/[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 149 by ph10, Mon Apr 16 15:28:08 2007 UTC revision 1394 by ph10, Sat Nov 9 09:17:20 2013 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2007 University of Cambridge             Copyright (c) 1997-2013 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 42  POSSIBILITY OF SUCH DAMAGE. Line 42  POSSIBILITY OF SUCH DAMAGE.
42  supporting internal functions that are not used by other modules. */  supporting internal functions that are not used by other modules. */
43    
44    
45    #ifdef HAVE_CONFIG_H
46    #include "config.h"
47    #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 processed string start */
51  #define PSEND   end_pattern    /* Field containing processed string end */  #define PSEND   end_pattern    /* Field containing processed string end */
52    
   
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56  /* When DEBUG is defined, we need the pcre_printint() function, which is also  /* When PCRE_DEBUG is defined, we need the pcre(16|32)_printint() function, which
57  used by pcretest. DEBUG is not defined when building a production library. */  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
59    COMPILE_PCREx macro will already be appropriately set. */
60    
61  #ifdef DEBUG  #ifdef PCRE_DEBUG
62  #include "pcre_printint.src"  /* pcre_printint.c should not include any headers */
63    #define PCRE_INCLUDED
64    #include "pcre_printint.c"
65    #undef PCRE_INCLUDED
66  #endif  #endif
67    
68    
69    /* Macro for setting individual bits in class bitmaps. */
70    
71    #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
74    holds the compiled pattern length does not overflow. We make it a bit less than
75    INT_MAX to allow for adding in group terminating bytes, so that we don't have
76    to check them every time. */
77    
78    #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         *
95  *************************************************/  *************************************************/
# Line 72  so this number is very generous. Line 104  so this number is very generous.
104  The same workspace is used during the second, actual compile phase for  The same workspace is used during the second, actual compile phase for
105  remembering forward references to groups so that they can be filled in at the  remembering forward references to groups so that they can be filled in at the
106  end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE  end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE
107  is 4 there is plenty of room. */  is 4 there is plenty of room for most patterns. However, the memory can get
108    filled up by repetitions of forward references, for example patterns like
109    /(?1){0,1999}(b)/, and one user did hit the limit. The code has been changed so
110    that the workspace is expanded using malloc() in this situation. The value
111    below is therefore a minimum, and we put a maximum on it for safety. The
112    minimum is now also defined in terms of LINK_SIZE so that the use of malloc()
113    kicks in at the same number of forward references in all cases. */
114    
115    #define COMPILE_WORK_SIZE (2048*LINK_SIZE)
116    #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 COMPILE_WORK_SIZE (4096)  #define NAMED_GROUP_LIST_SIZE  20
124    
125    /* 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. */
127    
128    #define WORK_SIZE_SAFETY_MARGIN (100)
129    
130    /* Private flags added to firstchar and reqchar. */
131    
132    #define REQ_CASELESS    (1 << 0)        /* Indicates caselessness */
133    #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. */
139    
140    #define UTF_LENGTH     0x10000000l      /* The char contains its length. */
141    
142  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
143  are simple data values; negative values are for special things like \d and so  are simple data values; negative values are for special things like \d and so
144  on. Zero means further processing is needed (for things like \x), or the escape  on. Zero means further processing is needed (for things like \x), or the escape
145  is invalid. */  is invalid. */
146    
147  #ifndef EBCDIC  /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC
148    
149    /* This is the "normal" table for ASCII systems or for EBCDIC systems running
150    in UTF-8 mode. */
151    
152  static const short int escapes[] = {  static const short int escapes[] = {
153       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,                       0,
154       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,                       0,
155     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */       0,                       0,
156       0,      0,      0,      0,      0,      0,      0,      0,   /* H - O */       0,                       0,
157  -ESC_P, -ESC_Q, -ESC_R, -ESC_S,      0,      0,      0, -ESC_W,   /* P - W */       0,                       0,
158  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */       CHAR_COLON,              CHAR_SEMICOLON,
159     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
160       0,      0,      0, -ESC_k,      0,      0,  ESC_n,      0,   /* h - o */       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
161  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0,      0, -ESC_w,   /* p - w */       CHAR_COMMERCIAL_AT,      -ESC_A,
162       0,      0, -ESC_z                                            /* x - z */       -ESC_B,                  -ESC_C,
163         -ESC_D,                  -ESC_E,
164         0,                       -ESC_G,
165         -ESC_H,                  0,
166         0,                       -ESC_K,
167         0,                       0,
168         -ESC_N,                  0,
169         -ESC_P,                  -ESC_Q,
170         -ESC_R,                  -ESC_S,
171         0,                       0,
172         -ESC_V,                  -ESC_W,
173         -ESC_X,                  0,
174         -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,
175         CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,
176         CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,
177         CHAR_GRAVE_ACCENT,       7,
178         -ESC_b,                  0,
179         -ESC_d,                  ESC_e,
180         ESC_f,                   0,
181         -ESC_h,                  0,
182         0,                       -ESC_k,
183         0,                       0,
184         ESC_n,                   0,
185         -ESC_p,                  0,
186         ESC_r,                   -ESC_s,
187         ESC_tee,                 0,
188         -ESC_v,                  -ESC_w,
189         0,                       0,
190         -ESC_z
191  };  };
192    
193  #else           /* This is the "abnormal" table for EBCDIC systems */  #else
194    
195    /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
196    
197  static const short int escapes[] = {  static const short int escapes[] = {
198  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
199  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 106  static const short int escapes[] = { Line 203  static const short int escapes[] = {
203  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,
204  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',
205  /*  80 */     0,     7, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,  /*  80 */     0,     7, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,
206  /*  88 */     0,     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,     'l',      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,     0, -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,
211  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,
212  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
213  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,
214  /*  C8 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
215  /*  D0 */   '}',     0,      0,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,-ESC_N,      0, -ESC_P,
216  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
217  /*  E0 */  '\\',     0, -ESC_S,       0,      0,     0, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
218  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
219  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,
220  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0
# Line 125  static const short int escapes[] = { Line 222  static const short int escapes[] = {
222  #endif  #endif
223    
224    
225  /* Tables of names of POSIX character classes and their lengths. The list is  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
226  terminated by a zero length entry. The first three must be alpha, lower, upper,  searched linearly. Put all the names into a single string, in order to reduce
227  as this is assumed for handling case independence. */  the number of relocations when a shared library is dynamically linked. The
228    string is built from string macros so that it works in UTF-8 mode on EBCDIC
229  static const char *const posix_names[] = {  platforms. */
230    "alpha", "lower", "upper",  
231    "alnum", "ascii", "blank", "cntrl", "digit", "graph",  typedef struct verbitem {
232    "print", "punct", "space", "word",  "xdigit" };    int   len;                 /* Length of verb name */
233      int   op;                  /* Op when no arg, or -1 if arg mandatory */
234      int   op_arg;              /* Op when arg present, or -1 if not allowed */
235    } verbitem;
236    
237    static const char verbnames[] =
238      "\0"                       /* Empty name is a shorthand for MARK */
239      STRING_MARK0
240      STRING_ACCEPT0
241      STRING_COMMIT0
242      STRING_F0
243      STRING_FAIL0
244      STRING_PRUNE0
245      STRING_SKIP0
246      STRING_THEN;
247    
248    static const verbitem verbs[] = {
249      { 0, -1,        OP_MARK },
250      { 4, -1,        OP_MARK },
251      { 6, OP_ACCEPT, -1 },
252      { 6, OP_COMMIT, -1 },
253      { 1, OP_FAIL,   -1 },
254      { 4, OP_FAIL,   -1 },
255      { 5, OP_PRUNE,  OP_PRUNE_ARG },
256      { 4, OP_SKIP,   OP_SKIP_ARG  },
257      { 4, OP_THEN,   OP_THEN_ARG  }
258    };
259    
260    static const int verbcount = sizeof(verbs)/sizeof(verbitem);
261    
262    
263    /* Tables of names of POSIX character classes and their lengths. The names are
264    now all in a single string, to reduce the number of relocations when a shared
265    library is dynamically loaded. The list of lengths is terminated by a zero
266    length entry. The first three must be alpha, lower, upper, as this is assumed
267    for handling case independence. The indices for graph, print, and punct are
268    needed, so identify them. */
269    
270    static const char posix_names[] =
271      STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
272      STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
273      STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
274      STRING_word0  STRING_xdigit;
275    
276  static const uschar posix_name_lengths[] = {  static const pcre_uint8 posix_name_lengths[] = {
277    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
278    
279    #define PC_GRAPH  8
280    #define PC_PRINT  9
281    #define PC_PUNCT 10
282    
283    
284  /* Table of class bit maps for each POSIX class. Each class is formed from a  /* Table of class bit maps for each POSIX class. Each class is formed from a
285  base map, with an optional addition or removal of another map. Then, for some  base map, with an optional addition or removal of another map. Then, for some
286  classes, there is some additional tweaking: for [:blank:] the vertical space  classes, there is some additional tweaking: for [:blank:] the vertical space
# Line 164  static const int posix_class_maps[] = { Line 308  static const int posix_class_maps[] = {
308    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
309  };  };
310    
311    /* Table of substitutes for \d etc when PCRE_UCP is set. They are replaced by
312    Unicode property escapes. */
313    
314    #ifdef SUPPORT_UCP
315    static const pcre_uchar string_PNd[]  = {
316      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
317      CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
318    static const pcre_uchar string_pNd[]  = {
319      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
320      CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
321    static const pcre_uchar string_PXsp[] = {
322      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
323      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
324    static const pcre_uchar string_pXsp[] = {
325      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
326      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
327    static const pcre_uchar string_PXwd[] = {
328      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
329      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
330    static const pcre_uchar string_pXwd[] = {
331      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
332      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
333    
334    static const pcre_uchar *substitutes[] = {
335      string_PNd,           /* \D */
336      string_pNd,           /* \d */
337      string_PXsp,          /* \S */   /* Xsp is Perl space, but from 8.34, Perl */
338      string_pXsp,          /* \s */   /* space and POSIX space are the same. */
339      string_PXwd,          /* \W */
340      string_pXwd           /* \w */
341    };
342    
343    /* The POSIX class substitutes must be in the order of the POSIX class names,
344    defined above, and there are both positive and negative cases. NULL means no
345    general substitute of a Unicode property escape (\p or \P). However, for some
346    POSIX classes (e.g. graph, print, punct) a special property code is compiled
347    directly. */
348    
349    static const pcre_uchar string_pL[] =   {
350      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
351      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
352    static const pcre_uchar string_pLl[] =  {
353      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
354      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
355    static const pcre_uchar string_pLu[] =  {
356      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
357      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
358    static const pcre_uchar string_pXan[] = {
359      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
360      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
361    static const pcre_uchar string_h[] =    {
362      CHAR_BACKSLASH, CHAR_h, '\0' };
363    static const pcre_uchar string_pXps[] = {
364      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
365      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
366    static const pcre_uchar string_PL[] =   {
367      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
368      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
369    static const pcre_uchar string_PLl[] =  {
370      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
371      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
372    static const pcre_uchar string_PLu[] =  {
373      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
374      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
375    static const pcre_uchar string_PXan[] = {
376      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
377      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
378    static const pcre_uchar string_H[] =    {
379      CHAR_BACKSLASH, CHAR_H, '\0' };
380    static const pcre_uchar string_PXps[] = {
381      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
382      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
383    
384    static const pcre_uchar *posix_substitutes[] = {
385      string_pL,            /* alpha */
386      string_pLl,           /* lower */
387      string_pLu,           /* upper */
388      string_pXan,          /* alnum */
389      NULL,                 /* ascii */
390      string_h,             /* blank */
391      NULL,                 /* cntrl */
392      string_pNd,           /* digit */
393      NULL,                 /* graph */
394      NULL,                 /* print */
395      NULL,                 /* punct */
396      string_pXps,          /* space */   /* Xps is POSIX space, but from 8.34 */
397      string_pXwd,          /* word  */   /* Perl and POSIX space are the same */
398      NULL,                 /* xdigit */
399      /* Negated cases */
400      string_PL,            /* ^alpha */
401      string_PLl,           /* ^lower */
402      string_PLu,           /* ^upper */
403      string_PXan,          /* ^alnum */
404      NULL,                 /* ^ascii */
405      string_H,             /* ^blank */
406      NULL,                 /* ^cntrl */
407      string_PNd,           /* ^digit */
408      NULL,                 /* ^graph */
409      NULL,                 /* ^print */
410      NULL,                 /* ^punct */
411      string_PXps,          /* ^space */  /* Xps is POSIX space, but from 8.34 */
412      string_PXwd,          /* ^word */   /* Perl and POSIX space are the same */
413      NULL                  /* ^xdigit */
414    };
415    #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))
416    #endif
417    
418  #define STRING(a)  # a  #define STRING(a)  # a
419  #define XSTRING(s) STRING(s)  #define XSTRING(s) STRING(s)
# Line 171  static const int posix_class_maps[] = { Line 421  static const int posix_class_maps[] = {
421  /* The texts of compile-time error messages. These are "char *" because they  /* The texts of compile-time error messages. These are "char *" because they
422  are passed to the outside world. Do not ever re-use any error number, because  are passed to the outside world. Do not ever re-use any error number, because
423  they are documented. Always add a new error instead. Messages marked DEAD below  they are documented. Always add a new error instead. Messages marked DEAD below
424  are no longer used. */  are no longer used. This used to be a table of strings, but in order to reduce
425    the number of relocations needed when a shared library is loaded dynamically,
426  static const char *error_texts[] = {  it is now one long string. We cannot use a table of offsets, because the
427    "no error",  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
428    "\\ at end of pattern",  simply count through to the one we want - this isn't a performance issue
429    "\\c at end of pattern",  because these strings are used only when there is a compilation error.
430    "unrecognized character follows \\",  
431    "numbers out of order in {} quantifier",  Each substring ends with \0 to insert a null character. This includes the final
432    substring, so that the whole string ends with \0\0, which can be detected when
433    counting through. */
434    
435    static const char error_texts[] =
436      "no error\0"
437      "\\ at end of pattern\0"
438      "\\c at end of pattern\0"
439      "unrecognized character follows \\\0"
440      "numbers out of order in {} quantifier\0"
441    /* 5 */    /* 5 */
442    "number too big in {} quantifier",    "number too big in {} quantifier\0"
443    "missing terminating ] for character class",    "missing terminating ] for character class\0"
444    "invalid escape sequence in character class",    "invalid escape sequence in character class\0"
445    "range out of order in character class",    "range out of order in character class\0"
446    "nothing to repeat",    "nothing to repeat\0"
447    /* 10 */    /* 10 */
448    "operand of unlimited repeat could match the empty string",  /** DEAD **/    "operand of unlimited repeat could match the empty string\0"  /** DEAD **/
449    "internal error: unexpected repeat",    "internal error: unexpected repeat\0"
450    "unrecognized character after (?",    "unrecognized character after (? or (?-\0"
451    "POSIX named classes are supported only within a class",    "POSIX named classes are supported only within a class\0"
452    "missing )",    "missing )\0"
453    /* 15 */    /* 15 */
454    "reference to non-existent subpattern",    "reference to non-existent subpattern\0"
455    "erroffset passed as NULL",    "erroffset passed as NULL\0"
456    "unknown option bit(s) set",    "unknown option bit(s) set\0"
457    "missing ) after comment",    "missing ) after comment\0"
458    "parentheses nested too deeply",  /** DEAD **/    "parentheses nested too deeply\0"  /** DEAD **/
459    /* 20 */    /* 20 */
460    "regular expression too large",    "regular expression is too large\0"
461    "failed to get memory",    "failed to get memory\0"
462    "unmatched parentheses",    "unmatched parentheses\0"
463    "internal error: code overflow",    "internal error: code overflow\0"
464    "unrecognized character after (?<",    "unrecognized character after (?<\0"
465    /* 25 */    /* 25 */
466    "lookbehind assertion is not fixed length",    "lookbehind assertion is not fixed length\0"
467    "malformed number or name after (?(",    "malformed number or name after (?(\0"
468    "conditional group contains more than two branches",    "conditional group contains more than two branches\0"
469    "assertion expected after (?(",    "assertion expected after (?(\0"
470    "(?R or (?digits must be followed by )",    "(?R or (?[+-]digits must be followed by )\0"
471    /* 30 */    /* 30 */
472    "unknown POSIX class name",    "unknown POSIX class name\0"
473    "POSIX collating elements are not supported",    "POSIX collating elements are not supported\0"
474    "this version of PCRE is not compiled with PCRE_UTF8 support",    "this version of PCRE is compiled without UTF support\0"
475    "spare error",  /** DEAD **/    "spare error\0"  /** DEAD **/
476    "character value in \\x{...} sequence is too large",    "character value in \\x{} or \\o{} is too large\0"
477    /* 35 */    /* 35 */
478    "invalid condition (?(0)",    "invalid condition (?(0)\0"
479    "\\C not allowed in lookbehind assertion",    "\\C not allowed in lookbehind assertion\0"
480    "PCRE does not support \\L, \\l, \\N, \\U, or \\u",    "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
481    "number after (?C is > 255",    "number after (?C is > 255\0"
482    "closing ) for (?C expected",    "closing ) for (?C expected\0"
483    /* 40 */    /* 40 */
484    "recursive call could loop indefinitely",    "recursive call could loop indefinitely\0"
485    "unrecognized character after (?P",    "unrecognized character after (?P\0"
486    "syntax error in subpattern name (missing terminator)",    "syntax error in subpattern name (missing terminator)\0"
487    "two named subpatterns have the same name",    "two named subpatterns have the same name\0"
488    "invalid UTF-8 string",    "invalid UTF-8 string\0"
489    /* 45 */    /* 45 */
490    "support for \\P, \\p, and \\X has not been compiled",    "support for \\P, \\p, and \\X has not been compiled\0"
491    "malformed \\P or \\p sequence",    "malformed \\P or \\p sequence\0"
492    "unknown property name after \\P or \\p",    "unknown property name after \\P or \\p\0"
493    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)",    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
494    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")",    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
495    /* 50 */    /* 50 */
496    "repeated subpattern is too long",    "repeated subpattern is too long\0"    /** DEAD **/
497    "octal value is greater than \\377 (not in UTF-8 mode)",    "octal value is greater than \\377 in 8-bit non-UTF-8 mode\0"
498    "internal error: overran compiling workspace",    "internal error: overran compiling workspace\0"
499    "internal error: previously-checked referenced subpattern not found",    "internal error: previously-checked referenced subpattern not found\0"
500    "DEFINE group contains more than one branch",    "DEFINE group contains more than one branch\0"
501    /* 55 */    /* 55 */
502    "repeating a DEFINE group is not allowed",    "repeating a DEFINE group is not allowed\0"  /** DEAD **/
503    "inconsistent NEWLINE options",    "inconsistent NEWLINE options\0"
504    "\\g is not followed by an (optionally braced) non-zero number"    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
505  };    "a numbered reference must not be zero\0"
506      "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
507      /* 60 */
508      "(*VERB) not recognized or malformed\0"
509      "number is too big\0"
510      "subpattern name expected\0"
511      "digit expected after (?+\0"
512      "] is an invalid data character in JavaScript compatibility mode\0"
513      /* 65 */
514      "different names for subpatterns of the same number are not allowed\0"
515      "(*MARK) must have an argument\0"
516      "this version of PCRE is not compiled with Unicode property support\0"
517      "\\c must be followed by an ASCII character\0"
518      "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
519      /* 70 */
520      "internal error: unknown opcode in find_fixedlength()\0"
521      "\\N is not supported in a class\0"
522      "too many forward references\0"
523      "disallowed Unicode code point (>= 0xd800 && <= 0xdfff)\0"
524      "invalid UTF-16 string\0"
525      /* 75 */
526      "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"
527      "character value in \\u.... sequence is too large\0"
528      "invalid UTF-32 string\0"
529      "setting UTF is disabled by the application\0"
530      "non-hex character in \\x{} (closing brace missing?)\0"
531      /* 80 */
532      "non-octal character in \\o{} (closing brace missing?)\0"
533      "missing opening brace after \\o\0"
534      "parentheses are too deeply nested\0"
535      "invalid range in character class\0"
536      "group name must start with a non-digit\0"
537      ;
538    
539  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
540  patterns. Note that the tables in chartables are dependent on the locale, and  patterns. Note that the tables in chartables are dependent on the locale, and
# Line 262  For convenience, we use the same bit def Line 552  For convenience, we use the same bit def
552    
553  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
554    
555  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */  /* Using a simple comparison for decimal numbers rather than a memory read
556  static const unsigned char digitab[] =  is much faster, and the resulting code is simpler (the compiler turns it
557    into a subtraction and unsigned comparison). */
558    
559    #define IS_DIGIT(x) ((x) >= CHAR_0 && (x) <= CHAR_9)
560    
561    #ifndef EBCDIC
562    
563    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
564    UTF-8 mode. */
565    
566    static const pcre_uint8 digitab[] =
567    {    {
568    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
569    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15 */
# Line 298  static const unsigned char digitab[] = Line 598  static const unsigned char digitab[] =
598    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
599    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
600    
601  #else           /* This is the "abnormal" case, for EBCDIC systems */  #else
602  static const unsigned char digitab[] =  
603    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
604    
605    static const pcre_uint8 digitab[] =
606    {    {
607    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
608    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15    */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15    */
# Line 334  static const unsigned char digitab[] = Line 637  static const unsigned char digitab[] =
637    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */
638    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */
639    
640  static const unsigned char ebcdic_chartab[] = { /* chartable partial dup */  static const pcre_uint8 ebcdic_chartab[] = { /* chartable partial dup */
641    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */
642    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */
643    0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*  16- 23 */    0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*  16- 23 */
# Line 370  static const unsigned char ebcdic_charta Line 673  static const unsigned char ebcdic_charta
673  #endif  #endif
674    
675    
676  /* Definition to allow mutual recursion */  /* This table is used to check whether auto-possessification is possible
677    between adjacent character-type opcodes. The left-hand (repeated) opcode is
678    used to select the row, and the right-hand opcode is use to select the column.
679    A value of 1 means that auto-possessification is OK. For example, the second
680    value in the first row means that \D+\d can be turned into \D++\d.
681    
682    The Unicode property types (\P and \p) have to be present to fill out the table
683    because of what their opcode values are, but the table values should always be
