(OP_lengths)[*code];
1676      break;      break;
# Line 1668  for (;;) Line 1684  for (;;)
1684    
1685    
1686    
   
1687  /*************************************************  /*************************************************
1688  *        Find the fixed length of a branch       *  *        Find the fixed length of a branch       *
1689  *************************************************/  *************************************************/
# Line 1686  and doing the check at the end; a flag s Line 1701  and doing the check at the end; a flag s
1701    
1702  Arguments:  Arguments:
1703    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1704    utf      TRUE in UTF-8 / UTF-16 mode    utf      TRUE in UTF-8 / UTF-16 / UTF-32 mode
1705    atend    TRUE if called when the pattern is complete    atend    TRUE if called when the pattern is complete
1706    cd       the "compile data" structure    cd       the "compile data" structure
1707      recurses    chain of recurse_check to catch mutual recursion
1708    
1709  Returns:   the fixed length,  Returns:   the fixed length,
1710               or -1 if there is no fixed length,               or -1 if there is no fixed length,
# Line 1698  Returns:   the fixed length, Line 1714  Returns:   the fixed length,
1714  */  */
1715    
1716  static int  static int
1717  find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd)  find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd,
1718      recurse_check *recurses)
1719  {  {
1720  int length = -1;  int length = -1;
1721    recurse_check this_recurse;
1722  register int branchlength = 0;  register int branchlength = 0;
1723  register pcre_uchar *cc = code + 1 + LINK_SIZE;  register pcre_uchar *cc = code + 1 + LINK_SIZE;
1724    
# Line 1712  for (;;) Line 1729  for (;;)
1729    {    {
1730    int d;    int d;
1731    pcre_uchar *ce, *cs;    pcre_uchar *ce, *cs;
1732    register int op = *cc;    register pcre_uchar op = *cc;
1733    
1734    switch (op)    switch (op)
1735      {      {
# Line 1726  for (;;) Line 1743  for (;;)
1743      case OP_ONCE:      case OP_ONCE:
1744      case OP_ONCE_NC:      case OP_ONCE_NC:
1745      case OP_COND:      case OP_COND:
1746      d = find_fixedlength(cc + ((op == OP_CBRA)? IMM2_SIZE : 0), utf, atend, cd);      d = find_fixedlength(cc + ((op == OP_CBRA)? IMM2_SIZE : 0), utf, atend, cd,
1747          recurses);
1748      if (d < 0) return d;      if (d < 0) return d;
1749      branchlength += d;      branchlength += d;
1750      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1760  for (;;) Line 1778  for (;;)
1778      cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1);  /* Start subpattern */      cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1779      do ce += GET(ce, 1); while (*ce == OP_ALT);           /* End subpattern */      do ce += GET(ce, 1); while (*ce == OP_ALT);           /* End subpattern */
1780      if (cc > cs && cc < ce) return -1;                    /* Recursion */      if (cc > cs && cc < ce) return -1;                    /* Recursion */
1781      d = find_fixedlength(cs + IMM2_SIZE, utf, atend, cd);      else   /* Check for mutual recursion */
1782          {
1783          recurse_check *r = recurses;
1784          for (r = recurses; r != NULL; r = r->prev) if (r->group == cs) break;
1785          if (r != NULL) return -1;   /* Mutual recursion */
1786          }
1787        this_recurse.prev = recurses;
1788        this_recurse.group = cs;
1789        d = find_fixedlength(cs + IMM2_SIZE, utf, atend, cd, &this_recurse);
1790      if (d < 0) return d;      if (d < 0) return d;
1791      branchlength += d;      branchlength += d;
1792      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
# Line 1773  for (;;) Line 1799  for (;;)
1799      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1800      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1801      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1802      cc += PRIV(OP_lengths)[*cc];      cc += 1 + LINK_SIZE;
1803      break;      break;
1804    
1805      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
# Line 1792  for (;;) Line 1818  for (;;)
1818      case OP_COMMIT:      case OP_COMMIT:
1819      case OP_CREF:      case OP_CREF:
1820      case OP_DEF:      case OP_DEF:
1821        case OP_DNCREF:
1822        case OP_DNRREF:
1823      case OP_DOLL:      case OP_DOLL:
1824      case OP_DOLLM:      case OP_DOLLM:
1825      case OP_EOD:      case OP_EOD:
1826      case OP_EODN:      case OP_EODN:
1827      case OP_FAIL:      case OP_FAIL:
     case OP_NCREF:  
     case OP_NRREF:  
1828      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1829      case OP_PRUNE:      case OP_PRUNE:
1830      case OP_REVERSE:      case OP_REVERSE:
# Line 1832  for (;;) Line 1858  for (;;)
1858      case OP_EXACTI:      case OP_EXACTI:
1859      case OP_NOTEXACT:      case OP_NOTEXACT:
1860      case OP_NOTEXACTI:      case OP_NOTEXACTI:
1861      branchlength += GET2(cc,1);      branchlength += (int)GET2(cc,1);
1862      cc += 2 + IMM2_SIZE;      cc += 2 + IMM2_SIZE;
1863  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
1864      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 1867  for (;;)
1867    
1868      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1869      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1870      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)
1871          cc += 2;
1872      cc += 1 + IMM2_SIZE + 1;      cc += 1 + IMM2_SIZE + 1;
1873      break;      break;
1874    
# Line 1876  for (;;) Line 1903  for (;;)
1903    
1904      /* Check a class for variable quantification */      /* Check a class for variable quantification */
1905    
 #if defined SUPPORT_UTF || defined COMPILE_PCRE16  
     case OP_XCLASS:  
     cc += GET(cc, 1) - PRIV(OP_lengths)[OP_CLASS];  
     /* Fall through */  
 #endif  
   
1906      case OP_CLASS:      case OP_CLASS:
1907      case OP_NCLASS:      case OP_NCLASS:
1908    #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1909        case OP_XCLASS:
1910        /* The original code caused an unsigned overflow in 64 bit systems,
1911        so now we use a conditional statement. */
1912        if (op == OP_XCLASS)
1913          cc += GET(cc, 1);
1914        else
1915          cc += PRIV(OP_lengths)[OP_CLASS];
1916    #else
1917      cc += PRIV(OP_lengths)[OP_CLASS];      cc += PRIV(OP_lengths)[OP_CLASS];
1918    #endif
1919    
1920      switch (*cc)      switch (*cc)
1921        {        {
       case OP_CRPLUS:  
       case OP_CRMINPLUS:  
1922        case OP_CRSTAR:        case OP_CRSTAR:
1923        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1924          case OP_CRPLUS:
1925          case OP_CRMINPLUS:
1926        case OP_CRQUERY:        case OP_CRQUERY:
1927        case OP_CRMINQUERY:        case OP_CRMINQUERY:
1928          case OP_CRPOSSTAR:
1929          case OP_CRPOSPLUS:
1930          case OP_CRPOSQUERY:
1931        return -1;        return -1;
1932    
1933        case OP_CRRANGE:        case OP_CRRANGE:
1934        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1935          case OP_CRPOSRANGE:
1936        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1937        branchlength += GET2(cc,1);        branchlength += (int)GET2(cc,1);
1938        cc += 1 + 2 * IMM2_SIZE;        cc += 1 + 2 * IMM2_SIZE;
1939        break;        break;
1940    
# Line 1966  for (;;) Line 2001  for (;;)
2001      case OP_QUERYI:      case OP_QUERYI:
2002      case OP_REF:      case OP_REF:
2003      case OP_REFI:      case OP_REFI:
2004        case OP_DNREF:
2005        case OP_DNREFI:
2006      case OP_SBRA:      case OP_SBRA:
2007      case OP_SBRAPOS:      case OP_SBRAPOS:
2008      case OP_SCBRA:      case OP_SCBRA:
# Line 2002  for (;;) Line 2039  for (;;)
2039    
2040    
2041    
   
2042  /*************************************************  /*************************************************
2043  *    Scan compiled regex for specific bracket    *  *    Scan compiled regex for specific bracket    *
2044  *************************************************/  *************************************************/
# Line 2015  length. Line 2051  length.
2051    
2052  Arguments:  Arguments:
2053    code        points to start of expression    code        points to start of expression
2054    utf         TRUE in UTF-8 / UTF-16 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2055    number      the required bracket number or negative to find a lookbehind    number      the required bracket number or negative to find a lookbehind
2056    
2057  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 2062  PRIV(find_bracket)(const pcre_uchar *cod
2062  {  {
2063  for (;;)  for (;;)
2064    {    {
2065    register int c = *code;    register pcre_uchar c = *code;
2066    
2067    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2068    
# Line 2049  for (;;) Line 2085  for (;;)
2085    else if (c == OP_CBRA || c == OP_SCBRA ||    else if (c == OP_CBRA || c == OP_SCBRA ||
2086             c == OP_CBRAPOS || c == OP_SCBRAPOS)             c == OP_CBRAPOS || c == OP_SCBRAPOS)
2087      {      {
2088      int n = GET2(code, 1+LINK_SIZE);      int n = (int)GET2(code, 1+LINK_SIZE);
2089      if (n == number) return (pcre_uchar *)code;      if (n == number) return (pcre_uchar *)code;
2090      code += PRIV(OP_lengths)[c];      code += PRIV(OP_lengths)[c];
2091      }      }
# Line 2079  for (;;) Line 2115  for (;;)
2115        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2116        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2117        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
2118        if (code[1 + IMM2_SIZE] == OP_PROP        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2119          || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;          code += 2;
2120        break;        break;
2121    
2122        case OP_MARK:        case OP_MARK:
2123        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
2124        case OP_SKIP_ARG:        case OP_SKIP_ARG:
       code += code[1];  
       break;  
   
