/[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 93 by nigel, Sat Feb 24 21:41:42 2007 UTC revision 1613 by ph10, Fri Nov 27 17:41:04 2015 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2006 University of Cambridge             Copyright (c) 1997-2014 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 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  #define NLBLOCK cd             /* Block containing newline information */  #ifdef HAVE_CONFIG_H
46  #define PSSTART start_pattern  /* Field containing processed string start */  #include "config.h"
47  #define PSEND   end_pattern    /* Field containing processed string end */  #endif
48    
49    #define NLBLOCK cd             /* Block containing newline information */
50    #define PSSTART start_pattern  /* Field containing pattern start */
51    #define PSEND   end_pattern    /* Field containing pattern 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  #if !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,       ESC_a,
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 105  static const short int escapes[] = { Line 202  static const short int escapes[] = {
202  /*  68 */     0,     0,    '|',     ',',    '%',   '_',    '>',    '?',  /*  68 */     0,     0,    '|',     ',',    '%',   '_',    '>',    '?',
203  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,
204  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',
205  /*  80 */     0,     7, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,  /*  80 */     0, ESC_a, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,
206  /*  88 */     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,       0,      0, ESC_n,      0, -ESC_p,
208  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,
209  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,     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
221  };  };
222    
223    /* We also need a table of characters that may follow \c in an EBCDIC
224    environment for characters 0-31. */
225    
226    static unsigned char ebcdic_escape_c[] = "@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\\]^_";
227    
228  #endif  #endif
229    
230    
231  /* 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
232  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
233  as this is assumed for handling case independence. */  the number of relocations when a shared library is dynamically linked. The
234    string is built from string macros so that it works in UTF-8 mode on EBCDIC
235  static const char *const posix_names[] = {  platforms. */
236    "alpha", "lower", "upper",  
237    "alnum", "ascii", "blank", "cntrl", "digit", "graph",  typedef struct verbitem {
238    "print", "punct", "space", "word",  "xdigit" };    int   len;                 /* Length of verb name */
239      int   op;                  /* Op when no arg, or -1 if arg mandatory */
240      int   op_arg;              /* Op when arg present, or -1 if not allowed */
241    } verbitem;
242    
243    static const char verbnames[] =
244      "\0"                       /* Empty name is a shorthand for MARK */
245      STRING_MARK0
246      STRING_ACCEPT0
247      STRING_COMMIT0
248      STRING_F0
249      STRING_FAIL0
250      STRING_PRUNE0
251      STRING_SKIP0
252      STRING_THEN;
253    
254    static const verbitem verbs[] = {
255      { 0, -1,        OP_MARK },
256      { 4, -1,        OP_MARK },
257      { 6, OP_ACCEPT, -1 },
258      { 6, OP_COMMIT, -1 },
259      { 1, OP_FAIL,   -1 },
260      { 4, OP_FAIL,   -1 },
261      { 5, OP_PRUNE,  OP_PRUNE_ARG },
262      { 4, OP_SKIP,   OP_SKIP_ARG  },
263      { 4, OP_THEN,   OP_THEN_ARG  }
264    };
265    
266    static const int verbcount = sizeof(verbs)/sizeof(verbitem);
267    
268    
269  static const uschar posix_name_lengths[] = {  /* Substitutes for [[:<:]] and [[:>:]], which mean start and end of word in
270    another regex library. */
271    
272    static const pcre_uchar sub_start_of_word[] = {
273      CHAR_BACKSLASH, CHAR_b, CHAR_LEFT_PARENTHESIS, CHAR_QUESTION_MARK,
274      CHAR_EQUALS_SIGN, CHAR_BACKSLASH, CHAR_w, CHAR_RIGHT_PARENTHESIS, '\0' };
275    
276    static const pcre_uchar sub_end_of_word[] = {
277      CHAR_BACKSLASH, CHAR_b, CHAR_LEFT_PARENTHESIS, CHAR_QUESTION_MARK,
278      CHAR_LESS_THAN_SIGN, CHAR_EQUALS_SIGN, CHAR_BACKSLASH, CHAR_w,
279      CHAR_RIGHT_PARENTHESIS, '\0' };
280    
281    
282    /* Tables of names of POSIX character classes and their lengths. The names are
283    now all in a single string, to reduce the number of relocations when a shared
284    library is dynamically loaded. The list of lengths is terminated by a zero
285    length entry. The first three must be alpha, lower, upper, as this is assumed
286    for handling case independence. The indices for graph, print, and punct are
287    needed, so identify them. */
288    
289    static const char posix_names[] =
290      STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
291      STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
292      STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
293      STRING_word0  STRING_xdigit;
294    
295    static const pcre_uint8 posix_name_lengths[] = {
296    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
297    
298    #define PC_GRAPH  8
299    #define PC_PRINT  9
300    #define PC_PUNCT 10
301    
302    
303  /* Table of class bit maps for each POSIX class. Each class is formed from a  /* Table of class bit maps for each POSIX class. Each class is formed from a
304  base map, with an optional addition or removal of another map. Then, for some  base map, with an optional addition or removal of another map. Then, for some
305  classes, there is some additional tweaking: for [:blank:] the vertical space  classes, there is some additional tweaking: for [:blank:] the vertical space
# Line 164  static const int posix_class_maps[] = { Line 327  static const int posix_class_maps[] = {
327    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
328  };  };
329    
330    /* Table of substitutes for \d etc when PCRE_UCP is set. They are replaced by
331    Unicode property escapes. */
332    
333    #ifdef SUPPORT_UCP
334    static const pcre_uchar string_PNd[]  = {
335      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
336      CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
337    static const pcre_uchar string_pNd[]  = {
338      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
339      CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
340    static const pcre_uchar string_PXsp[] = {
341      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
342      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
343    static const pcre_uchar string_pXsp[] = {
344      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
345      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
346    static const pcre_uchar string_PXwd[] = {
347      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
348      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
349    static const pcre_uchar string_pXwd[] = {
350      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
351      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
352    
353    static const pcre_uchar *substitutes[] = {
354      string_PNd,           /* \D */
355      string_pNd,           /* \d */
356      string_PXsp,          /* \S */   /* Xsp is Perl space, but from 8.34, Perl */
357      string_pXsp,          /* \s */   /* space and POSIX space are the same. */
358      string_PXwd,          /* \W */
359      string_pXwd           /* \w */
360    };
361    
362    /* The POSIX class substitutes must be in the order of the POSIX class names,
363    defined above, and there are both positive and negative cases. NULL means no
364    general substitute of a Unicode property escape (\p or \P). However, for some
365    POSIX classes (e.g. graph, print, punct) a special property code is compiled
366    directly. */
367    
368    static const pcre_uchar string_pL[] =   {
369      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
370      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
371    static const pcre_uchar string_pLl[] =  {
372      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
373      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
374    static const pcre_uchar string_pLu[] =  {
375      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
376      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
377    static const pcre_uchar string_pXan[] = {
378      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
379      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
380    static const pcre_uchar string_h[] =    {
381      CHAR_BACKSLASH, CHAR_h, '\0' };
382    static const pcre_uchar string_pXps[] = {
383      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
384      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
385    static const pcre_uchar string_PL[] =   {
386      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
387      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
388    static const pcre_uchar string_PLl[] =  {
389      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
390      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
391    static const pcre_uchar string_PLu[] =  {
392      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
393      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
394    static const pcre_uchar string_PXan[] = {
395      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
396      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
397    static const pcre_uchar string_H[] =    {
398      CHAR_BACKSLASH, CHAR_H, '\0' };
399    static const pcre_uchar string_PXps[] = {
400      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
401      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
402    
403    static const pcre_uchar *posix_substitutes[] = {
404      string_pL,            /* alpha */
405      string_pLl,           /* lower */
406      string_pLu,           /* upper */
407      string_pXan,          /* alnum */
408      NULL,                 /* ascii */
409      string_h,             /* blank */
410      NULL,                 /* cntrl */
411      string_pNd,           /* digit */
412      NULL,                 /* graph */
413      NULL,                 /* print */
414      NULL,                 /* punct */
415      string_pXps,          /* space */   /* Xps is POSIX space, but from 8.34 */
416      string_pXwd,          /* word  */   /* Perl and POSIX space are the same */
417      NULL,                 /* xdigit */
418      /* Negated cases */
419      string_PL,            /* ^alpha */
420      string_PLl,           /* ^lower */
421      string_PLu,           /* ^upper */
422      string_PXan,          /* ^alnum */
423      NULL,                 /* ^ascii */
424      string_H,             /* ^blank */
425      NULL,                 /* ^cntrl */
426      string_PNd,           /* ^digit */
427      NULL,                 /* ^graph */
428      NULL,                 /* ^print */
429      NULL,                 /* ^punct */
430      string_PXps,          /* ^space */  /* Xps is POSIX space, but from 8.34 */
431      string_PXwd,          /* ^word */   /* Perl and POSIX space are the same */
432      NULL                  /* ^xdigit */
433    };
434    #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))
435    #endif
436    
437  #define STRING(a)  # a  #define STRING(a)  # a
438  #define XSTRING(s) STRING(s)  #define XSTRING(s) STRING(s)
# Line 171  static const int posix_class_maps[] = { Line 440  static const int posix_class_maps[] = {
440  /* The texts of compile-time error messages. These are "char *" because they  /* The texts of compile-time error messages. These are "char *" because they
441  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
442  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
443  are no longer used. */  are no longer used. This used to be a table of strings, but in order to reduce
444    the number of relocations needed when a shared library is loaded dynamically,
445  static const char *error_texts[] = {  it is now one long string. We cannot use a table of offsets, because the
446    "no error",  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
447    "\\ at end of pattern",  simply count through to the one we want - this isn't a performance issue
448    "\\c at end of pattern",  because these strings are used only when there is a compilation error.
449    "unrecognized character follows \\",  
450    "numbers out of order in {} quantifier",  Each substring ends with \0 to insert a null character. This includes the final
451    substring, so that the whole string ends with \0\0, which can be detected when
452    counting through. */
453    
454    static const char error_texts[] =
455      "no error\0"
456      "\\ at end of pattern\0"
457      "\\c at end of pattern\0"
458      "unrecognized character follows \\\0"
459      "numbers out of order in {} quantifier\0"
460    /* 5 */    /* 5 */
461    "number too big in {} quantifier",    "number too big in {} quantifier\0"
462    "missing terminating ] for character class",    "missing terminating ] for character class\0"
463    "invalid escape sequence in character class",    "invalid escape sequence in character class\0"
464    "range out of order in character class",    "range out of order in character class\0"
465    "nothing to repeat",    "nothing to repeat\0"
466    /* 10 */    /* 10 */
467    "operand of unlimited repeat could match the empty string",  /** DEAD **/    "internal error: invalid forward reference offset\0"
468    "internal error: unexpected repeat",    "internal error: unexpected repeat\0"
469    "unrecognized character after (?",    "unrecognized character after (? or (?-\0"
470    "POSIX named classes are supported only within a class",    "POSIX named classes are supported only within a class\0"
471    "missing )",    "missing )\0"
472    /* 15 */    /* 15 */
473    "reference to non-existent subpattern",    "reference to non-existent subpattern\0"
474    "erroffset passed as NULL",    "erroffset passed as NULL\0"
475    "unknown option bit(s) set",    "unknown option bit(s) set\0"
476    "missing ) after comment",    "missing ) after comment\0"
477    "parentheses nested too deeply",  /** DEAD **/    "parentheses nested too deeply\0"  /** DEAD **/
478    /* 20 */    /* 20 */
479    "regular expression too large",    "regular expression is too large\0"
480    "failed to get memory",    "failed to get memory\0"
481    "unmatched parentheses",    "unmatched parentheses\0"
482    "internal error: code overflow",    "internal error: code overflow\0"
483    "unrecognized character after (?<",    "unrecognized character after (?<\0"
484    /* 25 */    /* 25 */
485    "lookbehind assertion is not fixed length",    "lookbehind assertion is not fixed length\0"
486    "malformed number or name after (?(",    "malformed number or name after (?(\0"
487    "conditional group contains more than two branches",    "conditional group contains more than two branches\0"
488    "assertion expected after (?(",    "assertion expected after (?(\0"
489    "(?R or (?digits must be followed by )",    "(?R or (?[+-]digits must be followed by )\0"
490    /* 30 */    /* 30 */
491    "unknown POSIX class name",    "unknown POSIX class name\0"
492    "POSIX collating elements are not supported",    "POSIX collating elements are not supported\0"
493    "this version of PCRE is not compiled with PCRE_UTF8 support",    "this version of PCRE is compiled without UTF support\0"
494    "spare error",  /** DEAD **/    "spare error\0"  /** DEAD **/
495    "character value in \\x{...} sequence is too large",    "character value in \\x{} or \\o{} is too large\0"
496    /* 35 */    /* 35 */
497    "invalid condition (?(0)",    "invalid condition (?(0)\0"
498    "\\C not allowed in lookbehind assertion",    "\\C not allowed in lookbehind assertion\0"
499    "PCRE does not support \\L, \\l, \\N, \\U, or \\u",    "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
500    "number after (?C is > 255",    "number after (?C is > 255\0"
501    "closing ) for (?C expected",    "closing ) for (?C expected\0"
502    /* 40 */    /* 40 */
503    "recursive call could loop indefinitely",    "recursive call could loop indefinitely\0"
504    "unrecognized character after (?P",    "unrecognized character after (?P\0"
505    "syntax error in subpattern name (missing terminator)",    "syntax error in subpattern name (missing terminator)\0"
506    "two named subpatterns have the same name",    "two named subpatterns have the same name\0"
507    "invalid UTF-8 string",    "invalid UTF-8 string\0"
508    /* 45 */    /* 45 */
509    "support for \\P, \\p, and \\X has not been compiled",    "support for \\P, \\p, and \\X has not been compiled\0"
510    "malformed \\P or \\p sequence",    "malformed \\P or \\p sequence\0"
511    "unknown property name after \\P or \\p",    "unknown property name after \\P or \\p\0"
512    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)",    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
513    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")",    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
514    /* 50 */    /* 50 */
515    "repeated subpattern is too long",    "repeated subpattern is too long\0"    /** DEAD **/
516    "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"
517    "internal error: overran compiling workspace",    "internal error: overran compiling workspace\0"
518    "internal error: previously-checked referenced subpattern not found",    "internal error: previously-checked referenced subpattern not found\0"
