/[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 77 by nigel, Sat Feb 24 21:40:45 2007 UTC revision 1069 by chpe, Tue Oct 16 15:54:29 2012 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) 1997-2005 University of Cambridge             Copyright (c) 1997-2012 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 42  POSSIBILITY OF SUCH DAMAGE. Line 42  POSSIBILITY OF SUCH DAMAGE.
42  supporting internal functions that are not used by other modules. */  supporting internal functions that are not used by other modules. */
43    
44    
45    #ifdef HAVE_CONFIG_H
46    #include "config.h"
47    #endif
48    
49    #define NLBLOCK cd             /* Block containing newline information */
50    #define PSSTART start_pattern  /* Field containing processed string start */
51    #define PSEND   end_pattern    /* Field containing processed string end */
52    
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56    /* When PCRE_DEBUG is defined, we need the pcre(16|32)_printint() function, which
57    is also used by pcretest. PCRE_DEBUG is not defined when building a production
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 PCRE_DEBUG
62    /* pcre_printint.c should not include any headers */
63    #define PCRE_INCLUDED
64    #include "pcre_printint.c"
65    #undef PCRE_INCLUDED
66    #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,
88        int, int, int *, int *, branch_chain *, compile_data *, int *);
89    
90    
91    
92  /*************************************************  /*************************************************
93  *      Code parameters and static tables         *  *      Code parameters and static tables         *
94  *************************************************/  *************************************************/
95    
96  /* Maximum number of items on the nested bracket stacks at compile time. This  /* This value specifies the size of stack workspace that is used during the
97  applies to the nesting of all kinds of parentheses. It does not limit  first pre-compile phase that determines how much memory is required. The regex
98  un-nested, non-capturing parentheses. This number can be made bigger if  is partly compiled into this space, but the compiled parts are discarded as
99  necessary - it is used to dimension one int and one unsigned char vector at  soon as they can be, so that hopefully there will never be an overrun. The code
100  compile time. */  does, however, check for an overrun. The largest amount I've seen used is 218,
101    so this number is very generous.
102    
103    The same workspace is used during the second, actual compile phase for
104    remembering forward references to groups so that they can be filled in at the
105    end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE
106    is 4 there is plenty of room for most patterns. However, the memory can get
107    filled up by repetitions of forward references, for example patterns like
108    /(?1){0,1999}(b)/, and one user did hit the limit. The code has been changed so
109    that the workspace is expanded using malloc() in this situation. The value
110    below is therefore a minimum, and we put a maximum on it for safety. The
111    minimum is now also defined in terms of LINK_SIZE so that the use of malloc()
112    kicks in at the same number of forward references in all cases. */
113    
114    #define COMPILE_WORK_SIZE (2048*LINK_SIZE)
115    #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)
116    
117    /* The overrun tests check for a slightly smaller size so that they detect the
118    overrun before it actually does run off the end of the data block. */
119    
120    #define WORK_SIZE_SAFETY_MARGIN (100)
121    
122    /* Private flags added to firstchar and reqchar. */
123    
124    #define REQ_CASELESS   0x10000000l      /* Indicates caselessness */
125    #define REQ_VARY       0x20000000l      /* Reqchar followed non-literal item */
126    #define REQ_MASK       (REQ_CASELESS | REQ_VARY)
127    
128  #define BRASTACK_SIZE 200  /* Repeated character flags. */
129    
130    #define UTF_LENGTH     0x10000000l      /* The char contains its length. */
131    
132  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
133  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
134  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
135  is invalid. */  is invalid. */
136    
137  #if !EBCDIC   /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC
138    
139    /* This is the "normal" table for ASCII systems or for EBCDIC systems running
140    in UTF-8 mode. */
141    
142  static const short int escapes[] = {  static const short int escapes[] = {
143       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,                       0,
144       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,                       0,
145     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */       0,                       0,
146       0,      0,      0,      0,      0,      0,      0,      0,   /* H - O */       0,                       0,
147  -ESC_P, -ESC_Q,      0, -ESC_S,      0,      0,      0, -ESC_W,   /* P - W */       0,                       0,
148  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */       CHAR_COLON,              CHAR_SEMICOLON,
149     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
150       0,      0,      0,      0,      0,      0,  ESC_n,      0,   /* h - o */       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
151  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0,      0, -ESC_w,   /* p - w */       CHAR_COMMERCIAL_AT,      -ESC_A,
152       0,      0, -ESC_z                                            /* x - z */       -ESC_B,                  -ESC_C,
153         -ESC_D,                  -ESC_E,
154         0,                       -ESC_G,
155         -ESC_H,                  0,
156         0,                       -ESC_K,
157         0,                       0,
158         -ESC_N,                  0,
159         -ESC_P,                  -ESC_Q,
160         -ESC_R,                  -ESC_S,
161         0,                       0,
162         -ESC_V,                  -ESC_W,
163         -ESC_X,                  0,
164         -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,
165         CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,
166         CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,
167         CHAR_GRAVE_ACCENT,       7,
168         -ESC_b,                  0,
169         -ESC_d,                  ESC_e,
170         ESC_f,                   0,
171         -ESC_h,                  0,
172         0,                       -ESC_k,
173         0,                       0,
174         ESC_n,                   0,
175         -ESC_p,                  0,
176         ESC_r,                   -ESC_s,
177         ESC_tee,                 0,
178         -ESC_v,                  -ESC_w,
179         0,                       0,
180         -ESC_z
181  };  };
182    
183  #else         /* This is the "abnormal" table for EBCDIC systems */  #else
184    
185    /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
186    
187  static const short int escapes[] = {  static const short int escapes[] = {
188  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
189  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 87  static const short int escapes[] = { Line 193  static const short int escapes[] = {
193  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,
194  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',
195  /*  80 */     0,     7, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,  /*  80 */     0,     7, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,
196  /*  88 */     0,     0,      0,     '{',      0,     0,      0,      0,  /*  88 */-ESC_h,     0,      0,     '{',      0,     0,      0,      0,
197  /*  90 */     0,     0,      0,     'l',      0, ESC_n,      0, -ESC_p,  /*  90 */     0,     0, -ESC_k,     'l',      0, ESC_n,      0, -ESC_p,
198  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,
199  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,     0, -ESC_w,      0,  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,-ESC_v, -ESC_w,      0,
200  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,
201  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,
202  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
203  /*  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,
204  /*  C8 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
205  /*  D0 */   '}',     0,      0,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,-ESC_N,      0, -ESC_P,
206  /*  D8 */-ESC_Q,     0,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
207  /*  E0 */  '\\',     0, -ESC_S,       0,      0,     0, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
208  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
209  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,
210  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0
# Line 106  static const short int escapes[] = { Line 212  static const short int escapes[] = {
212  #endif  #endif
213    
214    
215  /* 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
216  terminated by a zero length entry. The first three must be alpha, upper, lower,  searched linearly. Put all the names into a single string, in order to reduce
217  as this is assumed for handling case independence. */  the number of relocations when a shared library is dynamically linked. The
218    string is built from string macros so that it works in UTF-8 mode on EBCDIC
219  static const char *const posix_names[] = {  platforms. */
220    "alpha", "lower", "upper",  
221    "alnum", "ascii", "blank", "cntrl", "digit", "graph",  typedef struct verbitem {
222    "print", "punct", "space", "word",  "xdigit" };    int   len;                 /* Length of verb name */
223      int   op;                  /* Op when no arg, or -1 if arg mandatory */
224      int   op_arg;              /* Op when arg present, or -1 if not allowed */
225    } verbitem;
226    
227    static const char verbnames[] =
228      "\0"                       /* Empty name is a shorthand for MARK */
229      STRING_MARK0
230      STRING_ACCEPT0
231      STRING_COMMIT0
232      STRING_F0
233      STRING_FAIL0
234      STRING_PRUNE0
235      STRING_SKIP0
236      STRING_THEN;
237    
238    static const verbitem verbs[] = {
239      { 0, -1,        OP_MARK },
240      { 4, -1,        OP_MARK },
241      { 6, OP_ACCEPT, -1 },
242      { 6, OP_COMMIT, -1 },
243      { 1, OP_FAIL,   -1 },
244      { 4, OP_FAIL,   -1 },
245      { 5, OP_PRUNE,  OP_PRUNE_ARG },
246      { 4, OP_SKIP,   OP_SKIP_ARG  },
247      { 4, OP_THEN,   OP_THEN_ARG  }
248    };
249    
250    static const int verbcount = sizeof(verbs)/sizeof(verbitem);
251    
252    
253  static const uschar posix_name_lengths[] = {  /* Tables of names of POSIX character classes and their lengths. The names are
254    now all in a single string, to reduce the number of relocations when a shared
255    library is dynamically loaded. The list of lengths is terminated by a zero
256    length entry. The first three must be alpha, lower, upper, as this is assumed
257    for handling case independence. */
258    
259    static const char posix_names[] =
260      STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
261      STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
262      STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
263      STRING_word0  STRING_xdigit;
264    
265    static const pcre_uint8 posix_name_lengths[] = {
266    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 };
267    
268  /* Table of class bit maps for each POSIX class; up to three may be combined  /* Table of class bit maps for each POSIX class. Each class is formed from a
269  to form the class. The table for [:blank:] is dynamically modified to remove  base map, with an optional addition or removal of another map. Then, for some
270  the vertical space characters. */  classes, there is some additional tweaking: for [:blank:] the vertical space
271    characters are removed, and for [:alpha:] and [:alnum:] the underscore
272    character is removed. The triples in the table consist of the base map offset,
273    second map offset or -1 if no second map, and a non-negative value for map
274    addition or a negative value for map subtraction (if there are two maps). The
275    absolute value of the third field has these meanings: 0 => no tweaking, 1 =>
276    remove vertical space characters, 2 => remove underscore. */
277    
278  static const int posix_class_maps[] = {  static const int posix_class_maps[] = {
279    cbit_lower, cbit_upper, -1,             /* alpha */    cbit_word,  cbit_digit, -2,             /* alpha */
280    cbit_lower, -1,         -1,             /* lower */    cbit_lower, -1,          0,             /* lower */
281    cbit_upper, -1,         -1,             /* upper */    cbit_upper, -1,          0,             /* upper */
282    cbit_digit, cbit_lower, cbit_upper,     /* alnum */    cbit_word,  -1,          2,             /* alnum - word without underscore */
283    cbit_print, cbit_cntrl, -1,             /* ascii */    cbit_print, cbit_cntrl,  0,             /* ascii */
284    cbit_space, -1,         -1,             /* blank - a GNU extension */    cbit_space, -1,          1,             /* blank - a GNU extension */
285    cbit_cntrl, -1,         -1,             /* cntrl */    cbit_cntrl, -1,          0,             /* cntrl */
286    cbit_digit, -1,         -1,             /* digit */    cbit_digit, -1,          0,             /* digit */
287    cbit_graph, -1,         -1,             /* graph */    cbit_graph, -1,          0,             /* graph */
288    cbit_print, -1,         -1,             /* print */    cbit_print, -1,          0,             /* print */
289    cbit_punct, -1,         -1,             /* punct */    cbit_punct, -1,          0,             /* punct */
290    cbit_space, -1,         -1,             /* space */    cbit_space, -1,          0,             /* space */
291    cbit_word,  -1,         -1,             /* word - a Perl extension */    cbit_word,  -1,          0,             /* word - a Perl extension */
292    cbit_xdigit,-1,         -1              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
293  };  };
294    
295    /* Table of substitutes for \d etc when PCRE_UCP is set. The POSIX class
296    substitutes must be in the order of the names, defined above, and there are
297    both positive and negative cases. NULL means no substitute. */
298    
299  /* The texts of compile-time error messages. These are "char *" because they  #ifdef SUPPORT_UCP
300  are passed to the outside world. */  static const pcre_uchar string_PNd[]  = {
301      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
302      CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
303    static const pcre_uchar string_pNd[]  = {
304      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
305      CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
306    static const pcre_uchar string_PXsp[] = {
307      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
308      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
309    static const pcre_uchar string_pXsp[] = {
310      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
311      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
312    static const pcre_uchar string_PXwd[] = {
313      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
314      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
315    static const pcre_uchar string_pXwd[] = {
316      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
317      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
318    
319    static const pcre_uchar *substitutes[] = {
320      string_PNd,           /* \D */
321      string_pNd,           /* \d */
322      string_PXsp,          /* \S */       /* NOTE: Xsp is Perl space */
323      string_pXsp,          /* \s */
324      string_PXwd,          /* \W */
325      string_pXwd           /* \w */
326    };
327    
328    static const pcre_uchar string_pL[] =   {
329      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
330      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
331    static const pcre_uchar string_pLl[] =  {
332      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
333      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
334    static const pcre_uchar string_pLu[] =  {
335      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
336      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
337    static const pcre_uchar string_pXan[] = {
338      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
339      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
340    static const pcre_uchar string_h[] =    {
341      CHAR_BACKSLASH, CHAR_h, '\0' };
342    static const pcre_uchar string_pXps[] = {
343      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
344      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
345    static const pcre_uchar string_PL[] =   {
346      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
347      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
348    static const pcre_uchar string_PLl[] =  {
349      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
350      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
351    static const pcre_uchar string_PLu[] =  {
352      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
353      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
354    static const pcre_uchar string_PXan[] = {
355      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
356      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
357    static const pcre_uchar string_H[] =    {
358      CHAR_BACKSLASH, CHAR_H, '\0' };
359    static const pcre_uchar string_PXps[] = {
360      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
361      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
362    
363    static const pcre_uchar *posix_substitutes[] = {
364      string_pL,            /* alpha */
365      string_pLl,           /* lower */
366      string_pLu,           /* upper */
367      string_pXan,          /* alnum */
368      NULL,                 /* ascii */
369      string_h,             /* blank */
370      NULL,                 /* cntrl */
371      string_pNd,           /* digit */
372      NULL,                 /* graph */
373      NULL,                 /* print */
374      NULL,                 /* punct */
375      string_pXps,          /* space */    /* NOTE: Xps is POSIX space */
376      string_pXwd,          /* word */
377      NULL,                 /* xdigit */
378      /* Negated cases */
379      string_PL,            /* ^alpha */
380      string_PLl,           /* ^lower */
381      string_PLu,           /* ^upper */
382      string_PXan,          /* ^alnum */
383      NULL,                 /* ^ascii */
384      string_H,             /* ^blank */
385      NULL,                 /* ^cntrl */
386      string_PNd,           /* ^digit */
387      NULL,                 /* ^graph */
388      NULL,                 /* ^print */