2125        case OP_THEN_ARG:        case OP_THEN_ARG:
2126        code += code[1];        code += code[1];
2127        break;        break;
# Line 2102  for (;;) Line 2135  for (;;)
2135    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
2136    arrange to skip the extra bytes. */    arrange to skip the extra bytes. */
2137    
2138  #ifdef SUPPORT_UTF  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2139      if (utf) switch(c)      if (utf) switch(c)
2140        {        {
2141        case OP_CHAR:        case OP_CHAR:
2142        case OP_CHARI:        case OP_CHARI:
2143          case OP_NOT:
2144          case OP_NOTI:
2145        case OP_EXACT:        case OP_EXACT:
2146        case OP_EXACTI:        case OP_EXACTI:
2147          case OP_NOTEXACT:
2148          case OP_NOTEXACTI:
2149        case OP_UPTO:        case OP_UPTO:
2150        case OP_UPTOI:        case OP_UPTOI:
2151          case OP_NOTUPTO:
2152          case OP_NOTUPTOI:
2153        case OP_MINUPTO:        case OP_MINUPTO:
2154        case OP_MINUPTOI:        case OP_MINUPTOI:
2155          case OP_NOTMINUPTO:
2156          case OP_NOTMINUPTOI:
2157        case OP_POSUPTO:        case OP_POSUPTO:
2158        case OP_POSUPTOI:        case OP_POSUPTOI:
2159          case OP_NOTPOSUPTO:
2160          case OP_NOTPOSUPTOI:
2161        case OP_STAR:        case OP_STAR:
2162        case OP_STARI:        case OP_STARI:
2163          case OP_NOTSTAR:
2164          case OP_NOTSTARI:
2165        case OP_MINSTAR:        case OP_MINSTAR:
2166        case OP_MINSTARI:        case OP_MINSTARI:
2167          case OP_NOTMINSTAR:
2168          case OP_NOTMINSTARI:
2169        case OP_POSSTAR:        case OP_POSSTAR:
2170        case OP_POSSTARI:        case OP_POSSTARI:
2171          case OP_NOTPOSSTAR:
2172          case OP_NOTPOSSTARI:
2173        case OP_PLUS:        case OP_PLUS:
2174        case OP_PLUSI:        case OP_PLUSI:
2175          case OP_NOTPLUS:
2176          case OP_NOTPLUSI:
2177        case OP_MINPLUS:        case OP_MINPLUS:
2178        case OP_MINPLUSI:        case OP_MINPLUSI:
2179          case OP_NOTMINPLUS:
2180          case OP_NOTMINPLUSI:
2181        case OP_POSPLUS:        case OP_POSPLUS:
2182        case OP_POSPLUSI:        case OP_POSPLUSI:
2183          case OP_NOTPOSPLUS:
2184          case OP_NOTPOSPLUSI:
2185        case OP_QUERY:        case OP_QUERY:
2186        case OP_QUERYI:        case OP_QUERYI:
2187          case OP_NOTQUERY:
2188          case OP_NOTQUERYI:
2189        case OP_MINQUERY:        case OP_MINQUERY:
2190        case OP_MINQUERYI:        case OP_MINQUERYI:
2191          case OP_NOTMINQUERY:
2192          case OP_NOTMINQUERYI:
2193        case OP_POSQUERY:        case OP_POSQUERY:
2194        case OP_POSQUERYI:        case OP_POSQUERYI:
2195          case OP_NOTPOSQUERY:
2196          case OP_NOTPOSQUERYI:
2197        if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);        if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2198        break;        break;
2199        }        }
# Line 2154  instance of OP_RECURSE. Line 2215  instance of OP_RECURSE.
2215    
2216  Arguments:  Arguments:
2217    code        points to start of expression    code        points to start of expression
2218    utf         TRUE in UTF-8 / UTF-16 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2219    
2220  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
2221  */  */
# Line 2164  find_recurse(const pcre_uchar *code, BOO Line 2225  find_recurse(const pcre_uchar *code, BOO
2225  {  {
2226  for (;;)  for (;;)
2227    {    {
2228    register int c = *code;    register pcre_uchar c = *code;
2229    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2230    if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
2231    
# Line 2199  for (;;) Line 2260  for (;;)
2260        case OP_TYPEUPTO:        case OP_TYPEUPTO:
2261        case OP_TYPEMINUPTO:        case OP_TYPEMINUPTO:
2262        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2263        if (code[1 + IMM2_SIZE] == OP_PROP        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2264          || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;          code += 2;
2265        break;        break;
2266    
2267        case OP_MARK:        case OP_MARK:
2268        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
2269        case OP_SKIP_ARG:        case OP_SKIP_ARG:
       code += code[1];  
       break;  
   
2270        case OP_THEN_ARG:        case OP_THEN_ARG:
2271        code += code[1];        code += code[1];
2272        break;        break;
# Line 2222  for (;;) Line 2280  for (;;)
2280      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
2281      to arrange to skip the extra bytes. */      to arrange to skip the extra bytes. */
2282    
2283  #ifdef SUPPORT_UTF  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2284      if (utf) switch(c)      if (utf) switch(c)
2285        {        {
2286        case OP_CHAR:        case OP_CHAR:
# Line 2308  bracket whose current branch will alread Line 2366  bracket whose current branch will alread
2366  Arguments:  Arguments:
2367    code        points to start of search    code        points to start of search
2368    endcode     points to where to stop    endcode     points to where to stop
2369    utf         TRUE if in UTF-8 / UTF-16 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2370    cd          contains pointers to tables etc.    cd          contains pointers to tables etc.
2371      recurses    chain of recurse_check to catch mutual recursion
2372    
2373  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2374  */  */
2375    
2376  static BOOL  static BOOL
2377  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2378    BOOL utf, compile_data *cd)    BOOL utf, compile_data *cd, recurse_check *recurses)
2379  {  {
2380  register int c;  register pcre_uchar c;
2381    recurse_check this_recurse;
2382    
2383  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2384       code < endcode;       code < endcode;
2385       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 2407  for (code = first_significant_code(code
2407    
2408    if (c == OP_RECURSE)    if (c == OP_RECURSE)
2409      {      {
2410      const pcre_uchar *scode;      const pcre_uchar *scode = cd->start_code + GET(code, 1);
2411        const pcre_uchar *endgroup = scode;
2412      BOOL empty_branch;      BOOL empty_branch;
2413    
2414      /* Test for forward reference */      /* Test for forward reference or uncompleted reference. This is disabled
2415        when called to scan a completed pattern by setting cd->start_workspace to
2416        NULL. */
2417    
2418      for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)      if (cd->start_workspace != NULL)
2419        if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;        {
2420          const pcre_uchar *tcode;
2421          for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)
2422            if ((int)GET(tcode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;
2423          if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2424          }
2425    
2426      /* Not a forward reference, test for completed backward reference */      /* If the reference is to a completed group, we need to detect whether this
2427        is a recursive call, as otherwise there will be an infinite loop. If it is
2428        a recursion, just skip over it. Simple recursions are easily detected. For
2429        mutual recursions we keep a chain on the stack. */
2430    
2431      empty_branch = FALSE;      do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);
2432      scode = cd->start_code + GET(code, 1);      if (code >= scode && code <= endgroup) continue;  /* Simple recursion */
2433      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */      else
2434          {
2435          recurse_check *r = recurses;
2436          for (r = recurses; r != NULL; r = r->prev)
2437            if (r->group == scode) break;
2438          if (r != NULL) continue;   /* Mutual recursion */
2439          }
2440    
2441        /* Completed reference; scan the referenced group, remembering it on the
2442        stack chain to detect mutual recursions. */
2443    
2444      /* Completed backwards reference */      empty_branch = FALSE;
2445        this_recurse.prev = recurses;
2446        this_recurse.group = scode;
2447    
2448      do      do
2449        {        {
2450        if (could_be_empty_branch(scode, endcode, utf, cd))        if (could_be_empty_branch(scode, endcode, utf, cd, &this_recurse))
2451          {          {
2452          empty_branch = TRUE;          empty_branch = TRUE;
2453          break;          break;
# Line 2420  for (code = first_significant_code(code Line 2503  for (code = first_significant_code(code
2503        empty_branch = FALSE;        empty_branch = FALSE;
2504        do        do
2505          {          {
2506          if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd))          if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd,
2507            empty_branch = TRUE;            recurses)) empty_branch = TRUE;
2508          code += GET(code, 1);          code += GET(code, 1);
2509          }          }
2510        while (*code == OP_ALT);        while (*code == OP_ALT);
# Line 2462  for (code = first_significant_code(code Line 2545  for (code = first_significant_code(code
2545        case OP_CRMINSTAR:        case OP_CRMINSTAR:
2546        case OP_CRQUERY:        case OP_CRQUERY:
2547        case OP_CRMINQUERY:        case OP_CRMINQUERY:
2548          case OP_CRPOSSTAR:
2549          case OP_CRPOSQUERY:
2550        break;        break;
2551    
2552        default:                   /* Non-repeat => class must match */        default:                   /* Non-repeat => class must match */
2553        case OP_CRPLUS:            /* These repeats aren't empty */        case OP_CRPLUS:            /* These repeats aren't empty */
2554        case OP_CRMINPLUS:        case OP_CRMINPLUS:
2555          case OP_CRPOSPLUS:
2556        return FALSE;        return FALSE;
2557    
2558        case OP_CRRANGE:        case OP_CRRANGE:
2559        case OP_CRMINRANGE:        case OP_CRMINRANGE:
2560          case OP_CRPOSRANGE:
2561        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */
2562        break;        break;
2563        }        }
# Line 2478  for (code = first_significant_code(code Line 2565  for (code = first_significant_code(code
2565    
2566      /* Opcodes that must match a character */      /* Opcodes that must match a character */
2567    
2568        case OP_ANY:
2569        case OP_ALLANY:
2570        case OP_ANYBYTE:
2571    
2572      case OP_PROP:      case OP_PROP:
2573      case OP_NOTPROP:      case OP_NOTPROP:
2574        case OP_ANYNL:
2575    
2576        case OP_NOT_HSPACE:
2577        case OP_HSPACE:
2578        case OP_NOT_VSPACE:
2579        case OP_VSPACE:
2580      case OP_EXTUNI:      case OP_EXTUNI:
2581    
2582      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2583      case OP_DIGIT:      case OP_DIGIT:
2584      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2585      case OP_WHITESPACE:      case OP_WHITESPACE:
2586      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2587      case OP_WORDCHAR:      case OP_WORDCHAR:
2588      case OP_ANY:  
     case OP_ALLANY:  
     case OP_ANYBYTE:  
2589      case OP_CHAR:      case OP_CHAR:
2590      case OP_CHARI:      case OP_CHARI:
2591      case OP_NOT:      case OP_NOT:
2592      case OP_NOTI:      case OP_NOTI:
2593    
2594      case OP_PLUS:      case OP_PLUS:
2595        case OP_PLUSI:
2596      case OP_MINPLUS:      case OP_MINPLUS:
2597      case OP_POSPLUS:      case OP_MINPLUSI:
2598      case OP_EXACT:  
2599      case OP_NOTPLUS:      case OP_NOTPLUS:
2600        case OP_NOTPLUSI:
2601      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
2602        case OP_NOTMINPLUSI:
2603    
2604        case OP_POSPLUS:
2605        case OP_POSPLUSI:
2606      case OP_NOTPOSPLUS:      case OP_NOTPOSPLUS:
2607        case OP_NOTPOSPLUSI:
2608    
2609        case OP_EXACT:
2610        case OP_EXACTI:
2611      case OP_NOTEXACT:      case OP_NOTEXACT:
2612        case OP_NOTEXACTI:
2613    
2614      case OP_TYPEPLUS:      case OP_TYPEPLUS:
2615      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
2616      case OP_TYPEPOSPLUS:      case OP_TYPEPOSPLUS:
2617      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2618    
2619      return FALSE;      return FALSE;
2620    
2621      /* 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 2635  for (code = first_significant_code(code
2635      case OP_TYPEUPTO:      case OP_TYPEUPTO:
2636      case OP_TYPEMINUPTO:      case OP_TYPEMINUPTO:
2637      case OP_TYPEPOSUPTO:      case OP_TYPEPOSUPTO:
2638      if (code[1 + IMM2_SIZE] == OP_PROP      if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2639        || code[1 + IMM2_SIZE] == OP_NOTPROP) code += 2;        code += 2;
2640      break;      break;
2641    
2642      /* End of branch */      /* End of branch */
# Line 2539  for (code = first_significant_code(code Line 2649  for (code = first_significant_code(code
2649      return TRUE;      return TRUE;
2650    
2651      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2652      MINUPTO, and POSUPTO may be followed by a multibyte character */      MINUPTO, and POSUPTO and their caseless and negative versions may be
2653        followed by a multibyte character. */
2654    
2655  #ifdef SUPPORT_UTF  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2656      case OP_STAR:      case OP_STAR:
2657      case OP_STARI:      case OP_STARI:
2658        case OP_NOTSTAR:
2659        case OP_NOTSTARI:
2660    
2661      case OP_MINSTAR:      case OP_MINSTAR:
2662      case OP_MINSTARI:      case OP_MINSTARI:
2663        case OP_NOTMINSTAR:
2664        case OP_NOTMINSTARI:
2665    
2666      case OP_POSSTAR:      case OP_POSSTAR:
2667      case OP_POSSTARI:      case OP_POSSTARI:
2668        case OP_NOTPOSSTAR:
2669        case OP_NOTPOSSTARI:
2670    
2671      case OP_QUERY:      case OP_QUERY:
2672      case OP_QUERYI:      case OP_QUERYI:
2673        case OP_NOTQUERY:
2674        case OP_NOTQUERYI:
2675    
2676      case OP_MINQUERY:      case OP_MINQUERY:
2677      case OP_MINQUERYI:      case OP_MINQUERYI:
2678        case OP_NOTMINQUERY:
2679        case OP_NOTMINQUERYI:
2680    
2681      case OP_POSQUERY:      case OP_POSQUERY:
2682      case OP_POSQUERYI:      case OP_POSQUERYI:
2683        case OP_NOTPOSQUERY:
2684        case OP_NOTPOSQUERYI:
2685    
2686      if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);      if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2687      break;      break;
2688    
2689      case OP_UPTO:      case OP_UPTO:
2690      case OP_UPTOI:      case OP_UPTOI:
2691        case OP_NOTUPTO:
2692        case OP_NOTUPTOI:
2693    
2694      case OP_MINUPTO:      case OP_MINUPTO:
2695      case OP_MINUPTOI:      case OP_MINUPTOI:
2696        case OP_NOTMINUPTO:
2697        case OP_NOTMINUPTOI:
2698    
2699      case OP_POSUPTO:      case OP_POSUPTO:
2700      case OP_POSUPTOI:      case OP_POSUPTOI:
2701        case OP_NOTPOSUPTO:
2702        case OP_NOTPOSUPTOI:
2703    
2704      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]);
2705      break;      break;
2706  #endif  #endif
# Line 2573  for (code = first_significant_code(code Line 2711  for (code = first_significant_code(code
2711      case OP_MARK:      case OP_MARK:
2712      case OP_PRUNE_ARG:      case OP_PRUNE_ARG:
2713      case OP_SKIP_ARG:      case OP_SKIP_ARG:
     code += code[1];  
     break;  
   