684    zero because property types are handled separately in the code. The last four
685    columns apply to items that cannot be repeated, so there is no need to have
686    rows for them. Note that OP_DIGIT etc. are generated only when PCRE_UCP is
687    *not* set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
688    
689    #define APTROWS (LAST_AUTOTAB_LEFT_OP - FIRST_AUTOTAB_OP + 1)
690    #define APTCOLS (LAST_AUTOTAB_RIGHT_OP - FIRST_AUTOTAB_OP + 1)
691    
692    static const pcre_uint8 autoposstab[APTROWS][APTCOLS] = {
693    /* \D \d \S \s \W \w  . .+ \C \P \p \R \H \h \V \v \X \Z \z  $ $M */
694      { 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \D */
695      { 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \d */
696      { 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \S */
697      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \s */
698      { 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \W */
699      { 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \w */
700      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .  */
701      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .+ */
702      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \C */
703      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \P */
704      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \p */
705      { 0, 1, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \R */
706      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \H */
707      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \h */
708      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \V */
709      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0 },  /* \v */
710      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }   /* \X */
711    };
712    
713    
714    /* This table is used to check whether auto-possessification is possible
715    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP). The
716    left-hand (repeated) opcode is used to select the row, and the right-hand
717    opcode is used to select the column. The values are as follows:
718    
719      0   Always return FALSE (never auto-possessify)
720      1   Character groups are distinct (possessify if both are OP_PROP)
721      2   Check character categories in the same group (general or particular)
722      3   TRUE if the two opcodes are not the same (PROP vs NOTPROP)
723    
724      4   Check left general category vs right particular category
725      5   Check right general category vs left particular category
726    
727      6   Left alphanum vs right general category
728      7   Left space vs right general category
729      8   Left word vs right general category
730    
731      9   Right alphanum vs left general category
732     10   Right space vs left general category
733     11   Right word vs left general category
734    
735     12   Left alphanum vs right particular category
736     13   Left space vs right particular category
737     14   Left word vs right particular category
738    
739     15   Right alphanum vs left particular category
740     16   Right space vs left particular category
741     17   Right word vs left particular category
742    */
743    
744    static const pcre_uint8 propposstab[PT_TABSIZE][PT_TABSIZE] = {
745    /* ANY LAMP GC  PC  SC ALNUM SPACE PXSPACE WORD CLIST UCNC */
746      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_ANY */
747      { 0,  3,  0,  0,  0,    3,    1,      1,   0,    0,   0 },  /* PT_LAMP */
748      { 0,  0,  2,  4,  0,    9,   10,     10,  11,    0,   0 },  /* PT_GC */
749      { 0,  0,  5,  2,  0,   15,   16,     16,  17,    0,   0 },  /* PT_PC */
750      { 0,  0,  0,  0,  2,    0,    0,      0,   0,    0,   0 },  /* PT_SC */
751      { 0,  3,  6, 12,  0,    3,    1,      1,   0,    0,   0 },  /* PT_ALNUM */
752      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_SPACE */
753      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_PXSPACE */
754      { 0,  0,  8, 14,  0,    0,    1,      1,   3,    0,   0 },  /* PT_WORD */
755      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_CLIST */
756      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   3 }   /* PT_UCNC */
757    };
758    
759    /* This table is used to check whether auto-possessification is possible
760    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP) when one
761    specifies a general category and the other specifies a particular category. The
762    row is selected by the general category and the column by the particular
763    category. The value is 1 if the particular category is not part of the general
764    category. */
765    
766    static const pcre_uint8 catposstab[7][30] = {
767    /* Cc Cf Cn Co Cs Ll Lm Lo Lt Lu Mc Me Mn Nd Nl No Pc Pd Pe Pf Pi Po Ps Sc Sk Sm So Zl Zp Zs */
768      { 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* C */
769      { 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* L */
770      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* M */
771      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* N */
772      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1 },  /* P */
773      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 1, 1, 1 },  /* S */
774      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0 }   /* Z */
775    };
776    
777    /* This table is used when checking ALNUM, (PX)SPACE, SPACE, and WORD against
778    a general or particular category. The properties in each row are those
779    that apply to the character set in question. Duplication means that a little
780    unnecessary work is done when checking, but this keeps things much simpler
781    because they can all use the same code. For more details see the comment where
782    this table is used.
783    
784    Note: SPACE and PXSPACE used to be different because Perl excluded VT from
785    "space", but from Perl 5.18 it's included, so both categories are treated the
786    same here. */
787    
788    static const pcre_uint8 posspropstab[3][4] = {
789      { ucp_L, ucp_N, ucp_N, ucp_Nl },  /* ALNUM, 3rd and 4th values redundant */
790      { ucp_Z, ucp_Z, ucp_C, ucp_Cc },  /* SPACE and PXSPACE, 2nd value redundant */
791      { ucp_L, ucp_N, ucp_P, ucp_Po }   /* WORD */
792    };
793    
794    /* This table is used when converting repeating opcodes into possessified
795    versions as a result of an explicit possessive quantifier such as ++. A zero
796    value means there is no possessified version - in those cases the item in
797    question must be wrapped in ONCE brackets. The table is truncated at OP_CALLOUT
798    because all relevant opcodes are less than that. */
799    
800    static const pcre_uint8 opcode_possessify[] = {
801      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 0 - 15  */
802      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 16 - 31 */
803    
804      0,                       /* NOTI */
805      OP_POSSTAR, 0,           /* STAR, MINSTAR */
806      OP_POSPLUS, 0,           /* PLUS, MINPLUS */
807      OP_POSQUERY, 0,          /* QUERY, MINQUERY */
808      OP_POSUPTO, 0,           /* UPTO, MINUPTO */
809      0,                       /* EXACT */
810      0, 0, 0, 0,              /* POS{STAR,PLUS,QUERY,UPTO} */
811    
812      OP_POSSTARI, 0,          /* STARI, MINSTARI */
813      OP_POSPLUSI, 0,          /* PLUSI, MINPLUSI */
814      OP_POSQUERYI, 0,         /* QUERYI, MINQUERYI */
815      OP_POSUPTOI, 0,          /* UPTOI, MINUPTOI */
816      0,                       /* EXACTI */
817      0, 0, 0, 0,              /* POS{STARI,PLUSI,QUERYI,UPTOI} */
818    
819      OP_NOTPOSSTAR, 0,        /* NOTSTAR, NOTMINSTAR */
820      OP_NOTPOSPLUS, 0,        /* NOTPLUS, NOTMINPLUS */
821      OP_NOTPOSQUERY, 0,       /* NOTQUERY, NOTMINQUERY */
822      OP_NOTPOSUPTO, 0,        /* NOTUPTO, NOTMINUPTO */
823      0,                       /* NOTEXACT */
824      0, 0, 0, 0,              /* NOTPOS{STAR,PLUS,QUERY,UPTO} */
825    
826      OP_NOTPOSSTARI, 0,       /* NOTSTARI, NOTMINSTARI */
827      OP_NOTPOSPLUSI, 0,       /* NOTPLUSI, NOTMINPLUSI */
828      OP_NOTPOSQUERYI, 0,      /* NOTQUERYI, NOTMINQUERYI */
829      OP_NOTPOSUPTOI, 0,       /* NOTUPTOI, NOTMINUPTOI */
830      0,                       /* NOTEXACTI */
831      0, 0, 0, 0,              /* NOTPOS{STARI,PLUSI,QUERYI,UPTOI} */
832    
833      OP_TYPEPOSSTAR, 0,       /* TYPESTAR, TYPEMINSTAR */
834      OP_TYPEPOSPLUS, 0,       /* TYPEPLUS, TYPEMINPLUS */
835      OP_TYPEPOSQUERY, 0,      /* TYPEQUERY, TYPEMINQUERY */
836      OP_TYPEPOSUPTO, 0,       /* TYPEUPTO, TYPEMINUPTO */
837      0,                       /* TYPEEXACT */
838      0, 0, 0, 0,              /* TYPEPOS{STAR,PLUS,QUERY,UPTO} */
839    
840      OP_CRPOSSTAR, 0,         /* CRSTAR, CRMINSTAR */
841      OP_CRPOSPLUS, 0,         /* CRPLUS, CRMINPLUS */
842      OP_CRPOSQUERY, 0,        /* CRQUERY, CRMINQUERY */
843      OP_CRPOSRANGE, 0,        /* CRRANGE, CRMINRANGE */
844      0, 0, 0, 0,              /* CRPOS{STAR,PLUS,QUERY,RANGE} */
845    
846      0, 0, 0,                 /* CLASS, NCLASS, XCLASS */
847      0, 0,                    /* REF, REFI */
848      0, 0,                    /* DNREF, DNREFI */
849      0, 0                     /* RECURSE, CALLOUT */
850    };
851    
852    
853    
854    /*************************************************
855    *            Find an error text                  *
856    *************************************************/
857    
858    /* The error texts are now all in one long string, to save on relocations. As
859    some of the text is of unknown length, we can't use a table of offsets.
860    Instead, just count through the strings. This is not a performance issue
861    because it happens only when there has been a compilation error.
862    
863    Argument:   the error number
864    Returns:    pointer to the error string
865    */
866    
867    static const char *
868    find_error_text(int n)
869    {
870    const char *s = error_texts;
871    for (; n > 0; n--)
872      {
873      while (*s++ != CHAR_NULL) {};
874      if (*s == CHAR_NULL) return "Error text not found (please report)";
875      }
876    return s;
877    }
878    
879    
880    
881    /*************************************************
882    *           Expand the workspace                 *
883    *************************************************/
884    
885    /* This function is called during the second compiling phase, if the number of
886    forward references fills the existing workspace, which is originally a block on
887    the stack. A larger block is obtained from malloc() unless the ultimate limit
888    has been reached or the increase will be rather small.
889    
890    Argument: pointer to the compile data block
891    Returns:  0 if all went well, else an error number
892    */
893    
894    static int
895    expand_workspace(compile_data *cd)
896    {
897    pcre_uchar *newspace;
898    int newsize = cd->workspace_size * 2;
899    
900    if (newsize > COMPILE_WORK_SIZE_MAX) newsize = COMPILE_WORK_SIZE_MAX;
901    if (cd->workspace_size >= COMPILE_WORK_SIZE_MAX ||
902        newsize - cd->workspace_size < WORK_SIZE_SAFETY_MARGIN)
903     return ERR72;
904    
905    newspace = (PUBL(malloc))(IN_UCHARS(newsize));
906    if (newspace == NULL) return ERR21;
907    memcpy(newspace, cd->start_workspace, cd->workspace_size * sizeof(pcre_uchar));
908    cd->hwm = (pcre_uchar *)newspace + (cd->hwm - cd->start_workspace);
909    if (cd->workspace_size > COMPILE_WORK_SIZE)
910      (PUBL(free))((void *)cd->start_workspace);
911    cd->start_workspace = newspace;
912    cd->workspace_size = newsize;
913    return 0;
914    }
915    
916    
917    
918    /*************************************************
919    *            Check for counted repeat            *
920    *************************************************/
921    
922    /* This function is called when a '{' is encountered in a place where it might
923    start a quantifier. It looks ahead to see if it really is a quantifier or not.
924    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
925    where the ddds are digits.
926    
927    Arguments:
928      p         pointer to the first char after '{'
929    
930    Returns:    TRUE or FALSE
931    */
932    
933  static BOOL  static BOOL
934    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, int, int *,  is_counted_repeat(const pcre_uchar *p)
935      int *, branch_chain *, compile_data *, int *);  {
936    if (!IS_DIGIT(*p)) return FALSE;
937    p++;
938    while (IS_DIGIT(*p)) p++;
939    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
940    
941    if (*p++ != CHAR_COMMA) return FALSE;
942    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
943    
944    if (!IS_DIGIT(*p)) return FALSE;
945    p++;
946    while (IS_DIGIT(*p)) p++;
947    
948    return (*p == CHAR_RIGHT_CURLY_BRACKET);
949    }
950    
951    
952    
# Line 383  static BOOL Line 955  static BOOL
955  *************************************************/  *************************************************/
956    
957  /* This function is called when a \ has been encountered. It either returns a  /* This function is called when a \ has been encountered. It either returns a
958  positive value for a simple escape such as \n, or a negative value which  positive value for a simple escape such as \n, or 0 for a data character which
959  encodes one of the more complicated things such as \d. A backreference to group  will be placed in chptr. A backreference to group n is returned as negative n.
960  n is returned as -(ESC_REF + n); ESC_REF is the highest ESC_xxx macro. When  When UTF-8 is enabled, a positive value greater than 255 may be returned in
961  UTF-8 is enabled, a positive value greater than 255 may be returned. On entry,  chptr. On entry, ptr is pointing at the \. On exit, it is on the final
962  ptr is pointing at the \. On exit, it is on the final character of the escape  character of the escape sequence.
 sequence.  
963    
964  Arguments:  Arguments:
965    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
966      chptr          points to a returned data character
967    errorcodeptr   points to the errorcode variable    errorcodeptr   points to the errorcode variable
968    bracount       number of previous extracting brackets    bracount       number of previous extracting brackets
969    options        the options bits    options        the options bits
970    isclass        TRUE if inside a character class    isclass        TRUE if inside a character class
971    
972  Returns:         zero or positive => a data character  Returns:         zero => a data character
973                   negative => a special escape sequence                   positive => a special escape sequence
974                   on error, errorptr is set                   negative => a back reference
975                     on error, errorcodeptr is set
976  */  */
977    
978  static int  static int
979  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,
980    int options, BOOL isclass)    int bracount, int options, BOOL isclass)
981  {  {
982  BOOL utf8 = (options & PCRE_UTF8) != 0;  /* PCRE_UTF16 has the same value as PCRE_UTF8. */
983  const uschar *ptr = *ptrptr + 1;  BOOL utf = (options & PCRE_UTF8) != 0;
984  int c, i;  const pcre_uchar *ptr = *ptrptr + 1;
985    pcre_uint32 c;
986    int escape = 0;
987    int i;
988    
989  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
990  ptr--;                            /* Set pointer back to the last byte */  ptr--;                            /* Set pointer back to the last byte */
991    
992  /* If backslash is at the end of the pattern, it's an error. */  /* If backslash is at the end of the pattern, it's an error. */
993    
994  if (c == 0) *errorcodeptr = ERR1;  if (c == CHAR_NULL) *errorcodeptr = ERR1;
995    
996  /* Non-alphamerics are literals. For digits or letters, do an initial lookup in  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
997  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.
998  Otherwise further processing may be required. */  Otherwise further processing may be required. */
999    
1000  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1001  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  /* Not alphanumeric */
1002  else if ((i = escapes[c - '0']) != 0) c = i;  else if (c < CHAR_0 || c > CHAR_z) {}
1003    else if ((i = escapes[c - CHAR_0]) != 0)
1004      { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1005    
1006  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1007  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  /* Not alphanumeric */
1008  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if (c < CHAR_a || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}
1009    else if ((i = escapes[c - 0x48]) != 0)  { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1010  #endif  #endif
1011    
1012  /* Escapes that need further processing, or are illegal. */  /* Escapes that need further processing, or are illegal. */
1013    
1014  else  else
1015    {    {
1016    const uschar *oldptr;    const pcre_uchar *oldptr;
1017    BOOL braced, negated;    BOOL braced, negated, overflow;
1018      int s;
1019    
1020    switch (c)    switch (c)
1021      {      {
1022      /* A number of Perl escapes are not handled by PCRE. We give an explicit      /* A number of Perl escapes are not handled by PCRE. We give an explicit
1023      error. */      error. */
1024    
1025      case 'l':      case CHAR_l:
1026      case 'L':      case CHAR_L:
     case 'N':  
     case 'u':  
     case 'U':  
1027      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
1028      break;      break;
1029    
1030      /* \g must be followed by a number, either plain or braced. If positive, it      case CHAR_u:
1031      is an absolute backreference. If negative, it is a relative backreference.      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1032      This is a Perl 5.10 feature. */        {
1033          /* In JavaScript, \u must be followed by four hexadecimal numbers.
1034          Otherwise it is a lowercase u letter. */
1035          if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1036            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0
1037            && MAX_255(ptr[3]) && (digitab[ptr[3]] & ctype_xdigit) != 0
1038            && MAX_255(ptr[4]) && (digitab[ptr[4]] & ctype_xdigit) != 0)
1039            {
1040            c = 0;
1041            for (i = 0; i < 4; ++i)
1042              {
1043              register pcre_uint32 cc = *(++ptr);
1044    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1045              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1046              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1047    #else           /* EBCDIC coding */
1048              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1049              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1050    #endif
1051              }
1052    
1053    #if defined COMPILE_PCRE8
1054            if (c > (utf ? 0x10ffffU : 0xffU))
1055    #elif defined COMPILE_PCRE16
1056            if (c > (utf ? 0x10ffffU : 0xffffU))
1057    #elif defined COMPILE_PCRE32
1058            if (utf && c > 0x10ffffU)
1059    #endif
1060              {
1061              *errorcodeptr = ERR76;
1062              }
1063            else if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1064            }
1065          }
1066        else
1067          *errorcodeptr = ERR37;
1068        break;
1069    
1070        case CHAR_U:
1071        /* In JavaScript, \U is an uppercase U letter. */
1072        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
1073        break;
1074    
1075        /* In a character class, \g is just a literal "g". Outside a character
1076        class, \g must be followed by one of a number of specific things:
1077    
1078        (1) A number, either plain or braced. If positive, it is an absolute
1079        backreference. If negative, it is a relative backreference. This is a Perl
1080        5.10 feature.
1081    
1082        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
1083        is part of Perl's movement towards a unified syntax for back references. As
1084        this is synonymous with \k{name}, we fudge it up by pretending it really
1085        was \k.
1086    
1087        (3) For Oniguruma compatibility we also support \g followed by a name or a
1088        number either in angle brackets or in single quotes. However, these are
1089        (possibly recursive) subroutine calls, _not_ backreferences. Just return
1090        the ESC_g code (cf \k). */
1091    
1092        case CHAR_g:
1093        if (isclass) break;
1094        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
1095          {
1096          escape = ESC_g;
1097          break;
1098          }
1099    
1100        /* Handle the Perl-compatible cases */
1101    
1102      case 'g':      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
     if (ptr[1] == '{')  
1103        {        {
1104          const pcre_uchar *p;
1105          for (p = ptr+2; *p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
1106            if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
1107          if (*p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET)
1108            {
1109            escape = ESC_k;
1110            break;
1111            }
1112        braced = TRUE;        braced = TRUE;
1113        ptr++;        ptr++;
1114        }        }
1115      else braced = FALSE;      else braced = FALSE;
1116    
1117      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
1118        {        {
1119        negated = TRUE;        negated = TRUE;
1120        ptr++;        ptr++;
1121        }        }
1122      else negated = FALSE;      else negated = FALSE;
1123    
1124      c = 0;      /* The integer range is limited by the machine's int representation. */
1125      while ((digitab[ptr[1]] & ctype_digit) != 0)      s = 0;
1126        c = c * 10 + *(++ptr) - '0';      overflow = FALSE;
1127        while (IS_DIGIT(ptr[1]))
1128          {
1129          if (s > INT_MAX / 10 - 1) /* Integer overflow */
1130            {
1131            overflow = TRUE;
1132            break;
1133            }
1134          s = s * 10 + (int)(*(++ptr) - CHAR_0);
1135          }
1136        if (overflow) /* Integer overflow */
1137          {
1138          while (IS_DIGIT(ptr[1]))
1139            ptr++;
1140          *errorcodeptr = ERR61;
1141          break;
1142          }
1143    
1144      if (c == 0 || (braced && *(++ptr) != '}'))      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
1145        {        {
1146        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
1147        return 0;        break;
1148          }
1149    
1150        if (s == 0)
1151          {
1152          *errorcodeptr = ERR58;
1153          break;
1154        }        }
1155    
1156      if (negated)      if (negated)
1157        {        {
1158        if (c > bracount)        if (s > bracount)
1159          {          {
1160          *errorcodeptr = ERR15;          *errorcodeptr = ERR15;
1161          return 0;          break;
1162          }          }
1163        c = bracount - (c - 1);        s = bracount - (s - 1);
1164        }        }
1165    
1166      c = -(ESC_REF + c);      escape = -s;
1167      break;      break;
1168    
1169      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
1170      starting with one that is not zero is not straightforward. By experiment,      starting with one that is not zero is not straightforward. Perl has changed
1171      the way Perl works seems to be as follows:      over the years. Nowadays \g{} for backreferences and \o{} for octal are
1172        recommended to avoid the ambiguities in the old syntax.
1173    
1174      Outside a character class, the digits are read as a decimal number. If the      Outside a character class, the digits are read as a decimal number. If the
1175      number is less than 10, or if there are that many previous extracting      number is less than 8 (used to be 10), or if there are that many previous
1176      left brackets, then it is a back reference. Otherwise, up to three octal      extracting left brackets, then it is a back reference. Otherwise, up to
1177      digits are read to form an escaped byte. Thus \123 is likely to be octal      three octal digits are read to form an escaped byte. Thus \123 is likely to
1178      123 (cf \0123, which is octal 012 followed by the literal 3). If the octal      be octal 123 (cf \0123, which is octal 012 followed by the literal 3). If
1179      value is greater than 377, the least significant 8 bits are taken. Inside a      the octal value is greater than 377, the least significant 8 bits are
1180      character class, \ followed by a digit is always an octal number. */      taken. \8 and \9 are treated as the literal characters 8 and 9.
1181    
1182      case '1': case '2': case '3': case '4': case '5':      Inside a character class, \ followed by a digit is always either a literal
1183      case '6': case '7': case '8': case '9':      8 or 9 or an octal number. */
1184    
1185        case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:
1186        case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
1187    
1188      if (!isclass)      if (!isclass)
1189        {        {
1190        oldptr = ptr;        oldptr = ptr;
1191        c -= '0';        /* The integer range is limited by the machine's int representation. */
1192        while ((digitab[ptr[1]] & ctype_digit) != 0)        s = (int)(c -CHAR_0);
1193          c = c * 10 + *(++ptr) - '0';        overflow = FALSE;
1194        if (c < 10 || c <= bracount)        while (IS_DIGIT(ptr[1]))
1195            {
1196            if (s > INT_MAX / 10 - 1) /* Integer overflow */
1197              {
1198              overflow = TRUE;
1199              break;
1200              }
1201            s = s * 10 + (int)(*(++ptr) - CHAR_0);
1202            }
1203          if (overflow) /* Integer overflow */
1204            {
1205            while (IS_DIGIT(ptr[1]))
1206              ptr++;
1207            *errorcodeptr = ERR61;
1208            break;
1209            }
1210          if (s < 8 || s <= bracount)  /* Check for back reference */
1211          {          {
1212          c = -(ESC_REF + c);          escape = -s;
1213          break;          break;
1214          }          }
1215        ptr = oldptr;      /* Put the pointer back and fall through */        ptr = oldptr;      /* Put the pointer back and fall through */
1216        }        }
1217    
1218      /* Handle an octal number following \. If the first digit is 8 or 9, Perl      /* Handle a digit following \ when the number is not a back reference. If
1219      generates a binary zero byte and treats the digit as a following literal.      the first digit is 8 or 9, Perl used to generate a binary zero byte and
1220      Thus we have to pull back the pointer by one. */      then treat the digit as a following literal. At least by Perl 5.18 this
1221        changed so as not to insert the binary zero. */
1222    
1223      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8) break;
1224        {  
1225        ptr--;      /* Fall through with a digit less than 8 */
       c = 0;  
       break;  
       }  
1226    
1227      /* \0 always starts an octal number, but we may drop through to here with a      /* \0 always starts an octal number, but we may drop through to here with a
1228      larger first octal digit. The original code used just to take the least      larger first octal digit. The original code used just to take the least
1229      significant 8 bits of octal numbers (I think this is what early Perls used      significant 8 bits of octal numbers (I think this is what early Perls used
1230      to do). Nowadays we allow for larger numbers in UTF-8 mode, but no more      to do). Nowadays we allow for larger numbers in UTF-8 mode and 16-bit mode,
1231      than 3 octal digits. */      but no more than 3 octal digits. */
1232    
1233      case '0':      case CHAR_0:
1234      c -= '0';      c -= CHAR_0;
1235      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
1236          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
1237      if (!utf8 && c > 255) *errorcodeptr = ERR51;  #ifdef COMPILE_PCRE8
1238        if (!utf && c > 0xff) *errorcodeptr = ERR51;
1239    #endif
1240      break;      break;
1241    
1242      /* \x is complicated. \x{ddd} is a character number which can be greater      /* \o is a relatively new Perl feature, supporting a more general way of
1243      than 0xff in utf8 mode, but only if the ddd are hex digits. If not, { is      specifying character codes in octal. The only supported form is \o{ddd}. */
     treated as a data character. */  
1244    
1245      case 'x':      case CHAR_o:
1246      if (ptr[1] == '{')      if (ptr[1] != CHAR_LEFT_CURLY_BRACKET) *errorcodeptr = ERR81; else
1247        {        {
1248        const uschar *pt = ptr + 2;        ptr += 2;
       int count = 0;  
   