519    "DEFINE group contains more than one branch",    "DEFINE group contains more than one branch\0"
520    /* 55 */    /* 55 */
521    "repeating a DEFINE group is not allowed",    "repeating a DEFINE group is not allowed\0"  /** DEAD **/
522    "inconsistent NEWLINE options",    "inconsistent NEWLINE options\0"
523    "\\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"
524  };    "a numbered reference must not be zero\0"
525      "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
526      /* 60 */
527      "(*VERB) not recognized or malformed\0"
528      "number is too big\0"
529      "subpattern name expected\0"
530      "digit expected after (?+\0"
531      "] is an invalid data character in JavaScript compatibility mode\0"
532      /* 65 */
533      "different names for subpatterns of the same number are not allowed\0"
534      "(*MARK) must have an argument\0"
535      "this version of PCRE is not compiled with Unicode property support\0"
536    #ifndef EBCDIC
537      "\\c must be followed by an ASCII character\0"
538    #else
539      "\\c must be followed by a letter or one of [\\]^_?\0"
540    #endif
541      "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
542      /* 70 */
543      "internal error: unknown opcode in find_fixedlength()\0"
544      "\\N is not supported in a class\0"
545      "too many forward references\0"
546      "disallowed Unicode code point (>= 0xd800 && <= 0xdfff)\0"
547      "invalid UTF-16 string\0"
548      /* 75 */
549      "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"
550      "character value in \\u.... sequence is too large\0"
551      "invalid UTF-32 string\0"
552      "setting UTF is disabled by the application\0"
553      "non-hex character in \\x{} (closing brace missing?)\0"
554      /* 80 */
555      "non-octal character in \\o{} (closing brace missing?)\0"
556      "missing opening brace after \\o\0"
557      "parentheses are too deeply nested\0"
558      "invalid range in character class\0"
559      "group name must start with a non-digit\0"
560      /* 85 */
561      "parentheses are too deeply nested (stack check)\0"
562      "digits missing in \\x{} or \\o{}\0"
563      ;
564    
565  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
566  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 578  For convenience, we use the same bit def
578    
579  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
580    
581  #if !EBCDIC    /* This is the "normal" case, for ASCII systems */  /* Using a simple comparison for decimal numbers rather than a memory read
582  static const unsigned char digitab[] =  is much faster, and the resulting code is simpler (the compiler turns it
583    into a subtraction and unsigned comparison). */
584    
585    #define IS_DIGIT(x) ((x) >= CHAR_0 && (x) <= CHAR_9)
586    
587    #ifndef EBCDIC
588    
589    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
590    UTF-8 mode. */
591    
592    static const pcre_uint8 digitab[] =
593    {    {
594    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
595    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 624  static const unsigned char digitab[] =
624    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
625    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
626    
627  #else          /* This is the "abnormal" case, for EBCDIC systems */  #else
628  static const unsigned char digitab[] =  
629    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
630    
631    static const pcre_uint8 digitab[] =
632    {    {
633    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
634    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15    */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15    */
# Line 312  static const unsigned char digitab[] = Line 641  static const unsigned char digitab[] =
641    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */
642    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */
643    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */
644    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88-     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88- 95    */
645    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */
646    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */
647    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */
# Line 334  static const unsigned char digitab[] = Line 663  static const unsigned char digitab[] =
663    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */
664    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */
665    
666  static const unsigned char ebcdic_chartab[] = { /* chartable partial dup */  static const pcre_uint8 ebcdic_chartab[] = { /* chartable partial dup */
667    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */
668    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */
669    0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*  16- 23 */    0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*  16- 23 */
# Line 346  static const unsigned char ebcdic_charta Line 675  static const unsigned char ebcdic_charta
675    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */
676    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */
677    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */
678    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88-  */    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88- 95 */
679    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */
680    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */
681    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */
# Line 370  static const unsigned char ebcdic_charta Line 699  static const unsigned char ebcdic_charta
699  #endif  #endif
700    
701    
702  /* Definition to allow mutual recursion */  /* This table is used to check whether auto-possessification is possible
703    between adjacent character-type opcodes. The left-hand (repeated) opcode is
704    used to select the row, and the right-hand opcode is use to select the column.
705    A value of 1 means that auto-possessification is OK. For example, the second
706    value in the first row means that \D+\d can be turned into \D++\d.
707    
708    The Unicode property types (\P and \p) have to be present to fill out the table
709    because of what their opcode values are, but the table values should always be
710    zero because property types are handled separately in the code. The last four
711    columns apply to items that cannot be repeated, so there is no need to have
712    rows for them. Note that OP_DIGIT etc. are generated only when PCRE_UCP is
713    *not* set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
714    
715    #define APTROWS (LAST_AUTOTAB_LEFT_OP - FIRST_AUTOTAB_OP + 1)
716    #define APTCOLS (LAST_AUTOTAB_RIGHT_OP - FIRST_AUTOTAB_OP + 1)
717    
718    static const pcre_uint8 autoposstab[APTROWS][APTCOLS] = {
719    /* \D \d \S \s \W \w  . .+ \C \P \p \R \H \h \V \v \X \Z \z  $ $M */
720      { 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \D */
721      { 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \d */
722      { 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \S */
723      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \s */
724      { 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \W */
725      { 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 },  /* \w */
726      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .  */
727      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* .+ */
728      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 },  /* \C */
729      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \P */
730      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },  /* \p */
731      { 0, 1, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \R */
732      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 },  /* \H */
733      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \h */
734      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0 },  /* \V */
735      { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0 },  /* \v */
736      { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }   /* \X */
737    };
738    
739    
740    /* This table is used to check whether auto-possessification is possible
741    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP). The
742    left-hand (repeated) opcode is used to select the row, and the right-hand
743    opcode is used to select the column. The values are as follows:
744    
745      0   Always return FALSE (never auto-possessify)
746      1   Character groups are distinct (possessify if both are OP_PROP)
747      2   Check character categories in the same group (general or particular)
748      3   TRUE if the two opcodes are not the same (PROP vs NOTPROP)
749    
750      4   Check left general category vs right particular category
751      5   Check right general category vs left particular category
752    
753      6   Left alphanum vs right general category
754      7   Left space vs right general category
755      8   Left word vs right general category
756    
757      9   Right alphanum vs left general category
758     10   Right space vs left general category
759     11   Right word vs left general category
760    
761     12   Left alphanum vs right particular category
762     13   Left space vs right particular category
763     14   Left word vs right particular category
764    
765     15   Right alphanum vs left particular category
766     16   Right space vs left particular category
767     17   Right word vs left particular category
768    */
769    
770    static const pcre_uint8 propposstab[PT_TABSIZE][PT_TABSIZE] = {
771    /* ANY LAMP GC  PC  SC ALNUM SPACE PXSPACE WORD CLIST UCNC */
772      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_ANY */
773      { 0,  3,  0,  0,  0,    3,    1,      1,   0,    0,   0 },  /* PT_LAMP */
774      { 0,  0,  2,  4,  0,    9,   10,     10,  11,    0,   0 },  /* PT_GC */
775      { 0,  0,  5,  2,  0,   15,   16,     16,  17,    0,   0 },  /* PT_PC */
776      { 0,  0,  0,  0,  2,    0,    0,      0,   0,    0,   0 },  /* PT_SC */
777      { 0,  3,  6, 12,  0,    3,    1,      1,   0,    0,   0 },  /* PT_ALNUM */
778      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_SPACE */
779      { 0,  1,  7, 13,  0,    1,    3,      3,   1,    0,   0 },  /* PT_PXSPACE */
780      { 0,  0,  8, 14,  0,    0,    1,      1,   3,    0,   0 },  /* PT_WORD */
781      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   0 },  /* PT_CLIST */
782      { 0,  0,  0,  0,  0,    0,    0,      0,   0,    0,   3 }   /* PT_UCNC */
783    };
784    
785    /* This table is used to check whether auto-possessification is possible
786    between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP) when one
787    specifies a general category and the other specifies a particular category. The
788    row is selected by the general category and the column by the particular
789    category. The value is 1 if the particular category is not part of the general
790    category. */
791    
792    static const pcre_uint8 catposstab[7][30] = {
793    /* Cc Cf Cn Co Cs Ll Lm Lo Lt Lu Mc Me Mn Nd Nl No Pc Pd Pe Pf Pi Po Ps Sc Sk Sm So Zl Zp Zs */
794      { 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* C */
795      { 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* L */
796      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* M */
797      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 },  /* N */
798      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1 },  /* P */
799      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 1, 1, 1 },  /* S */
800      { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0 }   /* Z */
801    };
802    
803    /* This table is used when checking ALNUM, (PX)SPACE, SPACE, and WORD against
804    a general or particular category. The properties in each row are those
805    that apply to the character set in question. Duplication means that a little
806    unnecessary work is done when checking, but this keeps things much simpler
807    because they can all use the same code. For more details see the comment where
808    this table is used.
809    
810    Note: SPACE and PXSPACE used to be different because Perl excluded VT from
811    "space", but from Perl 5.18 it's included, so both categories are treated the
812    same here. */
813    
814    static const pcre_uint8 posspropstab[3][4] = {
815      { ucp_L, ucp_N, ucp_N, ucp_Nl },  /* ALNUM, 3rd and 4th values redundant */
816      { ucp_Z, ucp_Z, ucp_C, ucp_Cc },  /* SPACE and PXSPACE, 2nd value redundant */
817      { ucp_L, ucp_N, ucp_P, ucp_Po }   /* WORD */
818    };
819    
820    /* This table is used when converting repeating opcodes into possessified
821    versions as a result of an explicit possessive quantifier such as ++. A zero
822    value means there is no possessified version - in those cases the item in
823    question must be wrapped in ONCE brackets. The table is truncated at OP_CALLOUT
824    because all relevant opcodes are less than that. */
825    
826    static const pcre_uint8 opcode_possessify[] = {
827      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 0 - 15  */
828      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 16 - 31 */
829    
830      0,                       /* NOTI */
831      OP_POSSTAR, 0,           /* STAR, MINSTAR */
832      OP_POSPLUS, 0,           /* PLUS, MINPLUS */
833      OP_POSQUERY, 0,          /* QUERY, MINQUERY */
834      OP_POSUPTO, 0,           /* UPTO, MINUPTO */
835      0,                       /* EXACT */
836      0, 0, 0, 0,              /* POS{STAR,PLUS,QUERY,UPTO} */
837    
838      OP_POSSTARI, 0,          /* STARI, MINSTARI */
839      OP_POSPLUSI, 0,          /* PLUSI, MINPLUSI */
840      OP_POSQUERYI, 0,         /* QUERYI, MINQUERYI */
841      OP_POSUPTOI, 0,          /* UPTOI, MINUPTOI */
842      0,                       /* EXACTI */
843      0, 0, 0, 0,              /* POS{STARI,PLUSI,QUERYI,UPTOI} */
844    
845      OP_NOTPOSSTAR, 0,        /* NOTSTAR, NOTMINSTAR */
846      OP_NOTPOSPLUS, 0,        /* NOTPLUS, NOTMINPLUS */
847      OP_NOTPOSQUERY, 0,       /* NOTQUERY, NOTMINQUERY */
848      OP_NOTPOSUPTO, 0,        /* NOTUPTO, NOTMINUPTO */
849      0,                       /* NOTEXACT */
850      0, 0, 0, 0,              /* NOTPOS{STAR,PLUS,QUERY,UPTO} */
851    
852      OP_NOTPOSSTARI, 0,       /* NOTSTARI, NOTMINSTARI */
853      OP_NOTPOSPLUSI, 0,       /* NOTPLUSI, NOTMINPLUSI */
854      OP_NOTPOSQUERYI, 0,      /* NOTQUERYI, NOTMINQUERYI */
855      OP_NOTPOSUPTOI, 0,       /* NOTUPTOI, NOTMINUPTOI */
856      0,                       /* NOTEXACTI */
857      0, 0, 0, 0,              /* NOTPOS{STARI,PLUSI,QUERYI,UPTOI} */
858    
859      OP_TYPEPOSSTAR, 0,       /* TYPESTAR, TYPEMINSTAR */
860      OP_TYPEPOSPLUS, 0,       /* TYPEPLUS, TYPEMINPLUS */
861      OP_TYPEPOSQUERY, 0,      /* TYPEQUERY, TYPEMINQUERY */
862      OP_TYPEPOSUPTO, 0,       /* TYPEUPTO, TYPEMINUPTO */
863      0,                       /* TYPEEXACT */
864      0, 0, 0, 0,              /* TYPEPOS{STAR,PLUS,QUERY,UPTO} */
865    
866      OP_CRPOSSTAR, 0,         /* CRSTAR, CRMINSTAR */
867      OP_CRPOSPLUS, 0,         /* CRPLUS, CRMINPLUS */
868      OP_CRPOSQUERY, 0,        /* CRQUERY, CRMINQUERY */
869      OP_CRPOSRANGE, 0,        /* CRRANGE, CRMINRANGE */
870      0, 0, 0, 0,              /* CRPOS{STAR,PLUS,QUERY,RANGE} */
871    
872      0, 0, 0,                 /* CLASS, NCLASS, XCLASS */
873      0, 0,                    /* REF, REFI */
874      0, 0,                    /* DNREF, DNREFI */
875      0, 0                     /* RECURSE, CALLOUT */
876    };
877    
878    
879    
880    /*************************************************
881    *            Find an error text                  *
882    *************************************************/
883    
884    /* The error texts are now all in one long string, to save on relocations. As
885    some of the text is of unknown length, we can't use a table of offsets.
886    Instead, just count through the strings. This is not a performance issue
887    because it happens only when there has been a compilation error.
888    
889    Argument:   the error number
890    Returns:    pointer to the error string
891    */
892    
893    static const char *
894    find_error_text(int n)
895    {
896    const char *s = error_texts;
897    for (; n > 0; n--)
898      {
899      while (*s++ != CHAR_NULL) {};
900      if (*s == CHAR_NULL) return "Error text not found (please report)";
901      }
902    return s;
903    }
904    
905    
906    
907    /*************************************************
908    *           Expand the workspace                 *
909    *************************************************/
910    
911    /* This function is called during the second compiling phase, if the number of
912    forward references fills the existing workspace, which is originally a block on
913    the stack. A larger block is obtained from malloc() unless the ultimate limit
914    has been reached or the increase will be rather small.
915    
916    Argument: pointer to the compile data block
917    Returns:  0 if all went well, else an error number
918    */
919    
920    static int
921    expand_workspace(compile_data *cd)
922    {
923    pcre_uchar *newspace;
924    int newsize = cd->workspace_size * 2;
925    
926    if (newsize > COMPILE_WORK_SIZE_MAX) newsize = COMPILE_WORK_SIZE_MAX;
927    if (cd->workspace_size >= COMPILE_WORK_SIZE_MAX ||
928        newsize - cd->workspace_size < WORK_SIZE_SAFETY_MARGIN)