2714      case OP_THEN_ARG:      case OP_THEN_ARG:
2715      code += code[1];      code += code[1];
2716      break;      break;
# Line 2607  Arguments: Line 2742  Arguments:
2742    code        points to start of the recursion    code        points to start of the recursion
2743    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2744    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2745    utf         TRUE if in UTF-8 / UTF-16 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2746    cd          pointers to tables etc    cd          pointers to tables etc
2747    
2748  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
# Line 2619  could_be_empty(const pcre_uchar *code, c Line 2754  could_be_empty(const pcre_uchar *code, c
2754  {  {
2755  while (bcptr != NULL && bcptr->current_branch >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2756    {    {
2757    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd))    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd, NULL))
2758      return FALSE;      return FALSE;
2759    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2760    }    }
# Line 2629  return TRUE; Line 2764  return TRUE;
2764    
2765    
2766  /*************************************************  /*************************************************
2767  *           Check for POSIX class syntax         *  *        Base opcode of repeated opcodes         *
2768  *************************************************/  *************************************************/
2769    
2770  /* This function is called when the sequence "[:" or "[." or "[=" is  /* Returns the base opcode for repeated single character type opcodes. If the
2771  encountered in a character class. It checks whether this is followed by a  opcode is not a repeated character type, it returns with the original value.
 sequence of characters terminated by a matching ":]" or ".]" or "=]". If we  
 reach an unescaped ']' without the special preceding character, return FALSE.  
2772    
2773  Originally, this function only recognized a sequence of letters between the  Arguments:  c opcode
2774  terminators, but it seems that Perl recognizes any sequence of characters,  Returns:    base opcode for the type
2775  though of course unknown POSIX names are subsequently rejected. Perl gives an  */
 "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE  
 didn't consider this to be a POSIX class. Likewise for [:1234:].  
2776    
2777  The problem in trying to be exactly like Perl is in the handling of escapes. We  static pcre_uchar
2778  have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX  get_repeat_base(pcre_uchar c)
2779  class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code  {
2780  below handles the special case of \], but does not try to do any other escape  return (c > OP_TYPEPOSUPTO)? c :
2781  processing. This makes it different from Perl for cases such as [:l\ower:]         (c >= OP_TYPESTAR)?   OP_TYPESTAR :
2782  where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize         (c >= OP_NOTSTARI)?   OP_NOTSTARI :
2783  "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,         (c >= OP_NOTSTAR)?    OP_NOTSTAR :
2784  I think.         (c >= OP_STARI)?      OP_STARI :
2785                                 OP_STAR;
2786    }
2787    
 A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.  
 It seems that the appearance of a nested POSIX class supersedes an apparent  
 external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or  
 a digit.  
2788    
2789  In Perl, unescaped square brackets may also appear as part of class names. For  
2790  example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for  #ifdef SUPPORT_UCP
2791  [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not  /*************************************************
2792  seem right at all. PCRE does not allow closing square brackets in POSIX class  *        Check a character and a property        *
2793  names.  *************************************************/
2794    
2795    /* This function is called by check_auto_possessive() when a property item
2796    is adjacent to a fixed character.
2797    
2798  Arguments:  Arguments:
2799    ptr      pointer to the initial [    c            the character
2800    endptr   where to return the end pointer    ptype        the property type
2801      pdata        the data for the type
2802      negated      TRUE if it's a negated property (\P or \p{^)
2803    
2804  Returns:   TRUE or FALSE  Returns:       TRUE if auto-possessifying is OK
2805  */  */
2806    
2807  static BOOL  static BOOL
2808  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)  check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata,
2809      BOOL negated)
2810  {  {
2811  int terminator;          /* Don't combine these lines; the Solaris cc */  const pcre_uint32 *p;
2812  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  const ucd_record *prop = GET_UCD(c);
2813  for (++ptr; *ptr != 0; ptr++)  
2814    switch(ptype)
2815    {    {
2816    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)    case PT_LAMP:
2817      ptr++;    return (prop->chartype == ucp_Lu ||
2818    else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;            prop->chartype == ucp_Ll ||
2819    else            prop->chartype == ucp_Lt) == negated;
2820    
2821      case PT_GC:
2822      return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2823    
2824      case PT_PC:
2825      return (pdata == prop->chartype) == negated;
2826    
2827      case PT_SC:
2828      return (pdata == prop->script) == negated;
2829    
2830      /* These are specials */
2831    
2832      case PT_ALNUM:
2833      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2834              PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2835    
2836      /* Perl space used to exclude VT, but from Perl 5.18 it is included, which
2837      means that Perl space and POSIX space are now identical. PCRE was changed
2838      at release 8.34. */
2839    
2840      case PT_SPACE:    /* Perl space */
2841      case PT_PXSPACE:  /* POSIX space */
2842      switch(c)
2843      {      {
2844      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)      HSPACE_CASES:
2845        {      VSPACE_CASES:
2846        *endptr = ptr;      return negated;
2847        return TRUE;  
2848        }      default:
2849      if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z) == negated;
2850           (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||      }
2851            ptr[1] == CHAR_EQUALS_SIGN) &&    break;  /* Control never reaches here */
2852          check_posix_syntax(ptr, endptr))  
2853        return FALSE;    case PT_WORD:
2854      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2855              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2856              c == CHAR_UNDERSCORE) == negated;
2857    
2858      case PT_CLIST:
2859      p = PRIV(ucd_caseless_sets) + prop->caseset;
2860      for (;;)
2861        {
2862        if (c < *p) return !negated;
2863        if (c == *p++) return negated;
2864      }      }
2865      break;  /* Control never reaches here */
2866    }    }
2867    
2868  return FALSE;  return FALSE;
2869  }  }
2870    #endif  /* SUPPORT_UCP */
2871    
2872    
2873    
2874  /*************************************************  /*************************************************
2875  *          Check POSIX class name                *  *        Fill the character property list        *
2876  *************************************************/  *************************************************/
2877    
2878  /* This function is called to check the name given in a POSIX-style class entry  /* Checks whether the code points to an opcode that can take part in auto-
2879  such as [:alnum:].  possessification, and if so, fills a list with its properties.
2880    
2881  Arguments:  Arguments:
2882    ptr        points to the first letter    code        points to start of expression
2883    len        the length of the name    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2884      fcc         points to case-flipping table
2885      list        points to output list
2886                  list[0] will be filled with the opcode
2887                  list[1] will be non-zero if this opcode
2888                    can match an empty character string
2889                  list[2..7] depends on the opcode
2890    
2891  Returns:     a value representing the name, or -1 if unknown  Returns:      points to the start of the next opcode if *code is accepted
2892                  NULL if *code is not accepted
2893  */  */
2894    
2895  static int  static const pcre_uchar *
2896  check_posix_name(const pcre_uchar *ptr, int len)  get_chr_property_list(const pcre_uchar *code, BOOL utf,
2897      const pcre_uint8 *fcc, pcre_uint32 *list)
2898  {  {
2899  const char *pn = posix_names;  pcre_uchar c = *code;
2900  register int yield = 0;  pcre_uchar base;
2901  while (posix_name_lengths[yield] != 0)  const pcre_uchar *end;
2902    pcre_uint32 chr;
2903    
2904    #ifdef SUPPORT_UCP
2905    pcre_uint32 *clist_dest;
2906    const pcre_uint32 *clist_src;
2907    #else
2908    utf = utf;  /* Suppress "unused parameter" compiler warning */
2909    #endif
2910    
2911    list[0] = c;
2912    list[1] = FALSE;
2913    code++;
2914    
2915    if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
2916    {    {
2917    if (len == posix_name_lengths[yield] &&    base = get_repeat_base(c);
2918      STRNCMP_UC_C8(ptr, pn, len) == 0) return yield;    c -= (base - OP_STAR);
   pn += posix_name_lengths[yield] + 1;  
   yield++;  
   }  
 return -1;  
 }  
2919    
2920      if (c == OP_UPTO || c == OP_MINUPTO || c == OP_EXACT || c == OP_POSUPTO)
2921        code += IMM2_SIZE;
2922    
2923  /*************************************************    list[1] = (c != OP_PLUS && c != OP_MINPLUS && c != OP_EXACT && c != OP_POSPLUS);
 *    Adjust OP_RECURSE items in repeated group   *  
 *************************************************/  
2924    
2925  /* OP_RECURSE items contain an offset from the start of the regex to the group    switch(base)
2926  that is referenced. This means that groups can be replicated for fixed      {
2927  repetition simply by copying (because the recursion is allowed to refer to      case OP_STAR:
2928  earlier groups that are outside the current group). However, when a group is      list[0] = OP_CHAR;
2929  optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is      break;
2930    
2931        case OP_STARI:
2932        list[0] = OP_CHARI;
2933        break;
2934    
2935        case OP_NOTSTAR:
2936        list[0] = OP_NOT;
2937        break;
2938    
2939        case OP_NOTSTARI:
2940        list[0] = OP_NOTI;
2941        break;
2942    
2943        case OP_TYPESTAR:
2944        list[0] = *code;
2945        code++;
2946        break;
2947        }
2948      c = list[0];
2949      }
2950    
2951    switch(c)
2952      {
2953      case OP_NOT_DIGIT:
2954      case OP_DIGIT:
2955      case OP_NOT_WHITESPACE:
2956      case OP_WHITESPACE:
2957      case OP_NOT_WORDCHAR:
2958      case OP_WORDCHAR:
2959      case OP_ANY:
2960      case OP_ALLANY:
2961      case OP_ANYNL:
2962      case OP_NOT_HSPACE:
2963      case OP_HSPACE:
2964      case OP_NOT_VSPACE:
2965      case OP_VSPACE:
2966      case OP_EXTUNI:
2967      case OP_EODN:
2968      case OP_EOD:
2969      case OP_DOLL:
2970      case OP_DOLLM:
2971      return code;
2972    
2973      case OP_CHAR:
2974      case OP_NOT:
2975      GETCHARINCTEST(chr, code);
2976      list[2] = chr;
2977      list[3] = NOTACHAR;
2978      return code;
2979    
2980      case OP_CHARI:
2981      case OP_NOTI:
2982      list[0] = (c == OP_CHARI) ? OP_CHAR : OP_NOT;
2983      GETCHARINCTEST(chr, code);
2984      list[2] = chr;
2985    
2986    #ifdef SUPPORT_UCP
2987      if (chr < 128 || (chr < 256 && !utf))
2988        list[3] = fcc[chr];
2989      else
2990        list[3] = UCD_OTHERCASE(chr);
2991    #elif defined SUPPORT_UTF || !defined COMPILE_PCRE8
2992      list[3] = (chr < 256) ? fcc[chr] : chr;
2993    #else
2994      list[3] = fcc[chr];
2995    #endif
2996    
2997      /* The othercase might be the same value. */
2998    
2999      if (chr == list[3])
3000        list[3] = NOTACHAR;
3001      else
3002        list[4] = NOTACHAR;
3003      return code;
3004    
3005    #ifdef SUPPORT_UCP
3006      case OP_PROP:
3007      case OP_NOTPROP:
3008      if (code[0] != PT_CLIST)
3009        {
3010        list[2] = code[0];
3011        list[3] = code[1];
3012        return code + 2;
3013        }
3014    
3015      /* Convert only if we have enough space. */
3016    
3017      clist_src = PRIV(ucd_caseless_sets) + code[1];
3018      clist_dest = list + 2;
3019      code += 2;
3020    
3021      do {
3022         if (clist_dest >= list + 8)
3023           {
3024           /* Early return if there is not enough space. This should never
3025           happen, since all clists are shorter than 5 character now. */
3026           list[2] = code[0];
3027           list[3] = code[1];
3028           return code;
3029           }
3030         *clist_dest++ = *clist_src;
3031         }
3032      while(*clist_src++ != NOTACHAR);
3033    
3034      /* All characters are stored. The terminating NOTACHAR
3035      is copied form the clist itself. */
3036    
3037      list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
3038      return code;
3039    #endif
3040    
3041      case OP_NCLASS:
3042      case OP_CLASS:
3043    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3044      case OP_XCLASS:
3045      if (c == OP_XCLASS)
3046        end = code + GET(code, 0) - 1;
3047      else
3048    #endif
3049        end = code + 32 / sizeof(pcre_uchar);
3050    
3051      switch(*end)
3052        {
3053        case OP_CRSTAR:
3054        case OP_CRMINSTAR:
3055        case OP_CRQUERY:
3056        case OP_CRMINQUERY:
3057        case OP_CRPOSSTAR:
3058        case OP_CRPOSQUERY:
3059        list[1] = TRUE;
3060        end++;
3061        break;
3062    
3063        case OP_CRPLUS:
3064        case OP_CRMINPLUS:
3065        case OP_CRPOSPLUS:
3066        end++;
3067        break;
3068    
3069        case OP_CRRANGE:
3070        case OP_CRMINRANGE:
3071        case OP_CRPOSRANGE:
3072        list[1] = (GET2(end, 1) == 0);
3073        end += 1 + 2 * IMM2_SIZE;
3074        break;
3075        }
3076      list[2] = (pcre_uint32)(end - code);
3077      return end;
3078      }
3079    return NULL;    /* Opcode not accepted */
3080    }
3081    
3082    
3083    
3084    /*************************************************
3085    *    Scan further character sets for match       *
3086    *************************************************/
3087    
3088    /* Checks whether the base and the current opcode have a common character, in
3089    which case the base cannot be possessified.
3090    
3091    Arguments:
3092      code        points to the byte code
3093      utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3094      cd          static compile data
3095      base_list   the data list of the base opcode
3096    
3097    Returns:      TRUE if the auto-possessification is possible
3098    */
3099    
3100    static BOOL
3101    compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
3102      const pcre_uint32 *base_list, const pcre_uchar *base_end, int *rec_limit)
3103    {
3104    pcre_uchar c;
3105    pcre_uint32 list[8];
3106    const pcre_uint32 *chr_ptr;
3107    const pcre_uint32 *ochr_ptr;
3108    const pcre_uint32 *list_ptr;
3109    const pcre_uchar *next_code;
3110    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3111    const pcre_uchar *xclass_flags;
3112    #endif
3113    const pcre_uint8 *class_bitset;
3114    const pcre_uint8 *set1, *set2, *set_end;
3115    pcre_uint32 chr;
3116    BOOL accepted, invert_bits;
3117    BOOL entered_a_group = FALSE;
3118    
3119    if (*rec_limit == 0) return FALSE;
3120    --(*rec_limit);
3121    
3122    /* Note: the base_list[1] contains whether the current opcode has greedy
3123    (represented by a non-zero value) quantifier. This is a different from
3124    other character type lists, which stores here that the character iterator
3125    matches to an empty string (also represented by a non-zero value). */
3126    
3127    for(;;)
3128      {
3129      /* All operations move the code pointer forward.
3130      Therefore infinite recursions are not possible. */
3131    
3132      c = *code;
3133    
3134      /* Skip over callouts */
3135    
3136      if (c == OP_CALLOUT)
3137        {
3138        code += PRIV(OP_lengths)[c];
3139        continue;
3140        }
3141    
3142      if (c == OP_ALT)
3143        {
3144        do code += GET(code, 1); while (*code == OP_ALT);
3145        c = *code;
3146        }
3147    
3148      switch(c)
3149        {
3150        case OP_END:
3151        case OP_KETRPOS:
3152        /* TRUE only in greedy case. The non-greedy case could be replaced by
3153        an OP_EXACT, but it is probably not worth it. (And note that OP_EXACT
3154        uses more memory, which we cannot get at this stage.) */
3155    
3156        return base_list[1] != 0;
3157    
3158        case OP_KET:
3159        /* If the bracket is capturing, and referenced by an OP_RECURSE, or
3160        it is an atomic sub-pattern (assert, once, etc.) the non-greedy case
3161        cannot be converted to a possessive form. */
3162    
3163        if (base_list[1] == 0) return FALSE;
3164    
3165        switch(*(code - GET(code, 1)))
3166          {
3167          case OP_ASSERT:
3168          case OP_ASSERT_NOT:
3169          case OP_ASSERTBACK:
3170          case OP_ASSERTBACK_NOT:
3171          case OP_ONCE:
3172          case OP_ONCE_NC:
3173          /* Atomic sub-patterns and assertions can always auto-possessify their
3174          last iterator. However, if the group was entered as a result of checking
3175          a previous iterator, this is not possible. */
3176    
3177          return !entered_a_group;
3178          }
3179    
3180        code += PRIV(OP_lengths)[c];
3181        continue;
3182    
3183        case OP_ONCE:
3184        case OP_ONCE_NC:
3185        case OP_BRA:
3186        case OP_CBRA:
3187        next_code = code + GET(code, 1);
3188        code += PRIV(OP_lengths)[c];
3189    
3190        while (*next_code == OP_ALT)
3191          {
3192          if (!compare_opcodes(code, utf, cd, base_list, base_end, rec_limit))
3193            return FALSE;
3194          code = next_code + 1 + LINK_SIZE;
3195          next_code += GET(next_code, 1);
3196          }
3197    
3198        entered_a_group = TRUE;
3199        continue;
3200    
3201        case OP_BRAZERO:
3202        case OP_BRAMINZERO:
3203    
3204        next_code = code + 1;
3205        if (*next_code != OP_BRA && *next_code != OP_CBRA
3206            && *next_code != OP_ONCE && *next_code != OP_ONCE_NC) return FALSE;
3207    
3208        do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
3209    
3210        /* The bracket content will be checked by the
3211        OP_BRA/OP_CBRA case above. */
3212        next_code += 1 + LINK_SIZE;
3213        if (!compare_opcodes(next_code, utf, cd, base_list, base_end, rec_limit))
3214          return FALSE;
3215    
3216        code += PRIV(OP_lengths)[c];
3217        continue;
3218    
3219        default:
3220        break;
3221        }
3222    
3223      /* Check for a supported opcode, and load its properties. */
3224    
3225      code = get_chr_property_list(code, utf, cd->fcc, list);
3226      if (code == NULL) return FALSE;    /* Unsupported */
3227    
3228      /* If either opcode is a small character list, set pointers for comparing
3229      characters from that list with another list, or with a property. */
3230    
3231      if (base_list[0] == OP_CHAR)
3232        {
3233        chr_ptr = base_list + 2;
3234        list_ptr = list;
3235        }
3236      else if (list[0] == OP_CHAR)
3237        {
3238        chr_ptr = list + 2;
3239        list_ptr = base_list;
3240        }
3241    
3242      /* Character bitsets can also be compared to certain opcodes. */
3243    
3244      else if (base_list[0] == OP_CLASS || list[0] == OP_CLASS
3245    #ifdef COMPILE_PCRE8
3246          /* In 8 bit, non-UTF mode, OP_CLASS and OP_NCLASS are the same. */
3247          || (!utf && (base_list[0] == OP_NCLASS || list[0] == OP_NCLASS))
3248    #endif
3249          )
3250        {
3251    #ifdef COMPILE_PCRE8
3252        if (base_list[0] == OP_CLASS || (!utf && base_list[0] == OP_NCLASS))
3253    #else
3254        if (base_list[0] == OP_CLASS)
3255    #endif
3256          {
3257          set1 = (pcre_uint8 *)(base_end - base_list[2]);
3258          list_ptr = list;
3259          }
3260        else
3261          {
3262          set1 = (pcre_uint8 *)(code - list[2]);
3263          list_ptr = base_list;
3264          }
3265    
3266        invert_bits = FALSE;
3267        switch(list_ptr[0])
3268          {
3269          case OP_CLASS:
3270          case OP_NCLASS:
3271          set2 = (pcre_uint8 *)
3272            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3273          break;
3274    
3275    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3276          case OP_XCLASS:
3277          xclass_flags = (list_ptr == list ? code : base_end) - list_ptr[2] + LINK_SIZE;
3278          if ((*xclass_flags & XCL_HASPROP) != 0) return FALSE;
3279          if ((*xclass_flags & XCL_MAP) == 0)
3280            {
3281            /* No bits are set for characters < 256. */
3282            if (list[1] == 0) return TRUE;
3283            /* Might be an empty repeat. */
3284            continue;
3285            }
3286          set2 = (pcre_uint8 *)(xclass_flags + 1);
3287          break;
3288    #endif
3289    
3290          case OP_NOT_DIGIT:
3291          invert_bits = TRUE;
3292          /* Fall through */
3293          case OP_DIGIT:
3294          set2 = (pcre_uint8 *)(cd->cbits + cbit_digit);
3295          break;
3296    
3297          case OP_NOT_WHITESPACE:
3298          invert_bits = TRUE;
3299          /* Fall through */
3300          case OP_WHITESPACE:
3301          set2 = (pcre_uint8 *)(cd->cbits + cbit_space);
3302          break;
3303    
3304          case OP_NOT_WORDCHAR:
3305          invert_bits = TRUE;
3306          /* Fall through */
3307          case OP_WORDCHAR:
3308          set2 = (pcre_uint8 *)(cd->cbits + cbit_word);
3309          break;
3310    
3311          default:
3312          return FALSE;
3313          }
3314    
3315        /* Because the sets are unaligned, we need
3316        to perform byte comparison here. */
3317        set_end = set1 + 32;
3318        if (invert_bits)
3319          {
3320          do
3321            {
3322            if ((*set1++ & ~(*set2++)) != 0) return FALSE;
3323            }
3324          while (set1 < set_end);
3325          }
3326        else
3327          {
3328          do
3329            {
3330            if ((*set1++ & *set2++) != 0) return FALSE;
3331            }
3332          while (set1 < set_end);
3333          }
3334    
3335        if (list[1] == 0) return TRUE;
3336        /* Might be an empty repeat. */
3337        continue;
3338        }
3339    
3340      /* Some property combinations also acceptable. Unicode property opcodes are
3341      processed specially; the rest can be handled with a lookup table. */
3342    
3343      else
3344        {
3345        pcre_uint32 leftop, rightop;
3346    
3347        leftop = base_list[0];
3348        rightop = list[0];
3349    
3350    #ifdef SUPPORT_UCP
3351        accepted = FALSE; /* Always set in non-unicode case. */
3352        if (leftop == OP_PROP || leftop == OP_NOTPROP)
3353          {
3354          if (rightop == OP_EOD)
3355            accepted = TRUE;
3356          else if (rightop == OP_PROP || rightop == OP_NOTPROP)
3357            {
3358            int n;
3359            const pcre_uint8 *p;
3360            BOOL same = leftop == rightop;
3361            BOOL lisprop = leftop == OP_PROP;
3362            BOOL risprop = rightop == OP_PROP;
3363            BOOL bothprop = lisprop && risprop;
3364    
3365            /* There's a table that specifies how each combination is to be
3366            processed:
3367              0   Always return FALSE (never auto-possessify)
3368              1   Character groups are distinct (possessify if both are OP_PROP)
3369              2   Check character categories in the same group (general or particular)
3370              3   Return TRUE if the two opcodes are not the same
3371              ... see comments below
3372            */
3373    
3374            n = propposstab[base_list[2]][list[2]];
3375            switch(n)
3376              {
3377              case 0: break;
3378              case 1: accepted = bothprop; break;
3379              case 2: accepted = (base_list[3] == list[3]) != same; break;
3380              case 3: accepted = !same; break;
3381    
3382              case 4:  /* Left general category, right particular category */
3383              accepted = risprop && catposstab[base_list[3]][list[3]] == same;
3384              break;
3385    
3386              case 5:  /* Right general category, left particular category */
3387              accepted = lisprop && catposstab[list[3]][base_list[3]] == same;
3388              break;
3389    
3390              /* This code is logically tricky. Think hard before fiddling with it.
3391              The posspropstab table has four entries per row. Each row relates to
3392              one of PCRE's special properties such as ALNUM or SPACE or WORD.
3393              Only WORD actually needs all four entries, but using repeats for the
3394              others means they can all use the same code below.
3395    
3396              The first two entries in each row are Unicode general categories, and
3397              apply always, because all the characters they include are part of the
3398              PCRE character set. The third and fourth entries are a general and a
3399              particular category, respectively, that include one or more relevant
3400              characters. One or the other is used, depending on whether the check
3401              is for a general or a particular category. However, in both cases the
3402              category contains more characters than the specials that are defined
3403              for the property being tested against. Therefore, it cannot be used
3404              in a NOTPROP case.
3405    
3406              Example: the row for WORD contains ucp_L, ucp_N, ucp_P, ucp_Po.
3407              Underscore is covered by ucp_P or ucp_Po. */
3408    
3409              case 6:  /* Left alphanum vs right general category */
3410              case 7:  /* Left space vs right general category */
3411              case 8:  /* Left word vs right general category */
3412              p = posspropstab[n-6];
3413              accepted = risprop && lisprop ==
3414                (list[3] != p[0] &&
3415                 list[3] != p[1] &&
3416                (list[3] != p[2] || !lisprop));
3417              break;
3418    
3419              case 9:   /* Right alphanum vs left general category */
3420              case 10:  /* Right space vs left general category */
3421              case 11:  /* Right word vs left general category */
3422              p = posspropstab[n-9];
3423              accepted = lisprop && risprop ==
3424                (base_list[3] != p[0] &&
3425                 base_list[3] != p[1] &&
3426                (base_list[3] != p[2] || !risprop));
3427              break;
3428    
3429              case 12:  /* Left alphanum vs right particular category */
3430              case 13:  /* Left space vs right particular category */
3431              case 14:  /* Left word vs right particular category */
3432              p = posspropstab[n-12];
3433              accepted = risprop && lisprop ==
3434                (catposstab[p[0]][list[3]] &&
3435                 catposstab[p[1]][list[3]] &&
3436                (list[3] != p[3] || !lisprop));
3437              break;
3438    
3439              case 15:  /* Right alphanum vs left particular category */
3440              case 16:  /* Right space vs left particular category */
3441              case 17:  /* Right word vs left particular category */
3442              p = posspropstab[n-15];
3443              accepted = lisprop && risprop ==
3444                (catposstab[p[0]][base_list[3]] &&
3445                 catposstab[p[1]][base_list[3]] &&
3446                (base_list[3] != p[3] || !risprop));
3447              break;
3448              }
3449            }
3450          }
3451    
3452        else
3453    #endif  /* SUPPORT_UCP */
3454    
3455        accepted = leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&
3456               rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&
3457               autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];
3458    
3459        if (!accepted) return FALSE;
3460    
3461        if (list[1] == 0) return TRUE;
3462        /* Might be an empty repeat. */
3463        continue;
3464        }
3465    
3466      /* Control reaches here only if one of the items is a small character list.
3467      All characters are checked against the other side. */
3468    
3469      do
3470        {
3471        chr = *chr_ptr;
3472    
3473        switch(list_ptr[0])
3474          {
3475          case OP_CHAR:
3476          ochr_ptr = list_ptr + 2;
3477          do
3478            {
3479            if (chr == *ochr_ptr) return FALSE;
3480            ochr_ptr++;
3481            }
3482          while(*ochr_ptr != NOTACHAR);
3483          break;
3484    
3485          case OP_NOT:
3486          ochr_ptr = list_ptr + 2;
3487          do
3488            {
3489            if (chr == *ochr_ptr)
3490              break;
3491            ochr_ptr++;
3492            }
3493          while(*ochr_ptr != NOTACHAR);
3494          if (*ochr_ptr == NOTACHAR) return FALSE;   /* Not found */
3495          break;
3496    
3497          /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not*
3498          set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3499    
3500          case OP_DIGIT:
3501          if (chr < 256 && (cd->ctypes[chr] & ctype_digit) != 0) return FALSE;
3502          break;
3503    
3504          case OP_NOT_DIGIT:
3505          if (chr > 255 || (cd->ctypes[chr] & ctype_digit) == 0) return FALSE;
3506          break;
3507    
3508          case OP_WHITESPACE:
3509          if (chr < 256 && (cd->ctypes[chr] & ctype_space) != 0) return FALSE;
3510          break;
3511    
3512          case OP_NOT_WHITESPACE:
3513          if (chr > 255 || (cd->ctypes[chr] & ctype_space) == 0) return FALSE;
3514          break;
3515    
3516          case OP_WORDCHAR:
3517          if (chr < 255 && (cd->ctypes[chr] & ctype_word) != 0) return FALSE;
3518          break;
3519    
3520          case OP_NOT_WORDCHAR:
3521          if (chr > 255 || (cd->ctypes[chr] & ctype_word) == 0) return FALSE;
3522          break;
3523    
3524          case OP_HSPACE:
3525          switch(chr)
3526            {
3527            HSPACE_CASES: return FALSE;
3528            default: break;
3529            }
3530          break;
3531    
3532          case OP_NOT_HSPACE:
3533          switch(chr)
3534            {
3535            HSPACE_CASES: break;
3536            default: return FALSE;
3537            }
3538          break;
3539    
3540          case OP_ANYNL:
3541          case OP_VSPACE:
3542          switch(chr)
3543            {
3544            VSPACE_CASES: return FALSE;
3545            default: break;
3546            }
3547          break;
3548    
3549          case OP_NOT_VSPACE:
3550          switch(chr)
3551            {
3552            VSPACE_CASES: break;
3553            default: return FALSE;
3554            }
3555          break;
3556    
3557          case OP_DOLL:
3558          case OP_EODN:
3559          switch (chr)
3560            {
3561            case CHAR_CR:
3562            case CHAR_LF:
3563            case CHAR_VT:
3564            case CHAR_FF:
3565            case CHAR_NEL:
3566    #ifndef EBCDIC
3567            case 0x2028:
3568            case 0x2029:
3569    #endif  /* Not EBCDIC */
3570            return FALSE;
3571            }
3572          break;
3573    
3574          case OP_EOD:    /* Can always possessify before \z */
3575          break;
3576    
3577    #ifdef SUPPORT_UCP
3578          case OP_PROP:
3579          case OP_NOTPROP:
3580          if (!check_char_prop(chr, list_ptr[2], list_ptr[3],
3581                list_ptr[0] == OP_NOTPROP))
3582            return FALSE;
3583          break;
3584    #endif
3585    
3586          case OP_NCLASS:
3587          if (chr > 255) return FALSE;
3588          /* Fall through */
3589    
3590          case OP_CLASS:
3591          if (chr > 255) break;
3592          class_bitset = (pcre_uint8 *)
3593            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3594          if ((class_bitset[chr >> 3] & (1 << (chr & 7))) != 0) return FALSE;
3595          break;
3596    
3597    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3598          case OP_XCLASS:
3599          if (PRIV(xclass)(chr, (list_ptr == list ? code : base_end) -
3600              list_ptr[2] + LINK_SIZE, utf)) return FALSE;
3601          break;
3602    #endif
3603    
3604          default:
3605          return FALSE;
3606          }
3607    
3608        chr_ptr++;
3609        }
3610      while(*chr_ptr != NOTACHAR);
3611    
3612      /* At least one character must be matched from this opcode. */
3613    
3614      if (list[1] == 0) return TRUE;
3615      }
3616    
3617    /* Control never reaches here. There used to be a fail-save return FALSE; here,
3618    but some compilers complain about an unreachable statement. */
3619    
3620    }
3621    
3622    
3623    
3624    /*************************************************
3625    *    Scan compiled regex for auto-possession     *
3626    *************************************************/
3627    
3628    /* Replaces single character iterations with their possessive alternatives
3629    if appropriate. This function modifies the compiled opcode!
3630    
3631    Arguments:
3632      code        points to start of the byte code
3633      utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3634      cd          static compile data
3635    
3636    Returns:      nothing
3637    */
3638    
3639    static void
3640    auto_possessify(pcre_uchar *code, BOOL utf, const compile_data *cd)
3641    {
3642    register pcre_uchar c;
3643    const pcre_uchar *end;
3644    pcre_uchar *repeat_opcode;
3645    pcre_uint32 list[8];
3646    int rec_limit;
3647    
3648    for (;;)
3649      {
3650      c = *code;
3651    
3652      /* When a pattern with bad UTF-8 encoding is compiled with NO_UTF_CHECK,
3653      it may compile without complaining, but may get into a loop here if the code
3654      pointer points to a bad value. This is, of course a documentated possibility,
3655      when NO_UTF_CHECK is set, so it isn't a bug, but we can detect this case and
3656      just give up on this optimization. */
3657    
3658      if (c >= OP_TABLE_LENGTH) return;
3659    
3660      if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
3661        {
3662        c -= get_repeat_base(c) - OP_STAR;
3663        end = (c <= OP_MINUPTO) ?
3664          get_chr_property_list(code, utf, cd->fcc, list) : NULL;
3665        list[1] = c == OP_STAR || c == OP_PLUS || c == OP_QUERY || c == OP_UPTO;
3666    
3667        rec_limit = 1000;
3668        if (end != NULL && compare_opcodes(end, utf, cd, list, end, &rec_limit))
3669          {
3670          switch(c)
3671            {
3672            case OP_STAR:
3673            *code += OP_POSSTAR - OP_STAR;
3674            break;
3675    
3676            case OP_MINSTAR:
3677            *code += OP_POSSTAR - OP_MINSTAR;
3678            break;
3679    
3680            case OP_PLUS:
3681            *code += OP_POSPLUS - OP_PLUS;
3682            break;
3683    
3684            case OP_MINPLUS:
3685            *code += OP_POSPLUS - OP_MINPLUS;
3686            break;
3687    
3688            case OP_QUERY:
3689            *code += OP_POSQUERY - OP_QUERY;
3690            break;
3691    
3692            case OP_MINQUERY:
3693            *code += OP_POSQUERY - OP_MINQUERY;
3694            break;
3695    
3696            case OP_UPTO:
3697            *code += OP_POSUPTO - OP_UPTO;
3698            break;
3699    
3700            case OP_MINUPTO:
3701            *code += OP_POSUPTO - OP_MINUPTO;
3702            break;
3703            }
3704          }
3705        c = *code;
3706        }
3707      else if (c == OP_CLASS || c == OP_NCLASS || c == OP_XCLASS)
3708        {
3709    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3710        if (c == OP_XCLASS)
3711          repeat_opcode = code + GET(code, 1);
3712        else
3713    #endif
3714          repeat_opcode = code + 1 + (32 / sizeof(pcre_uchar));
3715    
3716        c = *repeat_opcode;
3717        if (c >= OP_CRSTAR && c <= OP_CRMINRANGE)
3718          {
3719          /* end must not be NULL. */
3720          end = get_chr_property_list(code, utf, cd->fcc, list);
3721    
3722          list[1] = (c & 1) == 0;
3723    
3724          rec_limit = 1000;
3725          if (compare_opcodes(end, utf, cd, list, end, &rec_limit))
3726            {
3727            switch (c)
3728              {
3729              case OP_CRSTAR:
3730              case OP_CRMINSTAR:
3731              *repeat_opcode = OP_CRPOSSTAR;
3732              break;
3733    
3734              case OP_CRPLUS:
3735              case OP_CRMINPLUS:
3736              *repeat_opcode = OP_CRPOSPLUS;
3737              break;
3738    
3739              case OP_CRQUERY:
3740              case OP_CRMINQUERY:
3741              *repeat_opcode = OP_CRPOSQUERY;
3742              break;
3743    
3744              case OP_CRRANGE:
3745              case OP_CRMINRANGE:
3746              *repeat_opcode = OP_CRPOSRANGE;
3747              break;
3748              }
3749            }
3750          }
3751        c = *code;
3752        }
3753    
3754      switch(c)
3755        {
3756        case OP_END:
3757        return;
3758    
3759        case OP_TYPESTAR:
3760        case OP_TYPEMINSTAR:
3761        case OP_TYPEPLUS:
3762        case OP_TYPEMINPLUS:
3763        case OP_TYPEQUERY:
3764        case OP_TYPEMINQUERY:
3765        case OP_TYPEPOSSTAR:
3766        case OP_TYPEPOSPLUS:
3767        case OP_TYPEPOSQUERY:
3768        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
3769        break;
3770    
3771        case OP_TYPEUPTO:
3772        case OP_TYPEMINUPTO:
3773        case OP_TYPEEXACT:
3774        case OP_TYPEPOSUPTO:
3775        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
3776          code += 2;
3777        break;
3778    
3779    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3780        case OP_XCLASS:
3781        code += GET(code, 1);
3782        break;
3783    #endif
3784    
3785        case OP_MARK:
3786        case OP_PRUNE_ARG:
3787        case OP_SKIP_ARG:
3788        case OP_THEN_ARG:
3789        code += code[1];
3790        break;
3791        }
3792    
3793      /* Add in the fixed length from the table */
3794    
3795      code += PRIV(OP_lengths)[c];
3796    
3797      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
3798      a multi-byte character. The length in the table is a minimum, so we have to
3799      arrange to skip the extra bytes. */
3800    
3801    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3802      if (utf) switch(c)
3803        {
3804        case OP_CHAR:
3805        case OP_CHARI:
3806        case OP_NOT:
3807        case OP_NOTI:
3808        case OP_STAR:
3809        case OP_MINSTAR:
3810        case OP_PLUS:
3811        case OP_MINPLUS:
3812        case OP_QUERY:
3813        case OP_MINQUERY:
3814        case OP_UPTO:
3815        case OP_MINUPTO:
3816        case OP_EXACT:
3817        case OP_POSSTAR:
3818        case OP_POSPLUS:
3819        case OP_POSQUERY:
3820        case OP_POSUPTO:
3821        case OP_STARI:
3822        case OP_MINSTARI:
3823        case OP_PLUSI:
3824        case OP_MINPLUSI:
3825        case OP_QUERYI:
3826        case OP_MINQUERYI:
3827        case OP_UPTOI:
3828        case OP_MINUPTOI:
3829        case OP_EXACTI:
3830        case OP_POSSTARI:
3831        case OP_POSPLUSI:
3832        case OP_POSQUERYI:
3833        case OP_POSUPTOI:
3834        case OP_NOTSTAR:
3835        case OP_NOTMINSTAR:
3836        case OP_NOTPLUS:
3837        case OP_NOTMINPLUS:
3838        case OP_NOTQUERY:
3839        case OP_NOTMINQUERY:
3840        case OP_NOTUPTO:
3841        case OP_NOTMINUPTO:
3842        case OP_NOTEXACT:
3843        case OP_NOTPOSSTAR:
3844        case OP_NOTPOSPLUS:
3845        case OP_NOTPOSQUERY:
3846        case OP_NOTPOSUPTO:
3847        case OP_NOTSTARI:
3848        case OP_NOTMINSTARI:
3849        case OP_NOTPLUSI:
3850        case OP_NOTMINPLUSI:
3851        case OP_NOTQUERYI:
3852        case OP_NOTMINQUERYI:
3853        case OP_NOTUPTOI:
3854        case OP_NOTMINUPTOI:
3855        case OP_NOTEXACTI:
3856        case OP_NOTPOSSTARI:
3857        case OP_NOTPOSPLUSI:
3858        case OP_NOTPOSQUERYI:
3859        case OP_NOTPOSUPTOI:
3860        if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
3861        break;
3862        }
3863    #else
3864      (void)(utf);  /* Keep compiler happy by referencing function argument */
3865    #endif
3866      }
3867    }
3868    
3869    
3870    
3871    /*************************************************
3872    *           Check for POSIX class syntax         *
3873    *************************************************/
3874    
3875    /* This function is called when the sequence "[:" or "[." or "[=" is
3876    encountered in a character class. It checks whether this is followed by a
3877    sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
3878    reach an unescaped ']' without the special preceding character, return FALSE.
3879    
3880    Originally, this function only recognized a sequence of letters between the
3881    terminators, but it seems that Perl recognizes any sequence of characters,
3882    though of course unknown POSIX names are subsequently rejected. Perl gives an
3883    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
3884    didn't consider this to be a POSIX class. Likewise for [:1234:].
3885    
3886    The problem in trying to be exactly like Perl is in the handling of escapes. We