1249        c = 0;        c = 0;
1250        while ((digitab[*pt] & ctype_xdigit) != 0)        overflow = FALSE;
1251          while (*ptr >= CHAR_0 && *ptr <= CHAR_7)
1252          {          {
1253          register int cc = *pt++;          register pcre_uint32 cc = *ptr++;
1254          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1255          count++;  #ifdef COMPILE_PCRE32
1256            if (c >= 0x20000000l) { overflow = TRUE; break; }
1257  #ifndef EBCDIC  /* ASCII coding */  #endif
1258          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          c = (c << 3) + cc - CHAR_0 ;
1259          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));  #if defined COMPILE_PCRE8
1260  #else           /* EBCDIC coding */          if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1261          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */  #elif defined COMPILE_PCRE16
1262          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1263    #elif defined COMPILE_PCRE32
1264            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1265  #endif  #endif
1266          }          }
1267          if (overflow)
       if (*pt == '}')  
1268          {          {
1269          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          while (*ptr >= CHAR_0 && *ptr <= CHAR_7) ptr++;
1270          ptr = pt;          *errorcodeptr = ERR34;
         break;  
1271          }          }
1272          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1273        /* If the sequence of hex digits does not end with '}', then we don't          {
1274        recognize this construct; fall through to the normal \x handling. */          if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1275            }
1276          else *errorcodeptr = ERR80;
1277        }        }
1278        break;
1279    
1280      /* Read just a single-byte hex-defined char */      /* \x is complicated. In JavaScript, \x must be followed by two hexadecimal
1281        numbers. Otherwise it is a lowercase x letter. */
1282    
1283      c = 0;      case CHAR_x:
1284      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1285        {        {
1286        int cc;                               /* Some compilers don't like ++ */        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1287        cc = *(++ptr);                        /* in initializers */          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1288  #ifndef EBCDIC  /* ASCII coding */          {
1289        if (cc >= 'a') cc -= 32;              /* Convert to upper case */          c = 0;
1290        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));          for (i = 0; i < 2; ++i)
1291              {
1292              register pcre_uint32 cc = *(++ptr);
1293    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1294              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1295              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1296  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1297        if (cc <= 'z') cc += 64;              /* Convert to upper case */            if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1298        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1299  #endif  #endif
1300        }            }
1301            }
1302          }    /* End JavaScript handling */
1303    
1304        /* Handle \x in Perl's style. \x{ddd} is a character number which can be
1305        greater than 0xff in utf or non-8bit mode, but only if the ddd are hex
1306        digits. If not, { used to be treated as a data character. However, Perl
1307        seems to read hex digits up to the first non-such, and ignore the rest, so
1308        that, for example \x{zz} matches a binary zero. This seems crazy, so PCRE
1309        now gives an error. */
1310    
1311        else
1312          {
1313          if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1314            {
1315            ptr += 2;
1316            c = 0;
1317            overflow = FALSE;
1318            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0)
1319              {
1320              register pcre_uint32 cc = *ptr++;
1321              if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1322    
1323    #ifdef COMPILE_PCRE32
1324              if (c >= 0x10000000l) { overflow = TRUE; break; }
1325    #endif
1326    
1327    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1328              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1329              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1330    #else           /* EBCDIC coding */
1331              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1332              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1333    #endif
1334    
1335    #if defined COMPILE_PCRE8
1336              if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1337    #elif defined COMPILE_PCRE16
1338              if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1339    #elif defined COMPILE_PCRE32
1340              if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1341    #endif
1342              }
1343    
1344            if (overflow)
1345              {
1346              while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0) ptr++;
1347              *errorcodeptr = ERR34;
1348              }
1349    
1350            else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1351              {
1352              if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1353              }
1354    
1355            /* If the sequence of hex digits does not end with '}', give an error.
1356            We used just to recognize this construct and fall through to the normal
1357            \x handling, but nowadays Perl gives an error, which seems much more
1358            sensible, so we do too. */
1359    
1360            else *errorcodeptr = ERR79;
1361            }   /* End of \x{} processing */
1362    
1363          /* Read a single-byte hex-defined char (up to two hex digits after \x) */
1364    
1365          else
1366            {
1367            c = 0;
1368            while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1369              {
1370              pcre_uint32 cc;                          /* Some compilers don't like */
1371              cc = *(++ptr);                           /* ++ in initializers */
1372    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1373              if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1374              c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1375    #else           /* EBCDIC coding */
1376              if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1377              c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1378    #endif
1379              }
1380            }     /* End of \xdd handling */
1381          }       /* End of Perl-style \x handling */
1382      break;      break;
1383    
1384      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
1385      This coding is ASCII-specific, but then the whole concept of \cx is      An error is given if the byte following \c is not an ASCII character. This
1386        coding is ASCII-specific, but then the whole concept of \cx is
1387      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
1388    
1389      case 'c':      case CHAR_c:
1390      c = *(++ptr);      c = *(++ptr);
1391      if (c == 0)      if (c == CHAR_NULL)
1392        {        {
1393        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
1394        return 0;        break;
1395        }        }
1396    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
1397  #ifndef EBCDIC  /* ASCII coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
1398      if (c >= 'a' && c <= 'z') c -= 32;        {
1399          *errorcodeptr = ERR68;
1400          break;
1401          }
1402        if (c >= CHAR_a && c <= CHAR_z) c -= 32;
1403      c ^= 0x40;      c ^= 0x40;
1404  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
1405      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
1406      c ^= 0xC0;      c ^= 0xC0;
1407  #endif  #endif
1408      break;      break;
1409    
1410      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
1411      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
1412      for Perl compatibility, it is a literal. This code looks a bit odd, but      otherwise, for Perl compatibility, it is a literal. This code looks a bit
1413      there used to be some cases other than the default, and there may be again      odd, but there used to be some cases other than the default, and there may
1414      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
1415    
1416      default:      default:
1417      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 637  else Line 1424  else
1424      }      }
1425    }    }
1426    
1427    /* Perl supports \N{name} for character names, as well as plain \N for "not
1428    newline". PCRE does not support \N{name}. However, it does support
1429    quantification such as \N{2,3}. */
1430    
1431    if (escape == ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1432         !is_counted_repeat(ptr+2))
1433      *errorcodeptr = ERR37;
1434    
1435    /* If PCRE_UCP is set, we change the values for \d etc. */
1436    
1437    if ((options & PCRE_UCP) != 0 && escape >= ESC_D && escape <= ESC_w)
1438      escape += (ESC_DU - ESC_D);
1439    
1440    /* Set the pointer to the final character before returning. */
1441    
1442  *ptrptr = ptr;  *ptrptr = ptr;
1443  return c;  *chptr = c;
1444    return escape;
1445  }  }
1446    
1447    
# Line 656  escape sequence. Line 1459  escape sequence.
1459  Argument:  Argument:
1460    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
1461    negptr         points to a boolean that is set TRUE for negation else FALSE    negptr         points to a boolean that is set TRUE for negation else FALSE
1462    dptr           points to an int that is set to the detailed property value    ptypeptr       points to an unsigned int that is set to the type value
1463      pdataptr       points to an unsigned int that is set to the detailed property value
1464    errorcodeptr   points to the error code variable    errorcodeptr   points to the error code variable
1465    
1466  Returns:         type value from ucp_type_table, or -1 for an invalid type  Returns:         TRUE if the type value was found, or FALSE for an invalid type
1467  */  */
1468    
1469  static int  static BOOL
1470  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, unsigned int *ptypeptr,
1471      unsigned int *pdataptr, int *errorcodeptr)
1472  {  {
1473  int c, i, bot, top;  pcre_uchar c;
1474  const uschar *ptr = *ptrptr;  int i, bot, top;
1475  char name[32];  const pcre_uchar *ptr = *ptrptr;
1476    pcre_uchar name[32];
1477    
1478  c = *(++ptr);  c = *(++ptr);
1479  if (c == 0) goto ERROR_RETURN;  if (c == CHAR_NULL) goto ERROR_RETURN;
1480    
1481  *negptr = FALSE;  *negptr = FALSE;
1482    
1483  /* \P or \p can be followed by a name in {}, optionally preceded by ^ for  /* \P or \p can be followed by a name in {}, optionally preceded by ^ for
1484  negation. */  negation. */
1485    
1486  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
1487    {    {
1488    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1489      {      {
1490      *negptr = TRUE;      *negptr = TRUE;
1491      ptr++;      ptr++;
1492      }      }
1493    for (i = 0; i < sizeof(name) - 1; i++)    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1494      {      {
1495      c = *(++ptr);      c = *(++ptr);
1496      if (c == 0) goto ERROR_RETURN;      if (c == CHAR_NULL) goto ERROR_RETURN;
1497      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1498      name[i] = c;      name[i] = c;
1499      }      }
1500    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
1501    name[i] = 0;    name[i] = 0;
1502    }    }
1503    
# Line 708  else Line 1514  else
1514  /* Search for a recognized property name using binary chop */  /* Search for a recognized property name using binary chop */
1515    
1516  bot = 0;  bot = 0;
1517  top = _pcre_utt_size;  top = PRIV(utt_size);
1518    
1519  while (bot < top)  while (bot < top)
1520    {    {
1521      int r;
1522    i = (bot + top) >> 1;    i = (bot + top) >> 1;
1523    c = strcmp(name, _pcre_utt[i].name);    r = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1524    if (c == 0)    if (r == 0)
1525      {      {
1526      *dptr = _pcre_utt[i].value;      *ptypeptr = PRIV(utt)[i].type;
1527      return _pcre_utt[i].type;      *pdataptr = PRIV(utt)[i].value;
1528        return TRUE;
1529      }      }
1530    if (c > 0) bot = i + 1; else top = i;    if (r > 0) bot = i + 1; else top = i;
1531    }    }
1532    
1533  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
1534  *ptrptr = ptr;  *ptrptr = ptr;
1535  return -1;  return FALSE;
1536    
1537  ERROR_RETURN:  ERROR_RETURN:
1538  *errorcodeptr = ERR46;  *errorcodeptr = ERR46;
1539  *ptrptr = ptr;  *ptrptr = ptr;
1540  return -1;  return FALSE;
1541  }  }
1542  #endif  #endif
1543    
1544    
1545    
   