929     return ERR72;
930    
931    newspace = (PUBL(malloc))(IN_UCHARS(newsize));
932    if (newspace == NULL) return ERR21;
933    memcpy(newspace, cd->start_workspace, cd->workspace_size * sizeof(pcre_uchar));
934    cd->hwm = (pcre_uchar *)newspace + (cd->hwm - cd->start_workspace);
935    if (cd->workspace_size > COMPILE_WORK_SIZE)
936      (PUBL(free))((void *)cd->start_workspace);
937    cd->start_workspace = newspace;
938    cd->workspace_size = newsize;
939    return 0;
940    }
941    
942    
943    
944    /*************************************************
945    *            Check for counted repeat            *
946    *************************************************/
947    
948    /* This function is called when a '{' is encountered in a place where it might
949    start a quantifier. It looks ahead to see if it really is a quantifier or not.
950    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
951    where the ddds are digits.
952    
953    Arguments:
954      p         pointer to the first char after '{'
955    
956    Returns:    TRUE or FALSE
957    */
958    
959  static BOOL  static BOOL
960    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, int, int *,  is_counted_repeat(const pcre_uchar *p)
961      int *, branch_chain *, compile_data *, int *);  {
962    if (!IS_DIGIT(*p)) return FALSE;
963    p++;
964    while (IS_DIGIT(*p)) p++;
965    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
966    
967    if (*p++ != CHAR_COMMA) return FALSE;
968    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
969    
970    if (!IS_DIGIT(*p)) return FALSE;
971    p++;
972    while (IS_DIGIT(*p)) p++;
973    
974    return (*p == CHAR_RIGHT_CURLY_BRACKET);
975    }
976    
977    
978    
# Line 383  static BOOL Line 981  static BOOL
981  *************************************************/  *************************************************/
982    
983  /* This function is called when a \ has been encountered. It either returns a  /* This function is called when a \ has been encountered. It either returns a
984  positive value for a simple escape such as \n, or a negative value which  positive value for a simple escape such as \n, or 0 for a data character which
985  encodes one of the more complicated things such as \d. A backreference to group  will be placed in chptr. A backreference to group n is returned as negative n.
986  n is returned as -(ESC_REF + n); ESC_REF is the highest ESC_xxx macro. When  When UTF-8 is enabled, a positive value greater than 255 may be returned in
987  UTF-8 is enabled, a positive value greater than 255 may be returned. On entry,  chptr. On entry, ptr is pointing at the \. On exit, it is on the final
988  ptr is pointing at the \. On exit, it is on the final character of the escape  character of the escape sequence.
 sequence.  
989    
990  Arguments:  Arguments:
991    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
992      chptr          points to a returned data character
993    errorcodeptr   points to the errorcode variable    errorcodeptr   points to the errorcode variable
994    bracount       number of previous extracting brackets    bracount       number of previous extracting brackets
995    options        the options bits    options        the options bits
996    isclass        TRUE if inside a character class    isclass        TRUE if inside a character class
997    
998  Returns:         zero or positive => a data character  Returns:         zero => a data character
999                   negative => a special escape sequence                   positive => a special escape sequence
1000                   on error, errorptr is set                   negative => a back reference
1001                     on error, errorcodeptr is set
1002  */  */
1003    
1004  static int  static int
1005  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,
1006    int options, BOOL isclass)    int bracount, int options, BOOL isclass)
1007  {  {
1008  BOOL utf8 = (options & PCRE_UTF8) != 0;  /* PCRE_UTF16 has the same value as PCRE_UTF8. */
1009  const uschar *ptr = *ptrptr + 1;  BOOL utf = (options & PCRE_UTF8) != 0;
1010  int c, i;  const pcre_uchar *ptr = *ptrptr + 1;
1011    pcre_uint32 c;
1012    int escape = 0;
1013    int i;
1014    
1015  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */  GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
1016  ptr--;                            /* Set pointer back to the last byte */  ptr--;                            /* Set pointer back to the last byte */
1017    
1018  /* If backslash is at the end of the pattern, it's an error. */  /* If backslash is at the end of the pattern, it's an error. */
1019    
1020  if (c == 0) *errorcodeptr = ERR1;  if (c == CHAR_NULL) *errorcodeptr = ERR1;
1021    
1022  /* Non-alphamerics are literals. For digits or letters, do an initial lookup in  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
1023  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.
1024  Otherwise further processing may be required. */  Otherwise further processing may be required. */
1025    
1026  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1027  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  /* Not alphanumeric */
1028  else if ((i = escapes[c - '0']) != 0) c = i;  else if (c < CHAR_0 || c > CHAR_z) {}
1029    else if ((i = escapes[c - CHAR_0]) != 0)
1030  #else          /* EBCDIC coding */    { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1031  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  
1032  else if ((i = escapes[c - 0x48]) != 0)  c = i;  #else           /* EBCDIC coding */
1033    /* Not alphanumeric */
1034    else if (c < CHAR_a || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}
1035    else if ((i = escapes[c - 0x48]) != 0)  { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1036  #endif  #endif
1037    
1038  /* Escapes that need further processing, or are illegal. */  /* Escapes that need further processing, or are illegal. */
1039    
1040  else  else
1041    {    {
1042    const uschar *oldptr;    const pcre_uchar *oldptr;
1043    BOOL braced, negated;    BOOL braced, negated, overflow;
1044      int s;
1045    
1046    switch (c)    switch (c)
1047      {      {
1048      /* 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
1049      error. */      error. */
1050    
1051      case 'l':      case CHAR_l:
1052      case 'L':      case CHAR_L:
     case 'N':  
     case 'u':  
     case 'U':  
1053      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
1054      break;      break;
1055    
1056      /* \g must be followed by a number, either plain or braced. If positive, it      case CHAR_u:
1057      is an absolute backreference. If negative, it is a relative backreference.      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1058      This is a Perl 5.10 feature. */        {
1059          /* In JavaScript, \u must be followed by four hexadecimal numbers.
1060          Otherwise it is a lowercase u letter. */
1061          if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1062            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0
1063            && MAX_255(ptr[3]) && (digitab[ptr[3]] & ctype_xdigit) != 0
1064            && MAX_255(ptr[4]) && (digitab[ptr[4]] & ctype_xdigit) != 0)
1065            {
1066            c = 0;
1067            for (i = 0; i < 4; ++i)
1068              {
1069              register pcre_uint32 cc = *(++ptr);
1070    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1071              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1072              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1073    #else           /* EBCDIC coding */
1074              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1075              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1076    #endif
1077              }
1078    
1079    #if defined COMPILE_PCRE8
1080            if (c > (utf ? 0x10ffffU : 0xffU))
1081    #elif defined COMPILE_PCRE16
1082            if (c > (utf ? 0x10ffffU : 0xffffU))
1083    #elif defined COMPILE_PCRE32
1084            if (utf && c > 0x10ffffU)
1085    #endif
1086              {
1087              *errorcodeptr = ERR76;
1088              }
1089            else if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1090            }
1091          }
1092        else
1093          *errorcodeptr = ERR37;
1094        break;
1095    
1096        case CHAR_U:
1097        /* In JavaScript, \U is an uppercase U letter. */
1098        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
1099        break;
1100    
1101        /* In a character class, \g is just a literal "g". Outside a character
1102        class, \g must be followed by one of a number of specific things:
1103    
1104        (1) A number, either plain or braced. If positive, it is an absolute
1105        backreference. If negative, it is a relative backreference. This is a Perl
1106        5.10 feature.
1107    
1108        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
1109        is part of Perl's movement towards a unified syntax for back references. As
1110        this is synonymous with \k{name}, we fudge it up by pretending it really
1111        was \k.
1112    
1113        (3) For Oniguruma compatibility we also support \g followed by a name or a
1114        number either in angle brackets or in single quotes. However, these are
1115        (possibly recursive) subroutine calls, _not_ backreferences. Just return
1116        the ESC_g code (cf \k). */
1117    
1118        case CHAR_g:
1119        if (isclass) break;
1120        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
1121          {
1122          escape = ESC_g;
1123          break;
1124          }
1125    
1126        /* Handle the Perl-compatible cases */
1127    
1128      case 'g':      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
     if (ptr[1] == '{')  
1129        {        {
1130          const pcre_uchar *p;
1131          for (p = ptr+2; *p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
1132            if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
1133          if (*p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET)
1134            {
1135            escape = ESC_k;
1136            break;
1137            }
1138        braced = TRUE;        braced = TRUE;
1139        ptr++;        ptr++;
1140        }        }
1141      else braced = FALSE;      else braced = FALSE;
1142    
1143      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
1144        {        {
1145        negated = TRUE;        negated = TRUE;
1146        ptr++;        ptr++;
1147        }        }
1148      else negated = FALSE;      else negated = FALSE;
1149    
1150      c = 0;      /* The integer range is limited by the machine's int representation. */
1151      while ((digitab[ptr[1]] & ctype_digit) != 0)      s = 0;
1152        c = c * 10 + *(++ptr) - '0';      overflow = FALSE;
1153        while (IS_DIGIT(ptr[1]))
1154          {
1155          if (s > INT_MAX / 10 - 1) /* Integer overflow */
1156            {
1157            overflow = TRUE;
1158            break;
1159            }
1160          s = s * 10 + (int)(*(++ptr) - CHAR_0);
1161          }
1162        if (overflow) /* Integer overflow */
1163          {
1164          while (IS_DIGIT(ptr[1]))
1165            ptr++;
1166          *errorcodeptr = ERR61;
1167          break;
1168          }
1169    
1170      if (c == 0 || (braced && *(++ptr) != '}'))      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
1171        {        {
1172        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
1173        return 0;        break;
1174          }
1175    
1176        if (s == 0)
1177          {
1178          *errorcodeptr = ERR58;
1179          break;
1180        }        }
1181    
1182      if (negated)      if (negated)
1183        {        {
1184        if (c > bracount)        if (s > bracount)
1185          {          {
1186          *errorcodeptr = ERR15;          *errorcodeptr = ERR15;
1187          return 0;          break;
1188          }          }
1189        c = bracount - (c - 1);        s = bracount - (s - 1);
1190        }        }
1191    
1192      c = -(ESC_REF + c);      escape = -s;
1193      break;      break;
1194    
1195      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
1196      starting with one that is not zero is not straightforward. By experiment,      starting with one that is not zero is not straightforward. Perl has changed
1197      the way Perl works seems to be as follows:      over the years. Nowadays \g{} for backreferences and \o{} for octal are
1198        recommended to avoid the ambiguities in the old syntax.
1199    
1200      Outside a character class, the digits are read as a decimal number. If the      Outside a character class, the digits are read as a decimal number. If the
1201      number is less than 10, or if there are that many previous extracting      number is less than 8 (used to be 10), or if there are that many previous
1202      left brackets, then it is a back reference. Otherwise, up to three octal      extracting left brackets, then it is a back reference. Otherwise, up to
1203      digits are read to form an escaped byte. Thus \123 is likely to be octal      three octal digits are read to form an escaped byte. Thus \123 is likely to
1204      123 (cf \0123, which is octal 012 followed by the literal 3). If the octal      be octal 123 (cf \0123, which is octal 012 followed by the literal 3). If
1205      value is greater than 377, the least significant 8 bits are taken. Inside a      the octal value is greater than 377, the least significant 8 bits are
1206      character class, \ followed by a digit is always an octal number. */      taken. \8 and \9 are treated as the literal characters 8 and 9.
1207    
1208        Inside a character class, \ followed by a digit is always either a literal
1209        8 or 9 or an octal number. */
1210    
1211      case '1': case '2': case '3': case '4': case '5':      case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:
1212      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
1213    
1214      if (!isclass)      if (!isclass)
1215        {        {
1216        oldptr = ptr;        oldptr = ptr;
1217        c -= '0';        /* The integer range is limited by the machine's int representation. */
1218        while ((digitab[ptr[1]] & ctype_digit) != 0)        s = (int)(c -CHAR_0);
1219          c = c * 10 + *(++ptr) - '0';        overflow = FALSE;
1220        if (c < 10 || c <= bracount)        while (IS_DIGIT(ptr[1]))
1221            {
1222            if (s > INT_MAX / 10 - 1) /* Integer overflow */
1223              {
1224              overflow = TRUE;
1225              break;
1226              }
1227            s = s * 10 + (int)(*(++ptr) - CHAR_0);
1228            }
1229          if (overflow) /* Integer overflow */
1230            {
1231            while (IS_DIGIT(ptr[1]))
1232              ptr++;
1233            *errorcodeptr = ERR61;
1234            break;
1235            }
1236          if (s < 8 || s <= bracount)  /* Check for back reference */
1237          {          {
1238          c = -(ESC_REF + c);          escape = -s;
1239          break;          break;
1240          }          }
1241        ptr = oldptr;      /* Put the pointer back and fall through */        ptr = oldptr;      /* Put the pointer back and fall through */
1242        }        }
1243    
1244      /* Handle an octal number following \. If the first digit is 8 or 9, Perl      /* Handle a digit following \ when the number is not a back reference. If
1245      generates a binary zero byte and treats the digit as a following literal.      the first digit is 8 or 9, Perl used to generate a binary zero byte and
1246      Thus we have to pull back the pointer by one. */      then treat the digit as a following literal. At least by Perl 5.18 this
1247        changed so as not to insert the binary zero. */
1248    
1249      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8) break;
1250        {  
1251        ptr--;      /* Fall through with a digit less than 8 */
       c = 0;  
       break;  
       }  
1252    
1253      /* \0 always starts an octal number, but we may drop through to here with a      /* \0 always starts an octal number, but we may drop through to here with a
1254      larger first octal digit. The original code used just to take the least      larger first octal digit. The original code used just to take the least
1255      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
1256      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,
1257      than 3 octal digits. */      but no more than 3 octal digits. */
1258    
1259      case '0':      case CHAR_0:
1260      c -= '0';      c -= CHAR_0;
1261      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
1262          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
1263      if (!utf8 && c > 255) *errorcodeptr = ERR51;  #ifdef COMPILE_PCRE8
1264        if (!utf && c > 0xff) *errorcodeptr = ERR51;
1265    #endif
1266      break;      break;
1267    
1268      /* \x is complicated. \x{ddd} is a character number which can be greater      /* \o is a relatively new Perl feature, supporting a more general way of
1269      than 0xff in 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. */  
1270    
1271      case 'x':      case CHAR_o:
1272      if (ptr[1] == '{')      if (ptr[1] != CHAR_LEFT_CURLY_BRACKET) *errorcodeptr = ERR81; else
1273        if (ptr[2] == CHAR_RIGHT_CURLY_BRACKET) *errorcodeptr = ERR86; else
1274        {        {
1275        const uschar *pt = ptr + 2;        ptr += 2;
       int count = 0;  
   
1276        c = 0;        c = 0;
1277        while ((digitab[*pt] & ctype_xdigit) != 0)        overflow = FALSE;
1278          while (*ptr >= CHAR_0 && *ptr <= CHAR_7)
1279          {          {
1280          register int cc = *pt++;          register pcre_uint32 cc = *ptr++;
1281          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1282          count++;  #ifdef COMPILE_PCRE32
1283            if (c >= 0x20000000l) { overflow = TRUE; break; }
1284  #if !EBCDIC    /* ASCII coding */  #endif
1285          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          c = (c << 3) + cc - CHAR_0 ;
1286          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));  #if defined COMPILE_PCRE8
1287  #else          /* EBCDIC coding */          if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1288          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */  #elif defined COMPILE_PCRE16
1289          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1290    #elif defined COMPILE_PCRE32
1291            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1292  #endif  #endif
1293          }          }
1294          if (overflow)
       if (*pt == '}')  
1295          {          {
1296          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          while (*ptr >= CHAR_0 && *ptr <= CHAR_7) ptr++;
1297          ptr = pt;          *errorcodeptr = ERR34;
         break;  
1298          }          }
1299          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1300        /* If the sequence of hex digits does not end with '}', then we don't          {
1301        recognize this construct; fall through to the normal \x handling. */          if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1302            }
1303          else *errorcodeptr = ERR80;
1304        }        }
1305        break;
1306    
1307      /* Read just a single-byte hex-defined char */      /* \x is complicated. In JavaScript, \x must be followed by two hexadecimal
1308        numbers. Otherwise it is a lowercase x letter. */
1309    
1310      c = 0;      case CHAR_x:
1311      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1312        {        {
1313        int cc;                               /* Some compilers don't like ++ */        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1314        cc = *(++ptr);                        /* in initializers */          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1315  #if !EBCDIC    /* ASCII coding */          {
1316        if (cc >= 'a') cc -= 32;              /* Convert to upper case */          c = 0;
1317        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));          for (i = 0; i < 2; ++i)
1318  #else          /* EBCDIC coding */            {
1319        if (cc <= 'z') cc += 64;              /* Convert to upper case */            register pcre_uint32 cc = *(++ptr);
1320        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1321              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1322              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1323    #else           /* EBCDIC coding */
1324              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1325              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1326  #endif  #endif
1327        }            }
1328            }
1329          }    /* End JavaScript handling */
1330    
1331        /* Handle \x in Perl's style. \x{ddd} is a character number which can be
1332        greater than 0xff in utf or non-8bit mode, but only if the ddd are hex
1333        digits. If not, { used to be treated as a data character. However, Perl
1334        seems to read hex digits up to the first non-such, and ignore the rest, so
1335        that, for example \x{zz} matches a binary zero. This seems crazy, so PCRE
1336        now gives an error. */
1337    
1338        else
1339          {
1340          if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1341            {
1342            ptr += 2;
1343            if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1344              {
1345              *errorcodeptr = ERR86;
1346              break;
1347              }
1348            c = 0;
1349            overflow = FALSE;
1350            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0)
1351              {
1352              register pcre_uint32 cc = *ptr++;
1353              if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1354    
1355    #ifdef COMPILE_PCRE32
1356              if (c >= 0x10000000l) { overflow = TRUE; break; }
1357    #endif
1358    
1359    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1360              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1361              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1362    #else           /* EBCDIC coding */
1363              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1364              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1365    #endif
1366    
1367    #if defined COMPILE_PCRE8
1368              if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1369    #elif defined COMPILE_PCRE16
1370              if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1371    #elif defined COMPILE_PCRE32
1372              if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1373    #endif
1374              }
1375    
1376            if (overflow)
1377              {
1378              while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0) ptr++;
1379              *errorcodeptr = ERR34;
1380              }
1381    
1382            else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1383              {
1384              if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1385              }
1386    
1387            /* If the sequence of hex digits does not end with '}', give an error.
1388            We used just to recognize this construct and fall through to the normal
1389            \x handling, but nowadays Perl gives an error, which seems much more
1390            sensible, so we do too. */
1391    
1392            else *errorcodeptr = ERR79;
1393            }   /* End of \x{} processing */
1394    
1395          /* Read a single-byte hex-defined char (up to two hex digits after \x) */
1396    
1397          else
1398            {
1399            c = 0;
1400            while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1401              {
1402              pcre_uint32 cc;                          /* Some compilers don't like */
1403              cc = *(++ptr);                           /* ++ in initializers */
1404    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1405              if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1406              c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1407    #else           /* EBCDIC coding */
1408              if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1409              c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1410    #endif
1411              }
1412            }     /* End of \xdd handling */
1413          }       /* End of Perl-style \x handling */
1414      break;      break;
1415    
1416      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
1417      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
1418        coding is ASCII-specific, but then the whole concept of \cx is
1419      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
1420    
1421      case 'c':      case CHAR_c:
1422      c = *(++ptr);      c = *(++ptr);
1423      if (c == 0)      if (c == CHAR_NULL)
1424        {        {
1425        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
1426        return 0;        break;
1427        }        }
1428    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
1429  #if !EBCDIC    /* ASCII coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
1430      if (c >= 'a' && c <= 'z') c -= 32;        {
1431          *errorcodeptr = ERR68;
1432          break;
1433          }
1434        if (c >= CHAR_a && c <= CHAR_z) c -= 32;
1435      c ^= 0x40;      c ^= 0x40;
1436  #else          /* EBCDIC coding */  #else             /* EBCDIC coding */
1437      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
1438      c ^= 0xC0;      if (c == CHAR_QUESTION_MARK)
1439          c = ('\\' == 188 && '`' == 74)? 0x5f : 0xff;
1440        else
1441          {
1442          for (i = 0; i < 32; i++)
1443            {
1444            if (c == ebcdic_escape_c[i]) break;
1445            }
1446          if (i < 32) c = i; else *errorcodeptr = ERR68;
1447          }
1448  #endif  #endif
1449      break;      break;
1450    
1451      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
1452      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
1453      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
1454      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
1455      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
1456    
1457      default:      default:
1458      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 637  else Line 1465  else
1465      }      }
1466    }    }
1467    
1468    /* Perl supports \N{name} for character names, as well as plain \N for "not
1469    newline". PCRE does not support \N{name}. However, it does support
1470    quantification such as \N{2,3}. */
1471    
1472    if (escape == ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1473         !is_counted_repeat(ptr+2))
1474      *errorcodeptr = ERR37;
1475    
1476    /* If PCRE_UCP is set, we change the values for \d etc. */
1477    
1478    if ((options & PCRE_UCP) != 0 && escape >= ESC_D && escape <= ESC_w)
1479      escape += (ESC_DU - ESC_D);
1480    
1481    /* Set the pointer to the final character before returning. */
1482    
1483  *ptrptr = ptr;  *ptrptr = ptr;
1484  return c;  *chptr = c;
1485    return escape;
1486  }  }
1487    
1488    
# Line 656  escape sequence. Line 1500  escape sequence.
1500  Argument:  Argument:
1501    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
1502    negptr         points to a boolean that is set TRUE for negation else FALSE    negptr         points to a boolean that is set TRUE for negation else FALSE
1503    dptr           points to an int that is set to the detailed property value    ptypeptr       points to an unsigned int that is set to the type value
1504      pdataptr       points to an unsigned int that is set to the detailed property value
1505    errorcodeptr   points to the error code variable    errorcodeptr   points to the error code variable
1506    
1507  Returns:         type value from ucp_type_table, or -1 for an invalid type  Returns:         TRUE if the type value was found, or FALSE for an invalid type
1508  */  */
1509    
1510  static int  static BOOL
1511  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, unsigned int *ptypeptr,
1512      unsigned int *pdataptr, int *errorcodeptr)
1513  {  {
1514  int c, i, bot, top;  pcre_uchar c;
1515  const uschar *ptr = *ptrptr;  int i, bot, top;
1516  char name[32];  const pcre_uchar *ptr = *ptrptr;
1517    pcre_uchar name[32];
1518    
1519  c = *(++ptr);  c = *(++ptr);
1520  if (c == 0) goto ERROR_RETURN;  if (c == CHAR_NULL) goto ERROR_RETURN;
1521    
1522  *negptr = FALSE;  *negptr = FALSE;
1523    
1524  /* \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
1525  negation. */  negation. */
1526    
1527  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
1528    {    {
1529    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1530      {      {
1531      *negptr = TRUE;      *negptr = TRUE;
1532      ptr++;      ptr++;
1533      }      }
1534    for (i = 0; i < sizeof(name) - 1; i++)    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1535      {      {
1536      c = *(++ptr);      c = *(++ptr);
1537      if (c == 0) goto ERROR_RETURN;      if (c == CHAR_NULL) goto ERROR_RETURN;
1538      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1539      name[i] = c;      name[i] = c;
1540      }      }
1541    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
1542    name[i] = 0;    name[i] = 0;
1543    }    }
1544    
# Line 708  else Line 1555  else
1555  /* Search for a recognized property name using binary chop */  /* Search for a recognized property name using binary chop */
1556    
1557  bot = 0;  bot = 0;
1558  top = _pcre_utt_size;  top = PRIV(utt_size);
1559    
1560  while (bot < top)  while (bot < top)
1561    {    {
1562      int r;
1563    i = (bot + top) >> 1;    i = (bot + top) >> 1;
1564    c = strcmp(name, _pcre_utt[i].name);    r = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1565    if (c == 0)    if (r == 0)
1566      {      {
1567      *dptr = _pcre_utt[i].value;      *ptypeptr = PRIV(utt)[i].type;
1568      return _pcre_utt[i].type;      *pdataptr = PRIV(utt)[i].value;
1569        return TRUE;
1570      }      }
1571    if (c > 0) bot = i + 1; else top = i;    if (r > 0) bot = i + 1; else top = i;
1572    }    }
1573    
1574  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
1575  *ptrptr = ptr;  *ptrptr = ptr;
1576  return -1;  return FALSE;
1577    
1578  ERROR_RETURN:  ERROR_RETURN:
1579  *errorcodeptr = ERR46;  *errorcodeptr = ERR46;
1580  *ptrptr = ptr;  *ptrptr = ptr;
1581  return -1;  return FALSE;
1582  }  }
1583  #endif  #endif
1584    
1585    
1586    
   
1587  /*************************************************  /*************************************************
1588  *            Check for counted repeat            *  *         Read repeat counts                     *
1589  *************************************************/  *************************************************/
1590    
1591  /* This function is called when a '{' is encountered in a place where it might  /* Read an item of the form {n,m} and return the values. This is called only
1592  start a quantifier. It looks ahead to see if it really is a quantifier or not.  after is_counted_repeat() has confirmed that a repeat-count quantifier exists,
1593  It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}  so the syntax is guaranteed to be correct, but we need to check the values.
 where the ddds are digits.  
1594    
1595  Arguments:  Arguments:
1596    p         pointer to the first char after '{'    p              pointer to first char after '{'
1597      minp           pointer to int for min
1598  Returns:    TRUE or FALSE    maxp           pointer to int for max
1599  */                   returned as -1 if no max
1600      errorcodeptr   points to error code variable
 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 == '}');  
 }  
   
   
   
 /*************************************************  
 *         Read repeat counts                     *  
 *************************************************/  
   
 /* Read an item of the form {n,m} and return the values. This is called only  
 after is_counted_repeat() has confirmed that a repeat-count quantifier exists,  
 so the syntax is guaranteed to be correct, but we need to check the values.  
   
 Arguments:  
   p              pointer to first char after '{'  
   minp           pointer to int for min  
   maxp           pointer to int for max  
                  returned as -1 if no max  
   errorcodeptr   points to error code variable  
1601    
1602  Returns:         pointer to '}' on success;  Returns:         pointer to '}' on success;
1603                   current ptr on error, with errorcodeptr set non-zero                   current ptr on error, with errorcodeptr set non-zero
1604  */  */
1605    
1606  static const uschar *  static const pcre_uchar *
1607  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)
1608  {  {
1609  int min = 0;  int min = 0;
1610  int max = -1;  int max = -1;
1611    
1612  /* Read the minimum value and do a paranoid check: a negative value indicates  while (IS_DIGIT(*p))
 an integer overflow. */  
   
 while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  
 if (min < 0 || min > 65535)  
1613    {    {
1614    *errorcodeptr = ERR5;    min = min * 10 + (int)(*p++ - CHAR_0);
1615    return p;    if (min > 65535)
1616        {
1617        *errorcodeptr = ERR5;
1618        return p;
1619        }
1620    }    }
1621    
1622  /* Read the maximum value if there is one, and again do a paranoid on its size.  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
 Also, max must not be less than min. */  
   
 if (*p == '}') max = min; else  
1623    {    {
1624    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1625      {      {
1626      max = 0;      max = 0;
1627      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while(IS_DIGIT(*p))
     if (max < 0 || max > 65535)  
1628        {        {
1629        *errorcodeptr = ERR5;        max = max * 10 + (int)(*p++ - CHAR_0);
1630        return p;        if (max > 65535)
1631            {
1632            *errorcodeptr = ERR5;
1633            return p;
1634            }
1635        }        }
1636      if (max < min)      if (max < min)
1637        {        {
# Line 826  if (*p == '}') max = min; else Line 1641  if (*p == '}') max = min; else
1641      }      }
1642    }    }
1643    
 /* Fill in the required variables, and pass back the pointer to the terminating  
 '}'. */  
   
1644  *minp = min;  *minp = min;
1645  *maxp = max;  *maxp = max;
1646  return p;  return p;
# Line 837  return p; Line 1649  return p;
1649    
1650    
1651  /*************************************************  /*************************************************
 *       Find forward referenced subpattern       *  
 *************************************************/  
   
 /* This function scans along a pattern's text looking for capturing  
 subpatterns, and counting them. If it finds a named pattern that matches the  
 name it is given, it returns its number. Alternatively, if the name is NULL, it  
 returns when it reaches a given numbered subpattern. This is used for forward  
 references to subpatterns. We know that if (?P< is encountered, the name will  
 be terminated by '>' because that is checked in the first pass.  
   
 Arguments:  
   ptr          current position in the pattern  
   count        current count of capturing parens so far encountered  
   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  
   
 Returns:       the number of the named subpattern, or -1 if not found  
 */  
   
 static int  
 find_parens(const uschar *ptr, int count, const uschar *name, int lorn,  
   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;  
 }  
   
   
   
 /*************************************************  
1652  *      Find first significant op code            *  *      Find first significant op code            *
1653  *************************************************/  *************************************************/
1654    
1655  /* This is called by several functions that scan a compiled expression looking  /* This is called by several functions that scan a compiled expression looking
1656  for a fixed first character, or an anchoring op code etc. It skips over things  for a fixed first character, or an anchoring op code etc. It skips over things
1657  that do not influence this. For some calls, a change of option is important.  that do not influence this. For some calls, it makes sense to skip negative
1658  For some calls, it makes sense to skip negative forward and all backward  forward and all backward assertions, and also the \b assertion; for others it
1659  assertions, and also the \b assertion; for others it does not.  does not.
1660    
1661  Arguments:  Arguments:
1662    code         pointer to the start of the group    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  
1663    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1664    
1665  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1666  */  */
1667    
1668  static const uschar*  static const pcre_uchar*
1669  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const pcre_uchar *code, BOOL skipassert)
   BOOL skipassert)  
1670  {  {
1671  for (;;)  for (;;)
1672    {    {
1673    switch ((int)*code)    switch ((int)*code)
1674      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1675      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1676      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1677      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1678      if (!skipassert) return code;      if (!skipassert) return code;
1679      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
1680      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1681      break;      break;
1682    
1683      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
# Line 996  for (;;) Line 1687  for (;;)
1687    
1688      case OP_CALLOUT:      case OP_CALLOUT:
1689      case OP_CREF:      case OP_CREF:
1690        case OP_DNCREF:
1691      case OP_RREF:      case OP_RREF:
1692        case OP_DNRREF:
1693      case OP_DEF:      case OP_DEF:
1694      code += _pcre_OP_lengths[*code];      code += PRIV(OP_lengths)[*code];
1695      break;      break;
1696    
1697      default:      default:
# Line 1010  for (;;) Line 1703  for (;;)
1703    
1704    
1705    
   
1706  /*************************************************  /*************************************************
1707  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1708  *************************************************/  *************************************************/
1709    
1710  /* 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,
1711  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.
1712  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
1713    temporarily terminated with OP_END when this function is called.
1714    
1715    This function is called when a backward assertion is encountered, so that if it
1716    fails, the error message can point to the correct place in the pattern.
1717    However, we cannot do this when the assertion contains subroutine calls,
1718    because they can be forward references. We solve this by remembering this case
1719    and doing the check at the end; a flag specifies which mode we are running in.
1720    
1721  Arguments:  Arguments:
1722    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1723    options  the compiling options    utf      TRUE in UTF-8 / UTF-16 / UTF-32 mode
1724      atend    TRUE if called when the pattern is complete
1725  Returns:   the fixed length, or -1 if there is no fixed length,    cd       the "compile data" structure
1726               or -2 if \C was encountered    recurses    chain of recurse_check to catch mutual recursion
1727    
1728    Returns:   the fixed length,
1729                 or -1 if there is no fixed length,
1730                 or -2 if \C was encountered (in UTF-8 mode only)
1731                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1732                 or -4 if an unknown opcode was encountered (internal error)
1733  */  */
1734    
1735  static int  static int
1736  find_fixedlength(uschar *code, int options)  find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd,
1737      recurse_check *recurses)
1738  {  {
1739  int length = -1;  int length = -1;
1740    recurse_check this_recurse;
1741  register int branchlength = 0;  register int branchlength = 0;
1742  register uschar *cc = code + 1 + LINK_SIZE;  register pcre_uchar *cc = code + 1 + LINK_SIZE;
1743    
1744  /* 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
1745  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 1747  branch, check the length against that of
1747  for (;;)  for (;;)
1748    {    {
1749    int d;    int d;
1750    register int op = *cc;    pcre_uchar *ce, *cs;
1751      register pcre_uchar op = *cc;
1752    
1753    switch (op)    switch (op)
1754      {      {
1755        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1756        OP_BRA (normal non-capturing bracket) because the other variants of these
1757        opcodes are all concerned with unlimited repeated groups, which of course
1758        are not of fixed length. */
1759    
1760      case OP_CBRA:      case OP_CBRA:
1761      case OP_BRA:      case OP_BRA:
1762      case OP_ONCE:      case OP_ONCE:
1763        case OP_ONCE_NC:
1764      case OP_COND:      case OP_COND:
1765      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? IMM2_SIZE : 0), utf, atend, cd,
1766          recurses);
1767      if (d < 0) return d;      if (d < 0) return d;
1768      branchlength += d;      branchlength += d;
1769      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1770      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1771      break;      break;
1772    
1773      /* 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.
1774      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
1775      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
1776        the same code. Note that we must not include the OP_KETRxxx opcodes here,
1777        because they all imply an unlimited repeat. */
1778    
1779      case OP_ALT:      case OP_ALT:
1780      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1781      case OP_END:      case OP_END:
1782        case OP_ACCEPT:
1783        case OP_ASSERT_ACCEPT:
1784      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1785        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
1786      if (*cc != OP_ALT) return length;      if (*cc != OP_ALT) return length;
# Line 1072  for (;;) Line 1788  for (;;)
1788      branchlength = 0;      branchlength = 0;
1789      break;      break;
1790    
1791        /* A true recursion implies not fixed length, but a subroutine call may
1792        be OK. If the subroutine is a forward reference, we can't deal with
1793        it until the end of the pattern, so return -3. */
1794    
1795        case OP_RECURSE:
1796        if (!atend) return -3;
1797        cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1798        do ce += GET(ce, 1); while (*ce == OP_ALT);           /* End subpattern */
1799        if (cc > cs && cc < ce) return -1;                    /* Recursion */
1800        else   /* Check for mutual recursion */
1801          {
1802          recurse_check *r = recurses;
1803          for (r = recurses; r != NULL; r = r->prev) if (r->group == cs) break;
1804          if (r != NULL) return -1;   /* Mutual recursion */
1805          }
1806        this_recurse.prev = recurses;
1807        this_recurse.group = cs;
1808        d = find_fixedlength(cs + IMM2_SIZE, utf, atend, cd, &this_recurse);
1809        if (d < 0) return d;
1810        branchlength += d;
1811        cc += 1 + LINK_SIZE;
1812        break;
1813    
1814      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1815    
1816      case OP_ASSERT:      case OP_ASSERT:
# Line 1079  for (;;) Line 1818  for (;;)
1818      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1819      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1820      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1821      /* Fall through */      cc += 1 + LINK_SIZE;
1822        break;
1823    
1824      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1825    
1826      case OP_REVERSE:      case OP_MARK:
1827        case OP_PRUNE_ARG:
1828        case OP_SKIP_ARG:
1829        case OP_THEN_ARG:
1830        cc += cc[1] + PRIV(OP_lengths)[*cc];
1831        break;
1832    
1833        case OP_CALLOUT:
1834        case OP_CIRC:
1835        case OP_CIRCM:
1836        case OP_CLOSE:
1837        case OP_COMMIT:
1838      case OP_CREF:      case OP_CREF:
     case OP_RREF:  
1839      case OP_DEF:      case OP_DEF:
1840      case OP_OPT:      case OP_DNCREF:
1841      case OP_CALLOUT:      case OP_DNRREF:
1842      case OP_SOD:      case OP_DOLL:
1843      case OP_SOM:      case OP_DOLLM:
1844      case OP_EOD:      case OP_EOD:
1845      case OP_EODN:      case OP_EODN:
1846      case OP_CIRC:      case OP_FAIL:
     case OP_DOLL:  
1847      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1848        case OP_PRUNE:
1849        case OP_REVERSE:
1850        case OP_RREF:
1851        case OP_SET_SOM:
1852        case OP_SKIP:
1853        case OP_SOD:
1854        case OP_SOM:
1855        case OP_THEN:
1856      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1857      cc += _pcre_OP_lengths[*cc];      cc += PRIV(OP_lengths)[*cc];
1858      break;      break;
1859    
1860      /* Handle literal characters */      /* Handle literal characters */
1861    
1862      case OP_CHAR:      case OP_CHAR:
1863      case OP_CHARNC:      case OP_CHARI:
1864      case OP_NOT:      case OP_NOT:
1865        case OP_NOTI:
1866      branchlength++;      branchlength++;
1867      cc += 2;      cc += 2;
1868  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1869      if ((options & PCRE_UTF8) != 0)      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
       {  
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1870  #endif  #endif
1871      break;      break;
1872    
# Line 1119  for (;;) Line 1874  for (;;)
1874      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1875    
1876      case OP_EXACT:      case OP_EXACT:
1877      branchlength += GET2(cc,1);      case OP_EXACTI:
1878      cc += 4;      case OP_NOTEXACT:
1879  #ifdef SUPPORT_UTF8      case OP_NOTEXACTI:
1880      if ((options & PCRE_UTF8) != 0)      branchlength += (int)GET2(cc,1);
1881        {      cc += 2 + IMM2_SIZE;
1882        while((*cc & 0x80) == 0x80) cc++;  #ifdef SUPPORT_UTF
1883        }      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1884  #endif  #endif
1885      break;      break;
1886    
1887      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1888      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1889      cc += 4;      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)
1890          cc += 2;
1891        cc += 1 + IMM2_SIZE + 1;
1892      break;      break;
1893    
1894      /* Handle single-char matchers */      /* Handle single-char matchers */
# Line 1141  for (;;) Line 1898  for (;;)
1898      cc += 2;      cc += 2;
1899      /* Fall through */      /* Fall through */
1900    
1901        case OP_HSPACE:
1902        case OP_VSPACE:
1903        case OP_NOT_HSPACE:
1904        case OP_NOT_VSPACE:
1905      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
1906      case OP_DIGIT:      case OP_DIGIT:
1907      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
# Line 1148  for (;;) Line 1909  for (;;)
1909      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1910      case OP_WORDCHAR:      case OP_WORDCHAR:
1911      case OP_ANY:      case OP_ANY:
1912        case OP_ALLANY:
1913      branchlength++;      branchlength++;
1914      cc++;      cc++;
1915      break;      break;
1916    
1917      /* The single-byte matcher isn't allowed */      /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1918        otherwise \C is coded as OP_ALLANY. */
1919    
1920      case OP_ANYBYTE:      case OP_ANYBYTE:
1921      return -2;      return -2;
1922    
1923      /* Check a class for variable quantification */      /* Check a class for variable quantification */
1924    
 #ifdef SUPPORT_UTF8  
     case OP_XCLASS:  
     cc += GET(cc, 1) - 33;  
     /* Fall through */  
 #endif  
   
1925      case OP_CLASS:      case OP_CLASS:
1926      case OP_NCLASS:      case OP_NCLASS:
1927      cc += 33;  #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1928        case OP_XCLASS:
1929        /* The original code caused an unsigned overflow in 64 bit systems,
1930        so now we use a conditional statement. */
1931        if (op == OP_XCLASS)
1932          cc += GET(cc, 1);
1933        else
1934          cc += PRIV(OP_lengths)[OP_CLASS];
1935    #else
1936        cc += PRIV(OP_lengths)[OP_CLASS];
1937    #endif
1938    
1939      switch (*cc)      switch (*cc)
1940        {        {
1941        case OP_CRSTAR:        case OP_CRSTAR:
1942        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1943          case OP_CRPLUS:
1944          case OP_CRMINPLUS:
1945        case OP_CRQUERY:        case OP_CRQUERY:
1946        case OP_CRMINQUERY:        case OP_CRMINQUERY:
1947          case OP_CRPOSSTAR:
1948          case OP_CRPOSPLUS:
1949          case OP_CRPOSQUERY:
1950        return -1;        return -1;
1951    
1952        case OP_CRRANGE:        case OP_CRRANGE:
1953        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1954        if (GET2(cc,1) != GET2(cc,3)) return -1;        case OP_CRPOSRANGE:
1955        branchlength += GET2(cc,1);        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1956        cc += 5;        branchlength += (int)GET2(cc,1);
1957          cc += 1 + 2 * IMM2_SIZE;
1958        break;        break;
1959    
1960        default:        default:
# Line 1191  for (;;) Line 1964  for (;;)
1964    
1965      /* Anything else is variable length */      /* Anything else is variable length */
1966    
1967      default:      case OP_ANYNL:
1968        case OP_BRAMINZERO:
1969        case OP_BRAPOS:
1970        case OP_BRAPOSZERO:
1971        case OP_BRAZERO:
1972        case OP_CBRAPOS:
1973        case OP_EXTUNI:
1974        case OP_KETRMAX:
1975        case OP_KETRMIN:
1976        case OP_KETRPOS:
1977        case OP_MINPLUS:
1978        case OP_MINPLUSI:
1979        case OP_MINQUERY:
1980        case OP_MINQUERYI:
1981        case OP_MINSTAR:
1982        case OP_MINSTARI:
1983        case OP_MINUPTO:
1984        case OP_MINUPTOI:
1985        case OP_NOTMINPLUS:
1986        case OP_NOTMINPLUSI:
1987        case OP_NOTMINQUERY:
1988        case OP_NOTMINQUERYI:
1989        case OP_NOTMINSTAR:
1990        case OP_NOTMINSTARI:
1991        case OP_NOTMINUPTO:
1992        case OP_NOTMINUPTOI:
1993        case OP_NOTPLUS:
1994        case OP_NOTPLUSI:
1995        case OP_NOTPOSPLUS:
1996        case OP_NOTPOSPLUSI:
1997        case OP_NOTPOSQUERY:
1998        case OP_NOTPOSQUERYI:
1999        case OP_NOTPOSSTAR:
2000        case OP_NOTPOSSTARI:
2001        case OP_NOTPOSUPTO:
2002        case OP_NOTPOSUPTOI:
2003        case OP_NOTQUERY:
2004        case OP_NOTQUERYI:
2005        case OP_NOTSTAR:
2006        case OP_NOTSTARI:
2007        case OP_NOTUPTO:
2008        case OP_NOTUPTOI:
2009        case OP_PLUS:
2010        case OP_PLUSI:
2011        case OP_POSPLUS:
2012        case OP_POSPLUSI:
2013        case OP_POSQUERY:
2014        case OP_POSQUERYI:
2015        case OP_POSSTAR:
2016        case OP_POSSTARI:
2017        case OP_POSUPTO:
2018        case OP_POSUPTOI:
2019        case OP_QUERY:
2020        case OP_QUERYI:
2021        case OP_REF:
2022        case OP_REFI:
2023        case OP_DNREF:
2024        case OP_DNREFI:
2025        case OP_SBRA:
2026        case OP_SBRAPOS:
2027        case OP_SCBRA:
2028        case OP_SCBRAPOS:
2029        case OP_SCOND:
2030        case OP_SKIPZERO:
2031        case OP_STAR:
2032        case OP_STARI:
2033        case OP_TYPEMINPLUS:
2034        case OP_TYPEMINQUERY:
2035        case OP_TYPEMINSTAR:
2036        case OP_TYPEMINUPTO:
2037        case OP_TYPEPLUS:
2038        case OP_TYPEPOSPLUS:
2039        case OP_TYPEPOSQUERY:
2040        case OP_TYPEPOSSTAR:
2041        case OP_TYPEPOSUPTO:
2042        case OP_TYPEQUERY:
2043        case OP_TYPESTAR:
2044        case OP_TYPEUPTO:
2045        case OP_UPTO:
2046        case OP_UPTOI:
2047      return -1;      return -1;
2048    
2049        /* Catch unrecognized opcodes so that when new ones are added they
2050        are not forgotten, as has happened in the past. */
2051    
2052        default:
2053        return -4;
2054      }      }
2055    }    }
2056  /* Control never gets here */  /* Control never gets here */
# Line 1200  for (;;) Line 2058  for (;;)
2058    
2059    
2060    
   
2061  /*************************************************  /*************************************************
2062  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
2063  *************************************************/  *************************************************/
2064    
2065  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
2066  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
2067    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
2068    so that it can be called from pcre_study() when finding the minimum matching
2069    length.
2070    
2071  Arguments:  Arguments:
2072    code        points to start of expression    code        points to start of expression
2073    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2074    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
2075    
2076  Returns:      pointer to the opcode for the bracket, or NULL if not found  Returns:      pointer to the opcode for the bracket, or NULL if not found
2077  */  */
2078    
2079  static const uschar *  const pcre_uchar *
2080  find_bracket(const uschar *code, BOOL utf8, int number)  PRIV(find_bracket)(const pcre_uchar *code, BOOL utf, int number)
2081  {  {
2082  for (;;)  for (;;)
2083    {    {
2084    register int c = *code;    register pcre_uchar c = *code;
2085    
2086    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2087    
2088    /* 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 2091  for (;;)
2091    
2092    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
2093    
2094      /* Handle recursion */
2095    
2096      else if (c == OP_REVERSE)
2097        {
2098        if (number < 0) return (pcre_uchar *)code;
2099        code += PRIV(OP_lengths)[c];
2100        }
2101    
2102    /* Handle capturing bracket */    /* Handle capturing bracket */
2103    
2104    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
2105               c == OP_CBRAPOS || c == OP_SCBRAPOS)
2106      {      {
2107      int n = GET2(code, 1+LINK_SIZE);      int n = (int)GET2(code, 1+LINK_SIZE);
2108      if (n == number) return (uschar *)code;      if (n == number) return (pcre_uchar *)code;
2109      code += _pcre_OP_lengths[c];      code += PRIV(OP_lengths)[c];
2110      }      }
2111    
2112      /* Otherwise, we can get the item's length from the table, except that for
2113      repeated character types, we have to test for \p and \P, which have an extra
2114      two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2115      must add in its length. */
2116    
2117      else
2118        {
2119        switch(c)
2120          {
2121          case OP_TYPESTAR:
2122          case OP_TYPEMINSTAR:
2123          case OP_TYPEPLUS:
2124          case OP_TYPEMINPLUS:
2125          case OP_TYPEQUERY:
2126          case OP_TYPEMINQUERY:
2127          case OP_TYPEPOSSTAR:
2128          case OP_TYPEPOSPLUS:
2129          case OP_TYPEPOSQUERY:
2130          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2131          break;
2132    
2133          case OP_TYPEUPTO:
2134          case OP_TYPEMINUPTO:
2135          case OP_TYPEEXACT:
2136          case OP_TYPEPOSUPTO:
2137          if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2138            code += 2;
2139          break;
2140    
2141          case OP_MARK:
2142          case OP_PRUNE_ARG:
2143          case OP_SKIP_ARG:
2144          case OP_THEN_ARG:
2145          code += code[1];
2146          break;
2147          }
2148    
2149        /* Add in the fixed length from the table */
2150    
2151        code += PRIV(OP_lengths)[c];
2152    
2153    /* 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
2154    a multi-byte character. The length in the table is a minimum, so we have to    a multi-byte character. The length in the table is a minimum, so we have to
2155    arrange to skip the extra bytes. */    arrange to skip the extra bytes. */
2156    
2157    else  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2158      {      if (utf) switch(c)
     code += _pcre_OP_lengths[c];  
     if (utf8) switch(c)  
2159        {        {
2160        case OP_CHAR:        case OP_CHAR:
2161        case OP_CHARNC:        case OP_CHARI:
2162          case OP_NOT:
2163          case OP_NOTI:
2164        case OP_EXACT:        case OP_EXACT:
2165          case OP_EXACTI:
2166          case OP_NOTEXACT:
2167          case OP_NOTEXACTI:
2168        case OP_UPTO:        case OP_UPTO:
2169          case OP_UPTOI:
2170          case OP_NOTUPTO:
2171          case OP_NOTUPTOI:
2172        case OP_MINUPTO:        case OP_MINUPTO:
2173          case OP_MINUPTOI:
2174          case OP_NOTMINUPTO:
2175          case OP_NOTMINUPTOI:
2176        case OP_POSUPTO:        case OP_POSUPTO:
2177          case OP_POSUPTOI:
2178          case OP_NOTPOSUPTO:
2179          case OP_NOTPOSUPTOI:
2180        case OP_STAR:        case OP_STAR:
2181          case OP_STARI:
2182          case OP_NOTSTAR:
2183          case OP_NOTSTARI:
2184        case OP_MINSTAR:        case OP_MINSTAR:
2185          case OP_MINSTARI:
2186          case OP_NOTMINSTAR:
2187          case OP_NOTMINSTARI:
2188        case OP_POSSTAR:        case OP_POSSTAR:
2189          case OP_POSSTARI:
2190          case OP_NOTPOSSTAR:
2191          case OP_NOTPOSSTARI:
2192        case OP_PLUS:        case OP_PLUS:
2193          case OP_PLUSI:
2194          case OP_NOTPLUS:
2195          case OP_NOTPLUSI:
2196        case OP_MINPLUS:        case OP_MINPLUS:
2197          case OP_MINPLUSI:
2198          case OP_NOTMINPLUS:
2199          case OP_NOTMINPLUSI:
2200        case OP_POSPLUS:        case OP_POSPLUS:
2201          case OP_POSPLUSI:
2202          case OP_NOTPOSPLUS:
2203          case OP_NOTPOSPLUSI:
2204        case OP_QUERY:        case OP_QUERY:
2205          case OP_QUERYI:
2206          case OP_NOTQUERY:
2207          case OP_NOTQUERYI:
2208        case OP_MINQUERY:        case OP_MINQUERY:
2209          case OP_MINQUERYI:
2210          case OP_NOTMINQUERY:
2211          case OP_NOTMINQUERYI:
2212        case OP_POSQUERY:        case OP_POSQUERY:
2213        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2214          case OP_NOTPOSQUERY:
2215          case OP_NOTPOSQUERYI:
2216          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2217        break;        break;
2218        }        }
2219    #else
2220        (void)(utf);  /* Keep compiler happy by referencing function argument */
2221    #endif
2222      }      }
2223    }    }
2224  }  }
# Line 1281  instance of OP_RECURSE. Line 2234  instance of OP_RECURSE.
2234    
2235  Arguments:  Arguments:
2236    code        points to start of expression    code        points to start of expression
2237    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2238    
2239  Returns:      pointer to the opcode for OP_RECURSE, or NULL if not found  Returns:      pointer to the opcode for OP_RECURSE, or NULL if not found
2240  */  */
2241    
2242  static const uschar *  static const pcre_uchar *
2243  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const pcre_uchar *code, BOOL utf)
2244  {  {
2245  for (;;)  for (;;)
2246    {    {
2247    register int c = *code;    register pcre_uchar c = *code;
2248    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2249    if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
2250    
# Line 1301  for (;;) Line 2254  for (;;)
2254    
2255    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
2256    
2257    /* 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
2258    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
2259    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
2260    bytes. */    must add in its length. */
2261    
2262    else    else
2263      {      {
2264      code += _pcre_OP_lengths[c];      switch(c)
2265      if (utf8) switch(c)        {
2266          case OP_TYPESTAR:
2267          case OP_TYPEMINSTAR:
2268          case OP_TYPEPLUS:
2269          case OP_TYPEMINPLUS:
2270          case OP_TYPEQUERY:
2271          case OP_TYPEMINQUERY:
2272          case OP_TYPEPOSSTAR:
2273          case OP_TYPEPOSPLUS:
2274          case OP_TYPEPOSQUERY:
2275          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2276          break;
2277    
2278          case OP_TYPEPOSUPTO:
2279          case OP_TYPEUPTO:
2280          case OP_TYPEMINUPTO:
2281          case OP_TYPEEXACT:
2282          if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2283            code += 2;
2284          break;
2285    
2286          case OP_MARK:
2287          case OP_PRUNE_ARG:
2288          case OP_SKIP_ARG:
2289          case OP_THEN_ARG:
2290          code += code[1];
2291          break;
2292          }
2293    
2294        /* Add in the fixed length from the table */
2295    
2296        code += PRIV(OP_lengths)[c];
2297    
2298        /* In UTF-8 mode, opcodes that are followed by a character may be followed
2299        by a multi-byte character. The length in the table is a minimum, so we have
2300        to arrange to skip the extra bytes. */
2301    
2302    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2303        if (utf) switch(c)
2304        {        {
2305        case OP_CHAR:        case OP_CHAR:
2306        case OP_CHARNC:        case OP_CHARI:
2307          case OP_NOT:
2308          case OP_NOTI:
2309        case OP_EXACT:        case OP_EXACT:
2310          case OP_EXACTI:
2311          case OP_NOTEXACT:
2312          case OP_NOTEXACTI:
2313        case OP_UPTO:        case OP_UPTO:
2314          case OP_UPTOI:
2315          case OP_NOTUPTO:
2316          case OP_NOTUPTOI:
2317        case OP_MINUPTO:        case OP_MINUPTO:
2318          case OP_MINUPTOI:
2319          case OP_NOTMINUPTO:
2320          case OP_NOTMINUPTOI:
2321        case OP_POSUPTO:        case OP_POSUPTO:
2322          case OP_POSUPTOI:
2323          case OP_NOTPOSUPTO:
2324          case OP_NOTPOSUPTOI:
2325        case OP_STAR:        case OP_STAR:
2326          case OP_STARI:
2327          case OP_NOTSTAR:
2328          case OP_NOTSTARI:
2329        case OP_MINSTAR:        case OP_MINSTAR:
2330          case OP_MINSTARI:
2331          case OP_NOTMINSTAR:
2332          case OP_NOTMINSTARI:
2333        case OP_POSSTAR:        case OP_POSSTAR:
2334          case OP_POSSTARI:
2335          case OP_NOTPOSSTAR:
2336          case OP_NOTPOSSTARI:
2337        case OP_PLUS:        case OP_PLUS:
2338          case OP_PLUSI:
2339          case OP_NOTPLUS:
2340          case OP_NOTPLUSI:
2341        case OP_MINPLUS:        case OP_MINPLUS:
2342          case OP_MINPLUSI:
2343          case OP_NOTMINPLUS:
2344          case OP_NOTMINPLUSI:
2345        case OP_POSPLUS:        case OP_POSPLUS:
2346          case OP_POSPLUSI:
2347          case OP_NOTPOSPLUS:
2348          case OP_NOTPOSPLUSI:
2349        case OP_QUERY:        case OP_QUERY:
2350          case OP_QUERYI:
2351          case OP_NOTQUERY:
2352          case OP_NOTQUERYI:
2353        case OP_MINQUERY:        case OP_MINQUERY:
2354          case OP_MINQUERYI:
2355          case OP_NOTMINQUERY:
2356          case OP_NOTMINQUERYI:
2357        case OP_POSQUERY:        case OP_POSQUERY:
2358        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        case OP_POSQUERYI:
2359          case OP_NOTPOSQUERY:
2360          case OP_NOTPOSQUERYI:
2361          if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2362        break;        break;
2363        }        }
2364    #else
2365        (void)(utf);  /* Keep compiler happy by referencing function argument */
2366    #endif
2367      }      }
2368    }    }
2369  }  }
# Line 1343  for (;;) Line 2378  for (;;)
2378  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()
2379  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
2380  group that can match nothing. Note that first_significant_code() skips over  group that can match nothing. Note that first_significant_code() skips over
2381  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
2382  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
2383    bracket whose current branch will already have been scanned.
2384    
2385  Arguments:  Arguments:
2386    code        points to start of search    code        points to start of search
2387    endcode     points to where to stop    endcode     points to where to stop
2388    utf8        TRUE if in UTF8 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2389      cd          contains pointers to tables etc.
2390      recurses    chain of recurse_check to catch mutual recursion
2391    
2392  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2393  */  */
2394    
2395  static BOOL  static BOOL
2396  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2397      BOOL utf, compile_data *cd, recurse_check *recurses)
2398  {  {
2399  register int c;  register pcre_uchar c;
2400  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  recurse_check this_recurse;
2401    
2402    for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2403       code < endcode;       code < endcode;
2404       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
2405    {    {
2406    const uschar *ccode;    const pcre_uchar *ccode;
2407    
2408    c = *code;    c = *code;
2409    
2410    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE)    /* Skip over forward assertions; the other assertions are skipped by
2411      first_significant_code() with a TRUE final argument. */
2412    
2413      if (c == OP_ASSERT)
2414      {      {
2415        do code += GET(code, 1); while (*code == OP_ALT);
2416        c = *code;
2417        continue;
2418        }
2419    
2420      /* For a recursion/subroutine call, if its end has been reached, which
2421      implies a backward reference subroutine call, we can scan it. If it's a
2422      forward reference subroutine call, we can't. To detect forward reference
2423      we have to scan up the list that is kept in the workspace. This function is
2424      called only when doing the real compile, not during the pre-compile that
2425      measures the size of the compiled pattern. */
2426    
2427      if (c == OP_RECURSE)
2428        {
2429        const pcre_uchar *scode = cd->start_code + GET(code, 1);
2430        const pcre_uchar *endgroup = scode;
2431      BOOL empty_branch;      BOOL empty_branch;
     if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */  
2432    
2433      /* Scan a closed bracket */      /* Test for forward reference or uncompleted reference. This is disabled
2434        when called to scan a completed pattern by setting cd->start_workspace to
2435        NULL. */
2436    
2437        if (cd->start_workspace != NULL)
2438          {
2439          const pcre_uchar *tcode;
2440          for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)
2441            if ((int)GET(tcode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;
2442          if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2443          }
2444    
2445        /* If the reference is to a completed group, we need to detect whether this
2446        is a recursive call, as otherwise there will be an infinite loop. If it is
2447        a recursion, just skip over it. Simple recursions are easily detected. For
2448        mutual recursions we keep a chain on the stack. */
2449    
2450        do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);
2451        if (code >= scode && code <= endgroup) continue;  /* Simple recursion */
2452        else
2453          {
2454          recurse_check *r = recurses;
2455          for (r = recurses; r != NULL; r = r->prev)
2456            if (r->group == scode) break;
2457          if (r != NULL) continue;   /* Mutual recursion */
2458          }
2459    
2460        /* Completed reference; scan the referenced group, remembering it on the
2461        stack chain to detect mutual recursions. */
2462    
2463      empty_branch = FALSE;      empty_branch = FALSE;
2464        this_recurse.prev = recurses;
2465        this_recurse.group = scode;
2466    
2467      do      do
2468        {        {
2469        if (!empty_branch && could_be_empty_branch(code, endcode, utf8))        if (could_be_empty_branch(scode, endcode, utf, cd, &this_recurse))
2470            {
2471          empty_branch = TRUE;          empty_branch = TRUE;
2472        code += GET(code, 1);          break;
2473            }
2474          scode += GET(scode, 1);
2475        }        }
2476      while (*code == OP_ALT);      while (*scode == OP_ALT);
2477      if (!empty_branch) return FALSE;   /* All branches are non-empty */  
2478        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2479        continue;
2480        }
2481    
2482      /* Move past the KET and fudge things so that the increment in the "for"    /* Groups with zero repeats can of course be empty; skip them. */
     above has no effect. */  
2483    
2484      c = OP_END;    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2485      code += 1 + LINK_SIZE - _pcre_OP_lengths[c];        c == OP_BRAPOSZERO)
2486        {
2487        code += PRIV(OP_lengths)[c];
2488        do code += GET(code, 1); while (*code == OP_ALT);
2489        c = *code;
2490        continue;
2491        }
2492    
2493      /* A nested group that is already marked as "could be empty" can just be
2494      skipped. */
2495    
2496      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2497          c == OP_SCBRA || c == OP_SCBRAPOS)
2498        {
2499        do code += GET(code, 1); while (*code == OP_ALT);
2500        c = *code;
2501        continue;
2502        }
2503    
2504      /* For other groups, scan the branches. */
2505    
2506      if (c == OP_BRA  || c == OP_BRAPOS ||
2507          c == OP_CBRA || c == OP_CBRAPOS ||
2508          c == OP_ONCE || c == OP_ONCE_NC ||
2509          c == OP_COND || c == OP_SCOND)
2510        {
2511        BOOL empty_branch;
2512        if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
2513    
2514        /* If a conditional group has only one branch, there is a second, implied,
2515        empty branch, so just skip over the conditional, because it could be empty.
2516        Otherwise, scan the individual branches of the group. */
2517    
2518        if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
2519          code += GET(code, 1);
2520        else
2521          {
2522          empty_branch = FALSE;
2523          do
2524            {
2525            if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd,
2526              recurses)) empty_branch = TRUE;
2527            code += GET(code, 1);
2528            }
2529          while (*code == OP_ALT);
2530          if (!empty_branch) return FALSE;   /* All branches are non-empty */
2531          }
2532    
2533        c = *code;
2534      continue;      continue;
2535      }      }
2536    
# Line 1395  for (code = first_significant_code(code Line 2538  for (code = first_significant_code(code
2538    
2539    switch (c)    switch (c)
2540      {      {
2541      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
2542        cannot be represented just by a bit map. This includes negated single
2543        high-valued characters. The length in PRIV(OP_lengths)[] is zero; the
2544        actual length is stored in the compiled code, so we must update "code"
2545        here. */
2546    
2547  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2548      case OP_XCLASS:      case OP_XCLASS:
2549      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
2550      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
2551  #endif  #endif
2552    
2553      case OP_CLASS:      case OP_CLASS:
2554      case OP_NCLASS:      case OP_NCLASS:
2555      ccode = code + 33;      ccode = code + PRIV(OP_lengths)[OP_CLASS];
2556    
2557  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2558      CHECK_CLASS_REPEAT:      CHECK_CLASS_REPEAT:
2559  #endif  #endif
2560    
# Line 1417  for (code = first_significant_code(code Line 2564  for (code = first_significant_code(code
2564        case OP_CRMINSTAR:        case OP_CRMINSTAR:
2565        case OP_CRQUERY:        case OP_CRQUERY:
2566        case OP_CRMINQUERY:        case OP_CRMINQUERY:
2567          case OP_CRPOSSTAR:
2568          case OP_CRPOSQUERY:
2569        break;        break;
2570    
2571        default:                   /* Non-repeat => class must match */        default:                   /* Non-repeat => class must match */
2572        case OP_CRPLUS:            /* These repeats aren't empty */        case OP_CRPLUS:            /* These repeats aren't empty */
2573        case OP_CRMINPLUS:        case OP_CRMINPLUS:
2574          case OP_CRPOSPLUS:
2575        return FALSE;        return FALSE;
2576    
2577        case OP_CRRANGE:        case OP_CRRANGE:
2578        case OP_CRMINRANGE:        case OP_CRMINRANGE:
2579          case OP_CRPOSRANGE:
2580        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */
2581        break;        break;
2582        }        }
# Line 1433  for (code = first_significant_code(code Line 2584  for (code = first_significant_code(code
2584    
2585      /* Opcodes that must match a character */      /* Opcodes that must match a character */
2586    
2587        case OP_ANY:
2588        case OP_ALLANY:
2589        case OP_ANYBYTE:
2590    
2591      case OP_PROP:      case OP_PROP:
2592      case OP_NOTPROP:      case OP_NOTPROP:
2593        case OP_ANYNL:
2594    
2595        case OP_NOT_HSPACE:
2596        case OP_HSPACE:
2597        case OP_NOT_VSPACE:
2598        case OP_VSPACE:
2599      case OP_EXTUNI:      case OP_EXTUNI:
2600    
2601      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
2602      case OP_DIGIT:      case OP_DIGIT:
2603      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
2604      case OP_WHITESPACE:      case OP_WHITESPACE:
2605      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2606      case OP_WORDCHAR:      case OP_WORDCHAR:
2607      case OP_ANY:  
     case OP_ANYBYTE:  
2608      case OP_CHAR:      case OP_CHAR:
2609      case OP_CHARNC:      case OP_CHARI:
2610      case OP_NOT:      case OP_NOT:
2611        case OP_NOTI:
2612    
2613      case OP_PLUS:      case OP_PLUS:
2614        case OP_PLUSI:
2615      case OP_MINPLUS:      case OP_MINPLUS:
2616      case OP_POSPLUS:      case OP_MINPLUSI:
2617      case OP_EXACT:  
2618      case OP_NOTPLUS:      case OP_NOTPLUS:
2619        case OP_NOTPLUSI:
2620      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
2621        case OP_NOTMINPLUSI:
2622    
2623        case OP_POSPLUS:
2624        case OP_POSPLUSI:
2625      case OP_NOTPOSPLUS:      case OP_NOTPOSPLUS:
2626        case OP_NOTPOSPLUSI:
2627    
2628        case OP_EXACT:
2629        case OP_EXACTI:
2630      case OP_NOTEXACT:      case OP_NOTEXACT:
2631        case OP_NOTEXACTI:
2632    
2633      case OP_TYPEPLUS:      case OP_TYPEPLUS:
2634      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
2635      case OP_TYPEPOSPLUS:      case OP_TYPEPOSPLUS:
2636      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2637    
2638        return FALSE;
2639    
2640        /* These are going to continue, as they may be empty, but we have to
2641        fudge the length for the \p and \P cases. */
2642    
2643        case OP_TYPESTAR:
2644        case OP_TYPEMINSTAR:
2645        case OP_TYPEPOSSTAR:
2646        case OP_TYPEQUERY:
2647        case OP_TYPEMINQUERY:
2648        case OP_TYPEPOSQUERY:
2649        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2650        break;
2651    
2652        /* Same for these */
2653    
2654        case OP_TYPEUPTO:
2655        case OP_TYPEMINUPTO:
2656        case OP_TYPEPOSUPTO:
2657        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2658          code += 2;
2659        break;
2660    
2661        /* End of branch */
2662    
2663        case OP_KET:
2664        case OP_KETRMAX:
2665        case OP_KETRMIN:
2666        case OP_KETRPOS:
2667        case OP_ALT:
2668        return TRUE;
2669    
2670        /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2671        MINUPTO, and POSUPTO and their caseless and negative versions may be
2672        followed by a multibyte character. */
2673    
2674    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2675        case OP_STAR:
2676        case OP_STARI:
2677        case OP_NOTSTAR:
2678        case OP_NOTSTARI:
2679    
2680        case OP_MINSTAR:
2681        case OP_MINSTARI:
2682        case OP_NOTMINSTAR:
2683        case OP_NOTMINSTARI:
2684    
2685        case OP_POSSTAR:
2686        case OP_POSSTARI:
2687        case OP_NOTPOSSTAR:
2688        case OP_NOTPOSSTARI:
2689    
2690        case OP_QUERY:
2691        case OP_QUERYI:
2692        case OP_NOTQUERY:
2693        case OP_NOTQUERYI:
2694    
2695        case OP_MINQUERY:
2696        case OP_MINQUERYI:
2697        case OP_NOTMINQUERY:
2698        case OP_NOTMINQUERYI:
2699    
2700        case OP_POSQUERY:
2701        case OP_POSQUERYI:
2702        case OP_NOTPOSQUERY:
2703        case OP_NOTPOSQUERYI:
2704    
2705        if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2706        break;
2707    
2708        case OP_UPTO:
2709        case OP_UPTOI:
2710        case OP_NOTUPTO:
2711        case OP_NOTUPTOI:
2712    
2713        case OP_MINUPTO:
2714        case OP_MINUPTOI:
2715        case OP_NOTMINUPTO:
2716        case OP_NOTMINUPTOI:
2717    
2718        case OP_POSUPTO:
2719        case OP_POSUPTOI:
2720        case OP_NOTPOSUPTO:
2721        case OP_NOTPOSUPTOI:
2722    
2723        if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2724        break;
2725    #endif
2726    
2727        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2728        string. */
2729    
2730        case OP_MARK:
2731        case OP_PRUNE_ARG:
2732        case OP_SKIP_ARG:
2733        case OP_THEN_ARG:
2734        code += code[1];
2735        break;
2736    
2737        /* None of the remaining opcodes are required to match a character. */
2738    
2739        default:
2740        break;
2741        }
2742      }
2743    
2744    return TRUE;
2745    }
2746    
2747    
2748    
2749    /*************************************************
2750    *    Scan compiled regex for non-emptiness       *
2751    *************************************************/
2752    
2753    /* This function is called to check for left recursive calls. We want to check
2754    the current branch of the current pattern to see if it could match the empty
2755    string. If it could, we must look outwards for branches at other levels,
2756    stopping when we pass beyond the bracket which is the subject of the recursion.
2757    This function is called only during the real compile, not during the
2758    pre-compile.
2759    
2760    Arguments:
2761      code        points to start of the recursion
2762      endcode     points to where to stop (current RECURSE item)
2763      bcptr       points to the chain of current (unclosed) branch starts
2764      utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2765      cd          pointers to tables etc
2766    
2767    Returns:      TRUE if what is matched could be empty
2768    */
2769    
2770    static BOOL
2771    could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
2772      branch_chain *bcptr, BOOL utf, compile_data *cd)
2773    {
2774    while (bcptr != NULL && bcptr->current_branch >= code)
2775      {
2776      if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd, NULL))
2777      return FALSE;      return FALSE;
2778      bcptr = bcptr->outer;
2779      }
2780    return TRUE;
2781    }
2782    
2783    
2784    
2785    /*************************************************
2786    *        Base opcode of repeated opcodes         *
2787    *************************************************/
2788    
2789    /* Returns the base opcode for repeated single character type opcodes. If the
2790    opcode is not a repeated character type, it returns with the original value.
2791    
2792    Arguments:  c opcode
2793    Returns:    base opcode for the type
2794    */
2795    
2796    static pcre_uchar
2797    get_repeat_base(pcre_uchar c)
2798    {
2799    return (c > OP_TYPEPOSUPTO)? c :
2800           (c >= OP_TYPESTAR)?   OP_TYPESTAR :
2801           (c >= OP_NOTSTARI)?   OP_NOTSTARI :
2802           (c >= OP_NOTSTAR)?    OP_NOTSTAR :
2803           (c >= OP_STARI)?      OP_STARI :
2804                                 OP_STAR;
2805    }
2806    
2807    
2808    
2809    #ifdef SUPPORT_UCP
2810    /*************************************************
2811    *        Check a character and a property        *
2812    *************************************************/
2813    
2814    /* This function is called by check_auto_possessive() when a property item
2815    is adjacent to a fixed character.
2816    
2817    Arguments:
2818      c            the character
2819      ptype        the property type
2820      pdata        the data for the type
2821      negated      TRUE if it's a negated property (\P or \p{^)
2822    
2823    Returns:       TRUE if auto-possessifying is OK
2824    */
2825    
2826    static BOOL
2827    check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata,
2828      BOOL negated)
2829    {
2830    const pcre_uint32 *p;
2831    const ucd_record *prop = GET_UCD(c);
2832    
2833    switch(ptype)
2834      {
2835      case PT_LAMP:
2836      return (prop->chartype == ucp_Lu ||
2837              prop->chartype == ucp_Ll ||
2838              prop->chartype == ucp_Lt) == negated;
2839    
2840      case PT_GC:
2841      return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2842    
2843      case PT_PC:
2844      return (pdata == prop->chartype) == negated;
2845    
2846      case PT_SC:
2847      return (pdata == prop->script) == negated;
2848    
2849      /* These are specials */
2850    
2851      case PT_ALNUM:
2852      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2853              PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2854    
2855      /* Perl space used to exclude VT, but from Perl 5.18 it is included, which
2856      means that Perl space and POSIX space are now identical. PCRE was changed
2857      at release 8.34. */
2858    
2859      case PT_SPACE:    /* Perl space */
2860      case PT_PXSPACE:  /* POSIX space */
2861      switch(c)
2862        {
2863        HSPACE_CASES:
2864        VSPACE_CASES:
2865        return negated;
2866    
2867        default:
2868        return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z) == negated;
2869        }
2870      break;  /* Control never reaches here */
2871    
2872      case PT_WORD:
2873      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2874              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2875              c == CHAR_UNDERSCORE) == negated;
2876    
2877      case PT_CLIST:
2878      p = PRIV(ucd_caseless_sets) + prop->caseset;
2879      for (;;)
2880        {
2881        if (c < *p) return !negated;
2882        if (c == *p++) return negated;
2883        }
2884      break;  /* Control never reaches here */
2885      }
2886    
2887    return FALSE;
2888    }
2889    #endif  /* SUPPORT_UCP */
2890    
2891    
2892    
2893    /*************************************************
2894    *        Fill the character property list        *
2895    *************************************************/
2896    
2897    /* Checks whether the code points to an opcode that can take part in auto-
2898    possessification, and if so, fills a list with its properties.
2899    
2900    Arguments:
2901      code        points to start of expression
2902      utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2903      fcc         points to case-flipping table
2904      list        points to output list
2905                  list[0] will be filled with the opcode
2906                  list[1] will be non-zero if this opcode
2907                    can match an empty character string
2908                  list[2..7] depends on the opcode
2909    
2910    Returns:      points to the start of the next opcode if *code is accepted
2911                  NULL if *code is not accepted
2912    */
2913    
2914    static const pcre_uchar *
2915    get_chr_property_list(const pcre_uchar *code, BOOL utf,
2916      const pcre_uint8 *fcc, pcre_uint32 *list)
2917    {
2918    pcre_uchar c = *code;
2919    pcre_uchar base;
2920    const pcre_uchar *end;
2921    pcre_uint32 chr;
2922    
2923    #ifdef SUPPORT_UCP
2924    pcre_uint32 *clist_dest;
2925    const pcre_uint32 *clist_src;
2926    #else
2927    utf = utf;  /* Suppress "unused parameter" compiler warning */
2928    #endif
2929    
2930    list[0] = c;
2931    list[1] = FALSE;
2932    code++;
2933    
2934    if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
2935      {
2936      base = get_repeat_base(c);
2937      c -= (base - OP_STAR);
2938    
2939      if (c == OP_UPTO || c == OP_MINUPTO || c == OP_EXACT || c == OP_POSUPTO)
2940        code += IMM2_SIZE;
2941    
2942      list[1] = (c != OP_PLUS && c != OP_MINPLUS && c != OP_EXACT && c != OP_POSPLUS);
2943    
2944      switch(base)
2945        {
2946        case OP_STAR:
2947        list[0] = OP_CHAR;
2948        break;
2949    
2950        case OP_STARI:
2951        list[0] = OP_CHARI;
2952        break;
2953    
2954        case OP_NOTSTAR:
2955        list[0] = OP_NOT;
2956        break;
2957    
2958        case OP_NOTSTARI:
2959        list[0] = OP_NOTI;
2960        break;
2961    
2962        case OP_TYPESTAR:
2963        list[0] = *code;
2964        code++;
2965        break;
2966        }
2967      c = list[0];
2968      }
2969    
2970    switch(c)
2971      {
2972      case OP_NOT_DIGIT:
2973      case OP_DIGIT:
2974      case OP_NOT_WHITESPACE:
2975      case OP_WHITESPACE:
2976      case OP_NOT_WORDCHAR:
2977      case OP_WORDCHAR:
2978      case OP_ANY:
2979      case OP_ALLANY:
2980      case OP_ANYNL:
2981      case OP_NOT_HSPACE:
2982      case OP_HSPACE:
2983      case OP_NOT_VSPACE:
2984      case OP_VSPACE:
2985      case OP_EXTUNI:
2986      case OP_EODN:
2987      case OP_EOD:
2988      case OP_DOLL:
2989      case OP_DOLLM:
2990      return code;
2991    
2992      case OP_CHAR:
2993      case OP_NOT:
2994      GETCHARINCTEST(chr, code);
2995      list[2] = chr;
2996      list[3] = NOTACHAR;
2997      return code;
2998    
2999      case OP_CHARI:
3000      case OP_NOTI:
3001      list[0] = (c == OP_CHARI) ? OP_CHAR : OP_NOT;
3002      GETCHARINCTEST(chr, code);
3003      list[2] = chr;
3004    
3005    #ifdef SUPPORT_UCP
3006      if (chr < 128 || (chr < 256 && !utf))
3007        list[3] = fcc[chr];
3008      else
3009        list[3] = UCD_OTHERCASE(chr);
3010    #elif defined SUPPORT_UTF || !defined COMPILE_PCRE8
3011      list[3] = (chr < 256) ? fcc[chr] : chr;
3012    #else
3013      list[3] = fcc[chr];
3014    #endif
3015    
3016      /* The othercase might be the same value. */
3017    
3018      if (chr == list[3])
3019        list[3] = NOTACHAR;
3020      else
3021        list[4] = NOTACHAR;
3022      return code;
3023    
3024    #ifdef SUPPORT_UCP
3025      case OP_PROP:
3026      case OP_NOTPROP:
3027      if (code[0] != PT_CLIST)
3028        {
3029        list[2] = code[0];
3030        list[3] = code[1];
3031        return code + 2;
3032        }
3033    
3034      /* Convert only if we have enough space. */
3035    
3036      clist_src = PRIV(ucd_caseless_sets) + code[1];
3037      clist_dest = list + 2;
3038      code += 2;
3039    
3040      do {
3041         if (clist_dest >= list + 8)
3042           {
3043           /* Early return if there is not enough space. This should never
3044           happen, since all clists are shorter than 5 character now. */
3045           list[2] = code[0];
3046           list[3] = code[1];
3047           return code;
3048           }
3049         *clist_dest++ = *clist_src;
3050         }
3051      while(*clist_src++ != NOTACHAR);
3052    
3053      /* All characters are stored. The terminating NOTACHAR
3054      is copied form the clist itself. */
3055    
3056      list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
3057      return code;
3058    #endif
3059    
3060      case OP_NCLASS:
3061      case OP_CLASS:
3062    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3063      case OP_XCLASS:
3064      if (c == OP_XCLASS)
3065        end = code + GET(code, 0) - 1;
3066      else
3067    #endif
3068        end = code + 32 / sizeof(pcre_uchar);
3069    
3070      switch(*end)
3071        {
3072        case OP_CRSTAR:
3073        case OP_CRMINSTAR:
3074        case OP_CRQUERY:
3075        case OP_CRMINQUERY:
3076        case OP_CRPOSSTAR:
3077        case OP_CRPOSQUERY:
3078        list[1] = TRUE;
3079        end++;
3080        break;
3081    
3082        case OP_CRPLUS:
3083        case OP_CRMINPLUS:
3084        case OP_CRPOSPLUS:
3085        end++;
3086        break;
3087    
3088        case OP_CRRANGE:
3089        case OP_CRMINRANGE:
3090        case OP_CRPOSRANGE:
3091        list[1] = (GET2(end, 1) == 0);
3092        end += 1 + 2 * IMM2_SIZE;
3093        break;
3094        }
3095      list[2] = (pcre_uint32)(end - code);
3096      return end;
3097      }
3098    return NULL;    /* Opcode not accepted */
3099    }
3100    
3101    
3102    
3103    /*************************************************
3104    *    Scan further character sets for match       *
3105    *************************************************/
3106    
3107    /* Checks whether the base and the current opcode have a common character, in
3108    which case the base cannot be possessified.
3109    
3110    Arguments:
3111      code        points to the byte code
3112      utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
3113      cd          static compile data
3114      base_list   the data list of the base opcode
3115    
3116    Returns:      TRUE if the auto-possessification is possible
3117    */
3118    
3119    static BOOL
3120    compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
3121      const pcre_uint32 *base_list, const pcre_uchar *base_end, int *rec_limit)
3122    {
3123    pcre_uchar c;
3124    pcre_uint32 list[8];
3125    const pcre_uint32 *chr_ptr;
3126    const pcre_uint32 *ochr_ptr;
3127    const pcre_uint32 *list_ptr;
3128    const pcre_uchar *next_code;
3129    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3130    const pcre_uchar *xclass_flags;
3131    #endif
3132    const pcre_uint8 *class_bitset;
3133    const pcre_uint8 *set1, *set2, *set_end;
3134    pcre_uint32 chr;
3135    BOOL accepted, invert_bits;
3136    BOOL entered_a_group = FALSE;
3137    
3138    if (*rec_limit == 0) return FALSE;
3139    --(*rec_limit);
3140    
3141    /* Note: the base_list[1] contains whether the current opcode has greedy
3142    (represented by a non-zero value) quantifier. This is a different from
3143    other character type lists, which stores here that the character iterator
3144    matches to an empty string (also represented by a non-zero value). */
3145    
3146    for(;;)
3147      {
3148      /* All operations move the code pointer forward.
3149      Therefore infinite recursions are not possible. */
3150    
3151      c = *code;
3152    
3153      /* Skip over callouts */
3154    
3155      if (c == OP_CALLOUT)
3156        {
3157        code += PRIV(OP_lengths)[c];
3158        continue;
3159        }
3160    
3161      if (c == OP_ALT)
3162        {
3163        do code += GET(code, 1); while (*code == OP_ALT);
3164        c = *code;
3165        }
3166    
3167      switch(c)
3168        {
3169        case OP_END:
3170        case OP_KETRPOS:
3171        /* TRUE only in greedy case. The non-greedy case could be replaced by
3172        an OP_EXACT, but it is probably not worth it. (And note that OP_EXACT
3173        uses more memory, which we cannot get at this stage.) */
3174    
3175        return base_list[1] != 0;
3176    
3177        case OP_KET:
3178        /* If the bracket is capturing, and referenced by an OP_RECURSE, or
3179        it is an atomic sub-pattern (assert, once, etc.) the non-greedy case
3180        cannot be converted to a possessive form. */
3181    
3182        if (base_list[1] == 0) return FALSE;
3183    
3184        switch(*(code - GET(code, 1)))
3185          {
3186          case OP_ASSERT:
3187          case OP_ASSERT_NOT:
3188          case OP_ASSERTBACK:
3189          case OP_ASSERTBACK_NOT:
3190          case OP_ONCE:
3191          case OP_ONCE_NC:
3192          /* Atomic sub-patterns and assertions can always auto-possessify their
3193          last iterator. However, if the group was entered as a result of checking
3194          a previous iterator, this is not possible. */
3195    
3196          return !entered_a_group;
3197          }
3198    
3199        code += PRIV(OP_lengths)[c];
3200        continue;
3201    
3202        case OP_ONCE:
3203        case OP_ONCE_NC:
3204        case OP_BRA:
3205        case OP_CBRA:
3206        next_code = code + GET(code, 1);
3207        code += PRIV(OP_lengths)[c];
3208    
3209        while (*next_code == OP_ALT)
3210          {
3211          if (!compare_opcodes(code, utf, cd, base_list, base_end, rec_limit))
3212            return FALSE;
3213          code = next_code + 1 + LINK_SIZE;
3214          next_code += GET(next_code, 1);
3215          }
3216    
3217        entered_a_group = TRUE;
3218        continue;
3219    
3220        case OP_BRAZERO:
3221        case OP_BRAMINZERO:
3222    
3223        next_code = code + 1;
3224        if (*next_code != OP_BRA && *next_code != OP_CBRA
3225            && *next_code != OP_ONCE && *next_code != OP_ONCE_NC) return FALSE;
3226    
3227        do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
3228    
3229        /* The bracket content will be checked by the
3230        OP_BRA/OP_CBRA case above. */
3231        next_code += 1 + LINK_SIZE;
3232        if (!compare_opcodes(next_code, utf, cd, base_list, base_end, rec_limit))
3233          return FALSE;
3234    
3235        code += PRIV(OP_lengths)[c];
3236        continue;
3237    
3238        default:
3239        break;
3240        }
3241    
3242      /* Check for a supported opcode, and load its properties. */
3243    
3244      code = get_chr_property_list(code, utf, cd->fcc, list);
3245      if (code == NULL) return FALSE;    /* Unsupported */
3246    
3247      /* If either opcode is a small character list, set pointers for comparing
3248      characters from that list with another list, or with a property. */
3249    
3250      if (base_list[0] == OP_CHAR)
3251        {
3252        chr_ptr = base_list + 2;
3253        list_ptr = list;
3254        }
3255      else if (list[0] == OP_CHAR)
3256        {
3257        chr_ptr = list + 2;
3258        list_ptr = base_list;
3259        }
3260    
3261      /* Character bitsets can also be compared to certain opcodes. */
3262    
3263      else if (base_list[0] == OP_CLASS || list[0] == OP_CLASS
3264    #ifdef COMPILE_PCRE8
3265          /* In 8 bit, non-UTF mode, OP_CLASS and OP_NCLASS are the same. */
3266          || (!utf && (base_list[0] == OP_NCLASS || list[0] == OP_NCLASS))
3267    #endif
3268          )
3269        {
3270    #ifdef COMPILE_PCRE8
3271        if (base_list[0] == OP_CLASS || (!utf && base_list[0] == OP_NCLASS))
3272    #else
3273        if (base_list[0] == OP_CLASS)
3274    #endif
3275          {
3276          set1 = (pcre_uint8 *)(base_end - base_list[2]);
3277          list_ptr = list;
3278          }
3279        else
3280          {
3281          set1 = (pcre_uint8 *)(code - list[2]);
3282          list_ptr = base_list;
3283          }
3284    
3285        invert_bits = FALSE;
3286        switch(list_ptr[0])
3287          {
3288          case OP_CLASS:
3289          case OP_NCLASS:
3290          set2 = (pcre_uint8 *)
3291            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3292          break;
3293    
3294    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3295          case OP_XCLASS:
3296          xclass_flags = (list_ptr == list ? code : base_end) - list_ptr[2] + LINK_SIZE;
3297          if ((*xclass_flags & XCL_HASPROP) != 0) return FALSE;
3298          if ((*xclass_flags & XCL_MAP) == 0)
3299            {
3300            /* No bits are set for characters < 256. */
3301            if (list[1] == 0) return TRUE;
3302            /* Might be an empty repeat. */
3303            continue;
3304            }
3305          set2 = (pcre_uint8 *)(xclass_flags + 1);
3306          break;
3307    #endif
3308    
3309          case OP_NOT_DIGIT:
3310          invert_bits = TRUE;
3311          /* Fall through */
3312