3887    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
3888    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
3889    below handles the special case of \], but does not try to do any other escape
3890    processing. This makes it different from Perl for cases such as [:l\ower:]
3891    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
3892    "l\ower". This is a lesser evil than not diagnosing bad classes when Perl does,
3893    I think.
3894    
3895    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
3896    It seems that the appearance of a nested POSIX class supersedes an apparent
3897    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
3898    a digit.
3899    
3900    In Perl, unescaped square brackets may also appear as part of class names. For
3901    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
3902    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
3903    seem right at all. PCRE does not allow closing square brackets in POSIX class
3904    names.
3905    
3906    Arguments:
3907      ptr      pointer to the initial [
3908      endptr   where to return the end pointer
3909    
3910    Returns:   TRUE or FALSE
3911    */
3912    
3913    static BOOL
3914    check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
3915    {
3916    pcre_uchar terminator;          /* Don't combine these lines; the Solaris cc */
3917    terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
3918    for (++ptr; *ptr != CHAR_NULL; ptr++)
3919      {
3920      if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3921        ptr++;
3922      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
3923      else
3924        {
3925        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3926          {
3927          *endptr = ptr;
3928          return TRUE;
3929          }
3930        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
3931             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3932              ptr[1] == CHAR_EQUALS_SIGN) &&
3933            check_posix_syntax(ptr, endptr))
3934          return FALSE;
3935        }
3936      }
3937    return FALSE;
3938    }
3939    
3940    
3941    
3942    
3943    /*************************************************
3944    *          Check POSIX class name                *
3945    *************************************************/
3946    
3947    /* This function is called to check the name given in a POSIX-style class entry
3948    such as [:alnum:].
3949    
3950    Arguments:
3951      ptr        points to the first letter
3952      len        the length of the name
3953    
3954    Returns:     a value representing the name, or -1 if unknown
3955    */
3956    
3957    static int
3958    check_posix_name(const pcre_uchar *ptr, int len)
3959    {
3960    const char *pn = posix_names;
3961    register int yield = 0;
3962    while (posix_name_lengths[yield] != 0)
3963      {
3964      if (len == posix_name_lengths[yield] &&
3965        STRNCMP_UC_C8(ptr, pn, (unsigned int)len) == 0) return yield;
3966      pn += posix_name_lengths[yield] + 1;
3967      yield++;
3968      }
3969    return -1;
3970    }
3971    
3972    
3973    /*************************************************
3974    *    Adjust OP_RECURSE items in repeated group   *
3975    *************************************************/
3976    
3977    /* OP_RECURSE items contain an offset from the start of the regex to the group
3978    that is referenced. This means that groups can be replicated for fixed
3979    repetition simply by copying (because the recursion is allowed to refer to
3980    earlier groups that are outside the current group). However, when a group is
3981    optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is
3982  inserted before it, after it has been compiled. This means that any OP_RECURSE  inserted before it, after it has been compiled. This means that any OP_RECURSE
3983  items within it that refer to the group itself or any contained groups have to  items within it that refer to the group itself or any contained groups have to
3984  have their offsets adjusted. That one of the jobs of this function. Before it  have their offsets adjusted. That one of the jobs of this function. Before it
3985  is called, the partially compiled regex must be temporarily terminated with  is called, the partially compiled regex must be temporarily terminated with
3986  OP_END.  OP_END.
3987    
3988  This function has been extended with the possibility of forward references for  This function has been extended to cope with forward references for recursions
3989  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
3990  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
3991  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
3992  value in the reference (which is a group number).  (which is a group number). After the group has been scanned, all the offsets in
3993    the forward reference list for the group are adjusted.
3994    
3995  Arguments:  Arguments:
3996    group      points to the start of the group    group      points to the start of the group
3997    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
3998    utf        TRUE in UTF-8 / UTF-16 mode    utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode
3999    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
4000    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
4001    
4002  Returns:     nothing  Returns:     nothing
4003  */  */
4004    
4005  static void  static void
4006  adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,  adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
4007    pcre_uchar *save_hwm)    size_t save_hwm_offset)
4008  {  {
4009    int offset;
4010    pcre_uchar *hc;
4011  pcre_uchar *ptr = group;  pcre_uchar *ptr = group;
4012    
4013  while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)  while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
4014    {    {
4015    int offset;    for (hc = (pcre_uchar *)cd->start_workspace + save_hwm_offset; hc < cd->hwm;
4016    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)  
4017      {      {
4018      offset = GET(hc, 0);      offset = (int)GET(hc, 0);
4019      if (cd->start_code + offset == ptr + 1)      if (cd->start_code + offset == ptr + 1) break;
       {  
       PUT(hc, 0, offset + adjust);  
       break;  
       }  
4020      }      }
4021    
4022    /* 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
4023    group. */    the recursion's offset if it's after the start of this group. */
4024    
4025    if (hc >= cd->hwm)    if (hc >= cd->hwm)
4026      {      {
4027      offset = GET(ptr, 1);      offset = (int)GET(ptr, 1);
4028      if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);      if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
4029      }      }
4030    
4031    ptr += 1 + LINK_SIZE;    ptr += 1 + LINK_SIZE;
4032    }    }
4033    
4034    /* Now adjust all forward reference offsets for the group. */
4035    
4036    for (hc = (pcre_uchar *)cd->start_workspace + save_hwm_offset; hc < cd->hwm;
4037         hc += LINK_SIZE)
4038      {
4039      offset = (int)GET(hc, 0);
4040      PUT(hc, 0, offset + adjust);
4041      }
4042  }  }
4043    
4044    
# Line 2858  PUT(previous_callout, 2 + LINK_SIZE, len Line 4101  PUT(previous_callout, 2 + LINK_SIZE, len
4101  *************************************************/  *************************************************/
4102    
4103  /* 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
4104  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
4105  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
4106  start address.  start address. A character with multiple other cases is returned on its own
4107    with a special return value.
4108    
4109  Arguments:  Arguments:
4110    cptr        points to starting character value; updated    cptr        points to starting character value; updated
# Line 2868  Arguments: Line 4112  Arguments:
4112    ocptr       where to put start of othercase range    ocptr       where to put start of othercase range
4113    odptr       where to put end of othercase range    odptr       where to put end of othercase range
4114    
4115  Yield:        TRUE when range returned; FALSE when no more  Yield:        -1 when no more
4116                   0 when a range is returned
4117                  >0 the CASESET offset for char with multiple other cases
4118                    in this case, ocptr contains the original
4119  */  */
4120    
4121  static BOOL  static int
4122  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,  get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,
4123    unsigned int *odptr)    pcre_uint32 *odptr)
4124  {  {
4125  unsigned int c, othercase, next;  pcre_uint32 c, othercase, next;
4126    unsigned int co;
4127    
4128    /* Find the first character that has an other case. If it has multiple other
4129    cases, return its case offset value. */
4130    
4131  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
4132    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }    {
4133      if ((co = UCD_CASESET(c)) != 0)
4134        {
4135        *ocptr = c++;   /* Character that has the set */
4136        *cptr = c;      /* Rest of input range */
4137        return (int)co;
4138        }
4139      if ((othercase = UCD_OTHERCASE(c)) != c) break;
4140      }
4141    
4142    if (c > d) return -1;  /* Reached end of range */
4143    
4144  if (c > d) return FALSE;  /* Found a character that has a single other case. Search for the end of the
4145    range, which is either the end of the input range, or a character that has zero
4146    or more than one other cases. */
4147    
4148  *ocptr = othercase;  *ocptr = othercase;
4149  next = othercase + 1;  next = othercase + 1;
4150    
4151  for (++c; c <= d; c++)  for (++c; c <= d; c++)
4152    {    {
4153    if (UCD_OTHERCASE(c) != next) break;    if ((co = UCD_CASESET(c)) != 0 || UCD_OTHERCASE(c) != next) break;
4154    next++;    next++;
4155    }    }
4156    
4157  *odptr = next - 1;  *odptr = next - 1;     /* End of othercase range */
4158  *cptr = c;  *cptr = c;             /* Rest of input range */
4159    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;  
4160  }  }
4161  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
4162    
4163    
4164    
4165  /*************************************************  /*************************************************
4166  *     Check if auto-possessifying is possible    *  *        Add a character or range to a class     *
4167  *************************************************/  *************************************************/
   