 /*************************************************  
 *            Check for counted repeat            *  
 *************************************************/  
   
 /* This function is called when a '{' is encountered in a place where it might  
 start a quantifier. It looks ahead to see if it really is a quantifier or not.  
 It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}  
 where the ddds are digits.  
   
 Arguments:  
   p         pointer to the first char after '{'  
   
 Returns:    TRUE or FALSE  
 */  
   
 static BOOL  
 is_counted_repeat(const uschar *p)  
 {  
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
 if (*p == '}') return TRUE;  
   
 if (*p++ != ',') return FALSE;  
 if (*p == '}') return TRUE;  
   
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
   
 return (*p == '}');  
 }  
   
   
   
1546  /*************************************************  /*************************************************
1547  *         Read repeat counts                     *  *         Read repeat counts                     *
1548  *************************************************/  *************************************************/
# Line 788  Returns:         pointer to '}' on succe Line 1562  Returns:         pointer to '}' on succe
1562                   current ptr on error, with errorcodeptr set non-zero                   current ptr on error, with errorcodeptr set non-zero
1563  */  */
1564    
1565  static const uschar *  static const pcre_uchar *
1566  read_repeat_counts(const uschar *p, int *minp, int *maxp, int *errorcodeptr)  read_repeat_counts(const pcre_uchar *p, int *minp, int *maxp, int *errorcodeptr)
1567  {  {
1568  int min = 0;  int min = 0;
1569  int max = -1;  int max = -1;
# Line 797  int max = -1; Line 1571  int max = -1;
1571  /* Read the minimum value and do a paranoid check: a negative value indicates  /* Read the minimum value and do a paranoid check: a negative value indicates
1572  an integer overflow. */  an integer overflow. */
1573    
1574  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while (IS_DIGIT(*p)) min = min * 10 + (int)(*p++ - CHAR_0);
1575  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1576    {    {
1577    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 807  if (min < 0 || min > 65535) Line 1581  if (min < 0 || min > 65535)
1581  /* Read the maximum value if there is one, and again do a paranoid on its size.  /* Read the maximum value if there is one, and again do a paranoid on its size.
1582  Also, max must not be less than min. */  Also, max must not be less than min. */
1583    
1584  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1585    {    {
1586    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1587      {      {
1588      max = 0;      max = 0;
1589      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while(IS_DIGIT(*p)) max = max * 10 + (int)(*p++ - CHAR_0);
1590      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1591        {        {
1592        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 837  return p; Line 1611  return p;
1611    
1612    
1613  /*************************************************  /*************************************************
1614  *       Find forward referenced subpattern       *  *      Find first significant op code            *
1615  *************************************************/  *************************************************/
1616    
1617  /* This function scans along a pattern's text looking for capturing  /* This is called by several functions that scan a compiled expression looking
1618  subpatterns, and counting them. If it finds a named pattern that matches the  for a fixed first character, or an anchoring op code etc. It skips over things
1619  name it is given, it returns its number. Alternatively, if the name is NULL, it  that do not influence this. For some calls, it makes sense to skip negative
1620  returns when it reaches a given numbered subpattern. This is used for forward  forward and all backward assertions, and also the \b assertion; for others it
1621  references to subpatterns. We know that if (?P< is encountered, the name will  does not.
 be terminated by '>' because that is checked in the first pass.  
1622    
1623  Arguments:  Arguments:
1624    ptr          current position in the pattern    code         pointer to the start of the group
1625    count        current count of capturing parens so far encountered    skipassert   TRUE if certain assertions are to be skipped
   name         name to seek, or NULL if seeking a numbered subpattern  
   lorn         name length, or subpattern number if name is NULL  
   xmode        TRUE if we are in /x mode  
1626    
1627  Returns:       the number of the named subpattern, or -1 if not found  Returns:       pointer to the first significant opcode
1628  */  */
1629    
1630  static int  static const pcre_uchar*
1631  find_parens(const uschar *ptr, int count, const uschar *name, int lorn,  first_significant_code(const pcre_uchar *code, BOOL skipassert)
   BOOL xmode)  
 {  
 const uschar *thisname;  
   
 for (; *ptr != 0; ptr++)  
   {  
   int term;  
   
   /* Skip over backslashed characters and also entire \Q...\E */  
   
   if (*ptr == '\\')  
     {  
     if (*(++ptr) == 0) return -1;  
     if (*ptr == 'Q') for (;;)  
       {  
       while (*(++ptr) != 0 && *ptr != '\\');  
       if (*ptr == 0) return -1;  
       if (*(++ptr) == 'E') break;  
       }  
     continue;  
     }  
   
   /* Skip over character classes */  
   
   if (*ptr == '[')  
     {  
     while (*(++ptr) != ']')  
       {  
       if (*ptr == '\\')  
         {  
         if (*(++ptr) == 0) return -1;  
         if (*ptr == 'Q') for (;;)  
           {  
           while (*(++ptr) != 0 && *ptr != '\\');  
           if (*ptr == 0) return -1;  
           if (*(++ptr) == 'E') break;  
           }  
         continue;  
         }  
       }  
     continue;  
     }  
   
   /* Skip comments in /x mode */  
   
   if (xmode && *ptr == '#')  
     {  
     while (*(++ptr) != 0 && *ptr != '\n');  
     if (*ptr == 0) return -1;  
     continue;  
     }  
   
   /* An opening parens must now be a real metacharacter */  
   
   if (*ptr != '(') continue;  
   if (ptr[1] != '?')  
     {  
     count++;  
     if (name == NULL && count == lorn) return count;  
     continue;  
     }  
   
   ptr += 2;  
   if (*ptr == 'P') ptr++;                      /* Allow optional P */  
   
   /* We have to disambiguate (?<! and (?<= from (?<name> */  
   
   if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  
        *ptr != '\'')  
     continue;  
   
   count++;  
   
   if (name == NULL && count == lorn) return count;  
   term = *ptr++;  
   if (term == '<') term = '>';  
   thisname = ptr;  
   while (*ptr != term) ptr++;  
   if (name != NULL && lorn == ptr - thisname &&  
       strncmp((const char *)name, (const char *)thisname, lorn) == 0)  
     return count;  
   }  
   
 return -1;  
 }  
   
   
   
 /*************************************************  
 *      Find first significant op code            *  
 *************************************************/  
   
 /* This is called by several functions that scan a compiled expression looking  
 for a fixed first character, or an anchoring op code etc. It skips over things  
 that do not influence this. For some calls, a change of option is important.  
 For some calls, it makes sense to skip negative forward and all backward  
 assertions, and also the \b assertion; for others it does not.  
   
 Arguments:  
   code         pointer to the start of the group  
   options      pointer to external options  
   optbit       the option bit whose changing is significant, or  
                  zero if none are  
   skipassert   TRUE if certain assertions are to be skipped  
   
 Returns:       pointer to the first significant opcode  
 */  
   
 static const uschar*  
 first_significant_code(const uschar *code, int *options, int optbit,  
   BOOL skipassert)  
1632  {  {
1633  for (;;)  for (;;)
1634    {    {
1635    switch ((int)*code)    switch ((int)*code)
1636      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1637      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1638      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1639      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1640      if (!skipassert) return code;      if (!skipassert) return code;
1641      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
1642      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1643      break;      break;
1644    
1645      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
# Line 996  for (;;) Line 1649  for (;;)
1649    
1650      case OP_CALLOUT:      case OP_CALLOUT:
1651      case OP_CREF:      case OP_CREF:
1652        case OP_DNCREF:
1653      case OP_RREF:      case OP_RREF:
1654        case OP_DNRREF:
1655      case OP_DEF:      case OP_DEF:
1656      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1657      break;      break;
1658    
1659      default:      default:
# Line 1010  for (;;) Line 1665  for (;;)
1665    
1666    
1667    
   
1668  /*************************************************  /*************************************************
1669  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1670  *************************************************/  *************************************************/
1671    
1672  /* Scan a pattern and compute the fixed length of subject that will match it,  /* Scan a branch and compute the fixed length of subject that will match it,
1673  if the length is fixed. This is needed for dealing with backward assertions.  if the length is fixed. This is needed for dealing with backward assertions.
1674  In UTF8 mode, the result is in characters rather than bytes.  In UTF8 mode, the result is in characters rather than bytes. The branch is
1675    temporarily terminated with OP_END when this function is called.
1676    
1677    This function is called when a backward assertion is encountered, so that if it
1678    fails, the error message can point to the correct place in the pattern.
1679    However, we cannot do this when the assertion contains subroutine calls,
1680    because they can be forward references. We solve this by remembering this case
1681    and doing the check at the end; a flag specifies which mode we are running in.
1682    
1683  Arguments:  Arguments:
1684    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1685    options  the compiling options    utf      TRUE in UTF-8 / UTF-16 / UTF-32 mode
1686      atend    TRUE if called when the pattern is complete
1687  Returns:   the fixed length, or -1 if there is no fixed length,    cd       the "compile data" structure
1688               or -2 if \C was encountered  
1689    Returns:   the fixed length,
1690                 or -1 if there is no fixed length,
1691                 or -2 if \C was encountered (in UTF-8 mode only)
1692                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1693                 or -4 if an unknown opcode was encountered (internal error)
1694  */  */
1695    
1696  static int  static int
1697  find_fixedlength(uschar *code, int options)  find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd)
1698  {  {
1699  int length = -1;  int length = -1;
1700    
1701  register int branchlength = 0;  register int branchlength = 0;
1702  register uschar *cc = code + 1 + LINK_SIZE;  register pcre_uchar *cc = code + 1 + LINK_SIZE;
1703    
1704  /* Scan along the opcodes for this branch. If we get to the end of the  /* Scan along the opcodes for this branch. If we get to the end of the
1705  branch, check the length against that of the other branches. */  branch, check the length against that of the other branches. */
# Line 1041  branch, check the length against that of Line 1707  branch, check the length against that of
1707  for (;;)  for (;;)
1708    {    {
1709    int d;    int d;
1710    register int op = *cc;    pcre_uchar *ce, *cs;
1711      register pcre_uchar op = *cc;
1712    
1713    switch (op)    switch (op)
1714      {      {
1715        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1716        OP_BRA (normal non-capturing bracket) because the other variants of these
1717        opcodes are all concerned with unlimited repeated groups, which of course
1718        are not of fixed length. */
1719    
1720      case OP_CBRA:      case OP_CBRA:
1721      case OP_BRA:      case OP_BRA:
1722      case OP_ONCE:      case OP_ONCE:
1723        case OP_ONCE_NC:
1724      case OP_COND:      case OP_COND:
1725      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? IMM2_SIZE : 0), utf, atend, cd);
1726      if (d < 0) return d;      if (d < 0) return d;
1727      branchlength += d;      branchlength += d;
1728      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1729      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1730      break;      break;
1731    
1732      /* Reached end of a branch; if it's a ket it is the end of a nested      /* Reached end of a branch; if it's a ket it is the end of a nested call.
1733      call. If it's ALT it is an alternation in a nested call. If it is      If it's ALT it is an alternation in a nested call. An ACCEPT is effectively
1734      END it's the end of the outer call. All can be handled by the same code. */      an ALT. If it is END it's the end of the outer call. All can be handled by
1735        the same code. Note that we must not include the OP_KETRxxx opcodes here,
1736        because they all imply an unlimited repeat. */
1737    
1738      case OP_ALT:      case OP_ALT:
1739      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1740      case OP_END:      case OP_END:
1741        case OP_ACCEPT:
1742        case OP_ASSERT_ACCEPT:
1743      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1744        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
1745      if (*cc != OP_ALT) return length;      if (*cc != OP_ALT) return length;
# Line 1072  for (;;) Line 1747  for (;;)
1747      branchlength = 0;      branchlength = 0;
1748      break;      break;
1749    
1750        /* A true recursion implies not fixed length, but a subroutine call may
1751        be OK. If the subroutine is a forward reference, we can't deal with
1752        it until the end of the pattern, so return -3. */
1753    
1754        case OP_RECURSE:
1755        if (!atend) return -3;
1756        cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1757        do ce += GET(ce, 1); while (*ce == OP_ALT);           /* End subpattern */
1758        if (cc > cs && cc < ce) return -1;                    /* Recursion */
1759        d = find_fixedlength(cs + IMM2_SIZE, utf, atend, cd);
1760        if (d < 0) return d;
1761        branchlength += d;
1762        cc += 1 + LINK_SIZE;
1763        break;
1764    
1765      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1766    
1767      case OP_ASSERT:      case OP_ASSERT:
# Line 1079  for (;;) Line 1769  for (;;)
1769      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1770      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1771      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1772      /* Fall through */      cc += PRIV(OP_lengths)[*cc];
1773        break;
1774    
1775      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1776    
1777      case OP_REVERSE:      case OP_MARK:
1778        case OP_PRUNE_ARG:
1779        case OP_SKIP_ARG:
1780        case OP_THEN_ARG:
1781        cc += cc[1] + PRIV(OP_lengths)[*cc];
1782        break;
1783    
1784        case OP_CALLOUT:
1785        case OP_CIRC:
1786        case OP_CIRCM:
1787        case OP_CLOSE:
1788        case OP_COMMIT:
1789      case OP_CREF:      case OP_CREF:
     case OP_RREF:  
1790      case OP_DEF:      case OP_DEF:
1791      case OP_OPT:      case OP_DNCREF:
1792      case OP_CALLOUT:      case OP_DNRREF:
1793      case OP_SOD:      case OP_DOLL:
1794      case OP_SOM:      case OP_DOLLM:
1795      case OP_EOD:      case OP_EOD:
1796      case OP_EODN:      case OP_EODN:
1797      case OP_CIRC:      case OP_FAIL:
     case OP_DOLL:  
1798      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1799        case OP_PRUNE:
1800        case OP_REVERSE:
1801        case OP_RREF:
1802        case OP_SET_SOM:
1803        case OP_SKIP:
1804        case OP_SOD:
1805        case OP_SOM:
1806        case OP_THEN:
1807      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1808      cc += _pcre_OP_lengths[*cc];      cc += PRIV(OP_lengths)[*cc];
1809      break;      break;
1810    
1811      /* Handle literal characters */      /* Handle literal characters */
1812    
1813      case OP_CHAR:      case OP_CHAR:
1814      case OP_CHARNC:      case OP_CHARI:
1815      case OP_NOT:      case OP_NOT:
1816        case OP_NOTI:
1817      branchlength++;      branchlength++;
1818      cc += 2;      cc += 2;
1819  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1820      if ((options & PCRE_UTF8) != 0)      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
       {  
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1821  #endif  #endif
1822      break;      break;
1823    
# Line 1119  for (;;) Line 1825  for (;;)
1825      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1826    
1827      case OP_EXACT:      case OP_EXACT:
1828      branchlength += GET2(cc,1);      case OP_EXACTI:
1829      cc += 4;      case OP_NOTEXACT:
1830  #ifdef SUPPORT_UTF8      case OP_NOTEXACTI:
1831      if ((options & PCRE_UTF8) != 0)      branchlength += (int)GET2(cc,1);
1832        {      cc += 2 + IMM2_SIZE;
1833        while((*cc & 0x80) == 0x80) cc++;  #ifdef SUPPORT_UTF
1834        }      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1835  #endif  #endif
1836      break;      break;
1837    
1838      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1839      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1840      cc += 4;      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)
1841          cc += 2;
1842        cc += 1 + IMM2_SIZE + 1;
1843      break;      break;
1844    
1845      /* Handle single-char matchers */      /* Handle single-char matchers */
# Line 1141  for (;;) Line 1849  for (;;)
1849      cc += 2;      cc += 2;
1850      /* Fall through */      /* Fall through */
1851    
1852        case OP_HSPACE:
1853        case OP_VSPACE:
1854        case OP_NOT_HSPACE:
1855        case OP_NOT_VSPACE:
1856      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
1857      case OP_DIGIT:      case OP_DIGIT:
1858      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
# Line 1148  for (;;) Line 1860  for (;;)
1860      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1861      case OP_WORDCHAR:      case OP_WORDCHAR:
1862      case OP_ANY:      case OP_ANY:
1863        case OP_ALLANY:
1864      branchlength++;      branchlength++;
1865      cc++;      cc++;
1866      break;      break;
1867    
1868      /* The single-byte matcher isn't allowed */      /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1869        otherwise \C is coded as OP_ALLANY. */
1870    
1871      case OP_ANYBYTE:      case OP_ANYBYTE:
1872      return -2;      return -2;
1873    
1874      /* Check a class for variable quantification */      /* Check a class for variable quantification */
1875    
 #ifdef SUPPORT_UTF8  
     case OP_XCLASS:  
     cc += GET(cc, 1) - 33;  
     /* Fall through */  
 #endif  
   
1876      case OP_CLASS:      case OP_CLASS:
1877      case OP_NCLASS:      case OP_NCLASS:
1878      cc += 33;  #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1879        case OP_XCLASS:
1880        /* The original code caused an unsigned overflow in 64 bit systems,
1881        so now we use a conditional statement. */
1882        if (op == OP_XCLASS)
1883          cc += GET(cc, 1);
1884        else
1885          cc += PRIV(OP_lengths)[OP_CLASS];
1886    #else
1887        cc += PRIV(OP_lengths)[OP_CLASS];
1888    #endif
1889    
1890      switch (*cc)      switch (*cc)
1891        {        {
1892        case OP_CRSTAR:        case OP_CRSTAR:
1893        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1894          case OP_CRPLUS:
1895          case OP_CRMINPLUS:
1896        case OP_CRQUERY:        case OP_CRQUERY:
1897        case OP_CRMINQUERY:        case OP_CRMINQUERY:
1898          case OP_CRPOSSTAR:
1899          case OP_CRPOSPLUS:
1900          case OP_CRPOSQUERY:
1901        return -1;        return -1;
1902    
1903        case OP_CRRANGE:        case OP_CRRANGE:
1904        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1905        if (GET2(cc,1) != GET2(cc,3)) return -1;        case OP_CRPOSRANGE:
1906        branchlength += GET2(cc,1);        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1907        cc += 5;        branchlength += (int)GET2(cc,1);
1908          cc += 1 + 2 * IMM2_SIZE;
1909        break;        break;
1910    
1911        default:        default:
# Line 1191  for (;;) Line 1915  for (;;)
1915    
1916      /* Anything else is variable length */      /* Anything else is variable length */
1917    
1918      default:      case OP_ANYNL:
1919        case OP_BRAMINZERO:
1920        case OP_BRAPOS:
1921        case OP_BRAPOSZERO:
1922        case OP_BRAZERO:
1923        case OP_CBRAPOS:
1924        case OP_EXTUNI:
1925        case OP_KETRMAX:
1926        case OP_KETRMIN:
1927        case OP_KETRPOS:
1928        case OP_MINPLUS:
1929        case OP_MINPLUSI:
1930        case OP_MINQUERY:
1931        case OP_MINQUERYI:
1932        case OP_MINSTAR:
1933        case OP_MINSTARI:
1934        case OP_MINUPTO:
1935        case OP_MINUPTOI:
1936        case OP_NOTMINPLUS:
1937        case OP_NOTMINPLUSI:
1938        case OP_NOTMINQUERY:
1939        case OP_NOTMINQUERYI:
1940        case OP_NOTMINSTAR:
1941        case OP_NOTMINSTARI:
1942        case OP_NOTMINUPTO:
1943        case OP_NOTMINUPTOI:
1944        case OP_NOTPLUS:
1945        case OP_NOTPLUSI:
1946        case OP_NOTPOSPLUS:
1947        case OP_NOTPOSPLUSI:
1948        case OP_NOTPOSQUERY:
1949        case OP_NOTPOSQUERYI:
1950        case OP_NOTPOSSTAR:
1951        case OP_NOTPOSSTARI:
1952        case OP_NOTPOSUPTO:
1953        case OP_NOTPOSUPTOI:
1954        case OP_NOTQUERY:
1955        case OP_NOTQUERYI:
1956        case OP_NOTSTAR:
1957        case OP_NOTSTARI:
1958        case OP_NOTUPTO:
1959        case OP_NOTUPTOI:
1960        case OP_PLUS:
1961        case OP_PLUSI:
1962        case OP_POSPLUS:
1963        case OP_POSPLUSI:
1964        case OP_POSQUERY:
1965        case OP_POSQUERYI:
1966        case OP_POSSTAR:
1967        case OP_POSSTARI:
1968        case OP_POSUPTO:
1969        case OP_POSUPTOI:
1970        case OP_QUERY:
1971        case OP_QUERYI:
1972        case OP_REF:
1973        case OP_REFI:
1974        case OP_DNREF:
1975        case OP_DNREFI:
1976        case OP_SBRA:
1977        case OP_SBRAPOS:
1978        case OP_SCBRA:
1979        case OP_SCBRAPOS:
1980        case OP_SCOND:
1981        case OP_SKIPZERO:
1982        case OP_STAR:
1983        case OP_STARI:
1984        case OP_TYPEMINPLUS:
1985        case OP_TYPEMINQUERY:
1986        case OP_TYPEMINSTAR:
1987        case OP_TYPEMINUPTO:
1988        case OP_TYPEPLUS:
1989        case OP_TYPEPOSPLUS:
1990        case OP_TYPEPOSQUERY:
1991        case OP_TYPEPOSSTAR:
1992        case OP_TYPEPOSUPTO:
1993        case OP_TYPEQUERY:
1994        case OP_TYPESTAR:
1995        case OP_TYPEUPTO:
1996        case OP_UPTO:
1997        case OP_UPTOI:
1998      return -1;      return -1;
1999    
2000        /* Catch unrecognized opcodes so that when new ones are added they
2001        are not forgotten, as has happened in the past. */
2002    
2003        default:
2004        return -4;
2005      }      }
2006    }    }
2007  /* Control never gets here */  /* Control never gets here */
# Line 1200  for (;;) Line 2009  for (;;)
2009    
2010    
2011    
   
2012  /*************************************************  /*************************************************
2013  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
2014  *************************************************/  *************************************************/
2015    
2016  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
2017  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
2018    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
2019    so that it can be called from pcre_study() when finding the minimum matching
2020    length.
2021    
2022  Arguments:  Arguments:
2023    code        points to start of expression    code        points to start of expression
2024    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2025    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
2026    
2027  Returns:      pointer to the opcode for the bracket, or NULL if not found  Returns:      pointer to the opcode for the bracket, or NULL if not found
2028  */  */
2029    
2030  static const uschar *  const pcre_uchar *
2031  find_bracket(const uschar *code, BOOL utf8, int number)  PRIV(find_bracket)(const pcre_uchar *code, BOOL utf, int number)
2032  {  {
2033  for (;;)  for (;;)
2034    {    {
2035    register int c = *code;    register pcre_uchar c = *code;
2036    
2037    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2038    
2039    /* XCLASS is used for classes that cannot be represented just by a bit    /* XCLASS is used for classes that cannot be represented just by a bit
# Line 1230  for (;;) Line 2042  for (;;)
2042    
2043    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
2044    
2045      /* Handle recursion */
2046    
2047      else if (c == OP_REVERSE)
2048        {
2049        if (number < 0) return (pcre_uchar *)code;
2050        code += PRIV(OP_lengths)[c];
2051        }
2052    
2053    /* Handle capturing bracket */    /* Handle capturing bracket */
2054    
2055    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
2056               c == OP_CBRAPOS || c == OP_SCBRAPOS)
2057      {      {
2058      int n = GET2(code, 1+LINK_SIZE);      int n = (int)GET2(code, 1+LINK_SIZE);
2059      if (n == number) return (uschar *)code;      if (n == number) return (pcre_uchar *)code;
2060      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2061      }      }
2062    
2063      /* Otherwise, we can get the item's length from the table, except that for
2064      repeated character types, we have to test for \p and \P, which have an extra
2065      two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2066      must add in its length. */
2067    
2068      else
2069        {
2070        switch(c)
2071          {
2072          case OP_TYPESTAR:
2073          case OP_TYPEMINSTAR:
2074          case OP_TYPEPLUS:
2075          case OP_TYPEMINPLUS:
2076          case OP_TYPEQUERY:
2077          case OP_TYPEMINQUERY:
2078          case OP_TYPEPOSSTAR:
2079          case OP_TYPEPOSPLUS:
2080          case OP_TYPEPOSQUERY:
2081          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2082          break;
2083    
2084          case OP_TYPEUPTO:
2085          case OP_TYPEMINUPTO:
2086          case OP_TYPEEXACT:
2087          case OP_TYPEPOSUPTO:
2088          if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2089            code += 2;
2090          break;
2091    
2092          case OP_MARK:
2093          case OP_PRUNE_ARG:
2094          case OP_SKIP_ARG:
2095          case OP_THEN_ARG:
2096          code += code[1];
2097          break;
2098          }
2099    
2100        /* Add in the fixed length from the table */
2101    
2102        code += PRIV(OP_lengths)[c];
2103    
2104    /* In UTF-8 mode, opcodes that are followed by a character may be followed by    /* In UTF-8 mode, opcodes that are followed by a character may be followed by
2105    a multi-byte character. The length in the table is a minimum, so we have to    a multi-byte character. The length in the table is a minimum, so we have to
2106    arrange to skip the extra bytes. */    arrange to skip the extra bytes. */
2107    
2108    else  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2109      {      if (utf) switch(c)
     code += _pcre_OP_lengths[c];  
 #ifdef SUPPORT_UTF8  
     if (utf8) switch(c)  
2110        {        {
2111        case OP_CHAR:        case OP_CHAR:
2112        case OP_CHARNC:        case OP_CHARI:
2113        case OP_EXACT:        case OP_EXACT:
2114          case OP_EXACTI:
2115        case OP_UPTO:        case OP_UPTO:
2116          case OP_UPTOI:
2117        case OP_MINUPTO:        case OP_MINUPTO:
2118          case OP_MINUPTOI:
2119        case OP_POSUPTO:        case OP_POSUPTO:
2120          case OP_POSUPTOI:
2121        case OP_STAR:        case OP_STAR:
2122          case OP_STARI:
2123        case OP_MINSTAR:        case OP_MINSTAR:
2124          case OP_MINSTARI:
2125        case OP_POSSTAR:        case OP_POSSTAR:
2126          case OP_POSSTARI:
2127        case OP_PLUS:        case OP_PLUS:
2128          case OP_PLUSI:
2129        case OP_MINPLUS:        case OP_MINPLUS:
2130          case OP_MINPLUSI:
2131        case OP_POSPLUS:        case OP_POSPLUS:
2132          case OP_POSPLUSI:
2133        case OP_QUERY:        case OP_QUERY:
2134          case OP_QUERYI:
2135        case OP_MINQUERY:        case OP_MINQUERY:
2136          case OP_MINQUERYI:
2137        case OP_POSQUERY:        case OP_POSQUERY:
2138        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2139          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2140        break;        break;
2141        }        }
2142    #else
2143        (void)(utf);  /* Keep compiler happy by referencing function argument */
2144  #endif  #endif
2145      }      }
2146    }    }
# Line 1283  instance of OP_RECURSE. Line 2157  instance of OP_RECURSE.
2157    
2158  Arguments:  Arguments:
2159    code        points to start of expression    code        points to start of expression
2160    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2161    
2162  Returns:      pointer to the opcode for OP_RECURSE, or NULL if not found  Returns:      pointer to the opcode for OP_RECURSE, or NULL if not found
2163  */  */
2164    
2165  static const uschar *  static const pcre_uchar *
2166  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const pcre_uchar *code, BOOL utf)
2167  {  {
2168  for (;;)  for (;;)
2169    {    {
2170    register int c = *code;    register pcre_uchar c = *code;
2171    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2172    if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
2173    
# Line 1303  for (;;) Line 2177  for (;;)
2177    
2178    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
2179    
2180    /* Otherwise, we get the item's length from the table. In UTF-8 mode, opcodes    /* Otherwise, we can get the item's length from the table, except that for
2181    that are followed by a character may be followed by a multi-byte character.    repeated character types, we have to test for \p and \P, which have an extra
2182    The length in the table is a minimum, so we have to arrange to skip the extra    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2183    bytes. */    must add in its length. */
2184    
2185    else    else
2186      {      {
2187      code += _pcre_OP_lengths[c];      switch(c)
2188  #ifdef SUPPORT_UTF8        {
2189      if (utf8) switch(c)        case OP_TYPESTAR:
2190          case OP_TYPEMINSTAR:
2191          case OP_TYPEPLUS:
2192          case OP_TYPEMINPLUS:
2193          case OP_TYPEQUERY:
2194          case OP_TYPEMINQUERY:
2195          case OP_TYPEPOSSTAR:
2196          case OP_TYPEPOSPLUS:
2197          case OP_TYPEPOSQUERY:
2198          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2199          break;
2200    
2201          case OP_TYPEPOSUPTO:
2202          case OP_TYPEUPTO:
2203          case OP_TYPEMINUPTO:
2204          case OP_TYPEEXACT:
2205          if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2206            code += 2;
2207          break;
2208    
2209          case OP_MARK:
2210          case OP_PRUNE_ARG:
2211          case OP_SKIP_ARG:
2212          case OP_THEN_ARG:
2213          code += code[1];
2214          break;
2215          }
2216    
2217        /* Add in the fixed length from the table */
2218    
2219        code += PRIV(OP_lengths)[c];
2220    
2221        /* In UTF-8 mode, opcodes that are followed by a character may be followed
2222        by a multi-byte character. The length in the table is a minimum, so we have
2223        to arrange to skip the extra bytes. */
2224    
2225    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2226        if (utf) switch(c)
2227        {        {
2228        case OP_CHAR:        case OP_CHAR:
2229        case OP_CHARNC:        case OP_CHARI:
2230          case OP_NOT:
2231          case OP_NOTI:
2232        case OP_EXACT:        case OP_EXACT:
2233          case OP_EXACTI:
2234          case OP_NOTEXACT:
2235          case OP_NOTEXACTI:
2236        case OP_UPTO:        case OP_UPTO:
2237          case OP_UPTOI:
2238          case OP_NOTUPTO:
2239          case OP_NOTUPTOI:
2240        case OP_MINUPTO:        case OP_MINUPTO:
2241          case OP_MINUPTOI:
2242          case OP_NOTMINUPTO:
2243          case OP_NOTMINUPTOI:
2244        case OP_POSUPTO:        case OP_POSUPTO:
2245          case OP_POSUPTOI:
2246          case OP_NOTPOSUPTO:
2247          case OP_NOTPOSUPTOI:
2248        case OP_STAR:        case OP_STAR:
2249          case OP_STARI:
2250          case OP_NOTSTAR:
2251          case OP_NOTSTARI:
2252        case OP_MINSTAR:        case OP_MINSTAR:
2253          case OP_MINSTARI:
2254          case OP_NOTMINSTAR:
2255          case OP_NOTMINSTARI:
2256        case OP_POSSTAR:        case OP_POSSTAR:
2257          case OP_POSSTARI:
2258          case OP_NOTPOSSTAR:
2259          case OP_NOTPOSSTARI:
2260        case OP_PLUS:        case OP_PLUS:
2261          case OP_PLUSI:
2262          case OP_NOTPLUS:
2263          case OP_NOTPLUSI:
2264        case OP_MINPLUS:        case OP_MINPLUS:
2265          case OP_MINPLUSI:
2266          case OP_NOTMINPLUS:
2267          case OP_NOTMINPLUSI:
2268        case OP_POSPLUS:        case OP_POSPLUS:
2269          case OP_POSPLUSI:
2270          case OP_NOTPOSPLUS:
2271          case OP_NOTPOSPLUSI:
2272        case OP_QUERY:        case OP_QUERY:
2273          case OP_QUERYI:
2274          case OP_NOTQUERY:
2275          case OP_NOTQUERYI:
2276        case OP_MINQUERY:        case OP_MINQUERY:
2277          case OP_MINQUERYI:
2278          case OP_NOTMINQUERY:
2279          case OP_NOTMINQUERYI:
2280        case OP_POSQUERY:        case OP_POSQUERY:
2281        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2282          case OP_NOTPOSQUERY:
2283          case OP_NOTPOSQUERYI:
2284          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2285        break;        break;
2286        }        }
2287    #else
2288        (void)(utf);  /* Keep compiler happy by referencing function argument */
2289  #endif  #endif
2290      }      }
2291    }    }
# Line 1347  for (;;) Line 2301  for (;;)
2301  can match the empty string or not. It is called from could_be_empty()  can match the empty string or not. It is called from could_be_empty()
2302  below and from compile_branch() when checking for an unlimited repeat of a  below and from compile_branch() when checking for an unlimited repeat of a
2303  group that can match nothing. Note that first_significant_code() skips over  group that can match nothing. Note that first_significant_code() skips over
2304  assertions. If we hit an unclosed bracket, we return "empty" - this means we've  backward and negative forward assertions when its final argument is TRUE. If we
2305  struck an inner bracket whose current branch will already have been scanned.  hit an unclosed bracket, we return "empty" - this means we've struck an inner
2306    bracket whose current branch will already have been scanned.
2307    
2308  Arguments:  Arguments:
2309    code        points to start of search    code        points to start of search
2310    endcode     points to where to stop    endcode     points to where to stop
2311    utf8        TRUE if in UTF8 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2312      cd          contains pointers to tables etc.
2313      recurses    chain of recurse_check to catch mutual recursion
2314    
2315  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2316  */  */
2317    
2318    typedef struct recurse_check {
2319      struct recurse_check *prev;
2320      const pcre_uchar *group;
2321    } recurse_check;
2322    
2323  static BOOL  static BOOL
2324  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2325      BOOL utf, compile_data *cd, recurse_check *recurses)
2326  {  {
2327  register int c;  register pcre_uchar c;
2328  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  recurse_check this_recurse;
2329    
2330    for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2331       code < endcode;       code < endcode;
2332       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
2333    {    {
2334    const uschar *ccode;    const pcre_uchar *ccode;
2335    
2336    c = *code;    c = *code;
2337    
2338    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE)    /* Skip over forward assertions; the other assertions are skipped by
2339      first_significant_code() with a TRUE final argument. */
2340    
2341      if (c == OP_ASSERT)
2342        {
2343        do code += GET(code, 1); while (*code == OP_ALT);
2344        c = *code;
2345        continue;
2346        }
2347    
2348      /* For a recursion/subroutine call, if its end has been reached, which
2349      implies a backward reference subroutine call, we can scan it. If it's a
2350      forward reference subroutine call, we can't. To detect forward reference
2351      we have to scan up the list that is kept in the workspace. This function is
2352      called only when doing the real compile, not during the pre-compile that
2353      measures the size of the compiled pattern. */
2354    
2355      if (c == OP_RECURSE)
2356      {      {
2357        const pcre_uchar *scode = cd->start_code + GET(code, 1);
2358      BOOL empty_branch;      BOOL empty_branch;
     if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */  
2359    
2360      /* Scan a closed bracket */      /* Test for forward reference or uncompleted reference. This is disabled
2361        when called to scan a completed pattern by setting cd->start_workspace to
2362        NULL. */
2363    
2364        if (cd->start_workspace != NULL)
2365          {
2366          const pcre_uchar *tcode;
2367          for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)
2368            if ((int)GET(tcode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;
2369          if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2370          }
2371    
2372        /* If we are scanning a completed pattern, there are no forward references
2373        and all groups are complete. We need to detect whether this is a recursive
2374        call, as otherwise there will be an infinite loop. If it is a recursion,
2375        just skip over it. Simple recursions are easily detected. For mutual
2376        recursions we keep a chain on the stack. */
2377    
2378        else
2379          {
2380          recurse_check *r = recurses;
2381          const pcre_uchar *endgroup = scode;
2382    
2383          do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);
2384          if (code >= scode && code <= endgroup) continue;  /* Simple recursion */
2385    
2386          for (r = recurses; r != NULL; r = r->prev)
2387            if (r->group == scode) break;
2388          if (r != NULL) continue;   /* Mutual recursion */
2389          }
2390    
2391        /* Completed reference; scan the referenced group, remembering it on the
2392        stack chain to detect mutual recursions. */
2393    
2394      empty_branch = FALSE;      empty_branch = FALSE;
2395        this_recurse.prev = recurses;
2396        this_recurse.group = scode;
2397    
2398      do      do
2399        {        {
2400        if (!empty_branch && could_be_empty_branch(code, endcode, utf8))        if (could_be_empty_branch(scode, endcode, utf, cd, &this_recurse))
2401            {
2402          empty_branch = TRUE;          empty_branch = TRUE;
2403        code += GET(code, 1);          break;
2404            }
2405          scode += GET(scode, 1);
2406        }        }
2407      while (*code == OP_ALT);      while (*scode == OP_ALT);
2408      if (!empty_branch) return FALSE;   /* All branches are non-empty */  
2409        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2410        continue;
2411        }
2412    
2413      /* Groups with zero repeats can of course be empty; skip them. */
2414    
2415      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2416          c == OP_BRAPOSZERO)
2417        {
2418        code += PRIV(OP_lengths)[c];
2419        do code += GET(code, 1); while (*code == OP_ALT);
2420        c = *code;
2421        continue;
2422        }
2423    
2424      /* A nested group that is already marked as "could be empty" can just be
2425      skipped. */
2426    
2427      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2428          c == OP_SCBRA || c == OP_SCBRAPOS)
2429        {
2430        do code += GET(code, 1); while (*code == OP_ALT);
2431        c = *code;
2432        continue;
2433        }
2434    
2435      /* Move past the KET and fudge things so that the increment in the "for"    /* For other groups, scan the branches. */
2436      above has no effect. */  
2437      if (c == OP_BRA  || c == OP_BRAPOS ||
2438          c == OP_CBRA || c == OP_CBRAPOS ||
2439          c == OP_ONCE || c == OP_ONCE_NC ||
2440          c == OP_COND)
2441        {
2442        BOOL empty_branch;
2443        if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
2444    
2445        /* If a conditional group has only one branch, there is a second, implied,
2446        empty branch, so just skip over the conditional, because it could be empty.
2447        Otherwise, scan the individual branches of the group. */
2448    
2449        if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
2450          code += GET(code, 1);
2451        else
2452          {
2453          empty_branch = FALSE;
2454          do
2455            {
2456            if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd, NULL))
2457              empty_branch = TRUE;
2458            code += GET(code, 1);
2459            }
2460          while (*code == OP_ALT);
2461          if (!empty_branch) return FALSE;   /* All branches are non-empty */
2462          }
2463    
2464      c = OP_END;      c = *code;
     code += 1 + LINK_SIZE - _pcre_OP_lengths[c];  
2465      continue;      continue;
2466      }      }
2467    
# Line 1399  for (code = first_significant_code(code Line 2469  for (code = first_significant_code(code
2469    
2470    switch (c)    switch (c)
2471      {      {
2472      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
2473        cannot be represented just by a bit map. This includes negated single
2474        high-valued characters. The length in PRIV(OP_lengths)[] is zero; the
2475        actual length is stored in the compiled code, so we must update "code"
2476        here. */
2477    
2478  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2479      case OP_XCLASS:      case OP_XCLASS:
2480      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
2481      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
2482  #endif  #endif
2483    
2484      case OP_CLASS:      case OP_CLASS:
2485      case OP_NCLASS:      case OP_NCLASS:
2486      ccode = code + 33;      ccode = code + PRIV(OP_lengths)[OP_CLASS];
2487    
2488  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2489      CHECK_CLASS_REPEAT:      CHECK_CLASS_REPEAT:
2490  #endif  #endif
2491    
# Line 1421  for (code = first_significant_code(code Line 2495  for (code = first_significant_code(code
2495        case OP_CRMINSTAR:        case OP_CRMINSTAR:
2496        case OP_CRQUERY:        case OP_CRQUERY:
2497        case OP_CRMINQUERY:        case OP_CRMINQUERY:
2498          case OP_CRPOSSTAR:
2499          case OP_CRPOSQUERY:
2500        break;        break;
2501    
2502        default:                   /* Non-repeat => class must match */        default:                   /* Non-repeat => class must match */
2503        case OP_CRPLUS:            /* These repeats aren't empty */        case OP_CRPLUS:            /* These repeats aren't empty */
2504        case OP_CRMINPLUS:        case OP_CRMINPLUS:
2505          case OP_CRPOSPLUS:
2506        return FALSE;        return FALSE;
2507    
2508        case OP_CRRANGE:        case OP_CRRANGE:
2509        case OP_CRMINRANGE:        case OP_CRMINRANGE:
2510          case OP_CRPOSRANGE:
2511        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */
2512        break;        break;
2513        }        }
# Line 1437  for (code = first_significant_code(code Line 2515  for (code = first_significant_code(code
2515    
2516      /* Opcodes that must match a character */      /* Opcodes that must match a character */
2517    
2518        case OP_ANY:
2519        case OP_ALLANY:
2520        case OP_ANYBYTE:
2521    
2522      case OP_PROP:      case OP_PROP:
2523      case OP_NOTPROP:      case OP_NOTPROP:
2524        case OP_ANYNL:
2525    
2526        case OP_NOT_HSPACE:
2527        case OP_HSPACE:
2528        case OP_NOT_VSPACE:
2529        case OP_VSPACE:
2530      case OP_EXTUNI:      case OP_EXTUNI:
2531    
2532      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2533      case OP_DIGIT:      case OP_DIGIT:
2534      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2535      case OP_WHITESPACE:      case OP_WHITESPACE:
2536      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2537      case OP_WORDCHAR:      case OP_WORDCHAR:
2538      case OP_ANY:  
     case OP_ANYBYTE:  
2539      case OP_CHAR:      case OP_CHAR:
2540      case OP_CHARNC:      case OP_CHARI:
2541      case OP_NOT:      case OP_NOT:
2542        case OP_NOTI:
2543    
2544      case OP_PLUS:      case OP_PLUS:
2545        case OP_PLUSI:
2546      case OP_MINPLUS:      case OP_MINPLUS:
2547      case OP_POSPLUS:      case OP_MINPLUSI:
2548      case OP_EXACT:  
2549      case OP_NOTPLUS:      case OP_NOTPLUS:
2550        case OP_NOTPLUSI:
2551      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
2552        case OP_NOTMINPLUSI:
2553    
2554        case OP_POSPLUS:
2555        case OP_POSPLUSI:
2556      case OP_NOTPOSPLUS:      case OP_NOTPOSPLUS:
2557        case OP_NOTPOSPLUSI:
2558    
2559        case OP_EXACT:
2560        case OP_EXACTI:
2561      case OP_NOTEXACT:      case OP_NOTEXACT:
2562        case OP_NOTEXACTI:
2563    
2564      case OP_TYPEPLUS:      case OP_TYPEPLUS:
2565      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
2566      case OP_TYPEPOSPLUS:      case OP_TYPEPOSPLUS:
2567      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2568    
2569      return FALSE;      return FALSE;
2570    
2571        /* These are going to continue, as they may be empty, but we have to
2572        fudge the length for the \p and \P cases. */
2573    
2574        case OP_TYPESTAR:
2575        case OP_TYPEMINSTAR:
2576        case OP_TYPEPOSSTAR:
2577        case OP_TYPEQUERY:
2578        case OP_TYPEMINQUERY:
2579        case OP_TYPEPOSQUERY:
2580        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2581        break;
2582    
2583        /* Same for these */
2584    
2585        case OP_TYPEUPTO:
2586        case OP_TYPEMINUPTO:
2587        case OP_TYPEPOSUPTO:
2588        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2589          code += 2;
2590        break;
2591    
2592      /* End of branch */      /* End of branch */
2593    
2594      case OP_KET:      case OP_KET:
2595      case OP_KETRMAX:      case OP_KETRMAX:
2596      case OP_KETRMIN:      case OP_KETRMIN:
2597        case OP_KETRPOS:
2598      case OP_ALT:      case OP_ALT:
2599      return TRUE;      return TRUE;
2600    
2601      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2602      MINUPTO, and POSUPTO may be followed by a multibyte character */      MINUPTO, and POSUPTO and their caseless and negative versions may be
2603        followed by a multibyte character. */
2604    
2605  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2606      case OP_STAR:      case OP_STAR:
2607        case OP_STARI:
2608        case OP_NOTSTAR:
2609        case OP_NOTSTARI:
2610    
2611      case OP_MINSTAR:      case OP_MINSTAR:
2612        case OP_MINSTARI:
2613        case OP_NOTMINSTAR:
2614        case OP_NOTMINSTARI:
2615    
2616      case OP_POSSTAR:      case OP_POSSTAR:
2617        case OP_POSSTARI:
2618        case OP_NOTPOSSTAR:
2619        case OP_NOTPOSSTARI:
2620    
2621      case OP_QUERY:      case OP_QUERY:
2622        case OP_QUERYI:
2623        case OP_NOTQUERY:
2624        case OP_NOTQUERYI:
2625    
2626      case OP_MINQUERY:      case OP_MINQUERY:
2627        case OP_MINQUERYI:
2628        case OP_NOTMINQUERY:
2629        case OP_NOTMINQUERYI:
2630    
2631      case OP_POSQUERY:      case OP_POSQUERY:
2632        case OP_POSQUERYI:
2633        case OP_NOTPOSQUERY:
2634        case OP_NOTPOSQUERYI:
2635    
2636        if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2637        break;
2638    
2639      case OP_UPTO:      case OP_UPTO:
2640        case OP_UPTOI:
2641        case OP_NOTUPTO:
2642        case OP_NOTUPTOI:
2643    
2644      case OP_MINUPTO:      case OP_MINUPTO:
2645        case OP_MINUPTOI:
2646        case OP_NOTMINUPTO:
2647        case OP_NOTMINUPTOI:
2648    
2649      case OP_POSUPTO:      case OP_POSUPTO:
2650      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_POSUPTOI:
2651        case OP_NOTPOSUPTO:
2652        case OP_NOTPOSUPTOI:
2653    
2654        if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2655      break;      break;
2656  #endif  #endif
2657    
2658        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2659        string. */
2660    
2661        case OP_MARK:
2662        case OP_PRUNE_ARG:
2663        case OP_SKIP_ARG:
2664        case OP_THEN_ARG:
2665        code += code[1];
2666        break;
2667    
2668        /* None of the remaining opcodes are required to match a character. */
2669    
2670        default:
2671        break;
2672      }      }
2673    }    }
2674    
# Line 1505  return TRUE; Line 2685  return TRUE;
2685  the current branch of the current pattern to see if it could match the empty  the current branch of the current pattern to see if it could match the empty
2686  string. If it could, we must look outwards for branches at other levels,  string. If it could, we must look outwards for branches at other levels,
2687  stopping when we pass beyond the bracket which is the subject of the recursion.  stopping when we pass beyond the bracket which is the subject of the recursion.
2688    This function is called only during the real compile, not during the
2689    pre-compile.
2690    
2691  Arguments:  Arguments:
2692    code        points to start of the recursion    code        points to start of the recursion
2693    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2694    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2695    utf8        TRUE if in UTF-8 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2696      cd          pointers to tables etc
2697    
2698  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2699  */  */
2700    
2701  static BOOL  static BOOL
2702  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
2703    BOOL utf8)    branch_chain *bcptr, BOOL utf, compile_data *cd)
2704  {  {
2705  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2706    {    {
2707    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd, NULL))
2708        return FALSE;
2709    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2710    }    }
2711  return TRUE;  return TRUE;
2712    }
2713    
2714    
2715    
2716    /*************************************************
2717    *        Base opcode of repeated opcodes         *
2718    *************************************************/
2719    
2720    /* Returns the base opcode for repeated single character type opcodes. If the
2721    opcode is not a repeated character type, it returns with the original value.
2722    
2723    Arguments:  c opcode
2724    Returns:    base opcode for the type
2725    */
2726    
2727    static pcre_uchar
2728    get_repeat_base(pcre_uchar c)
2729    {
2730    return (c > OP_TYPEPOSUPTO)? c :
2731           (c >= OP_TYPESTAR)?   OP_TYPESTAR :
2732           (c >= OP_NOTSTARI)?   OP_NOTSTARI :
2733           (c >= OP_NOTSTAR)?    OP_NOTSTAR :
2734           (c >= OP_STARI)?      OP_STARI :
2735                                 OP_STAR;
2736    }
2737    
2738    
2739    
2740    #ifdef SUPPORT_UCP
2741    /*************************************************
2742    *        Check a character and a property        *
2743    *************************************************/
2744    
2745    /* This function is called by check_auto_possessive() when a property item
2746    is adjacent to a fixed character.
2747    
2748    Arguments:
2749      c            the character
2750      ptype        the property type
2751      pdata        the data for the type
2752      negated      TRUE if it's a negated property (\P or \p{^)
2753    
2754    Returns:       TRUE if auto-possessifying is OK
2755    */
2756    
2757    static BOOL
2758    check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata,
2759      BOOL negated)
2760    {
2761    const pcre_uint32 *p;
2762    const ucd_record *prop = GET_UCD(c);
2763    
2764    switch(ptype)
2765      {
2766      case PT_LAMP:
2767      return (prop->chartype == ucp_Lu ||
2768              prop->chartype == ucp_Ll ||
2769              prop->chartype == ucp_Lt) == negated;
2770    
2771      case PT_GC:
2772      return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2773    
2774      case PT_PC:
2775      return (pdata == prop->chartype) == negated;
2776    
2777      case PT_SC:
2778      return (pdata == prop->script) == negated;
2779    
2780      /* These are specials */
2781    
2782      case PT_ALNUM:
2783      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2784              PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2785    
2786      /* Perl space used to exclude VT, but from Perl 5.18 it is included, which
2787      means that Perl space and POSIX space are now identical. PCRE was changed
2788      at release 8.34. */
2789    
2790      case PT_SPACE:    /* Perl space */
2791      case PT_PXSPACE:  /* POSIX space */
2792      switch(c)
2793        {
2794        HSPACE_CASES:
2795        VSPACE_CASES:
2796        return negated;
2797    
2798        default:
2799        return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z) == negated;
2800        }
2801      break;  /* Control never reaches here */
2802    
2803      case PT_WORD:
2804      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2805              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2806              c == CHAR_UNDERSCORE) == negated;
2807    
2808      case PT_CLIST:
2809      p = PRIV(ucd_caseless_sets) + prop->caseset;
2810      for (;;)
2811        {
2812        if (c < *p) return !negated;
2813        if (c == *p++) return negated;
2814        }
2815      break;  /* Control never reaches here */
2816      }
2817    
2818    return FALSE;
2819    }
2820    #endif  /* SUPPORT_UCP */
2821    
2822    
2823    
2824    /*************************************************
2825    *        Fill the character property list        *
2826    *************************************************/
2827    
2828    /* Checks whether the code points to an opcode that can take part in auto-
2829    possessification, and if so, fills a list with its properties.
2830    
2831    Arguments:
2832      code        points to start of expression
2833      utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2834      fcc         points to case-flipping table
2835      list        points to output list
2836                  list[0] will be filled with the opcode
2837                  list[1] will be non-zero if this opcode
2838                    can match an empty character string
2839                  list[2..7] depends on the opcode
2840    
2841    Returns:      points to the start of the next opcode if *code is accepted
2842                  NULL if *code is not accepted
2843    */
2844    
2845    static const pcre_uchar *
2846    get_chr_property_list(const pcre_uchar *code, BOOL utf,
2847      const pcre_uint8 *fcc, pcre_uint32 *list)
2848    {
2849    pcre_uchar c = *code;
2850    const pcre_uchar *end;
2851    const pcre_uint32 *clist_src;
2852    pcre_uint32 *clist_dest;
2853    pcre_uint32 chr;
2854    pcre_uchar base;
2855    
2856    list[0] = c;
2857    list[1] = FALSE;
2858    code++;
2859    
2860    if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
2861      {
2862      base = get_repeat_base(c);
2863      c -= (base - OP_STAR);
2864    
2865      if (c == OP_UPTO || c == OP_MINUPTO || c == OP_EXACT || c == OP_POSUPTO)
2866        code += IMM2_SIZE;
2867    
2868      list[1] = (c != OP_PLUS && c != OP_MINPLUS && c != OP_EXACT && c != OP_POSPLUS);
2869    
2870      switch(base)
2871        {
2872        case OP_STAR:
2873        list[0] = OP_CHAR;
2874        break;
2875    
2876        case OP_STARI:
2877        list[0] = OP_CHARI;
2878        break;
2879    
2880        case OP_NOTSTAR:
2881        list[0] = OP_NOT;
2882        break;
2883    
2884        case OP_NOTSTARI:
2885        list[0] = OP_NOTI;
2886        break;
2887    
2888        case OP_TYPESTAR:
2889        list[0] = *code;
2890        code++;
2891        break;
2892        }
2893      c = list[0];
2894      }
2895    
2896    switch(c)
2897      {
2898      case OP_NOT_DIGIT:
2899      case OP_DIGIT:
2900      case OP_NOT_WHITESPACE:
2901      case OP_WHITESPACE:
2902      case OP_NOT_WORDCHAR:
2903      case OP_WORDCHAR:
2904      case OP_ANY:
2905      case OP_ALLANY:
2906      case OP_ANYNL:
2907      case OP_NOT_HSPACE:
2908      case OP_HSPACE:
2909      case OP_NOT_VSPACE:
2910      case OP_VSPACE:
2911      case OP_EXTUNI:
2912      case OP_EODN:
2913      case OP_EOD:
2914      case OP_DOLL:
2915      case OP_DOLLM:
2916      return code;
2917    
2918      case OP_CHAR:
2919      case OP_NOT:
2920      GETCHARINCTEST(chr, code);
2921      list[2] = chr;
2922      list[3] = NOTACHAR;
2923      return code;
2924    
2925      case OP_CHARI:
2926      case OP_NOTI:
2927      list[0] = (c == OP_CHARI) ? OP_CHAR : OP_NOT;
2928      GETCHARINCTEST(chr, code);
2929      list[2] = chr;
2930    
2931    #ifdef SUPPORT_UCP
2932      if (chr < 128 || (chr < 256 && !utf))
2933        list[3] = fcc[chr];
2934      else
2935        list[3] = UCD_OTHERCASE(chr);
2936    #elif defined SUPPORT_UTF || !defined COMPILE_PCRE8
2937      list[3] = (chr < 256) ? fcc[chr] : chr;
2938    #else
2939      list[3] = fcc[chr];
2940    #endif
2941    
2942      /* The othercase might be the same value. */
2943    
2944      if (chr == list[3])
2945        list[3] = NOTACHAR;
2946      else
2947        list[4] = NOTACHAR;
2948      return code;
2949    
2950    #ifdef SUPPORT_UCP
2951      case OP_PROP:
2952      case OP_NOTPROP:
2953      if (code[0] != PT_CLIST)
2954        {
2955        list[2] = code[0];
2956        list[3] = code[1];
2957        return code + 2;
2958        }
2959    
2960      /* Convert only if we have enough space. */
2961    
2962      clist_src = PRIV(ucd_caseless_sets) + code[1];
2963      clist_dest = list + 2;
2964      code += 2;
2965    
2966      do {
2967         if (clist_dest >= list + 8)
2968           {
2969           /* Early return if there is not enough space. This should never
2970           happen, since all clists are shorter than 5 character now. */
2971           list[2] = code[0];
2972           list[3] = code[1];
2973           return code;
2974           }
2975         *clist_dest++ = *clist_src;
2976         }
2977      while(*clist_src++ != NOTACHAR);
2978    
2979      /* All characters are stored. The terminating NOTACHAR
2980      is copied form the clist itself. */
2981    
2982      list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
2983      return code;
2984    #endif
2985    
2986      case OP_NCLASS:
2987      case OP_CLASS:
2988    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2989      case OP_XCLASS:
2990      if (c == OP_XCLASS)
2991        end = code + GET(code, 0) - 1;
2992      else
2993    #endif
2994        end = code + 32 / sizeof(pcre_uchar);
2995    
2996      switch(*end)
2997        {
2998        case OP_CRSTAR:
2999        case OP_CRMINSTAR:
3000        case OP_CRQUERY:
3001        case OP_CRMINQUERY:
3002        case OP_CRPOSSTAR:
3003        case OP_CRPOSQUERY:
3004        list[1] = TRUE;
3005        end++;
3006        break;
3007    
3008        case OP_CRPLUS:
3009        case OP_CRMINPLUS:
3010        case OP_CRPOSPLUS:
3011        end++;
3012        break;
3013    
3014        case OP_CRRANGE:
3015        case OP_CRMINRANGE:
3016        case OP_CRPOSRANGE:
3017        list[1] = (GET2(end, 1) == 0);
3018        end += 1 + 2 * IMM2_SIZE;
3019        break;
3020        }
3021      list[2] = end - code;
3022      return end;
3023      }
3024    return NULL;    /* Opcode not accepted */
3025    }
3026    
3027    
3028    
3029    /*************************************************
3030    *    Scan further character sets for match       *
3031    *************************************************/
3032    
3033    /* Checks whether the base and the current opcode have a common character, in
3034    which case the base cannot be possessified.
3035    
3036    Arguments:
3037      code        points to the byte code
3038      utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3039      cd          static compile data
3040      base_list   the data list of the base opcode
3041    
3042    Returns:      TRUE if the auto-possessification is possible
3043    */
3044    
3045    static BOOL
3046    compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
3047      const pcre_uint32 *base_list, const pcre_uchar *base_end)
3048    {
3049    pcre_uchar c;
3050    pcre_uint32 list[8];
3051    const pcre_uint32 *chr_ptr;
3052    const pcre_uint32 *ochr_ptr;
3053    const pcre_uint32 *list_ptr;
3054    const pcre_uchar *next_code;
3055    const pcre_uint8 *class_bitset;
3056    const pcre_uint32 *set1, *set2, *set_end;
3057    pcre_uint32 chr;
3058    BOOL accepted, invert_bits;
3059    
3060    /* Note: the base_list[1] contains whether the current opcode has greedy
3061    (represented by a non-zero value) quantifier. This is a different from
3062    other character type lists, which stores here that the character iterator
3063    matches to an empty string (also represented by a non-zero value). */
3064    
3065    for(;;)
3066      {
3067      /* All operations move the code pointer forward.
3068      Therefore infinite recursions are not possible. */
3069    
3070      c = *code;
3071    
3072      /* Skip over callouts */
3073    
3074      if (c == OP_CALLOUT)
3075        {
3076        code += PRIV(OP_lengths)[c];
3077        continue;
3078        }
3079    
3080      if (c == OP_ALT)
3081        {
3082        do code += GET(code, 1); while (*code == OP_ALT);
3083        c = *code;
3084        }
3085    
3086      switch(c)
3087        {
3088        case OP_END:
3089        case OP_KETRPOS:
3090        /* TRUE only in greedy case. The non-greedy case could be replaced by
3091        an OP_EXACT, but it is probably not worth it. (And note that OP_EXACT
3092        uses more memory, which we cannot get at this stage.) */
3093    
3094        return base_list[1] != 0;
3095    
3096        case OP_KET:
3097        /* If the bracket is capturing, and referenced by an OP_RECURSE, or
3098        it is an atomic sub-pattern (assert, once, etc.) the non-greedy case
3099        cannot be converted to a possessive form. */
3100    
3101        if (base_list[1] == 0) return FALSE;
3102    
3103        switch(*(code - GET(code, 1)))
3104          {
3105          case OP_ASSERT:
3106          case OP_ASSERT_NOT:
3107          case OP_ASSERTBACK:
3108          case OP_ASSERTBACK_NOT:
3109          case OP_ONCE:
3110          case OP_ONCE_NC:
3111          /* Atomic sub-patterns and assertions can always auto-possessify their
3112          last iterator. */
3113          return TRUE;
3114          }
3115    
3116        code += PRIV(OP_lengths)[c];
3117        continue;
3118    
3119        case OP_ONCE:
3120        case OP_ONCE_NC:
3121        case OP_BRA:
3122        case OP_CBRA:
3123        next_code = code + GET(code, 1);
3124        code += PRIV(OP_lengths)[c];
3125    
3126        while (*next_code == OP_ALT)
3127          {
3128          if (!compare_opcodes(code, utf, cd, base_list, base_end)) return FALSE;
3129          code = next_code + 1 + LINK_SIZE;
3130          next_code += GET(next_code, 1);
3131          }
3132        continue;
3133    
3134        case OP_BRAZERO:
3135        case OP_BRAMINZERO:
3136    
3137        next_code = code + 1;
3138        if (*next_code != OP_BRA && *next_code != OP_CBRA
3139            && *next_code != OP_ONCE && *next_code != OP_ONCE_NC) return FALSE;
3140    
3141        do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
3142    
3143        /* The bracket content will be checked by the
3144        OP_BRA/OP_CBRA case above. */
3145        next_code += 1 + LINK_SIZE;
3146        if (!compare_opcodes(next_code, utf, cd, base_list, base_end))
3147          return FALSE;
3148    
3149        code += PRIV(OP_lengths)[c];
3150        continue;
3151        }
3152    
3153      /* Check for a supported opcode, and load its properties. */
3154    
3155      code = get_chr_property_list(code, utf, cd->fcc, list);
3156      if (code == NULL) return FALSE;    /* Unsupported */
3157    
3158      /* If either opcode is a small character list, set pointers for comparing
3159      characters from that list with another list, or with a property. */
3160    
3161      if (base_list[0] == OP_CHAR)
3162        {
3163        chr_ptr = base_list + 2;
3164        list_ptr = list;
3165        }
3166      else if (list[0] == OP_CHAR)
3167        {
3168        chr_ptr = list + 2;
3169        list_ptr = base_list;
3170        }
3171    
3172      /* Character bitsets can also be compared to certain opcodes. */
3173    
3174      else if (base_list[0] == OP_CLASS || list[0] == OP_CLASS
3175    #ifdef COMPILE_PCRE8
3176          /* In 8 bit, non-UTF mode, OP_CLASS and OP_NCLASS are the same. */
3177          || (!utf && (base_list[0] == OP_NCLASS || list[0] == OP_NCLASS))
3178    #endif
3179          )
3180        {
3181    #ifdef COMPILE_PCRE8
3182        if (base_list[0] == OP_CLASS || (!utf && base_list[0] == OP_NCLASS))
3183    #else
3184        if (base_list[0] == OP_CLASS)
3185    #endif
3186          {
3187          set1 = (pcre_uint32 *)(base_end - base_list[2]);
3188          list_ptr = list;
3189          }
3190        else
3191          {
3192          set1 = (pcre_uint32 *)(code - list[2]);
3193          list_ptr = base_list;
3194          }
3195    
3196        invert_bits = FALSE;
3197        switch(list_ptr[0])
3198          {
3199          case OP_CLASS:
3200          case OP_NCLASS:
3201          set2 = (pcre_uint32 *)
3202            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3203          break;
3204    
3205          /* OP_XCLASS cannot be supported here, because its bitset
3206          is not necessarily complete. E.g: [a-\0x{200}] is stored
3207          as a character range, and the appropriate bits are not set. */
3208    
3209          case OP_NOT_DIGIT:
3210            invert_bits = TRUE;
3211            /* Fall through */
3212          case OP_DIGIT:
3213            set2 = (pcre_uint32 *)(cd->cbits + cbit_digit);
3214            break;
3215    
3216          case OP_NOT_WHITESPACE:
3217            invert_bits = TRUE;
3218            /* Fall through */
3219          case OP_WHITESPACE:
3220            set2 = (pcre_uint32 *)(cd->cbits + cbit_space);
3221            break;
3222    
3223          case OP_NOT_WORDCHAR:
3224            invert_bits = TRUE;
3225            /* Fall through */
3226          case OP_WORDCHAR:
3227            set2 = (pcre_uint32 *)(cd->cbits + cbit_word);
3228            break;
3229    
3230          default:
3231          return FALSE;
3232          }
3233    
3234        /* Compare 4 bytes to improve speed. */
3235        set_end = set1 + (32 / 4);
3236        if (invert_bits)
3237          {
3238          do
3239            {
3240            if ((*set1++ & ~(*set2++)) != 0) return FALSE;
3241            }
3242          while (set1 < set_end);
3243          }
3244        else
3245          {
3246          do
3247            {
3248            if ((*set1++ & *set2++) != 0) return FALSE;
3249            }
3250          while (set1 < set_end);
3251          }
3252    
3253        if (list[1] == 0) return TRUE;
3254        /* Might be an empty repeat. */
3255        continue;
3256        }
3257    
3258      /* Some property combinations also acceptable. Unicode property opcodes are
3259      processed specially; the rest can be handled with a lookup table. */
3260    
3261      else
3262        {
3263        pcre_uint32 leftop, rightop;
3264    
3265        leftop = base_list[0];
3266        rightop = list[0];
3267    
3268    #ifdef SUPPORT_UCP
3269        accepted = FALSE; /* Always set in non-unicode case. */
3270        if (leftop == OP_PROP || leftop == OP_NOTPROP)
3271          {
3272          if (rightop == OP_EOD)
3273            accepted = TRUE;
3274          else if (rightop == OP_PROP || rightop == OP_NOTPROP)
3275            {
3276            int n;
3277            const pcre_uint8 *p;
3278            BOOL same = leftop == rightop;
3279            BOOL lisprop = leftop == OP_PROP;
3280            BOOL risprop = rightop == OP_PROP;
3281            BOOL bothprop = lisprop && risprop;
3282    
3283            /* There's a table that specifies how each combination is to be
3284            processed:
3285              0   Always return FALSE (never auto-possessify)
3286              1   Character groups are distinct (possessify if both are OP_PROP)
3287              2   Check character categories in the same group (general or particular)
3288              3   Return TRUE if the two opcodes are not the same
3289              ... see comments below
3290            */
3291    
3292            n = propposstab[base_list[2]][list[2]];
3293            switch(n)
3294              {
3295              case 0: break;
3296              case 1: accepted = bothprop; break;
3297              case 2: accepted = (base_list[3] == list[3]) != same; break;
3298              case 3: accepted = !same; break;
3299    
3300              case 4:  /* Left general category, right particular category */
3301              accepted = risprop && catposstab[base_list[3]][list[3]] == same;
3302              break;
3303    
3304              case 5:  /* Right general category, left particular category */
3305              accepted = lisprop && catposstab[list[3]][base_list[3]] == same;
3306              break;
3307    
3308              /* This code is logically tricky. Think hard before fiddling with it.
3309              The posspropstab table has four entries per row. Each row relates to
3310              one of PCRE's special properties such as ALNUM or SPACE or WORD.
3311              Only WORD actually needs all four entries, but using repeats for the
3312              others means they can all use the same code below.
3313    
3314              The first two entries in each row are Unicode general categories, and
3315              apply always, because all the characters they include are part of the
3316              PCRE character set. The third and fourth entries are a general and a
3317              particular category, respectively, that include one or more relevant
3318              characters. One or the other is used, depending on whether the check
3319              is for a general or a particular category. However, in both cases the
3320              category contains more characters than the specials that are defined
3321              for the property being tested against. Therefore, it cannot be used
3322              in a NOTPROP case.
3323    
3324              Example: the row for WORD contains ucp_L, ucp_N, ucp_P, ucp_Po.
3325              Underscore is covered by ucp_P or ucp_Po. */
3326    
3327              case 6:  /* Left alphanum vs right general category */
3328              case 7:  /* Left space vs right general category */
3329              case 8:  /* Left word vs right general category */
3330              p = posspropstab[n-6];
3331              accepted = risprop && lisprop ==
3332                (list[3] != p[0] &&
3333                 list[3] != p[1] &&
3334                (list[3] != p[2] || !lisprop));
3335              break;
3336    
3337              case 9:   /* Right alphanum vs left general category */
3338              case 10:  /* Right space vs left general category */
3339              case 11:  /* Right word vs left general category */
3340              p = posspropstab[n-9];
3341              accepted = lisprop && risprop ==
3342                (base_list[3] != p[0] &&
3343                 base_list[3] != p[1] &&
3344                (base_list[3] != p[2] || !risprop));
3345              break;
3346    
3347              case 12:  /* Left alphanum vs right particular category */
3348              case 13:  /* Left space vs right particular category */
3349              case 14:  /* Left word vs right particular category */
3350              p = posspropstab[n-12];
3351              accepted = risprop && lisprop ==
3352                (catposstab[p[0]][list[3]] &&
3353                 catposstab[p[1]][list[3]] &&
3354                (list[3] != p[3] || !lisprop));
3355              break;
3356    
3357              case 15:  /* Right alphanum vs left particular category */
3358              case 16:  /* Right space vs left particular category */
3359              case 17:  /* Right word vs left particular category */
3360              p = posspropstab[n-15];
3361              accepted = lisprop && risprop ==
3362                (catposstab[p[0]][base_list[3]] &&
3363                 catposstab[p[1]][base_list[3]] &&
3364                (base_list[3] != p[3] || !risprop));
3365              break;
3366              }
3367            }
3368          }
3369    
3370        else
3371    #endif  /* SUPPORT_UCP */
3372    
3373        accepted = leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&
3374               rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&
3375               autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];
3376    
3377        if (!accepted)
3378          return FALSE;
3379    
3380        if (list[1] == 0) return TRUE;
3381        /* Might be an empty repeat. */
3382        continue;
3383        }
3384    
3385      /* Control reaches here only if one of the items is a small character list.
3386      All characters are checked against the other side. */
3387    
3388      do
3389        {
3390        chr = *chr_ptr;
3391    
3392        switch(list_ptr[0])
3393          {
3394          case OP_CHAR:
3395          ochr_ptr = list_ptr + 2;
3396          do
3397            {
3398            if (chr == *ochr_ptr) return FALSE;
3399            ochr_ptr++;
3400            }
3401          while(*ochr_ptr != NOTACHAR);
3402          break;
3403    
3404          case OP_NOT:
3405          ochr_ptr = list_ptr + 2;
3406          do
3407            {
3408            if (chr == *ochr_ptr)
3409              break;
3410            ochr_ptr++;
3411            }
3412          while(*ochr_ptr != NOTACHAR);
3413          if (*ochr_ptr == NOTACHAR) return FALSE;   /* Not found */
3414          break;
3415    
3416          /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not*
3417          set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3418    
3419          case OP_DIGIT:
3420          if (chr < 256 && (cd->ctypes[chr] & ctype_digit) != 0) return FALSE;
3421          break;
3422    
3423          case OP_NOT_DIGIT:
3424          if (chr > 255 || (cd->ctypes[chr] & ctype_digit) == 0) return FALSE;
3425          break;
3426    
3427          case OP_WHITESPACE:
3428          if (chr < 256 && (cd->ctypes[chr] & ctype_space) != 0) return FALSE;
3429          break;
3430    
3431          case OP_NOT_WHITESPACE:
3432          if (chr > 255 || (cd->ctypes[chr] & ctype_space) == 0) return FALSE;
3433          break;
3434    
3435          case OP_WORDCHAR:
3436          if (chr < 255 && (cd->ctypes[chr] & ctype_word) != 0) return FALSE;
3437          break;
3438    
3439          case OP_NOT_WORDCHAR:
3440          if (chr > 255 || (cd->ctypes[chr] & ctype_word) == 0) return FALSE;
3441          break;
3442    
3443          case OP_HSPACE:
3444          switch(chr)
3445            {
3446            HSPACE_CASES: return FALSE;
3447            default: break;
3448            }
3449          break;
3450    
3451          case OP_NOT_HSPACE:
3452          switch(chr)
3453            {
3454            HSPACE_CASES: break;
3455            default: return FALSE;
3456            }
3457          break;
3458    
3459          case OP_ANYNL:
3460          case OP_VSPACE:
3461          switch(chr)
3462            {
3463            VSPACE_CASES: return FALSE;
3464            default: break;
3465            }
3466          break;
3467    
3468          case OP_NOT_VSPACE:
3469          switch(chr)
3470            {
3471            VSPACE_CASES: break;
3472            default: return FALSE;
3473            }
3474          break;
3475    
3476          case OP_DOLL:
3477          case OP_EODN:
3478          switch (chr)
3479            {
3480            case CHAR_CR:
3481            case CHAR_LF:
3482            case CHAR_VT:
3483            case CHAR_FF:
3484            case CHAR_NEL:
3485    #ifndef EBCDIC
3486            case 0x2028:
3487            case 0x2029:
3488    #endif  /* Not EBCDIC */
3489            return FALSE;
3490            }
3491          break;
3492    
3493