389      NULL,                 /* ^punct */
390      string_PXps,          /* ^space */   /* NOTE: Xps is POSIX space */
391      string_PXwd,          /* ^word */
392      NULL                  /* ^xdigit */
393    };
394    #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))
395    #endif
396    
397    #define STRING(a)  # a
398    #define XSTRING(s) STRING(s)
399    
400  static const char *error_texts[] = {  /* The texts of compile-time error messages. These are "char *" because they
401    "no error",  are passed to the outside world. Do not ever re-use any error number, because
402    "\\ at end of pattern",  they are documented. Always add a new error instead. Messages marked DEAD below
403    "\\c at end of pattern",  are no longer used. This used to be a table of strings, but in order to reduce
404    "unrecognized character follows \\",  the number of relocations needed when a shared library is loaded dynamically,
405    "numbers out of order in {} quantifier",  it is now one long string. We cannot use a table of offsets, because the
406    lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
407    simply count through to the one we want - this isn't a performance issue
408    because these strings are used only when there is a compilation error.
409    
410    Each substring ends with \0 to insert a null character. This includes the final
411    substring, so that the whole string ends with \0\0, which can be detected when
412    counting through. */
413    
414    static const char error_texts[] =
415      "no error\0"
416      "\\ at end of pattern\0"
417      "\\c at end of pattern\0"
418      "unrecognized character follows \\\0"
419      "numbers out of order in {} quantifier\0"
420    /* 5 */    /* 5 */
421    "number too big in {} quantifier",    "number too big in {} quantifier\0"
422    "missing terminating ] for character class",    "missing terminating ] for character class\0"
423    "invalid escape sequence in character class",    "invalid escape sequence in character class\0"
424    "range out of order in character class",    "range out of order in character class\0"
425    "nothing to repeat",    "nothing to repeat\0"
426    /* 10 */    /* 10 */
427    "operand of unlimited repeat could match the empty string",    "operand of unlimited repeat could match the empty string\0"  /** DEAD **/
428    "internal error: unexpected repeat",    "internal error: unexpected repeat\0"
429    "unrecognized character after (?",    "unrecognized character after (? or (?-\0"
430    "POSIX named classes are supported only within a class",    "POSIX named classes are supported only within a class\0"
431    "missing )",    "missing )\0"
432    /* 15 */    /* 15 */
433    "reference to non-existent subpattern",    "reference to non-existent subpattern\0"
434    "erroffset passed as NULL",    "erroffset passed as NULL\0"
435    "unknown option bit(s) set",    "unknown option bit(s) set\0"
436    "missing ) after comment",    "missing ) after comment\0"
437    "parentheses nested too deeply",    "parentheses nested too deeply\0"  /** DEAD **/
438    /* 20 */    /* 20 */
439    "regular expression too large",    "regular expression is too large\0"
440    "failed to get memory",    "failed to get memory\0"
441    "unmatched parentheses",    "unmatched parentheses\0"
442    "internal error: code overflow",    "internal error: code overflow\0"
443    "unrecognized character after (?<",    "unrecognized character after (?<\0"
444    /* 25 */    /* 25 */
445    "lookbehind assertion is not fixed length",    "lookbehind assertion is not fixed length\0"
446    "malformed number after (?(",    "malformed number or name after (?(\0"
447    "conditional group contains more than two branches",    "conditional group contains more than two branches\0"
448    "assertion expected after (?(",    "assertion expected after (?(\0"
449    "(?R or (?digits must be followed by )",    "(?R or (?[+-]digits must be followed by )\0"
450    /* 30 */    /* 30 */
451    "unknown POSIX class name",    "unknown POSIX class name\0"
452    "POSIX collating elements are not supported",    "POSIX collating elements are not supported\0"
453    "this version of PCRE is not compiled with PCRE_UTF8 support",    "this version of PCRE is compiled without UTF support\0"
454    "spare error",    "spare error\0"  /** DEAD **/
455    "character value in \\x{...} sequence is too large",    "character value in \\x{...} sequence is too large\0"
456    /* 35 */    /* 35 */
457    "invalid condition (?(0)",    "invalid condition (?(0)\0"
458    "\\C not allowed in lookbehind assertion",    "\\C not allowed in lookbehind assertion\0"
459    "PCRE does not support \\L, \\l, \\N, \\U, or \\u",    "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
460    "number after (?C is > 255",    "number after (?C is > 255\0"
461    "closing ) for (?C expected",    "closing ) for (?C expected\0"
462    /* 40 */    /* 40 */
463    "recursive call could loop indefinitely",    "recursive call could loop indefinitely\0"
464    "unrecognized character after (?P",    "unrecognized character after (?P\0"
465    "syntax error after (?P",    "syntax error in subpattern name (missing terminator)\0"
466    "two named groups have the same name",    "two named subpatterns have the same name\0"
467    "invalid UTF-8 string",    "invalid UTF-8 string\0"
468    /* 45 */    /* 45 */
469    "support for \\P, \\p, and \\X has not been compiled",    "support for \\P, \\p, and \\X has not been compiled\0"
470    "malformed \\P or \\p sequence",    "malformed \\P or \\p sequence\0"
471    "unknown property name after \\P or \\p"    "unknown property name after \\P or \\p\0"
472  };    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
473      "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
474      /* 50 */
475      "repeated subpattern is too long\0"    /** DEAD **/
476      "octal value is greater than \\377 in 8-bit non-UTF-8 mode\0"
477      "internal error: overran compiling workspace\0"
478      "internal error: previously-checked referenced subpattern not found\0"
479      "DEFINE group contains more than one branch\0"
480      /* 55 */
481      "repeating a DEFINE group is not allowed\0"  /** DEAD **/
482      "inconsistent NEWLINE options\0"
483      "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
484      "a numbered reference must not be zero\0"
485      "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
486      /* 60 */
487      "(*VERB) not recognized\0"
488      "number is too big\0"
489      "subpattern name expected\0"
490      "digit expected after (?+\0"
491      "] is an invalid data character in JavaScript compatibility mode\0"
492      /* 65 */
493      "different names for subpatterns of the same number are not allowed\0"
494      "(*MARK) must have an argument\0"
495      "this version of PCRE is not compiled with Unicode property support\0"
496      "\\c must be followed by an ASCII character\0"
497      "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
498      /* 70 */
499      "internal error: unknown opcode in find_fixedlength()\0"
500      "\\N is not supported in a class\0"
501      "too many forward references\0"
502      "disallowed Unicode code point (>= 0xd800 && <= 0xdfff)\0"
503      "invalid UTF-16 string\0"
504      /* 75 */
505      "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"
506      "character value in \\u.... sequence is too large\0"
507      "invalid UTF-32 string\0"
508      ;
509    
510  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
511  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 220  For convenience, we use the same bit def Line 523  For convenience, we use the same bit def
523    
524  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
525    
526  #if !EBCDIC    /* This is the "normal" case, for ASCII systems */  /* Using a simple comparison for decimal numbers rather than a memory read
527  static const unsigned char digitab[] =  is much faster, and the resulting code is simpler (the compiler turns it
528    into a subtraction and unsigned comparison). */
529    
530    #define IS_DIGIT(x) ((x) >= CHAR_0 && (x) <= CHAR_9)
531    
532    #ifndef EBCDIC
533    
534    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
535    UTF-8 mode. */
536    
537    static const pcre_uint8 digitab[] =
538    {    {
539    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
540    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15 */
# Line 256  static const unsigned char digitab[] = Line 569  static const unsigned char digitab[] =
569    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
570    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
571    
572  #else          /* This is the "abnormal" case, for EBCDIC systems */  #else
573  static const unsigned char digitab[] =  
574    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
575    
576    static const pcre_uint8 digitab[] =
577    {    {
578    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
579    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15    */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15    */
# Line 270  static const unsigned char digitab[] = Line 586  static const unsigned char digitab[] =
586    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */
587    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */
588    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */
589    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88-     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88- 95    */
590    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */
591    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */
592    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */
# Line 292  static const unsigned char digitab[] = Line 608  static const unsigned char digitab[] =
608    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */
609    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */
610    
611  static const unsigned char ebcdic_chartab[] = { /* chartable partial dup */  static const pcre_uint8 ebcdic_chartab[] = { /* chartable partial dup */
612    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */
613    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */
614    0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*  16- 23 */    0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*  16- 23 */
# Line 304  static const unsigned char ebcdic_charta Line 620  static const unsigned char ebcdic_charta
620    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */
621    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */
622    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */
623    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88-  */    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88- 95 */
624    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */
625    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */
626    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */
# Line 328  static const unsigned char ebcdic_charta Line 644  static const unsigned char ebcdic_charta
644  #endif  #endif
645    
646    
647  /* Definition to allow mutual recursion */  
648    
649    /*************************************************
650    *            Find an error text                  *
651    *************************************************/
652    
653    /* The error texts are now all in one long string, to save on relocations. As
654    some of the text is of unknown length, we can't use a table of offsets.
655    Instead, just count through the strings. This is not a performance issue
656    because it happens only when there has been a compilation error.
657    
658    Argument:   the error number
659    Returns:    pointer to the error string
660    */
661    
662    static const char *
663    find_error_text(int n)
664    {
665    const char *s = error_texts;
666    for (; n > 0; n--)
667      {
668      while (*s++ != 0) {};
669      if (*s == 0) return "Error text not found (please report)";
670      }
671    return s;
672    }
673    
674    
675    /*************************************************
676    *           Expand the workspace                 *
677    *************************************************/
678    
679    /* This function is called during the second compiling phase, if the number of
680    forward references fills the existing workspace, which is originally a block on
681    the stack. A larger block is obtained from malloc() unless the ultimate limit
682    has been reached or the increase will be rather small.
683    
684    Argument: pointer to the compile data block
685    Returns:  0 if all went well, else an error number
686    */
687    
688    static int
689    expand_workspace(compile_data *cd)
690    {
691    pcre_uchar *newspace;
692    int newsize = cd->workspace_size * 2;
693    
694    if (newsize > COMPILE_WORK_SIZE_MAX) newsize = COMPILE_WORK_SIZE_MAX;
695    if (cd->workspace_size >= COMPILE_WORK_SIZE_MAX ||
696        newsize - cd->workspace_size < WORK_SIZE_SAFETY_MARGIN)
697     return ERR72;
698    
699    newspace = (PUBL(malloc))(IN_UCHARS(newsize));
700    if (newspace == NULL) return ERR21;
701    memcpy(newspace, cd->start_workspace, cd->workspace_size * sizeof(pcre_uchar));
702    cd->hwm = (pcre_uchar *)newspace + (cd->hwm - cd->start_workspace);
703    if (cd->workspace_size > COMPILE_WORK_SIZE)
704      (PUBL(free))((void *)cd->start_workspace);
705    cd->start_workspace = newspace;
706    cd->workspace_size = newsize;
707    return 0;
708    }
709    
710    
711    
712    /*************************************************
713    *            Check for counted repeat            *
714    *************************************************/
715    
716    /* This function is called when a '{' is encountered in a place where it might
717    start a quantifier. It looks ahead to see if it really is a quantifier or not.
718    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
719    where the ddds are digits.
720    
721    Arguments:
722      p         pointer to the first char after '{'
723    
724    Returns:    TRUE or FALSE
725    */
726    
727  static BOOL  static BOOL
728    compile_regex(int, int, int *, uschar **, const uschar **, int *, BOOL, int,  is_counted_repeat(const pcre_uchar *p)
729      int *, int *, branch_chain *, compile_data *);  {
730    if (!IS_DIGIT(*p)) return FALSE;
731    p++;
732    while (IS_DIGIT(*p)) p++;
733    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
734    
735    if (*p++ != CHAR_COMMA) return FALSE;
736    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
737    
738    if (!IS_DIGIT(*p)) return FALSE;
739    p++;
740    while (IS_DIGIT(*p)) p++;
741    
742    return (*p == CHAR_RIGHT_CURLY_BRACKET);
743    }
744    
745    
746    
# Line 341  static BOOL Line 749  static BOOL
749  *************************************************/  *************************************************/
750    
751  /* 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
752  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
753  encodes one of the more complicated things such as \d. When UTF-8 is enabled,  which will be placed in chptr. A backreference to group n is returned as
754  a positive value greater than 255 may be returned. On entry, ptr is pointing at  negative n. When UTF-8 is enabled, a positive value greater than 255 may
755  the \. On exit, it is on the final character of the escape sequence.  be returned in chptr.
756    On entry,ptr is pointing at the \. On exit, it is on the final character of the
757    escape sequence.
758    
759  Arguments:  Arguments:
760    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
761      chptr          points to the data character
762    errorcodeptr   points to the errorcode variable    errorcodeptr   points to the errorcode variable
763    bracount       number of previous extracting brackets    bracount       number of previous extracting brackets
764    options        the options bits    options        the options bits
765    isclass        TRUE if inside a character class    isclass        TRUE if inside a character class
766    
767  Returns:         zero or positive => a data character  Returns:         zero => a data character
768                   negative => a special escape sequence                   positive => a special escape sequence
769                   on error, errorptr is set                   negative => a back reference
770                     on error, errorcodeptr is set
771  */  */
772    
773  static int  static int
774  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,
775    int options, BOOL isclass)    int bracount, int options, BOOL isclass)
776  {  {
777  const uschar *ptr = *ptrptr;  /* PCRE_UTF16 has the same value as PCRE_UTF8. */
778  int c, i;  BOOL utf = (options & PCRE_UTF8) != 0;
779    const pcre_uchar *ptr = *ptrptr + 1;
780    pcre_uint32 c;
781    int escape = 0;
782    int i;
783    
784    GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
785    ptr--;                            /* Set pointer back to the last byte */
786    
787  /* 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. */
788    
 c = *(++ptr);  
789  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
790    
791  /* 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
792  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.
793  Otherwise further processing may be required. */  Otherwise further processing may be required. */
794    
795  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
796  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  /* Not alphanumeric */
797  else if ((i = escapes[c - '0']) != 0) c = i;  else if (c < CHAR_0 || c > CHAR_z) {}
798    else if ((i = escapes[c - CHAR_0]) != 0) { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
799  #else          /* EBCDIC coding */  
800  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  #else           /* EBCDIC coding */
801  else if ((i = escapes[c - 0x48]) != 0)  c = i;  /* Not alphanumeric */
802    else if (c < CHAR_a || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}
803    else if ((i = escapes[c - 0x48]) != 0)  { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
804  #endif  #endif
805    
806  /* Escapes that need further processing, or are illegal. */  /* Escapes that need further processing, or are illegal. */
807    
808  else  else
809    {    {
810    const uschar *oldptr;    const pcre_uchar *oldptr;
811      BOOL braced, negated, overflow;
812      int s;
813    
814    switch (c)    switch (c)
815      {      {
816      /* 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
817      error. */      error. */
818    
819      case 'l':      case CHAR_l:
820      case 'L':      case CHAR_L:
     case 'N':  
     case 'u':  
     case 'U':  
821      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
822      break;      break;
823    
824        case CHAR_u:
825        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
826          {
827          /* In JavaScript, \u must be followed by four hexadecimal numbers.
828          Otherwise it is a lowercase u letter. */
829          if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
830            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0
831            && MAX_255(ptr[3]) && (digitab[ptr[3]] & ctype_xdigit) != 0
832            && MAX_255(ptr[4]) && (digitab[ptr[4]] & ctype_xdigit) != 0)
833            {
834            c = 0;
835            for (i = 0; i < 4; ++i)
836              {
837              register pcre_uint32 cc = *(++ptr);
838    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
839              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
840              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
841    #else           /* EBCDIC coding */
842              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
843              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
844    #endif
845              }
846    
847    #if defined COMPILE_PCRE8
848            if (c > (utf ? 0x10ffff : 0xff))
849    #elif defined COMPILE_PCRE16
850            if (c > (utf ? 0x10ffff : 0xffff))
851    #elif defined COMPILE_PCRE32
852            if (utf && c > 0x10ffff)
853    #endif
854              {
855              *errorcodeptr = ERR76;
856              }
857            else if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
858            }
859          }
860        else
861          *errorcodeptr = ERR37;
862        break;
863    
864        case CHAR_U:
865        /* In JavaScript, \U is an uppercase U letter. */
866        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
867        break;
868    
869        /* In a character class, \g is just a literal "g". Outside a character
870        class, \g must be followed by one of a number of specific things:
871    
872        (1) A number, either plain or braced. If positive, it is an absolute
873        backreference. If negative, it is a relative backreference. This is a Perl
874        5.10 feature.
875    
876        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
877        is part of Perl's movement towards a unified syntax for back references. As
878        this is synonymous with \k{name}, we fudge it up by pretending it really
879        was \k.
880    
881        (3) For Oniguruma compatibility we also support \g followed by a name or a
882        number either in angle brackets or in single quotes. However, these are
883        (possibly recursive) subroutine calls, _not_ backreferences. Just return
884        the ESC_g code (cf \k). */
885    
886        case CHAR_g:
887        if (isclass) break;
888        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
889          {
890          escape = ESC_g;
891          break;
892          }
893    
894        /* Handle the Perl-compatible cases */
895    
896        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
897          {
898          const pcre_uchar *p;
899          for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
900            if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
901          if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
902            {
903            escape = ESC_k;
904            break;
905            }
906          braced = TRUE;
907          ptr++;
908          }
909        else braced = FALSE;
910    
911        if (ptr[1] == CHAR_MINUS)
912          {
913          negated = TRUE;
914          ptr++;
915          }
916        else negated = FALSE;
917    
918        /* The integer range is limited by the machine's int representation. */
919        s = 0;
920        overflow = FALSE;
921        while (IS_DIGIT(ptr[1]))
922          {
923          if (s > INT_MAX / 10 - 1) /* Integer overflow */
924            {
925            overflow = TRUE;
926            break;
927            }
928          s = s * 10 + (int)(*(++ptr) - CHAR_0);
929          }
930        if (overflow) /* Integer overflow */
931          {
932          while (IS_DIGIT(ptr[1]))
933            ptr++;
934          *errorcodeptr = ERR61;
935          break;
936          }
937    
938        if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
939          {
940          *errorcodeptr = ERR57;
941          break;
942          }
943    
944        if (s == 0)
945          {
946          *errorcodeptr = ERR58;
947          break;
948          }
949    
950        if (negated)
951          {
952          if (s > bracount)
953            {
954            *errorcodeptr = ERR15;
955            break;
956            }
957          s = bracount - (s - 1);
958          }
959    
960        escape = -s;
961        break;
962    
963      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
964      starting with one that is not zero is not straightforward. By experiment,      starting with one that is not zero is not straightforward. By experiment,
965      the way Perl works seems to be as follows:      the way Perl works seems to be as follows:
# Line 413  else Line 972  else
972      value is greater than 377, the least significant 8 bits are taken. Inside a      value is greater than 377, the least significant 8 bits are taken. Inside a
973      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
974    
975      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:
976      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
977    
978      if (!isclass)      if (!isclass)
979        {        {
980        oldptr = ptr;        oldptr = ptr;
981        c -= '0';        /* The integer range is limited by the machine's int representation. */
982        while ((digitab[ptr[1]] & ctype_digit) != 0)        s = (int)(c -CHAR_0);
983          c = c * 10 + *(++ptr) - '0';        overflow = FALSE;
984        if (c < 10 || c <= bracount)        while (IS_DIGIT(ptr[1]))
985            {
986            if (s > INT_MAX / 10 - 1) /* Integer overflow */
987              {
988              overflow = TRUE;
989              break;
990              }
991            s = s * 10 + (int)(*(++ptr) - CHAR_0);
992            }
993          if (overflow) /* Integer overflow */
994            {
995            while (IS_DIGIT(ptr[1]))
996              ptr++;
997            *errorcodeptr = ERR61;
998            break;
999            }
1000          if (s < 10 || s <= bracount)
1001          {          {
1002          c = -(ESC_REF + c);          escape = -s;
1003          break;          break;
1004          }          }
1005        ptr = oldptr;      /* Put the pointer back and fall through */        ptr = oldptr;      /* Put the pointer back and fall through */
# Line 434  else Line 1009  else
1009      generates a binary zero byte and treats the digit as a following literal.      generates a binary zero byte and treats the digit as a following literal.
1010      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
1011    
1012      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
1013        {        {
1014        ptr--;        ptr--;
1015        c = 0;        c = 0;
# Line 442  else Line 1017  else
1017        }        }
1018    
1019      /* \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
1020      larger first octal digit. */      larger first octal digit. The original code used just to take the least
1021        significant 8 bits of octal numbers (I think this is what early Perls used
1022      case '0':      to do). Nowadays we allow for larger numbers in UTF-8 mode and 16-bit mode,
1023      c -= '0';      but no more than 3 octal digits. */
1024      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')  
1025          c = c * 8 + *(++ptr) - '0';      case CHAR_0:
1026      c &= 255;     /* Take least significant 8 bits */      c -= CHAR_0;
1027        while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
1028            c = c * 8 + *(++ptr) - CHAR_0;
1029    #ifdef COMPILE_PCRE8
1030        if (!utf && c > 0xff) *errorcodeptr = ERR51;
1031    #endif
1032      break;      break;
1033    
1034      /* \x is complicated when UTF-8 is enabled. \x{ddd} is a character number      /* \x is complicated. \x{ddd} is a character number which can be greater
1035      which can be greater than 0xff, but only if the ddd are hex digits. */      than 0xff in utf or non-8bit mode, but only if the ddd are hex digits.
1036        If not, { is treated as a data character. */
1037    
1038        case CHAR_x:
1039        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1040          {
1041          /* In JavaScript, \x must be followed by two hexadecimal numbers.
1042          Otherwise it is a lowercase x letter. */
1043          if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1044            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1045            {
1046            c = 0;
1047            for (i = 0; i < 2; ++i)
1048              {
1049              register pcre_uint32 cc = *(++ptr);
1050    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1051              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1052              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1053    #else           /* EBCDIC coding */
1054              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1055              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1056    #endif
1057              }
1058            }
1059          break;
1060          }
1061    
1062      case 'x':      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
 #ifdef SUPPORT_UTF8  
     if (ptr[1] == '{' && (options & PCRE_UTF8) != 0)  
1063        {        {
1064        const uschar *pt = ptr + 2;        const pcre_uchar *pt = ptr + 2;
1065        register int count = 0;  
1066        c = 0;        c = 0;
1067        while ((digitab[*pt] & ctype_xdigit) != 0)        overflow = FALSE;
1068          while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0)
1069          {          {
1070          int cc = *pt++;          register pcre_uint32 cc = *pt++;
1071          count++;          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1072  #if !EBCDIC    /* ASCII coding */  
1073          if (cc >= 'a') cc -= 32;               /* Convert to upper case */  #ifdef COMPILE_PCRE32
1074          c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));          if (c >= 0x10000000l) { overflow = TRUE; break; }
1075  #else          /* EBCDIC coding */  #endif
1076          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */  
1077          c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1078            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1079            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1080    #else           /* EBCDIC coding */
1081            if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1082            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1083    #endif
1084    
1085    #if defined COMPILE_PCRE8
1086            if (c > (utf ? 0x10ffff : 0xff)) { overflow = TRUE; break; }
1087    #elif defined COMPILE_PCRE16
1088            if (c > (utf ? 0x10ffff : 0xffff)) { overflow = TRUE; break; }
1089    #elif defined COMPILE_PCRE32
1090            if (utf && c > 0x10ffff) { overflow = TRUE; break; }
1091  #endif  #endif
1092          }          }
1093        if (*pt == '}')  
1094          if (overflow)
1095            {
1096            while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0) pt++;
1097            *errorcodeptr = ERR34;
1098            }
1099    
1100          if (*pt == CHAR_RIGHT_CURLY_BRACKET)
1101          {          {
1102          if (c < 0 || count > 8) *errorcodeptr = ERR34;          if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1103          ptr = pt;          ptr = pt;
1104          break;          break;
1105          }          }
1106    
1107        /* If the sequence of hex digits does not end with '}', then we don't        /* If the sequence of hex digits does not end with '}', then we don't
1108        recognize this construct; fall through to the normal \x handling. */        recognize this construct; fall through to the normal \x handling. */
1109        }        }
 #endif  
1110    
1111      /* Read just a single hex char */      /* Read just a single-byte hex-defined char */
1112    
1113      c = 0;      c = 0;
1114      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1115        {        {
1116        int cc;                               /* Some compilers don't like ++ */        pcre_uint32 cc;                          /* Some compilers don't like */
1117        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
1118  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1119        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1120        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1121  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
1122        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1123        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1124  #endif  #endif
1125        }        }
1126      break;      break;
1127    
1128      /* Other special escapes not starting with a digit are straightforward */      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
1129        An error is given if the byte following \c is not an ASCII character. This
1130        coding is ASCII-specific, but then the whole concept of \cx is
1131        ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
1132    
1133      case 'c':      case CHAR_c:
1134      c = *(++ptr);      c = *(++ptr);
1135      if (c == 0)      if (c == 0)
1136        {        {
1137        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
1138        return 0;        break;
1139        }        }
1140    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
1141      /* A letter is upper-cased; then the 0x40 bit is flipped. This coding      if (c > 127)  /* Excludes all non-ASCII in either mode */
1142      is ASCII-specific, but then the whole concept of \cx is ASCII-specific.        {
1143      (However, an EBCDIC equivalent has now been added.) */        *errorcodeptr = ERR68;
1144          break;
1145  #if !EBCDIC    /* ASCII coding */        }
1146      if (c >= 'a' && c <= 'z') c -= 32;      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
1147      c ^= 0x40;      c ^= 0x40;
1148  #else          /* EBCDIC coding */  #else             /* EBCDIC coding */
1149      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
1150      c ^= 0xC0;      c ^= 0xC0;
1151  #endif  #endif
1152      break;      break;
1153    
1154      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
1155      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
1156      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
1157      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
1158      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
1159    
1160      default:      default:
1161      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 541  else Line 1168  else
1168      }      }
1169    }    }
1170    
1171  *ptrptr = ptr;  /* Perl supports \N{name} for character names, as well as plain \N for "not
1172  return c;  newline". PCRE does not support \N{name}. However, it does support
1173  }  quantification such as \N{2,3}. */
1174    
1175    if (escape == ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1176         !is_counted_repeat(ptr+2))
1177      *errorcodeptr = ERR37;
1178    
1179    /* If PCRE_UCP is set, we change the values for \d etc. */
1180    
1181    if ((options & PCRE_UCP) != 0 && escape >= ESC_D && escape <= ESC_w)
1182      escape += (ESC_DU - ESC_D);
1183    
1184    /* Set the pointer to the final character before returning. */
1185    
1186    *ptrptr = ptr;
1187    *chptr = c;
1188    return escape;
1189    }
1190    
1191  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
1192  /*************************************************  /*************************************************
# Line 560  escape sequence. Line 1201  escape sequence.
1201  Argument:  Argument:
1202    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
1203    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
1204      dptr           points to an int that is set to the detailed property value
1205    errorcodeptr   points to the error code variable    errorcodeptr   points to the error code variable
1206    
1207  Returns:     value from ucp_type_table, or -1 for an invalid type  Returns:         type value from ucp_type_table, or -1 for an invalid type
1208  */  */
1209    
1210  static int  static int
1211  get_ucp(const uschar **ptrptr, BOOL *negptr, int *errorcodeptr)  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)
1212  {  {
1213  int c, i, bot, top;  pcre_uchar c;
1214  const uschar *ptr = *ptrptr;  int i, bot, top;
1215  char name[4];  const pcre_uchar *ptr = *ptrptr;
1216    pcre_uchar name[32];
1217    
1218  c = *(++ptr);  c = *(++ptr);
1219  if (c == 0) goto ERROR_RETURN;  if (c == 0) goto ERROR_RETURN;
1220    
1221  *negptr = FALSE;  *negptr = FALSE;
1222    
1223  /* \P or \p can be followed by a one- or two-character name in {}, optionally  /* \P or \p can be followed by a name in {}, optionally preceded by ^ for
1224  preceded by ^ for negation. */  negation. */
1225    
1226  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
1227    {    {
1228    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1229      {      {
1230      *negptr = TRUE;      *negptr = TRUE;
1231      ptr++;      ptr++;
1232      }      }
1233    for (i = 0; i <= 2; i++)    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1234      {      {
1235      c = *(++ptr);      c = *(++ptr);
1236      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
1237      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1238      name[i] = c;      name[i] = c;
1239      }      }
1240    if (c !='}')   /* Try to distinguish error cases */    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
     {  
     while (*(++ptr) != 0 && *ptr != '}');  
     if (*ptr == '}') goto UNKNOWN_RETURN; else goto ERROR_RETURN;  
     }  
1241    name[i] = 0;    name[i] = 0;
1242    }    }
1243    
# Line 615  else Line 1254  else
1254  /* Search for a recognized property name using binary chop */  /* Search for a recognized property name using binary chop */
1255    
1256  bot = 0;  bot = 0;
1257  top = _pcre_utt_size;  top = PRIV(utt_size);
1258    
1259  while (bot < top)  while (bot < top)
1260    {    {
1261    i = (bot + top)/2;    int r;
1262    c = strcmp(name, _pcre_utt[i].name);    i = (bot + top) >> 1;
1263    if (c == 0) return _pcre_utt[i].value;    r = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1264    if (c > 0) bot = i + 1; else top = i;    if (r == 0)
1265        {
1266        *dptr = PRIV(utt)[i].value;
1267        return PRIV(utt)[i].type;
1268        }
1269      if (r > 0) bot = i + 1; else top = i;
1270    }    }
1271    
 UNKNOWN_RETURN:  
1272  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
1273  *ptrptr = ptr;  *ptrptr = ptr;
1274  return -1;  return -1;
# Line 641  return -1; Line 1284  return -1;
1284    
1285    
1286  /*************************************************  /*************************************************
 *            Check for counted repeat            *  
 *************************************************/  
   
 /* This function is called when a '{' is encountered in a place where it might  
 start a quantifier. It looks ahead to see if it really is a quantifier or not.  
 It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}  
 where the ddds are digits.  
   
 Arguments:  
   p         pointer to the first char after '{'  
   
 Returns:    TRUE or FALSE  
 */  
   
 static BOOL  
 is_counted_repeat(const uschar *p)  
 {  
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
 if (*p == '}') return TRUE;  
   
 if (*p++ != ',') return FALSE;  
 if (*p == '}') return TRUE;  
   
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
   
 return (*p == '}');  
 }  
   
   
   
 /*************************************************  
1287  *         Read repeat counts                     *  *         Read repeat counts                     *
1288  *************************************************/  *************************************************/
1289    
# Line 692  Returns:         pointer to '}' on succe Line 1302  Returns:         pointer to '}' on succe
1302                   current ptr on error, with errorcodeptr set non-zero                   current ptr on error, with errorcodeptr set non-zero
1303  */  */
1304    
1305  static const uschar *  static const pcre_uchar *
1306  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)
1307  {  {
1308  int min = 0;  int min = 0;
1309  int max = -1;  int max = -1;
1310    
1311  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  /* Read the minimum value and do a paranoid check: a negative value indicates
1312    an integer overflow. */
1313    
1314    while (IS_DIGIT(*p)) min = min * 10 + (int)(*p++ - CHAR_0);
1315    if (min < 0 || min > 65535)
1316      {
1317      *errorcodeptr = ERR5;
1318      return p;
1319      }
1320    
1321    /* Read the maximum value if there is one, and again do a paranoid on its size.
1322    Also, max must not be less than min. */
1323    
1324  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1325    {    {
1326    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1327      {      {
1328      max = 0;      max = 0;
1329      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while(IS_DIGIT(*p)) max = max * 10 + (int)(*p++ - CHAR_0);
1330        if (max < 0 || max > 65535)
1331          {
1332          *errorcodeptr = ERR5;
1333          return p;
1334          }
1335      if (max < min)      if (max < min)
1336        {        {
1337        *errorcodeptr = ERR4;        *errorcodeptr = ERR4;
# Line 714  if (*p == '}') max = min; else Line 1340  if (*p == '}') max = min; else
1340      }      }
1341    }    }
1342    
1343  /* Do paranoid checks, then fill in the required variables, and pass back the  /* Fill in the required variables, and pass back the pointer to the terminating
1344  pointer to the terminating '}'. */  '}'. */
1345    
1346  if (min > 65535 || max > 65535)  *minp = min;
1347    *errorcodeptr = ERR5;  *maxp = max;
 else  
   {  
   *minp = min;  
   *maxp = max;  
   }  
1348  return p;  return p;
1349  }  }
1350    
1351    
1352    
1353  /*************************************************  /*************************************************
1354  *      Find first significant op code            *  *  Subroutine for finding forward reference      *
1355  *************************************************/  *************************************************/
1356    
1357  /* This is called by several functions that scan a compiled expression looking  /* This recursive function is called only from find_parens() below. The
1358  for a fixed first character, or an anchoring op code etc. It skips over things  top-level call starts at the beginning of the pattern. All other calls must
1359  that do not influence this. For some calls, a change of option is important.  start at a parenthesis. It scans along a pattern's text looking for capturing
1360  For some calls, it makes sense to skip negative forward and all backward  subpatterns, and counting them. If it finds a named pattern that matches the
1361  assertions, and also the \b assertion; for others it does not.  name it is given, it returns its number. Alternatively, if the name is NULL, it
1362    returns when it reaches a given numbered subpattern. Recursion is used to keep
1363    track of subpatterns that reset the capturing group numbers - the (?| feature.
1364    
1365    This function was originally called only from the second pass, in which we know
1366    that if (?< or (?' or (?P< is encountered, the name will be correctly
1367    terminated because that is checked in the first pass. There is now one call to
1368    this function in the first pass, to check for a recursive back reference by
1369    name (so that we can make the whole group atomic). In this case, we need check
1370    only up to the current position in the pattern, and that is still OK because
1371    and previous occurrences will have been checked. To make this work, the test
1372    for "end of pattern" is a check against cd->end_pattern in the main loop,
1373    instead of looking for a binary zero. This means that the special first-pass
1374    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1375    processing items within the loop are OK, because afterwards the main loop will
1376    terminate.)
1377    
1378  Arguments:  Arguments:
1379    code         pointer to the start of the group    ptrptr       address of the current character pointer (updated)
1380    options      pointer to external options    cd           compile background data
1381    optbit       the option bit whose changing is significant, or    name         name to seek, or NULL if seeking a numbered subpattern
1382                   zero if none are    lorn         name length, or subpattern number if name is NULL
1383    skipassert   TRUE if certain assertions are to be skipped    xmode        TRUE if we are in /x mode
1384      utf          TRUE if we are in UTF-8 / UTF-16 / UTF-32 mode
1385      count        pointer to the current capturing subpattern number (updated)
1386    
1387  Returns:       pointer to the first significant opcode  Returns:       the number of the named subpattern, or -1 if not found
1388  */  */
1389    
1390  static const uschar*  static int
1391  first_significant_code(const uschar *code, int *options, int optbit,  find_parens_sub(pcre_uchar **ptrptr, compile_data *cd, const pcre_uchar *name, int lorn,
1392    BOOL skipassert)    BOOL xmode, BOOL utf, int *count)
1393  {  {
1394  for (;;)  pcre_uchar *ptr = *ptrptr;
1395    int start_count = *count;
1396    int hwm_count = start_count;
1397    BOOL dup_parens = FALSE;
1398    
1399    /* If the first character is a parenthesis, check on the type of group we are
1400    dealing with. The very first call may not start with a parenthesis. */
1401    
1402    if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1403    {    {
1404    switch ((int)*code)    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
     {  
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
1405    
1406      case OP_ASSERT_NOT:    if (ptr[1] == CHAR_ASTERISK) ptr += 2;
     case OP_ASSERTBACK:  
     case OP_ASSERTBACK_NOT:  
     if (!skipassert) return code;  
     do code += GET(code, 1); while (*code == OP_ALT);  
     code += _pcre_OP_lengths[*code];  
     break;  
1407    
1408      case OP_WORD_BOUNDARY:    /* Handle a normal, unnamed capturing parenthesis. */
1409      case OP_NOT_WORD_BOUNDARY:  
1410      if (!skipassert) return code;    else if (ptr[1] != CHAR_QUESTION_MARK)
1411      /* Fall through */      {
1412        *count += 1;
1413        if (name == NULL && *count == lorn) return *count;
1414        ptr++;
1415        }
1416    
1417      /* All cases now have (? at the start. Remember when we are in a group
1418      where the parenthesis numbers are duplicated. */
1419    
1420      else if (ptr[2] == CHAR_VERTICAL_LINE)
1421        {
1422        ptr += 3;
1423        dup_parens = TRUE;
1424        }
1425    
1426      /* Handle comments; all characters are allowed until a ket is reached. */
1427    
1428      else if (ptr[2] == CHAR_NUMBER_SIGN)
1429        {
1430        for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1431        goto FAIL_EXIT;
1432        }
1433    
1434      /* Handle a condition. If it is an assertion, just carry on so that it
1435      is processed as normal. If not, skip to the closing parenthesis of the
1436      condition (there can't be any nested parens). */
1437    
1438      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1439        {
1440        ptr += 2;
1441        if (ptr[1] != CHAR_QUESTION_MARK)
1442          {
1443          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1444          if (*ptr != 0) ptr++;
1445          }
1446        }
1447    
1448      /* Start with (? but not a condition. */
1449    
1450      else
1451        {
1452        ptr += 2;
1453        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1454    
1455        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1456    
1457        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1458            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1459          {
1460          pcre_uchar term;
1461          const pcre_uchar *thisname;
1462          *count += 1;
1463          if (name == NULL && *count == lorn) return *count;
1464          term = *ptr++;
1465          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1466          thisname = ptr;
1467          while (*ptr != term) ptr++;
1468          if (name != NULL && lorn == (int)(ptr - thisname) &&
1469              STRNCMP_UC_UC(name, thisname, (unsigned int)lorn) == 0)
1470            return *count;
1471          term++;
1472          }
1473        }
1474      }
1475    
1476    /* Past any initial parenthesis handling, scan for parentheses or vertical
1477    bars. Stop if we get to cd->end_pattern. Note that this is important for the
1478    first-pass call when this value is temporarily adjusted to stop at the current
1479    position. So DO NOT change this to a test for binary zero. */
1480    
1481    for (; ptr < cd->end_pattern; ptr++)
1482      {
1483      /* Skip over backslashed characters and also entire \Q...\E */
1484    
1485      if (*ptr == CHAR_BACKSLASH)
1486        {
1487        if (*(++ptr) == 0) goto FAIL_EXIT;
1488        if (*ptr == CHAR_Q) for (;;)
1489          {
1490          while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1491          if (*ptr == 0) goto FAIL_EXIT;
1492          if (*(++ptr) == CHAR_E) break;
1493          }
1494        continue;
1495        }
1496    
1497      /* Skip over character classes; this logic must be similar to the way they
1498      are handled for real. If the first character is '^', skip it. Also, if the
1499      first few characters (either before or after ^) are \Q\E or \E we skip them
1500      too. This makes for compatibility with Perl. Note the use of STR macros to
1501      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1502    
1503      if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1504        {
1505        BOOL negate_class = FALSE;
1506        for (;;)
1507          {
1508          if (ptr[1] == CHAR_BACKSLASH)
1509            {
1510            if (ptr[2] == CHAR_E)
1511              ptr+= 2;
1512            else if (STRNCMP_UC_C8(ptr + 2,
1513                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1514              ptr += 4;
1515            else
1516              break;
1517            }
1518          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1519            {
1520            negate_class = TRUE;
1521            ptr++;
1522            }
1523          else break;
1524          }
1525    
1526        /* If the next character is ']', it is a data character that must be
1527        skipped, except in JavaScript compatibility mode. */
1528    
1529        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1530            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1531          ptr++;
1532    
1533        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1534          {
1535          if (*ptr == 0) return -1;
1536          if (*ptr == CHAR_BACKSLASH)
1537            {
1538            if (*(++ptr) == 0) goto FAIL_EXIT;
1539            if (*ptr == CHAR_Q) for (;;)
1540              {
1541              while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1542              if (*ptr == 0) goto FAIL_EXIT;
1543              if (*(++ptr) == CHAR_E) break;
1544              }
1545            continue;
1546            }
1547          }
1548        continue;
1549        }
1550    
1551      /* Skip comments in /x mode */
1552    
1553      if (xmode && *ptr == CHAR_NUMBER_SIGN)
1554        {
1555        ptr++;
1556        while (*ptr != 0)
1557          {
1558          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1559          ptr++;
1560    #ifdef SUPPORT_UTF
1561          if (utf) FORWARDCHAR(ptr);
1562    #endif
1563          }
1564        if (*ptr == 0) goto FAIL_EXIT;
1565        continue;
1566        }
1567    
1568      /* Check for the special metacharacters */
1569    
1570      if (*ptr == CHAR_LEFT_PARENTHESIS)
1571        {
1572        int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, count);
1573        if (rc > 0) return rc;
1574        if (*ptr == 0) goto FAIL_EXIT;
1575        }
1576    
1577      else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1578        {
1579        if (dup_parens && *count < hwm_count) *count = hwm_count;
1580        goto FAIL_EXIT;
1581        }
1582    
1583      else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1584        {
1585        if (*count > hwm_count) hwm_count = *count;
1586        *count = start_count;
1587        }
1588      }
1589    
1590    FAIL_EXIT:
1591    *ptrptr = ptr;
1592    return -1;
1593    }
1594    
1595    
1596    
1597    
1598    /*************************************************
1599    *       Find forward referenced subpattern       *
1600    *************************************************/
1601    
1602    /* This function scans along a pattern's text looking for capturing
1603    subpatterns, and counting them. If it finds a named pattern that matches the
1604    name it is given, it returns its number. Alternatively, if the name is NULL, it
1605    returns when it reaches a given numbered subpattern. This is used for forward
1606    references to subpatterns. We used to be able to start this scan from the
1607    current compiling point, using the current count value from cd->bracount, and
1608    do it all in a single loop, but the addition of the possibility of duplicate
1609    subpattern numbers means that we have to scan from the very start, in order to
1610    take account of such duplicates, and to use a recursive function to keep track
1611    of the different types of group.
1612    
1613    Arguments:
1614      cd           compile background data
1615      name         name to seek, or NULL if seeking a numbered subpattern
1616      lorn         name length, or subpattern number if name is NULL
1617      xmode        TRUE if we are in /x mode
1618      utf          TRUE if we are in UTF-8 / UTF-16 / UTF-32 mode
1619    
1620    Returns:       the number of the found subpattern, or -1 if not found
1621    */
1622    
1623    static int
1624    find_parens(compile_data *cd, const pcre_uchar *name, int lorn, BOOL xmode,
1625      BOOL utf)
1626    {
1627    pcre_uchar *ptr = (pcre_uchar *)cd->start_pattern;
1628    int count = 0;
1629    int rc;
1630    
1631    /* If the pattern does not start with an opening parenthesis, the first call
1632    to find_parens_sub() will scan right to the end (if necessary). However, if it
1633    does start with a parenthesis, find_parens_sub() will return when it hits the
1634    matching closing parens. That is why we have to have a loop. */
1635    
1636    for (;;)
1637      {
1638      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, &count);
1639      if (rc > 0 || *ptr++ == 0) break;
1640      }
1641    
1642    return rc;
1643    }
1644    
1645    
1646    
1647    
1648    /*************************************************
1649    *      Find first significant op code            *
1650    *************************************************/
1651    
1652    /* This is called by several functions that scan a compiled expression looking
1653    for a fixed first character, or an anchoring op code etc. It skips over things
1654    that do not influence this. For some calls, it makes sense to skip negative
1655    forward and all backward assertions, and also the \b assertion; for others it
1656    does not.
1657    
1658    Arguments:
1659      code         pointer to the start of the group
1660      skipassert   TRUE if certain assertions are to be skipped
1661    
1662    Returns:       pointer to the first significant opcode
1663    */
1664    
1665    static const pcre_uchar*
1666    first_significant_code(const pcre_uchar *code, BOOL skipassert)
1667    {
1668    for (;;)
1669      {
1670      switch ((int)*code)
1671        {
1672        case OP_ASSERT_NOT:
1673        case OP_ASSERTBACK:
1674        case OP_ASSERTBACK_NOT:
1675        if (!skipassert) return code;
1676        do code += GET(code, 1); while (*code == OP_ALT);
1677        code += PRIV(OP_lengths)[*code];
1678        break;
1679    
1680        case OP_WORD_BOUNDARY:
1681        case OP_NOT_WORD_BOUNDARY:
1682        if (!skipassert) return code;
1683        /* Fall through */
1684    
1685      case OP_CALLOUT:      case OP_CALLOUT:
1686      case OP_CREF:      case OP_CREF:
1687      case OP_BRANUMBER:      case OP_NCREF:
1688      code += _pcre_OP_lengths[*code];      case OP_RREF:
1689        case OP_NRREF:
1690        case OP_DEF:
1691        code += PRIV(OP_lengths)[*code];
1692      break;      break;
1693    
1694      default:      default:
# Line 793  for (;;) Line 1702  for (;;)
1702    
1703    
1704  /*************************************************  /*************************************************
1705  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1706  *************************************************/  *************************************************/
1707    
1708  /* 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,
1709  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.
1710  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
1711    temporarily terminated with OP_END when this function is called.
1712    
1713    This function is called when a backward assertion is encountered, so that if it
1714    fails, the error message can point to the correct place in the pattern.
1715    However, we cannot do this when the assertion contains subroutine calls,
1716    because they can be forward references. We solve this by remembering this case
1717    and doing the check at the end; a flag specifies which mode we are running in.
1718    
1719  Arguments:  Arguments:
1720    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1721    options  the compiling options    utf      TRUE in UTF-8 / UTF-16 / UTF-32 mode
1722      atend    TRUE if called when the pattern is complete
1723  Returns:   the fixed length, or -1 if there is no fixed length,    cd       the "compile data" structure
1724               or -2 if \C was encountered  
1725    Returns:   the fixed length,
1726                 or -1 if there is no fixed length,
1727                 or -2 if \C was encountered (in UTF-8 mode only)
1728                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1729                 or -4 if an unknown opcode was encountered (internal error)
1730  */  */
1731    
1732  static int  static int
1733  find_fixedlength(uschar *code, int options)  find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd)
1734  {  {
1735  int length = -1;  int length = -1;
1736    
1737  register int branchlength = 0;  register int branchlength = 0;
1738  register uschar *cc = code + 1 + LINK_SIZE;  register pcre_uchar *cc = code + 1 + LINK_SIZE;
1739    
1740  /* 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
1741  branch, check the length against that of the other branches. */  branch, check the length against that of the other branches. */
# Line 822  branch, check the length against that of Line 1743  branch, check the length against that of
1743  for (;;)  for (;;)
1744    {    {
1745    int d;    int d;
1746    register int op = *cc;    pcre_uchar *ce, *cs;
1747    if (op >= OP_BRA) op = OP_BRA;    register pcre_uchar op = *cc;
1748    
1749    switch (op)    switch (op)
1750      {      {
1751        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1752        OP_BRA (normal non-capturing bracket) because the other variants of these
1753        opcodes are all concerned with unlimited repeated groups, which of course
1754        are not of fixed length. */
1755    
1756        case OP_CBRA:
1757      case OP_BRA:      case OP_BRA:
1758      case OP_ONCE:      case OP_ONCE:
1759        case OP_ONCE_NC:
1760      case OP_COND:      case OP_COND:
1761      d = find_fixedlength(cc, options);      d = find_fixedlength(cc + ((op == OP_CBRA)? IMM2_SIZE : 0), utf, atend, cd);
1762      if (d < 0) return d;      if (d < 0) return d;
1763      branchlength += d;      branchlength += d;
1764      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1765      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1766      break;      break;
1767    
1768      /* 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.
1769      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
1770      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
1771        the same code. Note that we must not include the OP_KETRxxx opcodes here,
1772        because they all imply an unlimited repeat. */
1773    
1774      case OP_ALT:      case OP_ALT:
1775      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1776      case OP_END:      case OP_END:
1777        case OP_ACCEPT:
1778        case OP_ASSERT_ACCEPT:
1779      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1780        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
1781      if (*cc != OP_ALT) return length;      if (*cc != OP_ALT) return length;
# Line 853  for (;;) Line 1783  for (;;)
1783      branchlength = 0;      branchlength = 0;
1784      break;      break;
1785    
1786        /* A true recursion implies not fixed length, but a subroutine call may
1787        be OK. If the subroutine is a forward reference, we can't deal with
1788        it until the end of the pattern, so return -3. */
1789    
1790        case OP_RECURSE:
1791        if (!atend) return -3;
1792        cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1793        do ce += GET(ce, 1); while (*ce == OP_ALT);           /* End subpattern */
1794        if (cc > cs && cc < ce) return -1;                    /* Recursion */
1795        d = find_fixedlength(cs + IMM2_SIZE, utf, atend, cd);
1796        if (d < 0) return d;
1797        branchlength += d;
1798        cc += 1 + LINK_SIZE;
1799        break;
1800    
1801      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1802    
1803      case OP_ASSERT:      case OP_ASSERT:
# Line 860  for (;;) Line 1805  for (;;)
1805      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1806      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1807      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1808      /* Fall through */      cc += PRIV(OP_lengths)[*cc];
1809        break;
1810    
1811      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1812    
1813      case OP_REVERSE:      case OP_MARK:
1814      case OP_BRANUMBER:      case OP_PRUNE_ARG:
1815      case OP_CREF:      case OP_SKIP_ARG:
1816      case OP_OPT:      case OP_THEN_ARG:
1817        cc += cc[1] + PRIV(OP_lengths)[*cc];
1818        break;
1819    
1820      case OP_CALLOUT:      case OP_CALLOUT:
     case OP_SOD:  
     case OP_SOM:  
     case OP_EOD:  
     case OP_EODN:  
1821      case OP_CIRC:      case OP_CIRC:
1822        case OP_CIRCM:
1823        case OP_CLOSE:
1824        case OP_COMMIT:
1825        case OP_CREF:
1826        case OP_DEF:
1827      case OP_DOLL:      case OP_DOLL:
1828        case OP_DOLLM:
1829        case OP_EOD:
1830        case OP_EODN:
1831        case OP_FAIL:
1832        case OP_NCREF:
1833        case OP_NRREF:
1834      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1835        case OP_PRUNE:
1836        case OP_REVERSE:
1837        case OP_RREF:
1838        case OP_SET_SOM:
1839        case OP_SKIP:
1840        case OP_SOD:
1841        case OP_SOM:
1842        case OP_THEN:
1843      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1844      cc += _pcre_OP_lengths[*cc];      cc += PRIV(OP_lengths)[*cc];
1845      break;      break;
1846    
1847      /* Handle literal characters */      /* Handle literal characters */
1848    
1849      case OP_CHAR:      case OP_CHAR:
1850      case OP_CHARNC:      case OP_CHARI:
1851        case OP_NOT:
1852        case OP_NOTI:
1853      branchlength++;      branchlength++;
1854      cc += 2;      cc += 2;
1855  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
1856      if ((options & PCRE_UTF8) != 0)      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
       {  
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1857  #endif  #endif
1858      break;      break;
1859    
# Line 898  for (;;) Line 1861  for (;;)
1861      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1862    
1863      case OP_EXACT:      case OP_EXACT:
1864      branchlength += GET2(cc,1);      case OP_EXACTI:
1865      cc += 4;      case OP_NOTEXACT:
1866  #ifdef SUPPORT_UTF8      case OP_NOTEXACTI:
1867      if ((options & PCRE_UTF8) != 0)      branchlength += (int)GET2(cc,1);
1868        {      cc += 2 + IMM2_SIZE;
1869        while((*cc & 0x80) == 0x80) cc++;  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
1870        }      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1871  #endif  #endif
1872      break;      break;
1873    
1874      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1875      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1876      cc += 4;      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)
1877          cc += 2;
1878        cc += 1 + IMM2_SIZE + 1;
1879      break;      break;
1880    
1881      /* Handle single-char matchers */      /* Handle single-char matchers */
1882    
1883      case OP_PROP:      case OP_PROP:
1884      case OP_NOTPROP:      case OP_NOTPROP:
1885      cc++;      cc += 2;
1886      /* Fall through */      /* Fall through */
1887    
1888        case OP_HSPACE:
1889        case OP_VSPACE:
1890        case OP_NOT_HSPACE:
1891        case OP_NOT_VSPACE:
1892      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
1893      case OP_DIGIT:      case OP_DIGIT:
1894      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
# Line 927  for (;;) Line 1896  for (;;)
1896      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1897      case OP_WORDCHAR:      case OP_WORDCHAR:
1898      case OP_ANY:      case OP_ANY:
1899        case OP_ALLANY:
1900      branchlength++;      branchlength++;
1901      cc++;      cc++;
1902      break;      break;
1903    
1904      /* The single-byte matcher isn't allowed */      /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1905        otherwise \C is coded as OP_ALLANY. */
1906    
1907      case OP_ANYBYTE:      case OP_ANYBYTE:
1908      return -2;      return -2;
1909    
1910      /* Check a class for variable quantification */      /* Check a class for variable quantification */
1911    
1912  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1913      case OP_XCLASS:      case OP_XCLASS:
1914      cc += GET(cc, 1) - 33;      cc += GET(cc, 1) - PRIV(OP_lengths)[OP_CLASS];
1915      /* Fall through */      /* Fall through */
1916  #endif  #endif
1917    
1918      case OP_CLASS:      case OP_CLASS:
1919      case OP_NCLASS:      case OP_NCLASS:
1920      cc += 33;      cc += PRIV(OP_lengths)[OP_CLASS];
1921    
1922      switch (*cc)      switch (*cc)
1923        {        {
1924          case OP_CRPLUS:
1925          case OP_CRMINPLUS:
1926        case OP_CRSTAR:        case OP_CRSTAR:
1927        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1928        case OP_CRQUERY:        case OP_CRQUERY:
# Line 958  for (;;) Line 1931  for (;;)
1931    
1932        case OP_CRRANGE:        case OP_CRRANGE:
1933        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1934        if (GET2(cc,1) != GET2(cc,3)) return -1;        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1935        branchlength += GET2(cc,1);        branchlength += (int)GET2(cc,1);
1936        cc += 5;        cc += 1 + 2 * IMM2_SIZE;
1937        break;        break;
1938    
1939        default:        default:
# Line 970  for (;;) Line 1943  for (;;)
1943    
1944      /* Anything else is variable length */      /* Anything else is variable length */
1945    
1946      default:      case OP_ANYNL:
1947        case OP_BRAMINZERO:
1948        case OP_BRAPOS:
1949        case OP_BRAPOSZERO:
1950        case OP_BRAZERO:
1951        case OP_CBRAPOS:
1952        case OP_EXTUNI:
1953        case OP_KETRMAX:
1954        case OP_KETRMIN:
1955        case OP_KETRPOS:
1956        case OP_MINPLUS:
1957        case OP_MINPLUSI:
1958        case OP_MINQUERY:
1959        case OP_MINQUERYI:
1960        case OP_MINSTAR:
1961        case OP_MINSTARI:
1962        case OP_MINUPTO:
1963        case OP_MINUPTOI:
1964        case OP_NOTMINPLUS:
1965        case OP_NOTMINPLUSI:
1966        case OP_NOTMINQUERY:
1967        case OP_NOTMINQUERYI:
1968        case OP_NOTMINSTAR:
1969        case OP_NOTMINSTARI:
1970        case OP_NOTMINUPTO:
1971        case OP_NOTMINUPTOI:
1972        case OP_NOTPLUS:
1973        case OP_NOTPLUSI:
1974        case OP_NOTPOSPLUS:
1975        case OP_NOTPOSPLUSI:
1976        case OP_NOTPOSQUERY:
1977        case OP_NOTPOSQUERYI:
1978        case OP_NOTPOSSTAR:
1979        case OP_NOTPOSSTARI:
1980        case OP_NOTPOSUPTO:
1981        case OP_NOTPOSUPTOI:
1982        case OP_NOTQUERY:
1983        case OP_NOTQUERYI:
1984        case OP_NOTSTAR:
1985        case OP_NOTSTARI:
1986        case OP_NOTUPTO:
1987        case OP_NOTUPTOI:
1988        case OP_PLUS:
1989        case OP_PLUSI:
1990        case OP_POSPLUS:
1991        case OP_POSPLUSI:
1992        case OP_POSQUERY:
1993        case OP_POSQUERYI:
1994        case OP_POSSTAR:
1995        case OP_POSSTARI:
1996        case OP_POSUPTO:
1997        case OP_POSUPTOI:
1998        case OP_QUERY:
1999        case OP_QUERYI:
2000        case OP_REF:
2001        case OP_REFI:
2002        case OP_SBRA:
2003        case OP_SBRAPOS:
2004        case OP_SCBRA:
2005        case OP_SCBRAPOS:
2006        case OP_SCOND:
2007        case OP_SKIPZERO:
2008        case OP_STAR:
2009        case OP_STARI:
2010        case OP_TYPEMINPLUS:
2011        case OP_TYPEMINQUERY:
2012        case OP_TYPEMINSTAR:
2013        case OP_TYPEMINUPTO:
2014        case OP_TYPEPLUS:
2015        case OP_TYPEPOSPLUS:
2016        case OP_TYPEPOSQUERY:
2017        case OP_TYPEPOSSTAR:
2018        case OP_TYPEPOSUPTO:
2019        case OP_TYPEQUERY:
2020        case OP_TYPESTAR:
2021        case OP_TYPEUPTO:
2022        case OP_UPTO:
2023        case OP_UPTOI:
2024      return -1;      return -1;
2025    
2026        /* Catch unrecognized opcodes so that when new ones are added they
2027        are not forgotten, as has happened in the past. */
2028    
2029        default:
2030        return -4;
2031      }      }
2032    }    }
2033  /* Control never gets here */  /* Control never gets here */
# Line 981  for (;;) Line 2037  for (;;)
2037    
2038    
2039  /*************************************************  /*************************************************
2040  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
2041  *************************************************/  *************************************************/
2042    
2043  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
2044  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
2045    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
2046    so that it can be called from pcre_study() when finding the minimum matching
2047    length.
2048    
2049  Arguments:  Arguments:
2050    code        points to start of expression    code        points to start of expression
2051    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2052    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
2053    
2054  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
2055  */  */
2056    
2057  static const uschar *  const pcre_uchar *
2058  find_bracket(const uschar *code, BOOL utf8, int number)  PRIV(find_bracket)(const pcre_uchar *code, BOOL utf, int number)
2059  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
2060  for (;;)  for (;;)
2061    {    {
2062    register int c = *code;    register pcre_uchar c = *code;
2063    
2064    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2065    else if (c > OP_BRA)  
2066      /* XCLASS is used for classes that cannot be represented just by a bit
2067      map. This includes negated single high-valued characters. The length in
2068      the table is zero; the actual length is stored in the compiled code. */
2069    
2070      if (c == OP_XCLASS) code += GET(code, 1);
2071    
2072      /* Handle recursion */
2073    
2074      else if (c == OP_REVERSE)
2075        {
2076        if (number < 0) return (pcre_uchar *)code;
2077        code += PRIV(OP_lengths)[c];
2078        }
2079    
2080      /* Handle capturing bracket */
2081    
2082      else if (c == OP_CBRA || c == OP_SCBRA ||
2083               c == OP_CBRAPOS || c == OP_SCBRAPOS)
2084      {      {
2085      int n = c - OP_BRA;      int n = (int)GET2(code, 1+LINK_SIZE);
2086      if (n > EXTRACT_BASIC_MAX) n = GET2(code, 2+LINK_SIZE);      if (n == number) return (pcre_uchar *)code;
2087      if (n == number) return (uschar *)code;      code += PRIV(OP_lengths)[c];
     code += _pcre_OP_lengths[OP_BRA];  
2088      }      }
2089    
2090      /* Otherwise, we can get the item's length from the table, except that for
2091      repeated character types, we have to test for \p and \P, which have an extra
2092      two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2093      must add in its length. */
2094    
2095    else    else
2096      {      {
2097      code += _pcre_OP_lengths[c];      switch(c)
2098          {
2099          case OP_TYPESTAR:
2100          case OP_TYPEMINSTAR:
2101          case OP_TYPEPLUS:
2102          case OP_TYPEMINPLUS:
2103          case OP_TYPEQUERY:
2104          case OP_TYPEMINQUERY:
2105          case OP_TYPEPOSSTAR:
2106          case OP_TYPEPOSPLUS:
2107          case OP_TYPEPOSQUERY:
2108          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2109          break;
2110    
2111  #ifdef SUPPORT_UTF8        case OP_TYPEUPTO:
2112          case OP_TYPEMINUPTO:
2113          case OP_TYPEEXACT:
2114          case OP_TYPEPOSUPTO:
2115          if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2116            code += 2;
2117          break;
2118    
2119      /* In UTF-8 mode, opcodes that are followed by a character may be followed        case OP_MARK:
2120      by a multi-byte character. The length in the table is a minimum, so we have        case OP_PRUNE_ARG:
2121      to scan along to skip the extra bytes. All opcodes are less than 128, so we        case OP_SKIP_ARG:
2122      can use relatively efficient code. */        code += code[1];
2123          break;
2124    
2125          case OP_THEN_ARG:
2126          code += code[1];
2127          break;
2128          }
2129    
2130      if (utf8) switch(c)      /* Add in the fixed length from the table */
2131    
2132        code += PRIV(OP_lengths)[c];
2133    
2134      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
2135      a multi-byte character. The length in the table is a minimum, so we have to
2136      arrange to skip the extra bytes. */
2137    
2138    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2139        if (utf) switch(c)
2140        {        {
2141        case OP_CHAR:        case OP_CHAR:
2142        case OP_CHARNC:        case OP_CHARI:
2143        case OP_EXACT:        case OP_EXACT:
2144          case OP_EXACTI:
2145        case OP_UPTO:        case OP_UPTO:
2146          case OP_UPTOI:
2147        case OP_MINUPTO:        case OP_MINUPTO:
2148          case OP_MINUPTOI:
2149          case OP_POSUPTO:
2150          case OP_POSUPTOI:
2151        case OP_STAR:        case OP_STAR:
2152          case OP_STARI:
2153        case OP_MINSTAR:        case OP_MINSTAR:
2154          case OP_MINSTARI:
2155          case OP_POSSTAR:
2156          case OP_POSSTARI:
2157        case OP_PLUS:        case OP_PLUS:
2158          case OP_PLUSI:
2159        case OP_MINPLUS:        case OP_MINPLUS:
2160          case OP_MINPLUSI:
2161          case OP_POSPLUS:
2162          case OP_POSPLUSI:
2163        case OP_QUERY:        case OP_QUERY:
2164          case OP_QUERYI:
2165        case OP_MINQUERY:        case OP_MINQUERY:
2166        while ((*code & 0xc0) == 0x80) code++;        case OP_MINQUERYI:
2167        break;        case OP_POSQUERY:
2168          case OP_POSQUERYI:
2169        /* XCLASS is used for classes that cannot be represented just by a bit        if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
       map. This includes negated single high-valued characters. The length in  
       the table is zero; the actual length is stored in the compiled code. */  
   
       case OP_XCLASS:  
       code += GET(code, 1) + 1;  
2170        break;        break;
2171        }        }
2172    #else
2173        (void)(utf);  /* Keep compiler happy by referencing function argument */
2174  #endif  #endif
2175      }      }
2176    }    }
# Line 1064  instance of OP_RECURSE. Line 2187  instance of OP_RECURSE.
2187    
2188  Arguments:  Arguments:
2189    code        points to start of expression    code        points to start of expression
2190    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2191    
2192  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
2193  */  */
2194    
2195  static const uschar *  static const pcre_uchar *
2196  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const pcre_uchar *code, BOOL utf)
2197  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
2198  for (;;)  for (;;)
2199    {    {
2200    register int c = *code;    register int c = *code;
2201    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2202    else if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
2203    else if (c > OP_BRA)  
2204      {    /* XCLASS is used for classes that cannot be represented just by a bit
2205      code += _pcre_OP_lengths[OP_BRA];    map. This includes negated single high-valued characters. The length in
2206      }    the table is zero; the actual length is stored in the compiled code. */
2207    
2208      if (c == OP_XCLASS) code += GET(code, 1);
2209    
2210      /* Otherwise, we can get the item's length from the table, except that for
2211      repeated character types, we have to test for \p and \P, which have an extra
2212      two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2213      must add in its length. */
2214    
2215    else    else
2216      {      {
2217      code += _pcre_OP_lengths[c];      switch(c)
2218          {
2219          case OP_TYPESTAR:
2220          case OP_TYPEMINSTAR:
2221          case OP_TYPEPLUS:
2222          case OP_TYPEMINPLUS:
2223          case OP_TYPEQUERY:
2224          case OP_TYPEMINQUERY:
2225          case OP_TYPEPOSSTAR:
2226          case OP_TYPEPOSPLUS:
2227          case OP_TYPEPOSQUERY:
2228          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2229          break;
2230    
2231          case OP_TYPEPOSUPTO:
2232          case OP_TYPEUPTO:
2233          case OP_TYPEMINUPTO:
2234          case OP_TYPEEXACT:
2235          if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2236            code += 2;
2237          break;
2238    
2239          case OP_MARK:
2240          case OP_PRUNE_ARG:
2241          case OP_SKIP_ARG:
2242          code += code[1];
2243          break;
2244    
2245          case OP_THEN_ARG:
2246          code += code[1];
2247          break;
2248          }
2249    
2250  #ifdef SUPPORT_UTF8      /* Add in the fixed length from the table */
2251    
2252        code += PRIV(OP_lengths)[c];
2253    
2254      /* In UTF-8 mode, opcodes that are followed by a character may be followed      /* In UTF-8 mode, opcodes that are followed by a character may be followed
2255      by a multi-byte character. The length in the table is a minimum, so we have      by a multi-byte character. The length in the table is a minimum, so we have
2256      to scan along to skip the extra bytes. All opcodes are less than 128, so we      to arrange to skip the extra bytes. */
     can use relatively efficient code. */  
2257    
2258      if (utf8) switch(c)  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2259        if (utf) switch(c)
2260        {        {
2261        case OP_CHAR:        case OP_CHAR:
2262        case OP_CHARNC:        case OP_CHARI:
2263          case OP_NOT:
2264          case OP_NOTI:
2265        case OP_EXACT:        case OP_EXACT:
2266          case OP_EXACTI:
2267          case OP_NOTEXACT:
2268          case OP_NOTEXACTI:
2269        case OP_UPTO:        case OP_UPTO:
2270          case OP_UPTOI:
2271          case OP_NOTUPTO:
2272          case OP_NOTUPTOI:
2273        case OP_MINUPTO:        case OP_MINUPTO:
2274          case OP_MINUPTOI:
2275          case OP_NOTMINUPTO:
2276          case OP_NOTMINUPTOI:
2277          case OP_POSUPTO:
2278          case OP_POSUPTOI:
2279          case OP_NOTPOSUPTO:
2280          case OP_NOTPOSUPTOI:
2281        case OP_STAR:        case OP_STAR:
2282          case OP_STARI:
2283          case OP_NOTSTAR:
2284          case OP_NOTSTARI:
2285        case OP_MINSTAR:        case OP_MINSTAR:
2286          case OP_MINSTARI:
2287          case OP_NOTMINSTAR:
2288          case OP_NOTMINSTARI:
2289          case OP_POSSTAR:
2290          case OP_POSSTARI:
2291          case OP_NOTPOSSTAR:
2292          case OP_NOTPOSSTARI:
2293        case OP_PLUS:        case OP_PLUS:
2294          case OP_PLUSI:
2295          case OP_NOTPLUS:
2296          case OP_NOTPLUSI:
2297        case OP_MINPLUS:        case OP_MINPLUS:
2298          case OP_MINPLUSI:
2299          case OP_NOTMINPLUS:
2300          case OP_NOTMINPLUSI:
2301          case OP_POSPLUS:
2302          case OP_POSPLUSI:
2303          case OP_NOTPOSPLUS:
2304          case OP_NOTPOSPLUSI:
2305        case OP_QUERY:        case OP_QUERY:
2306          case OP_QUERYI:
2307          case OP_NOTQUERY:
2308          case OP_NOTQUERYI:
2309        case OP_MINQUERY:        case OP_MINQUERY:
2310        while ((*code & 0xc0) == 0x80) code++;        case OP_MINQUERYI:
2311        break;        case OP_NOTMINQUERY:
2312          case OP_NOTMINQUERYI:
2313        /* XCLASS is used for classes that cannot be represented just by a bit        case OP_POSQUERY:
2314        map. This includes negated single high-valued characters. The length in        case OP_POSQUERYI:
2315        the table is zero; the actual length is stored in the compiled code. */        case OP_NOTPOSQUERY:
2316          case OP_NOTPOSQUERYI:
2317        case OP_XCLASS:        if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
       code += GET(code, 1) + 1;  
2318        break;        break;
2319        }        }
2320    #else
2321        (void)(utf);  /* Keep compiler happy by referencing function argument */
2322  #endif  #endif
2323      }      }
2324    }    }
# Line 1132  for (;;) Line 2331  for (;;)
2331  *************************************************/  *************************************************/
2332    
2333  /* This function scans through a branch of a compiled pattern to see whether it  /* This function scans through a branch of a compiled pattern to see whether it
2334  can match the empty string or not. It is called only from could_be_empty()  can match the empty string or not. It is called from could_be_empty()
2335  below. Note that first_significant_code() skips over assertions. If we hit an  below and from compile_branch() when checking for an unlimited repeat of a
2336  unclosed bracket, we return "empty" - this means we've struck an inner bracket  group that can match nothing. Note that first_significant_code() skips over
2337  whose current branch will already have been scanned.  backward and negative forward assertions when its final argument is TRUE. If we
2338    hit an unclosed bracket, we return "empty" - this means we've struck an inner
2339    bracket whose current branch will already have been scanned.
2340    
2341  Arguments:  Arguments:
2342    code        points to start of search    code        points to start of search
2343    endcode     points to where to stop    endcode     points to where to stop
2344    utf8        TRUE if in UTF8 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2345      cd          contains pointers to tables etc.
2346    
2347  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2348  */  */
2349    
2350  static BOOL  static BOOL
2351  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2352      BOOL utf, compile_data *cd)
2353  {  {
2354  register int c;  register int c;
2355  for (code = first_significant_code(code + 1 + LINK_SIZE, NULL, 0, TRUE);  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2356       code < endcode;       code < endcode;
2357       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
2358    {    {
2359    const uschar *ccode;    const pcre_uchar *ccode;
2360    
2361    c = *code;    c = *code;
2362    
2363    if (c >= OP_BRA)    /* Skip over forward assertions; the other assertions are skipped by
2364      first_significant_code() with a TRUE final argument. */
2365    
2366      if (c == OP_ASSERT)
2367      {      {
2368        do code += GET(code, 1); while (*code == OP_ALT);
2369        c = *code;
2370        continue;
2371        }
2372    
2373      /* For a recursion/subroutine call, if its end has been reached, which
2374      implies a backward reference subroutine call, we can scan it. If it's a
2375      forward reference subroutine call, we can't. To detect forward reference
2376      we have to scan up the list that is kept in the workspace. This function is
2377      called only when doing the real compile, not during the pre-compile that
2378      measures the size of the compiled pattern. */
2379    
2380      if (c == OP_RECURSE)
2381        {
2382        const pcre_uchar *scode;
2383      BOOL empty_branch;      BOOL empty_branch;
     if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */  
2384    
2385      /* Scan a closed bracket */      /* Test for forward reference */
2386    
2387        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
2388          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
2389    
2390        /* Not a forward reference, test for completed backward reference */
2391    
2392      empty_branch = FALSE;      empty_branch = FALSE;
2393        scode = cd->start_code + GET(code, 1);
2394        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2395    
2396        /* Completed backwards reference */
2397    
2398      do      do
2399        {        {
2400        if (!empty_branch && could_be_empty_branch(code, endcode, utf8))        if (could_be_empty_branch(scode, endcode, utf, cd))
2401            {
2402          empty_branch = TRUE;          empty_branch = TRUE;
2403            break;
2404            }
2405          scode += GET(scode, 1);
2406          }
2407        while (*scode == OP_ALT);
2408    
2409        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2410        continue;
2411        }
2412    
2413      /* Groups with zero repeats can of course be empty; skip them. */
2414    
2415      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2416          c == OP_BRAPOSZERO)
2417        {
2418        code += PRIV(OP_lengths)[c];
2419        do code += GET(code, 1); while (*code == OP_ALT);
2420        c = *code;
2421        continue;
2422        }
2423    
2424      /* A nested group that is already marked as "could be empty" can just be
2425      skipped. */
2426    
2427      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2428          c == OP_SCBRA || c == OP_SCBRAPOS)
2429        {
2430        do code += GET(code, 1); while (*code == OP_ALT);
2431        c = *code;
2432        continue;
2433        }
2434    
2435      /* For other groups, scan the branches. */
2436    
2437      if (c == OP_BRA  || c == OP_BRAPOS ||
2438          c == OP_CBRA || c == OP_CBRAPOS ||
2439          c == OP_ONCE || c == OP_ONCE_NC ||
2440          c == OP_COND)
2441        {
2442        BOOL empty_branch;
2443        if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
2444    
2445        /* If a conditional group has only one branch, there is a second, implied,
2446        empty branch, so just skip over the conditional, because it could be empty.
2447        Otherwise, scan the individual branches of the group. */
2448    
2449        if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
2450        code += GET(code, 1);        code += GET(code, 1);
2451        else
2452          {
2453          empty_branch = FALSE;
2454          do
2455            {
2456            if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd))
2457              empty_branch = TRUE;
2458            code += GET(code, 1);
2459            }
2460          while (*code == OP_ALT);
2461          if (!empty_branch) return FALSE;   /* All branches are non-empty */
2462        }        }
2463      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
     code += 1 + LINK_SIZE;  
2464      c = *code;      c = *code;
2465        continue;
2466      }      }
2467    
2468    else switch (c)    /* Handle the other opcodes */
2469    
2470      switch (c)
2471      {      {
2472      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
2473        cannot be represented just by a bit map. This includes negated single
2474        high-valued characters. The length in PRIV(OP_lengths)[] is zero; the
2475        actual length is stored in the compiled code, so we must update "code"
2476        here. */
2477    
2478  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2479      case OP_XCLASS:      case OP_XCLASS:
2480      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
2481      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
2482  #endif  #endif
2483    
2484      case OP_CLASS:      case OP_CLASS:
2485      case OP_NCLASS:      case OP_NCLASS:
2486      ccode = code + 33;      ccode = code + PRIV(OP_lengths)[OP_CLASS];
2487    
2488  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2489      CHECK_CLASS_REPEAT:      CHECK_CLASS_REPEAT:
2490  #endif  #endif
2491    
# Line 1227  for (code = first_significant_code(code Line 2521  for (code = first_significant_code(code
2521      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2522      case OP_WORDCHAR:      case OP_WORDCHAR:
2523      case OP_ANY:      case OP_ANY:
2524        case OP_ALLANY:
2525      case OP_ANYBYTE:      case OP_ANYBYTE:
2526      case OP_CHAR:      case OP_CHAR:
2527      case OP_CHARNC:      case OP_CHARI:
2528      case OP_NOT:      case OP_NOT:
2529        case OP_NOTI:
2530      case OP_PLUS:      case OP_PLUS:
2531      case OP_MINPLUS:      case OP_MINPLUS:
2532        case OP_POSPLUS:
2533      case OP_EXACT:      case OP_EXACT:
2534      case OP_NOTPLUS:      case OP_NOTPLUS:
2535      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
2536        case OP_NOTPOSPLUS:
2537      case OP_NOTEXACT:      case OP_NOTEXACT:
2538      case OP_TYPEPLUS:      case OP_TYPEPLUS:
2539      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
2540        case OP_TYPEPOSPLUS:
2541      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2542      return FALSE;      return FALSE;
2543    
2544        /* These are going to continue, as they may be empty, but we have to
2545        fudge the length for the \p and \P cases. */
2546    
2547        case OP_TYPESTAR:
2548        case OP_TYPEMINSTAR:
2549        case OP_TYPEPOSSTAR:
2550        case OP_TYPEQUERY:
2551        case OP_TYPEMINQUERY:
2552        case OP_TYPEPOSQUERY:
2553        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2554        break;
2555    
2556        /* Same for these */
2557    
2558        case OP_TYPEUPTO:
2559        case OP_TYPEMINUPTO:
2560        case OP_TYPEPOSUPTO:
2561        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2562          code += 2;
2563        break;
2564    
2565      /* End of branch */      /* End of branch */
2566    
2567      case OP_KET:      case OP_KET:
2568      case OP_KETRMAX:      case OP_KETRMAX:
2569      case OP_KETRMIN:      case OP_KETRMIN:
2570        case OP_KETRPOS:
2571      case OP_ALT:      case OP_ALT:
2572      return TRUE;      return TRUE;
2573    
2574      /* In UTF-8 mode, STAR, MINSTAR, QUERY, MINQUERY, UPTO, and MINUPTO  may be      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2575      followed by a multibyte character */      MINUPTO, and POSUPTO may be followed by a multibyte character */
2576    
2577  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2578      case OP_STAR:      case OP_STAR:
2579        case OP_STARI:
2580      case OP_MINSTAR:      case OP_MINSTAR:
2581        case OP_MINSTARI:
2582        case OP_POSSTAR:
2583        case OP_POSSTARI:
2584      case OP_QUERY:      case OP_QUERY:
2585        case OP_QUERYI:
2586      case OP_MINQUERY:      case OP_MINQUERY:
2587        case OP_MINQUERYI:
2588        case OP_POSQUERY:
2589        case OP_POSQUERYI:
2590        if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2591        break;
2592    
2593      case OP_UPTO:      case OP_UPTO:
2594        case OP_UPTOI:
2595      case OP_MINUPTO:      case OP_MINUPTO:
2596      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_MINUPTOI:
2597        case OP_POSUPTO:
2598        case OP_POSUPTOI:
2599        if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2600      break;      break;
2601  #endif  #endif
2602    
2603        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2604        string. */
2605    
2606        case OP_MARK:
2607        case OP_PRUNE_ARG:
2608        case OP_SKIP_ARG:
2609        code += code[1];
2610        break;
2611    
2612        case OP_THEN_ARG:
2613        code += code[1];
2614        break;
2615    
2616        /* None of the remaining opcodes are required to match a character. */
2617    
2618        default:
2619        break;
2620      }      }
2621    }    }
2622    
# Line 1279  return TRUE; Line 2633  return TRUE;
2633  the current branch of the current pattern to see if it could match the empty  the current branch of the current pattern to see if it could match the empty
2634  string. If it could, we must look outwards for branches at other levels,  string. If it could, we must look outwards for branches at other levels,
2635  stopping when we pass beyond the bracket which is the subject of the recursion.  stopping when we pass beyond the bracket which is the subject of the recursion.
2636    This function is called only during the real compile, not during the
2637    pre-compile.
2638    
2639  Arguments:  Arguments:
2640    code        points to start of the recursion    code        points to start of the recursion
2641    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2642    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2643    utf8        TRUE if in UTF-8 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2644      cd          pointers to tables etc
2645    
2646  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2647  */  */
2648    
2649  static BOOL  static BOOL
2650  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
2651    BOOL utf8)    branch_chain *bcptr, BOOL utf, compile_data *cd)
2652  {  {
2653  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2654    {    {
2655    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd))
2656        return FALSE;
2657    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2658    }    }
2659  return TRUE;  return TRUE;
# Line 1308  return TRUE; Line 2666  return TRUE;
2666  *************************************************/  *************************************************/
2667    
2668  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2669  encountered in a character class. It checks whether this is followed by an  encountered in a character class. It checks whether this is followed by a
2670  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2671  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2672    
2673    Originally, this function only recognized a sequence of letters between the
2674    terminators, but it seems that Perl recognizes any sequence of characters,
2675    though of course unknown POSIX names are subsequently rejected. Perl gives an
2676    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2677    didn't consider this to be a POSIX class. Likewise for [:1234:].
2678    
2679    The problem in trying to be exactly like Perl is in the handling of escapes. We
2680    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2681    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2682    below handles the special case of \], but does not try to do any other escape
2683    processing. This makes it different from Perl for cases such as [:l\ower:]
2684    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2685    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2686    I think.
2687    
2688    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2689    It seems that the appearance of a nested POSIX class supersedes an apparent
2690    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2691    a digit.
2692    
2693    In Perl, unescaped square brackets may also appear as part of class names. For
2694    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2695    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2696    seem right at all. PCRE does not allow closing square brackets in POSIX class
2697    names.
2698    
2699  Argument:  Arguments:
2700    ptr      pointer to the initial [    ptr      pointer to the initial [
2701    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2702    
2703  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2704  */  */
2705    
2706  static BOOL  static BOOL
2707  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
2708  {  {
2709  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2710  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2711  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2712    {    {
2713    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2714    return TRUE;      ptr++;
2715      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2716      else
2717        {
2718        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2719          {
2720          *endptr = ptr;
2721          return TRUE;
2722          }
2723        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2724             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2725              ptr[1] == CHAR_EQUALS_SIGN) &&
2726            check_posix_syntax(ptr, endptr))
2727          return FALSE;
2728        }
2729    }    }
2730  return FALSE;  return FALSE;
2731  }  }
# Line 1353  Returns:     a value representing the na Line 2748  Returns:     a value representing the na
2748  */  */
2749    
2750  static int  static int
2751  check_posix_name(const uschar *ptr, int len)  check_posix_name(const pcre_uchar *ptr, int len)
2752  {  {
2753    const char *pn = posix_names;
2754  register int yield = 0;  register int yield = 0;
2755  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2756    {    {
2757    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2758      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      STRNCMP_UC_C8(ptr, pn, len) == 0) return yield;
2759      pn += posix_name_lengths[yield] + 1;
2760    yield++;    yield++;
2761    }    }
2762  return -1;  return -1;
# Line 1374  return -1; Line 2771  return -1;
2771  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2772  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2773  earlier groups that are outside the current group). However, when a group is  earlier groups that are outside the current group). However, when a group is
2774  optional (i.e. the minimum quantifier is zero), OP_BRAZERO is inserted before  optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is
2775  it, after it has been compiled. This means that any OP_RECURSE items within it  inserted before it, after it has been compiled. This means that any OP_RECURSE
2776  that refer to the group itself or any contained groups have to have their  items within it that refer to the group itself or any contained groups have to
2777  offsets adjusted. That is the job of this function. Before it is called, the  have their offsets adjusted. That one of the jobs of this function. Before it
2778  partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2779    OP_END.
2780    
2781    This function has been extended with the possibility of forward references for
2782    recursions and subroutine calls. It must also check the list of such references
2783    for the group we are dealing with. If it finds that one of the recursions in
2784    the current group is on this list, it adjusts the offset in the list, not the
2785    value in the reference (which is a group number).
2786    
2787  Arguments:  Arguments:
2788    group      points to the start of the group    group      points to the start of the group
2789    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
2790    utf8       TRUE in UTF-8 mode    utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode
2791    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
2792      save_hwm   the hwm forward reference pointer at the start of the group
2793    
2794  Returns:     nothing  Returns:     nothing
2795  */  */
2796    
2797  static void  static void
2798  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd)  adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
2799      pcre_uchar *save_hwm)
2800  {  {
2801  uschar *ptr = group;  pcre_uchar *ptr = group;
2802  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  
2803    while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
2804    {    {
2805    int offset = GET(ptr, 1);    int offset;
2806    if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);    pcre_uchar *hc;
2807    
2808      /* See if this recursion is on the forward reference list. If so, adjust the
2809      reference. */
2810    
2811      for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
2812        {
2813        offset = GET(hc, 0);
2814        if (cd->start_code + offset == ptr + 1)
2815          {
2816          PUT(hc, 0, offset + adjust);
2817          break;
2818          }
2819        }
2820    
2821      /* Otherwise, adjust the recursion offset if it's after the start of this
2822      group. */
2823    
2824      if (hc >= cd->hwm)
2825        {
2826        offset = GET(ptr, 1);
2827        if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
2828        }
2829    
2830    ptr += 1 + LINK_SIZE;    ptr += 1 + LINK_SIZE;
2831    }    }
2832  }  }
# Line 1418  Arguments: Line 2848  Arguments:
2848  Returns:         new code pointer  Returns:         new code pointer
2849  */  */
2850    
2851  static uschar *  static pcre_uchar *
2852  auto_callout(uschar *code, const uschar *ptr, compile_data *cd)  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
2853  {  {
2854  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2855  *code++ = 255;  *code++ = 255;
2856  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2857  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2858  return code + 2*LINK_SIZE;  return code + 2 * LINK_SIZE;
2859  }  }
2860    
2861    
# Line 1447  Returns:             nothing Line 2877  Returns:             nothing
2877  */  */
2878    
2879  static void  static void
2880  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
2881  {  {
2882  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2883  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2884  }  }
2885    
# Line 1461  PUT(previous_callout, 2 + LINK_SIZE, len Line 2891  PUT(previous_callout, 2 + LINK_SIZE, len
2891  *************************************************/  *************************************************/
2892    
2893  /* This function is passed the start and end of a class range, in UTF-8 mode  /* This function is passed the start and end of a class range, in UTF-8 mode
2894  with UCP support. It searches up the characters, looking for internal ranges of  with UCP support. It searches up the characters, looking for ranges of
2895  characters in the "other" case. Each call returns the next one, updating the  characters in the "other" case. Each call returns the next one, updating the
2896  start address.  start address. A character with multiple other cases is returned on its own
2897    with a special return value.
2898    
2899  Arguments:  Arguments:
2900    cptr        points to starting character value; updated    cptr        points to starting character value; updated
# Line 1471  Arguments: Line 2902  Arguments:
2902    ocptr       where to put start of othercase range    ocptr       where to put start of othercase range
2903    odptr       where to put end of othercase range    odptr       where to put end of othercase range
2904    
2905  Yield:        TRUE when range returned; FALSE when no more  Yield:        -1 when no more
2906                   0 when a range is returned
2907                  >0 the CASESET offset for char with multiple other cases
2908                    in this case, ocptr contains the original
2909  */  */
2910    
2911  static BOOL  static int
2912  get_othercase_range(int *cptr, int d, int *ocptr, int *odptr)  get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,
2913      pcre_uint32 *odptr)
2914  {  {
2915  int c, chartype, othercase, next;  pcre_uint32 c, othercase, next;
2916    int co;
2917    
2918    /* Find the first character that has an other case. If it has multiple other
2919    cases, return its case offset value. */
2920    
2921  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2922    {    {
2923    if (_pcre_ucp_findchar(c, &chartype, &othercase) == ucp_L && othercase != 0)    if ((co = UCD_CASESET(c)) != 0)
2924      break;      {
2925        *ocptr = c++;   /* Character that has the set */
2926        *cptr = c;      /* Rest of input range */
2927        return co;
2928        }
2929      if ((othercase = UCD_OTHERCASE(c)) != c) break;
2930    }    }
2931    
2932  if (c > d) return FALSE;  if (c > d) return -1;  /* Reached end of range */
2933    
2934  *ocptr = othercase;  *ocptr = othercase;
2935  next = othercase + 1;  next = othercase + 1;
2936    
2937  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2938    {    {
2939    if (_pcre_ucp_findchar(c, &chartype, &othercase) != ucp_L ||    if (UCD_OTHERCASE(c) != next) break;
         othercase != next)  
     break;  
2940    next++;    next++;
2941    }    }
2942    
2943  *odptr = next - 1;  *odptr = next - 1;     /* End of othercase range */
2944  *cptr = c;  *cptr = c;             /* Rest of input range */
2945    return 0;
 return TRUE;  
2946  }  }
2947  #endif  /* SUPPORT_UCP */  
2948    
2949    
2950  /*************************************************  /*************************************************
2951  *           Compile one branch                   *  *        Check a character and a property        *
2952  *************************************************/  *************************************************/
2953    
2954  /* Scan the pattern, compiling it into the code vector. If the options are  /* This function is called by check_auto_possessive() when a property item
2955  changed during the branch, the pointer is used to change the external options  is adjacent to a fixed character.
 bits.  
2956    
2957  Arguments:  Arguments:
2958    optionsptr     pointer to the option bits    c            the character
2959    brackets       points to number of extracting brackets used    ptype        the property type
2960    codeptr        points to the pointer to the current code point    pdata        the data for the type
2961    ptrptr         points to the current pattern pointer    negated      TRUE if it's a negated property (\P or \p{^)
   errorcodeptr   points to error code variable  
   firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)  
   reqbyteptr     set to the last literal character required, else < 0  
   bcptr          points to current branch chain  
   cd             contains pointers to tables etc.  
2962    
2963  Returns:         TRUE on success  Returns:       TRUE if auto-possessifying is OK
                  FALSE, with *errorcodeptr set non-zero on error  
2964  */  */
2965    
2966  static BOOL  static BOOL
2967  compile_branch(int *optionsptr, int *brackets, uschar **codeptr,  check_char_prop(pcre_uint32 c, int ptype, int pdata, BOOL negated)
   const uschar **ptrptr, int *errorcodeptr, int *firstbyteptr,  
   int *reqbyteptr, branch_chain *bcptr, compile_data *cd)  
2968  {  {
2969  int repeat_type, op_type;  #ifdef SUPPORT_UCP
2970  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  const pcre_uint32 *p;
 int bravalue = 0;  
 int greedy_default, greedy_non_default;  
 int firstbyte, reqbyte;  
 int zeroreqbyte, zerofirstbyte;  
 int req_caseopt, reqvary, tempreqvary;  
 int condcount = 0;  
 int options = *optionsptr;  
 int after_manual_callout = 0;  
 register int c;  
 register uschar *code = *codeptr;  
 uschar *tempcode;  
 BOOL inescq = FALSE;  
 BOOL groupsetfirstbyte = FALSE;  
 const uschar *ptr = *ptrptr;  
 const uschar *tempptr;  
 uschar *previous = NULL;  
 uschar *previous_callout = NULL;  
 uschar classbits[32];  
   
 #ifdef SUPPORT_UTF8  
 BOOL class_utf8;  
 BOOL utf8 = (options & PCRE_UTF8) != 0;  
 uschar *class_utf8data;  
 uschar utf8_char[6];  
 #else  
 BOOL utf8 = FALSE;  
2971  #endif  #endif
2972    
2973  /* Set up the default and non-default settings for greediness */  const ucd_record *prop = GET_UCD(c);
   
 greedy_default = ((options & PCRE_UNGREEDY) != 0);  
 greedy_non_default = greedy_default ^ 1;  
2974    
2975  /* Initialize no first byte, no required byte. REQ_UNSET means "no char  switch(ptype)
2976  matching encountered yet". It gets changed to REQ_NONE if we hit something that    {
2977  matches a non-fixed char first char; reqbyte just remains unset if we never    case PT_LAMP:
2978  find one.    return (prop->chartype == ucp_Lu ||
2979              prop->chartype == ucp_Ll ||
2980              prop->chartype == ucp_Lt) == negated;
2981    
2982      case PT_GC:
2983      return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2984    
2985      case PT_PC:
2986      return (pdata == prop->chartype) == negated;
2987    
2988      case PT_SC:
2989      return (pdata == prop->script) == negated;
2990    
2991      /* These are specials */
2992    
2993      case PT_ALNUM:
2994      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2995              PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2996    
2997      case PT_SPACE:    /* Perl space */
2998      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2999              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
3000              == negated;
3001    
3002      case PT_PXSPACE:  /* POSIX space */
3003      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
3004              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
3005              c == CHAR_FF || c == CHAR_CR)
3006              == negated;
3007    
3008      case PT_WORD:
3009      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
3010              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
3011              c == CHAR_UNDERSCORE) == negated;
3012    
3013    #ifdef SUPPORT_UCP
3014      case PT_CLIST:
3015      p = PRIV(ucd_caseless_sets) + prop->caseset;
3016      for (;;)
3017        {
3018        if ((unsigned int)c < *p) return !negated;
3019        if ((unsigned int)c == *p++) return negated;
3020        }
3021      break;  /* Control never reaches here */
3022    #endif
3023      }
3024    
3025  When we hit a repeat whose minimum is zero, we may have to adjust these values  return FALSE;
3026  to take the zero repeat into account. This is implemented by setting them to  }
3027  zerofirstbyte and zeroreqbyte when such a repeat is encountered. The individual  #endif  /* SUPPORT_UCP */
 item types that can be repeated set these backoff variables appropriately. */  
3028    
 firstbyte = reqbyte = zerofirstbyte = zeroreqbyte = REQ_UNSET;  
3029    
 /* The variable req_caseopt contains either the REQ_CASELESS value or zero,  
 according to the current setting of the caseless flag. REQ_CASELESS is a bit  
 value > 255. It is added into the firstbyte or reqbyte variables to record the  
 case status of the value. This is used only for ASCII characters. */  
3030    
3031  req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;  /*************************************************
3032    *     Check if auto-possessifying is possible    *
3033    *************************************************/
3034    
3035  /* Switch on next character until the end of the branch */  /* This function is called for unlimited repeats of certain items, to see
3036    whether the next thing could possibly match the repeated item. If not, it makes
3037    sense to automatically possessify the repeated item.
3038    
3039  for (;; ptr++)  Arguments:
3040    {    previous      pointer to the repeated opcode
3041    BOOL negate_class;    utf           TRUE in UTF-8 / UTF-16 / UTF-32 mode
3042    BOOL possessive_quantifier;    ptr           next character in pattern
3043    BOOL is_quantifier;    options       options bits
3044    int class_charcount;    cd            contains pointers to tables etc.
   int class_lastchar;  
   int newoptions;  
   int recno;  
   int skipbytes;  
   int subreqbyte;  
   int subfirstbyte;  
   int mclength;  
   uschar mcbuffer[8];  
3045    
3046    /* Next byte in the pattern */  Returns:        TRUE if possessifying is wanted
3047    */
3048    
3049    c = *ptr;  static BOOL
3050    check_auto_possessive(const pcre_uchar *previous, BOOL utf,
3051      const pcre_uchar *ptr, int options, compile_data *cd)
3052    {
3053    pcre_uint32 c = NOTACHAR;
3054    pcre_uint32 next;
3055    int escape;
3056    int op_code = *previous++;
3057    
3058    /* If in \Q...\E, check for the end; if not, we have a literal */  /* Skip whitespace and comments in extended mode */
3059    
3060    if (inescq && c != 0)  if ((options & PCRE_EXTENDED) != 0)
3061      {
3062      for (;;)
3063      {      {
3064      if (c == '\\' && ptr[1] == 'E')      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
3065        if (*ptr == CHAR_NUMBER_SIGN)
3066        {        {
       inescq = FALSE;  
3067        ptr++;        ptr++;
3068        continue;        while (*ptr != 0)
       }  
     else  
       {  
       if (previous_callout != NULL)  
         {  
         complete_callout(previous_callout, ptr, cd);  
         previous_callout = NULL;  
         }  
       if ((options & PCRE_AUTO_CALLOUT) != 0)  
3069          {          {
3070          previous_callout = code;          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
3071          code = auto_callout(code, ptr, cd);          ptr++;
3072    #ifdef SUPPORT_UTF
3073            if (utf) FORWARDCHAR(ptr);
3074    #endif
3075          }          }
       goto NORMAL_CHAR;  
3076        }        }
3077        else break;
3078      }      }
3079      }
3080    
3081    /* Fill in length of a previous callout, except when the next thing is  /* If the next item is one that we can handle, get its value. A non-negative
3082    a quantifier. */  value is a character, a negative value is an escape value. */
   
   is_quantifier = c == '*' || c == '+' || c == '?' ||  
     (c == '{' && is_counted_repeat(ptr+1));  
3083    
3084    if (!is_quantifier && previous_callout != NULL &&  if (*ptr == CHAR_BACKSLASH)
3085         after_manual_callout-- <= 0)    {
3086      {    int temperrorcode = 0;
3087      complete_callout(previous_callout, ptr, cd);    escape = check_escape(&ptr, &next, &temperrorcode, cd->bracount, options, FALSE);
3088      previous_callout = NULL;    if (temperrorcode != 0) return FALSE;
3089      }    ptr++;    /* Point after the escape sequence */
3090      }
3091    else if (!MAX_255(*ptr) || (cd->ctypes[*ptr] & ctype_meta) == 0)
3092      {
3093      escape = 0;
3094    #ifdef SUPPORT_UTF
3095      if (utf) { GETCHARINC(next, ptr); } else
3096    #endif
3097      next = *ptr++;
3098      }
3099    else return FALSE;
3100    
3101    /* In extended mode, skip white space and comments */  /* Skip whitespace and comments in extended mode */
3102    
3103    if ((options & PCRE_EXTENDED) != 0)  if ((options & PCRE_EXTENDED) != 0)
3104      {
3105      for (;;)
3106      {      {
3107      if ((cd->ctypes[c] & ctype_space) != 0) continue;      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
3108      if (c == '#')      if (*ptr == CHAR_NUMBER_SIGN)
3109        {        {
3110        /* The space before the ; is to avoid a warning on a silly compiler        ptr++;
3111        on the Macintosh. */        while (*ptr != 0)
3112        while ((c = *(++ptr)) != 0 && c != NEWLINE) ;          {
3113        if (c != 0) continue;   /* Else fall through to handle end of string */          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
3114            ptr++;
3115    #ifdef SUPPORT_UTF
3116            if (utf) FORWARDCHAR(ptr);
3117    #endif
3118            }
3119        }        }
3120        else break;
3121      }      }
3122      }
3123    
3124    /* No auto callout for quantifiers. */  /* If the next thing is itself optional, we have to give up. */
3125    
3126    if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3127      STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3128        return FALSE;
3129    
3130    /* If the previous item is a character, get its value. */
3131    
3132    if (op_code == OP_CHAR || op_code == OP_CHARI ||
3133        op_code == OP_NOT || op_code == OP_NOTI)
3134      //if (escape == 0) switch(op_code)
3135      {
3136    #ifdef SUPPORT_UTF
3137      GETCHARTEST(c, previous);
3138    #else
3139      c = *previous;
3140    #endif
3141      }
3142    
3143    /* Now compare the next item with the previous opcode. First, handle cases when
3144    the next item is a character. */
3145    
3146    if (escape == 0)
3147      {
3148      /* For a caseless UTF match, the next character may have more than one other
3149      case, which maps to the special PT_CLIST property. Check this first. */
3150    
3151    #ifdef SUPPORT_UCP
3152      if (utf && c != NOTACHAR && (options & PCRE_CASELESS) != 0)
3153      {      {
3154      previous_callout = code;      int ocs = UCD_CASESET(next);
3155      code = auto_callout(code, ptr, cd);      if (ocs > 0) return check_char_prop(c, PT_CLIST, ocs, op_code >= OP_NOT);
3156      }      }
3157    #endif
3158    
3159    switch(c)    switch(op_code)
3160      {      {
3161      /* The branch terminates at end of string, |, or ). */      case OP_CHAR:
3162        return c != next;
     case 0:  
     case '|':  
     case ')':  
     *firstbyteptr = firstbyte;  
     *reqbyteptr = reqbyte;  
     *codeptr = code;  
     *ptrptr = ptr;  
     return TRUE;  
   
     /* Handle single-character metacharacters. In multiline mode, ^ disables  
     the setting of any following char as a first character. */  
3163    
3164      case '^':      /* For CHARI (caseless character) we must check the other case. If we have
3165      if ((options & PCRE_MULTILINE) != 0)      Unicode property support, we can use it to test the other case of
3166        high-valued characters. We know that next can have only one other case,
3167        because multi-other-case characters are dealt with above. */
3168    
3169        case OP_CHARI:
3170        if (c == next) return FALSE;
3171    #ifdef SUPPORT_UTF
3172        if (utf)
3173        {        {
3174        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        pcre_uint32 othercase;
3175          if (next < 128) othercase = cd->fcc[next]; else
3176    #ifdef SUPPORT_UCP
3177          othercase = UCD_OTHERCASE(next);
3178    #else
3179          othercase = NOTACHAR;
3180    #endif
3181          return c != othercase;
3182        }        }
3183      previous = NULL;      else
3184      *code++ = OP_CIRC;  #endif  /* SUPPORT_UTF */
3185      break;      return (c != TABLE_GET(next, cd->fcc, next));  /* Not UTF */
3186    
3187        case OP_NOT:
3188        return c == next;
3189    
3190        case OP_NOTI:
3191        if (c == next) return TRUE;
3192    #ifdef SUPPORT_UTF
3193        if (utf)
3194          {
3195          pcre_uint32 othercase;
3196          if (next < 128) othercase = cd->fcc[next]; else
3197    #ifdef SUPPORT_UCP
3198          othercase = UCD_OTHERCASE(next);
3199    #else
3200          othercase = NOTACHAR;
3201    #endif
3202          return c == othercase;
3203          }
3204        else
3205    #endif  /* SUPPORT_UTF */
3206        return (c == TABLE_GET(next, cd->fcc, next));  /* Not UTF */
3207    
3208      case '$':      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
3209      previous = NULL;      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
     *code++ = OP_DOLL;  
     break;  
3210    
3211      /* There can never be a first char if '.' is first, whatever happens about      case OP_DIGIT:
3212      repeats. The value of reqbyte doesn't change either. */      return next > 255 || (cd->ctypes[next] & ctype_digit) == 0;
3213    
3214      case '.':      case OP_NOT_DIGIT:
3215      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      return next <= 255 && (cd->ctypes[next] & ctype_digit) != 0;
     zerofirstbyte = firstbyte;  
     zeroreqbyte = reqbyte;  
     previous = code;  
     *code++ = OP_ANY;  
     break;  
3216    
3217      /* Character classes. If the included characters are all < 255 in value, we      case OP_WHITESPACE:
3218      build a 32-byte bitmap of the permitted characters, except in the special      return next > 255 || (cd->ctypes[next] & ctype_space) == 0;
     case where there is only one such character. For negated classes, we build  
     the map as usual, then invert it at the end. However, we use a different  
     opcode so that data characters > 255 can be handled correctly.  
3219    
3220      If the class contains characters outside the 0-255 range, a different      case OP_NOT_WHITESPACE:
3221      opcode is compiled. It may optionally have a bit map for characters < 256,      return next <= 255 && (cd->ctypes[next] & ctype_space) != 0;
     but those above are are explicitly listed afterwards. A flag byte tells  
     whether the bitmap is present, and whether this is a negated class or not.  
     */  
3222    
3223      case '[':      case OP_WORDCHAR:
3224      previous = code;      return next > 255 || (cd->ctypes[next] & ctype_word) == 0;
3225    
3226      /* PCRE supports POSIX class stuff inside a class. Perl gives an error if      case OP_NOT_WORDCHAR:
3227      they are encountered at the top level, so we'll do that too. */      return next <= 255 && (cd->ctypes[next] & ctype_word) != 0;
3228    
3229      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      case OP_HSPACE:
3230          check_posix_syntax(ptr, &tempptr, cd))      case OP_NOT_HSPACE:
3231        switch(next)
3232        {        {
3233        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        HSPACE_CASES:
3234        goto FAILED;        return op_code == OP_NOT_HSPACE;
       }  
   
     /* If the first character is '^', set the negation flag and skip it. */  
3235    
3236      if ((c = *(++ptr)) == '^')        default:
3237        {        return op_code != OP_NOT_HSPACE;
       negate_class = TRUE;  
       c = *(++ptr);  
3238        }        }
3239      else  
3240        case OP_ANYNL:
3241        case OP_VSPACE:
3242        case OP_NOT_VSPACE:
3243        switch(next)
3244        {        {
3245        negate_class = FALSE;        VSPACE_CASES:
3246        }        return op_code == OP_NOT_VSPACE;
3247    
3248      /* Keep a count of chars with values < 256 so that we can optimize the case        default:
3249      of just a single character (as long as it's < 256). For higher valued UTF-8        return op_code != OP_NOT_VSPACE;
3250      characters, we don't yet do any optimization. */        }
3251    
3252      class_charcount = 0;  #ifdef SUPPORT_UCP
3253      class_lastchar = -1;      case OP_PROP:
3254        return check_char_prop(next, previous[0], previous[1], FALSE);
3255    
3256  #ifdef SUPPORT_UTF8      case OP_NOTPROP:
3257      class_utf8 = FALSE;                       /* No chars >= 256 */      return check_char_prop(next, previous[0], previous[1], TRUE);
     class_utf8data = code + LINK_SIZE + 34;   /* For UTF-8 items */  
3258  #endif  #endif
3259    
3260      /* Initialize the 32-char bit map to all zeros. We have to build the      default:
3261      map in a temporary bit of store, in case the class contains only 1      return FALSE;
3262      character (< 256), because in that case the compiled code doesn't use the      }
3263      bit map. */    }
   
     memset(classbits, 0, 32 * sizeof(uschar));  
   
     /* Process characters until ] is reached. By writing this as a "do" it  
     means that an initial ] is taken as a data character. The first pass  
     through the regex checked the overall syntax, so we don't need to be very  
     strict here. At the start of the loop, c contains the first byte of the  
     character. */  
3264    
3265      do  /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP
3266        {  is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
3267  #ifdef SUPPORT_UTF8  generated only when PCRE_UCP is *not* set, that is, when only ASCII
3268        if (utf8 && c > 127)  characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
3269          {                           /* Braces are required because the */  replaced by OP_PROP codes when PCRE_UCP is set. */
         GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */  
         }  
 #endif  
3270    
3271        /* Inside \Q...\E everything is literal except \E */  switch(op_code)
3272      {
3273      case OP_CHAR:
3274      case OP_CHARI:
3275      switch(escape)
3276        {
3277        case ESC_d:
3278        return c > 255 || (cd->ctypes[c] & ctype_digit) == 0;
3279    
3280        if (inescq)      case ESC_D:
3281          {      return c <= 255 && (cd->ctypes[c] & ctype_digit) != 0;
         if (c == '\\' && ptr[1] == 'E')  
           {  
           inescq = FALSE;  
           ptr++;  
           continue;  
           }  
         else goto LONE_SINGLE_CHARACTER;  
         }  
3282    
3283        /* Handle POSIX class names. Perl allows a negation extension of the      case ESC_s:
3284        form [:^name:]. A square bracket that doesn't match the syntax is      return c > 255 || (cd->ctypes[c] & ctype_space) == 0;
       treated as a literal. We also recognize the POSIX constructions  
       [.ch.] and [=ch=] ("collating elements") and fault them, as Perl  
       5.6 and 5.8 do. */  
3285    
3286        if (c == '[' &&      case ESC_S:
3287            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      return c <= 255 && (cd->ctypes[c] & ctype_space) != 0;
           check_posix_syntax(ptr, &tempptr, cd))  
         {  
         BOOL local_negate = FALSE;  
         int posix_class, i;  
         register const uschar *cbits = cd->cbits;  
3288    
3289          if (ptr[1] != ':')      case ESC_w:
3290            {      return c > 255 || (cd->ctypes[c] & ctype_word) == 0;
           *errorcodeptr = ERR31;  
           goto FAILED;  
           }  
3291    
3292          ptr += 2;      case ESC_W:
3293          if (*ptr == '^')      return c <= 255 && (cd->ctypes[c] & ctype_word) != 0;
           {  
           local_negate = TRUE;  
           ptr++;  
           }  
3294    
3295          posix_class = check_posix_name(ptr, tempptr - ptr);      case ESC_h:
3296          if (posix_class < 0)      case ESC_H:
3297            {      switch(c)
3298            *errorcodeptr = ERR30;        {
3299            goto FAILED;        HSPACE_CASES:
3300            }        return escape != ESC_h;
3301    
3302          default:
3303          return escape == ESC_h;
3304          }
3305    
3306          /* If matching is caseless, upper and lower are converted to      case ESC_v:
3307          alpha. This relies on the fact that the class table starts with      case ESC_V:
3308          alpha, lower, upper as the first 3 entries. */      switch(c)
3309          {
3310          VSPACE_CASES:
3311          return escape != ESC_v;
3312    
3313          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)        default:
3314            posix_class = 0;        return escape == ESC_v;
3315          }
3316    
3317          /* Or into the map we are building up to 3 of the static class      /* When PCRE_UCP is set, these values get generated for \d etc. Find
3318          tables, or their negations. The [:blank:] class sets up the same      their substitutions and process them. The result will always be either
3319          chars as the [:space:] class (all white space). We remove the vertical      ESC_p or ESC_P. Then fall through to process those values. */
         white space chars afterwards. */  
3320    
3321          posix_class *= 3;  #ifdef SUPPORT_UCP
3322          for (i = 0; i < 3; i++)      case ESC_du:
3323            {      case ESC_DU:
3324            BOOL blankclass = strncmp((char *)ptr, "blank", 5) == 0;      case ESC_wu:
3325            int taboffset = posix_class_maps[posix_class + i];      case ESC_WU:
3326            if (taboffset < 0) break;      case ESC_su:
3327            if (local_negate)      case ESC_SU:
3328              {        {
3329              if (i == 0)        int temperrorcode = 0;
3330                for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+taboffset];        ptr = substitutes[escape - ESC_DU];
3331              else        escape = check_escape(&ptr, &next, &temperrorcode, 0, options, FALSE);
3332                for (c = 0; c < 32; c++) classbits[c] &= ~cbits[c+taboffset];        if (temperrorcode != 0) return FALSE;
3333              if (blankclass) classbits[1] |= 0x3c;        ptr++;    /* For compatibility */
3334              }        }
3335            else      /* Fall through */
             {  
             for (c = 0; c < 32; c++) classbits[c] |= cbits[c+taboffset];  
             if (blankclass) classbits[1] &= ~0x3c;  
             }  
           }  
3336    
3337          ptr = tempptr + 1;      case ESC_p:
3338          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */      case ESC_P:
3339          continue;    /* End of POSIX syntax handling */        {
3340          }        int ptype, pdata, errorcodeptr;
3341          BOOL negated;
3342    
3343        /* Backslash may introduce a single character, or it may introduce one        ptr--;      /* Make ptr point at the p or P */
3344        of the specials, which just set a flag. Escaped items are checked for        ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
3345        validity in the pre-compiling pass. The sequence \b is a special case.        if (ptype < 0) return FALSE;
3346        Inside a class (and only there) it is treated as backspace. Elsewhere        ptr++;      /* Point past the final curly ket */
3347        it marks a word boundary. Other escapes have preset maps ready to  
3348        or into the one we are building. We assume they have more than one        /* If the property item is optional, we have to give up. (When generated
3349        character in them, so set class_charcount bigger than one. */        from \d etc by PCRE_UCP, this test will have been applied much earlier,
3350          to the original \d etc. At this point, ptr will point to a zero byte. */
3351        if (c == '\\')  
3352          {        if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3353          c = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);          STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3354              return FALSE;
         if (-c == ESC_b) c = '\b';       /* \b is backslash in a class */  
         else if (-c == ESC_X) c = 'X';   /* \X is literal X in a class */  
         else if (-c == ESC_Q)            /* Handle start of quoted string */  
           {  
           if (ptr[1] == '\\' && ptr[2] == 'E')  
             {  
             ptr += 2; /* avoid empty string */  
             }  
           else inescq = TRUE;  
           continue;  
           }  
3355    
3356          if (c < 0)        /* Do the property check. */
           {  
           register const uschar *cbits = cd->cbits;  
           class_charcount += 2;     /* Greater than 1 is what matters */  
           switch (-c)  
             {  
             case ESC_d:  
             for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];  
             continue;  
3357    
3358              case ESC_D:        return check_char_prop(c, ptype, pdata, (escape == ESC_P) != negated);
3359              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];        }
3360              continue;  #endif
3361    
3362              case ESC_w:      default:
3363              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_word];      return FALSE;
3364              continue;      }
3365    
3366              case ESC_W:    /* In principle, support for Unicode properties should be integrated here as
3367              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];    well. It means re-organizing the above code so as to get hold of the property
3368              continue;    values before switching on the op-code. However, I wonder how many patterns
3369      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
3370      these op-codes are never generated.) */
3371    
3372              case ESC_s:    case OP_DIGIT:
3373              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];    return escape == ESC_D || escape == ESC_s || escape == ESC_W ||
3374              classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */           escape == ESC_h || escape == ESC_v || escape == ESC_R;
             continue;  
3375    
3376              case ESC_S:    case OP_NOT_DIGIT:
3377              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];    return escape == ESC_d;
             classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */  
             continue;  
3378    
3379  #ifdef SUPPORT_UCP    case OP_WHITESPACE:
3380              case ESC_p:    return escape == ESC_S || escape == ESC_d || escape == ESC_w;
             case ESC_P:  
               {  
               BOOL negated;  
               int property = get_ucp(&ptr, &negated, errorcodeptr);  
               if (property < 0) goto FAILED;  
               class_utf8 = TRUE;  
               *class_utf8data++ = ((-c == ESC_p) != negated)?  
                 XCL_PROP : XCL_NOTPROP;  
               *class_utf8data++ = property;  
               class_charcount -= 2;   /* Not a < 256 character */  
               }  
             continue;  
 #endif  
3381    
3382              /* Unrecognized escapes are faulted if PCRE is running in its    case OP_NOT_WHITESPACE:
3383              strict mode. By default, for compatibility with Perl, they are    return escape == ESC_s || escape == ESC_h || escape == ESC_v || escape == ESC_R;
             treated as literals. */  
3384    
3385              default:    case OP_HSPACE:
3386              if ((options & PCRE_EXTRA) != 0)    return escape == ESC_S || escape == ESC_H || escape == ESC_d ||
3387                {           escape == ESC_w || escape == ESC_v || escape == ESC_R;
               *errorcodeptr = ERR7;  
               goto FAILED;  
               }  
             c = *ptr;              /* The final character */  
             class_charcount -= 2;  /* Undo the default count from above */  
             }  
           }  
3388    
3389          /* Fall through if we have a single character (c >= 0). This may be    case OP_NOT_HSPACE:
3390          > 256 in UTF-8 mode. */    return escape == ESC_h;
3391    
3392          }   /* End of backslash handling */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
3393      case OP_ANYNL:
3394      case OP_VSPACE:
3395      return escape == ESC_V || escape == ESC_d || escape == ESC_w;
3396    
3397        /* A single character may be followed by '-' to form a range. However,    case OP_NOT_VSPACE:
3398        Perl does not permit ']' to be the end of the range. A '-' character    return escape == ESC_v || escape == ESC_R;
       here is treated as a literal. */  
3399    
3400        if (ptr[1] == '-' && ptr[2] != ']')    case OP_WORDCHAR:
3401          {    return escape == ESC_W || escape == ESC_s || escape == ESC_h ||
3402          int d;           escape == ESC_v || escape == ESC_R;
         ptr += 2;  
3403    
3404  #ifdef SUPPORT_UTF8    case OP_NOT_WORDCHAR:
3405          if (utf8)    return escape == ESC_w || escape == ESC_d;
           {                           /* Braces are required because the */  
           GETCHARLEN(d, ptr, ptr);    /* macro generates multiple statements */  
           }  
         else  
 #endif  
         d = *ptr;  /* Not UTF-8 mode */  
3406    
3407          /* The second part of a range can be a single-character escape, but    default:
3408          not any of the other escapes. Perl 5.6 treats a hyphen as a literal    return FALSE;
3409          in such circumstances. */    }
3410    
3411          if (d == '\\')  /* Control does not reach here */
3412            {  }
           const uschar *oldptr = ptr;  
           d = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);  
3413    
           /* \b is backslash; \X is literal X; any other special means the '-'  
           was literal */  
3414    
           if (d < 0)  
             {  
             if (d == -ESC_b) d = '\b';  
             else if (d == -ESC_X) d = 'X'; else  
               {  
               ptr = oldptr - 2;  
               goto LONE_SINGLE_CHARACTER;  /* A few lines below */  
               }  
             }  
           }  
3415    
3416          /* The check that the two values are in the correct order happens in  /*************************************************
3417          the pre-pass. Optimize one-character ranges */  *        Add a character or range to a class     *
3418    *************************************************/
3419    
3420          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */  /* This function packages up the logic of adding a character or range of
3421    characters to a class. The character values in the arguments will be within the
3422    valid values for the current mode (8-bit, 16-bit, UTF, etc). This function is
3423    mutually recursive with the function immediately below.
3424    
3425          /* In UTF-8 mode, if the upper limit is > 255, or > 127 for caseless  Arguments:
3426          matching, we have to use an XCLASS with extra data items. Caseless    classbits     the bit map for characters < 256
3427          matching for characters > 127 is available only if UCP support is    uchardptr     points to the pointer for extra data
3428          available. */    options       the options word
3429      cd            contains pointers to tables etc.
3430      start         start of range character
3431      end           end of range character
3432    
3433    Returns:        the number of < 256 characters added
3434                 &nb