 /* This function is called for unlimited repeats of certain items, to see  
 whether the next thing could possibly match the repeated item. If not, it makes  
 sense to automatically possessify the repeated item.  
   
 Arguments:  
   previous      pointer to the repeated opcode  
   utf           TRUE in UTF-8 / UTF-16 mode  
   ptr           next character in pattern  
   options       options bits  
   cd            contains pointers to tables etc.  
   
 Returns:        TRUE if possessifying is wanted  
 */  
   
 static BOOL  
 check_auto_possessive(const pcre_uchar *previous, BOOL utf,  
   const pcre_uchar *ptr, int options, compile_data *cd)  
 {  
 pcre_int32 c, next;  
 int op_code = *previous++;  
   
 /* Skip whitespace and comments in extended mode */  
   
 if ((options & PCRE_EXTENDED) != 0)  
   {  
   for (;;)  
     {  
     while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;  
     if (*ptr == CHAR_NUMBER_SIGN)  
       {  
       ptr++;  
       while (*ptr != 0)  
         {  
         if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }  
         ptr++;  
 #ifdef SUPPORT_UTF  
         if (utf) FORWARDCHAR(ptr);  
 #endif  
         }  
       }  
     else break;  
     }  
   }  
   
 /* If the next item is one that we can handle, get its value. A non-negative  
 value is a character, a negative value is an escape value. */  
   
 if (*ptr == CHAR_BACKSLASH)  
   {  
   int temperrorcode = 0;  
   next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);  
   if (temperrorcode != 0) return FALSE;  
   ptr++;    /* Point after the escape sequence */  
   }  
 else if (!MAX_255(*ptr) || (cd->ctypes[*ptr] & ctype_meta) == 0)  
   {  
 #ifdef SUPPORT_UTF  
   if (utf) { GETCHARINC(next, ptr); } else  
 #endif  
   next = *ptr++;  
   }  
 else return FALSE;  
   
 /* Skip whitespace and comments in extended mode */  
   
 if ((options & PCRE_EXTENDED) != 0)  
   {  
   for (;;)  
     {  
     while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;  
     if (*ptr == CHAR_NUMBER_SIGN)  
       {  
       ptr++;  
       while (*ptr != 0)  
         {  
         if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }  
         ptr++;  
 #ifdef SUPPORT_UTF  
         if (utf) FORWARDCHAR(ptr);  
 #endif  
         }  
       }  
     else break;  
     }  
   }  
   
 /* If the next thing is itself optional, we have to give up. */  
   
 if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||  
   STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)  
     return FALSE;  
   
 /* Now compare the next item with the previous opcode. First, handle cases when  
 the next item is a character. */  
   
 if (next >= 0) switch(op_code)  
   {  
   case OP_CHAR:  
 #ifdef SUPPORT_UTF  
   GETCHARTEST(c, previous);  
 #else  
   c = *previous;  
 #endif  
   return c != next;  
   
   /* For CHARI (caseless character) we must check the other case. If we have  
   Unicode property support, we can use it to test the other case of  
   high-valued characters. */  
   
   case OP_CHARI:  
 #ifdef SUPPORT_UTF  
   GETCHARTEST(c, previous);  
 #else  
   c = *previous;  
 #endif  
   if (c == next) return FALSE;  
 #ifdef SUPPORT_UTF  
   if (utf)  
     {  
     unsigned int othercase;  
     if (next < 128) othercase = cd->fcc[next]; else  
 #ifdef SUPPORT_UCP  
     othercase = UCD_OTHERCASE((unsigned int)next);  
 #else  
     othercase = NOTACHAR;  
 #endif  
     return (unsigned int)c != othercase;  
     }  
   else  
 #endif  /* SUPPORT_UTF */  
   return (c != TABLE_GET((unsigned int)next, cd->fcc, next));  /* Non-UTF-8 mode */  
   
   case OP_NOT:  
 #ifdef SUPPORT_UTF  
   GETCHARTEST(c, previous);  
 #else  
   c = *previous;  
 #endif  
   return c == next;  
4168    
4169    case OP_NOTI:  /* This function packages up the logic of adding a character or range of
4170  #ifdef SUPPORT_UTF  characters to a class. The character values in the arguments will be within the
4171    GETCHARTEST(c, previous);  valid values for the current mode (8-bit, 16-bit, UTF, etc). This function is
4172  #else  mutually recursive with the function immediately below.
   c = *previous;  
 #endif  
   if (c == next) return TRUE;  
 #ifdef SUPPORT_UTF  
   if (utf)  
     {  
     unsigned int othercase;  
     if (next < 128) othercase = cd->fcc[next]; else  
 #ifdef SUPPORT_UCP  
     othercase = UCD_OTHERCASE((unsigned int)next);  
 #else  
     othercase = NOTACHAR;  
 #endif  
     return (unsigned int)c == othercase;  
     }  
   else  
 #endif  /* SUPPORT_UTF */  
   return (c == TABLE_GET((unsigned int)next, cd->fcc, next));  /* Non-UTF-8 mode */  
4173    
4174    /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.  Arguments:
4175    When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */    classbits     the bit map for characters < 256
4176      uchardptr     points to the pointer for extra data
4177      options       the options word
4178      cd            contains pointers to tables etc.
4179      start         start of range character
4180      end           end of range character
4181    
4182    case OP_DIGIT:  Returns:        the number of < 256 characters added
4183    return next > 255 || (cd->ctypes[next] & ctype_digit) == 0;                  the pointer to extra data is updated
4184    */
4185    
4186    case OP_NOT_DIGIT:  static int
4187    return next <= 255 && (cd->ctypes[next] & ctype_digit) != 0;  add_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
4188      compile_data *cd, pcre_uint32 start, pcre_uint32 end)
4189    {
4190    pcre_uint32 c;
4191    pcre_uint32 classbits_end = (end <= 0xff ? end : 0xff);
4192    int n8 = 0;
4193    
4194    /* If caseless matching is required, scan the range and process alternate
4195    cases. In Unicode, there are 8-bit characters that have alternate cases that
4196    are greater than 255 and vice-versa. Sometimes we can just extend the original
4197    range. */
4198    
4199    case OP_WHITESPACE:  if ((options & PCRE_CASELESS) != 0)
4200    return next > 255 || (cd->ctypes[next] & ctype_space) == 0;    {
4201    #ifdef SUPPORT_UCP
4202      if ((options & PCRE_UTF8) != 0)
4203        {
4204        int rc;
4205        pcre_uint32 oc, od;
4206    
4207    case OP_NOT_WHITESPACE:      options &= ~PCRE_CASELESS;   /* Remove for recursive calls */
4208    return next <= 255 && (cd->ctypes[next] & ctype_space) != 0;      c = start;
4209    
4210    case OP_WORDCHAR:      while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)
4211    return next > 255 || (cd->ctypes[next] & ctype_word) == 0;        {
4212          /* Handle a single character that has more than one other case. */
4213    
4214    case OP_NOT_WORDCHAR:        if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,
4215    return next <= 255 && (cd->ctypes[next] & ctype_word) != 0;          PRIV(ucd_caseless_sets) + rc, oc);
4216    
4217    case OP_HSPACE:        /* Do nothing if the other case range is within the original range. */
   case OP_NOT_HSPACE:  
   switch(next)  
     {  
     case 0x09:  
     case 0x20:  
     case 0xa0:  
     case 0x1680:  
     case 0x180e:  
     case 0x2000:  
     case 0x2001: