/[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 79 by nigel, Sat Feb 24 21:40:52 2007 UTC revision 1046 by ph10, Tue Sep 25 16:27:58 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)_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    
127  #define BRASTACK_SIZE 200  /* Repeated character flags. */
128    
129    #define UTF_LENGTH     0x10000000l      /* The char contains its length. */
130    
131  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
132  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
133  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
134  is invalid. */  is invalid. */
135    
136  #if !EBCDIC   /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC
137    
138    /* This is the "normal" table for ASCII systems or for EBCDIC systems running
139    in UTF-8 mode. */
140    
141  static const short int escapes[] = {  static const short int escapes[] = {
142       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,                       0,
143       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,                       0,
144     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */       0,                       0,
145       0,      0,      0,      0,      0,      0,      0,      0,   /* H - O */       0,                       0,
146  -ESC_P, -ESC_Q,      0, -ESC_S,      0,      0,      0, -ESC_W,   /* P - W */       0,                       0,
147  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */       CHAR_COLON,              CHAR_SEMICOLON,
148     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
149       0,      0,      0,      0,      0,      0,  ESC_n,      0,   /* h - o */       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
150  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0,      0, -ESC_w,   /* p - w */       CHAR_COMMERCIAL_AT,      -ESC_A,
151       0,      0, -ESC_z                                            /* x - z */       -ESC_B,                  -ESC_C,
152         -ESC_D,                  -ESC_E,
153         0,                       -ESC_G,
154         -ESC_H,                  0,
155         0,                       -ESC_K,
156         0,                       0,
157         -ESC_N,                  0,
158         -ESC_P,                  -ESC_Q,
159         -ESC_R,                  -ESC_S,
160         0,                       0,
161         -ESC_V,                  -ESC_W,
162         -ESC_X,                  0,
163         -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,
164         CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,
165         CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,
166         CHAR_GRAVE_ACCENT,       7,
167         -ESC_b,                  0,
168         -ESC_d,                  ESC_e,
169         ESC_f,                   0,
170         -ESC_h,                  0,
171         0,                       -ESC_k,
172         0,                       0,
173         ESC_n,                   0,
174         -ESC_p,                  0,
175         ESC_r,                   -ESC_s,
176         ESC_tee,                 0,
177         -ESC_v,                  -ESC_w,
178         0,                       0,
179         -ESC_z
180  };  };
181    
182  #else         /* This is the "abnormal" table for EBCDIC systems */  #else
183    
184    /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
185    
186  static const short int escapes[] = {  static const short int escapes[] = {
187  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
188  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 87  static const short int escapes[] = { Line 192  static const short int escapes[] = {
192  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,
193  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',
194  /*  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,
195  /*  88 */     0,     0,      0,     '{',      0,     0,      0,      0,  /*  88 */-ESC_h,     0,      0,     '{',      0,     0,      0,      0,
196  /*  90 */     0,     0,      0,     'l',      0, ESC_n,      0, -ESC_p,  /*  90 */     0,     0, -ESC_k,     'l',      0, ESC_n,      0, -ESC_p,
197  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,
198  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,     0, -ESC_w,      0,  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,-ESC_v, -ESC_w,      0,
199  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,
200  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,
201  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
202  /*  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,
203  /*  C8 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
204  /*  D0 */   '}',     0,      0,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,-ESC_N,      0, -ESC_P,
205  /*  D8 */-ESC_Q,     0,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
206  /*  E0 */  '\\',     0, -ESC_S,       0,      0,     0, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
207  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
208  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,
209  /*  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 211  static const short int escapes[] = {
211  #endif  #endif
212    
213    
214  /* 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
215  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
216  as this is assumed for handling case independence. */  the number of relocations when a shared library is dynamically linked. The
217    string is built from string macros so that it works in UTF-8 mode on EBCDIC
218  static const char *const posix_names[] = {  platforms. */
219    "alpha", "lower", "upper",  
220    "alnum", "ascii", "blank", "cntrl", "digit", "graph",  typedef struct verbitem {
221    "print", "punct", "space", "word",  "xdigit" };    int   len;                 /* Length of verb name */
222      int   op;                  /* Op when no arg, or -1 if arg mandatory */
223      int   op_arg;              /* Op when arg present, or -1 if not allowed */
224    } verbitem;
225    
226    static const char verbnames[] =
227      "\0"                       /* Empty name is a shorthand for MARK */
228      STRING_MARK0
229      STRING_ACCEPT0
230      STRING_COMMIT0
231      STRING_F0
232      STRING_FAIL0
233      STRING_PRUNE0
234      STRING_SKIP0
235      STRING_THEN;
236    
237    static const verbitem verbs[] = {
238      { 0, -1,        OP_MARK },
239      { 4, -1,        OP_MARK },
240      { 6, OP_ACCEPT, -1 },
241      { 6, OP_COMMIT, -1 },
242      { 1, OP_FAIL,   -1 },
243      { 4, OP_FAIL,   -1 },
244      { 5, OP_PRUNE,  OP_PRUNE_ARG },
245      { 4, OP_SKIP,   OP_SKIP_ARG  },
246      { 4, OP_THEN,   OP_THEN_ARG  }
247    };
248    
249    static const int verbcount = sizeof(verbs)/sizeof(verbitem);
250    
251    
252  static const uschar posix_name_lengths[] = {  /* Tables of names of POSIX character classes and their lengths. The names are
253    now all in a single string, to reduce the number of relocations when a shared
254    library is dynamically loaded. The list of lengths is terminated by a zero
255    length entry. The first three must be alpha, lower, upper, as this is assumed
256    for handling case independence. */
257    
258    static const char posix_names[] =
259      STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
260      STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
261      STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
262      STRING_word0  STRING_xdigit;
263    
264    static const pcre_uint8 posix_name_lengths[] = {
265    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 };
266    
267  /* 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
268  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
269  the vertical space characters. */  classes, there is some additional tweaking: for [:blank:] the vertical space
270    characters are removed, and for [:alpha:] and [:alnum:] the underscore
271    character is removed. The triples in the table consist of the base map offset,
272    second map offset or -1 if no second map, and a non-negative value for map
273    addition or a negative value for map subtraction (if there are two maps). The
274    absolute value of the third field has these meanings: 0 => no tweaking, 1 =>
275    remove vertical space characters, 2 => remove underscore. */
276    
277  static const int posix_class_maps[] = {  static const int posix_class_maps[] = {
278    cbit_lower, cbit_upper, -1,             /* alpha */    cbit_word,  cbit_digit, -2,             /* alpha */
279    cbit_lower, -1,         -1,             /* lower */    cbit_lower, -1,          0,             /* lower */
280    cbit_upper, -1,         -1,             /* upper */    cbit_upper, -1,          0,             /* upper */
281    cbit_digit, cbit_lower, cbit_upper,     /* alnum */    cbit_word,  -1,          2,             /* alnum - word without underscore */
282    cbit_print, cbit_cntrl, -1,             /* ascii */    cbit_print, cbit_cntrl,  0,             /* ascii */
283    cbit_space, -1,         -1,             /* blank - a GNU extension */    cbit_space, -1,          1,             /* blank - a GNU extension */
284    cbit_cntrl, -1,         -1,             /* cntrl */    cbit_cntrl, -1,          0,             /* cntrl */
285    cbit_digit, -1,         -1,             /* digit */    cbit_digit, -1,          0,             /* digit */
286    cbit_graph, -1,         -1,             /* graph */    cbit_graph, -1,          0,             /* graph */
287    cbit_print, -1,         -1,             /* print */    cbit_print, -1,          0,             /* print */
288    cbit_punct, -1,         -1,             /* punct */    cbit_punct, -1,          0,             /* punct */
289    cbit_space, -1,         -1,             /* space */    cbit_space, -1,          0,             /* space */
290    cbit_word,  -1,         -1,             /* word - a Perl extension */    cbit_word,  -1,          0,             /* word - a Perl extension */
291    cbit_xdigit,-1,         -1              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
292  };  };
293    
294    /* Table of substitutes for \d etc when PCRE_UCP is set. The POSIX class
295    substitutes must be in the order of the names, defined above, and there are
296    both positive and negative cases. NULL means no substitute. */
297    
298  /* The texts of compile-time error messages. These are "char *" because they  #ifdef SUPPORT_UCP
299  are passed to the outside world. */  static const pcre_uchar string_PNd[]  = {
300      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
301      CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
302    static const pcre_uchar string_pNd[]  = {
303      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
304      CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
305    static const pcre_uchar string_PXsp[] = {
306      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
307      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
308    static const pcre_uchar string_pXsp[] = {
309      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
310      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
311    static const pcre_uchar string_PXwd[] = {
312      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
313      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
314    static const pcre_uchar string_pXwd[] = {
315      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
316      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
317    
318    static const pcre_uchar *substitutes[] = {
319      string_PNd,           /* \D */
320      string_pNd,           /* \d */
321      string_PXsp,          /* \S */       /* NOTE: Xsp is Perl space */
322      string_pXsp,          /* \s */
323      string_PXwd,          /* \W */
324      string_pXwd           /* \w */
325    };
326    
327  static const char *error_texts[] = {  static const pcre_uchar string_pL[] =   {
328    "no error",    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
329    "\\ at end of pattern",    CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
330    "\\c at end of pattern",  static const pcre_uchar string_pLl[] =  {
331    "unrecognized character follows \\",    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
332    "numbers out of order in {} quantifier",    CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
333    static const pcre_uchar string_pLu[] =  {
334      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
335      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
336    static const pcre_uchar string_pXan[] = {
337      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
338      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
339    static const pcre_uchar string_h[] =    {
340      CHAR_BACKSLASH, CHAR_h, '\0' };
341    static const pcre_uchar string_pXps[] = {
342      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
343      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
344    static const pcre_uchar string_PL[] =   {
345      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
346      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
347    static const pcre_uchar string_PLl[] =  {
348      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
349      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
350    static const pcre_uchar string_PLu[] =  {
351      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
352      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
353    static const pcre_uchar string_PXan[] = {
354      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
355      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
356    static const pcre_uchar string_H[] =    {
357      CHAR_BACKSLASH, CHAR_H, '\0' };
358    static const pcre_uchar string_PXps[] = {
359      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
360      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
361    
362    static const pcre_uchar *posix_substitutes[] = {
363      string_pL,            /* alpha */
364      string_pLl,           /* lower */
365      string_pLu,           /* upper */
366      string_pXan,          /* alnum */
367      NULL,                 /* ascii */
368      string_h,             /* blank */
369      NULL,                 /* cntrl */
370      string_pNd,           /* digit */
371      NULL,                 /* graph */
372      NULL,                 /* print */
373      NULL,                 /* punct */
374      string_pXps,          /* space */    /* NOTE: Xps is POSIX space */
375      string_pXwd,          /* word */
376      NULL,                 /* xdigit */
377      /* Negated cases */
378      string_PL,            /* ^alpha */
379      string_PLl,           /* ^lower */
380      string_PLu,           /* ^upper */
381      string_PXan,          /* ^alnum */
382      NULL,                 /* ^ascii */
383      string_H,             /* ^blank */
384      NULL,                 /* ^cntrl */
385      string_PNd,           /* ^digit */
386      NULL,                 /* ^graph */
387      NULL,                 /* ^print */
388      NULL,                 /* ^punct */
389      string_PXps,          /* ^space */   /* NOTE: Xps is POSIX space */
390      string_PXwd,          /* ^word */
391      NULL                  /* ^xdigit */
392    };
393    #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))
394    #endif
395    
396    #define STRING(a)  # a
397    #define XSTRING(s) STRING(s)
398    
399    /* The texts of compile-time error messages. These are "char *" because they
400    are passed to the outside world. Do not ever re-use any error number, because
401    they are documented. Always add a new error instead. Messages marked DEAD below
402    are no longer used. This used to be a table of strings, but in order to reduce
403    the number of relocations needed when a shared library is loaded dynamically,
404    it is now one long string. We cannot use a table of offsets, because the
405    lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
406    simply count through to the one we want - this isn't a performance issue
407    because these strings are used only when there is a compilation error.
408    
409    Each substring ends with \0 to insert a null character. This includes the final
410    substring, so that the whole string ends with \0\0, which can be detected when
411    counting through. */
412    
413    static const char error_texts[] =
414      "no error\0"
415      "\\ at end of pattern\0"
416      "\\c at end of pattern\0"
417      "unrecognized character follows \\\0"
418      "numbers out of order in {} quantifier\0"
419    /* 5 */    /* 5 */
420    "number too big in {} quantifier",    "number too big in {} quantifier\0"
421    "missing terminating ] for character class",    "missing terminating ] for character class\0"
422    "invalid escape sequence in character class",    "invalid escape sequence in character class\0"
423    "range out of order in character class",    "range out of order in character class\0"
424    "nothing to repeat",    "nothing to repeat\0"
425    /* 10 */    /* 10 */
426    "operand of unlimited repeat could match the empty string",    "operand of unlimited repeat could match the empty string\0"  /** DEAD **/
427    "internal error: unexpected repeat",    "internal error: unexpected repeat\0"
428    "unrecognized character after (?",    "unrecognized character after (? or (?-\0"
429    "POSIX named classes are supported only within a class",    "POSIX named classes are supported only within a class\0"
430    "missing )",    "missing )\0"
431    /* 15 */    /* 15 */
432    "reference to non-existent subpattern",    "reference to non-existent subpattern\0"
433    "erroffset passed as NULL",    "erroffset passed as NULL\0"
434    "unknown option bit(s) set",    "unknown option bit(s) set\0"
435    "missing ) after comment",    "missing ) after comment\0"
436    "parentheses nested too deeply",    "parentheses nested too deeply\0"  /** DEAD **/
437    /* 20 */    /* 20 */
438    "regular expression too large",    "regular expression is too large\0"
439    "failed to get memory",    "failed to get memory\0"
440    "unmatched parentheses",    "unmatched parentheses\0"
441    "internal error: code overflow",    "internal error: code overflow\0"
442    "unrecognized character after (?<",    "unrecognized character after (?<\0"
443    /* 25 */    /* 25 */
444    "lookbehind assertion is not fixed length",    "lookbehind assertion is not fixed length\0"
445    "malformed number after (?(",    "malformed number or name after (?(\0"
446    "conditional group contains more than two branches",    "conditional group contains more than two branches\0"
447    "assertion expected after (?(",    "assertion expected after (?(\0"
448    "(?R or (?digits must be followed by )",    "(?R or (?[+-]digits must be followed by )\0"
449    /* 30 */    /* 30 */
450    "unknown POSIX class name",    "unknown POSIX class name\0"
451    "POSIX collating elements are not supported",    "POSIX collating elements are not supported\0"
452    "this version of PCRE is not compiled with PCRE_UTF8 support",    "this version of PCRE is compiled without UTF support\0"
453    "spare error",    "spare error\0"  /** DEAD **/
454    "character value in \\x{...} sequence is too large",    "character value in \\x{...} sequence is too large\0"
455    /* 35 */    /* 35 */
456    "invalid condition (?(0)",    "invalid condition (?(0)\0"
457    "\\C not allowed in lookbehind assertion",    "\\C not allowed in lookbehind assertion\0"
458    "PCRE does not support \\L, \\l, \\N, \\U, or \\u",    "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
459    "number after (?C is > 255",    "number after (?C is > 255\0"
460    "closing ) for (?C expected",    "closing ) for (?C expected\0"
461    /* 40 */    /* 40 */
462    "recursive call could loop indefinitely",    "recursive call could loop indefinitely\0"
463    "unrecognized character after (?P",    "unrecognized character after (?P\0"
464    "syntax error after (?P",    "syntax error in subpattern name (missing terminator)\0"
465    "two named groups have the same name",    "two named subpatterns have the same name\0"
466    "invalid UTF-8 string",    "invalid UTF-8 string\0"
467    /* 45 */    /* 45 */
468    "support for \\P, \\p, and \\X has not been compiled",    "support for \\P, \\p, and \\X has not been compiled\0"
469    "malformed \\P or \\p sequence",    "malformed \\P or \\p sequence\0"
470    "unknown property name after \\P or \\p"    "unknown property name after \\P or \\p\0"
471  };    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
472      "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
473      /* 50 */
474      "repeated subpattern is too long\0"    /** DEAD **/
475      "octal value is greater than \\377 in 8-bit non-UTF-8 mode\0"
476      "internal error: overran compiling workspace\0"
477      "internal error: previously-checked referenced subpattern not found\0"
478      "DEFINE group contains more than one branch\0"
479      /* 55 */
480      "repeating a DEFINE group is not allowed\0"  /** DEAD **/
481      "inconsistent NEWLINE options\0"
482      "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
483      "a numbered reference must not be zero\0"
484      "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
485      /* 60 */
486      "(*VERB) not recognized\0"
487      "number is too big\0"
488      "subpattern name expected\0"
489      "digit expected after (?+\0"
490      "] is an invalid data character in JavaScript compatibility mode\0"
491      /* 65 */
492      "different names for subpatterns of the same number are not allowed\0"
493      "(*MARK) must have an argument\0"
494      "this version of PCRE is not compiled with Unicode property support\0"
495      "\\c must be followed by an ASCII character\0"
496      "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
497      /* 70 */
498      "internal error: unknown opcode in find_fixedlength()\0"
499      "\\N is not supported in a class\0"
500      "too many forward references\0"
501      "disallowed Unicode code point (>= 0xd800 && <= 0xdfff)\0"
502      "invalid UTF-16 string\0"
503      /* 75 */
504      "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"
505      "character value in \\u.... sequence is too large\0"
506      ;
507    
508  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
509  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 521  For convenience, we use the same bit def
521    
522  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
523    
524  #if !EBCDIC    /* This is the "normal" case, for ASCII systems */  /* Using a simple comparison for decimal numbers rather than a memory read
525  static const unsigned char digitab[] =  is much faster, and the resulting code is simpler (the compiler turns it
526    into a subtraction and unsigned comparison). */
527    
528    #define IS_DIGIT(x) ((x) >= CHAR_0 && (x) <= CHAR_9)
529    
530    #ifndef EBCDIC
531    
532    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
533    UTF-8 mode. */
534    
535    static const pcre_uint8 digitab[] =
536    {    {
537    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
538    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 567  static const unsigned char digitab[] =
567    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
568    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
569    
570  #else          /* This is the "abnormal" case, for EBCDIC systems */  #else
571  static const unsigned char digitab[] =  
572    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
573    
574    static const pcre_uint8 digitab[] =
575    {    {
576    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
577    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 584  static const unsigned char digitab[] =
584    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */
585    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */
586    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */
587    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88-     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88- 95    */
588    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */
589    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */
590    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 606  static const unsigned char digitab[] =
606    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */
607    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */
608    
609  static const unsigned char ebcdic_chartab[] = { /* chartable partial dup */  static const pcre_uint8 ebcdic_chartab[] = { /* chartable partial dup */
610    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */
611    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */
612    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 618  static const unsigned char ebcdic_charta
618    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */
619    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */
620    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */
621    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88-  */    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88- 95 */
622    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */
623    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */
624    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 642  static const unsigned char ebcdic_charta
642  #endif  #endif
643    
644    
645  /* Definition to allow mutual recursion */  
646    
647    /*************************************************
648    *            Find an error text                  *
649    *************************************************/
650    
651    /* The error texts are now all in one long string, to save on relocations. As
652    some of the text is of unknown length, we can't use a table of offsets.
653    Instead, just count through the strings. This is not a performance issue
654    because it happens only when there has been a compilation error.
655    
656    Argument:   the error number
657    Returns:    pointer to the error string
658    */
659    
660    static const char *
661    find_error_text(int n)
662    {
663    const char *s = error_texts;
664    for (; n > 0; n--)
665      {
666      while (*s++ != 0) {};
667      if (*s == 0) return "Error text not found (please report)";
668      }
669    return s;
670    }
671    
672    
673    /*************************************************
674    *           Expand the workspace                 *
675    *************************************************/
676    
677    /* This function is called during the second compiling phase, if the number of
678    forward references fills the existing workspace, which is originally a block on
679    the stack. A larger block is obtained from malloc() unless the ultimate limit
680    has been reached or the increase will be rather small.
681    
682    Argument: pointer to the compile data block
683    Returns:  0 if all went well, else an error number
684    */
685    
686    static int
687    expand_workspace(compile_data *cd)
688    {
689    pcre_uchar *newspace;
690    int newsize = cd->workspace_size * 2;
691    
692    if (newsize > COMPILE_WORK_SIZE_MAX) newsize = COMPILE_WORK_SIZE_MAX;
693    if (cd->workspace_size >= COMPILE_WORK_SIZE_MAX ||
694        newsize - cd->workspace_size < WORK_SIZE_SAFETY_MARGIN)
695     return ERR72;
696    
697    newspace = (PUBL(malloc))(IN_UCHARS(newsize));
698    if (newspace == NULL) return ERR21;
699    memcpy(newspace, cd->start_workspace, cd->workspace_size * sizeof(pcre_uchar));
700    cd->hwm = (pcre_uchar *)newspace + (cd->hwm - cd->start_workspace);
701    if (cd->workspace_size > COMPILE_WORK_SIZE)
702      (PUBL(free))((void *)cd->start_workspace);
703    cd->start_workspace = newspace;
704    cd->workspace_size = newsize;
705    return 0;
706    }
707    
708    
709    
710    /*************************************************
711    *            Check for counted repeat            *
712    *************************************************/
713    
714    /* This function is called when a '{' is encountered in a place where it might
715    start a quantifier. It looks ahead to see if it really is a quantifier or not.
716    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
717    where the ddds are digits.
718    
719    Arguments:
720      p         pointer to the first char after '{'
721    
722    Returns:    TRUE or FALSE
723    */
724    
725  static BOOL  static BOOL
726    compile_regex(int, int, int *, uschar **, const uschar **, int *, BOOL, int,  is_counted_repeat(const pcre_uchar *p)
727      int *, int *, branch_chain *, compile_data *);  {
728    if (!IS_DIGIT(*p)) return FALSE;
729    p++;
730    while (IS_DIGIT(*p)) p++;
731    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
732    
733    if (*p++ != CHAR_COMMA) return FALSE;
734    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
735    
736    if (!IS_DIGIT(*p)) return FALSE;
737    p++;
738    while (IS_DIGIT(*p)) p++;
739    
740    return (*p == CHAR_RIGHT_CURLY_BRACKET);
741    }
742    
743    
744    
# Line 342  static BOOL Line 748  static BOOL
748    
749  /* 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
750  positive value for a simple escape such as \n, or a negative value which  positive value for a simple escape such as \n, or a negative value which
751  encodes one of the more complicated things such as \d. When UTF-8 is enabled,  encodes one of the more complicated things such as \d. A backreference to group
752  a positive value greater than 255 may be returned. On entry, ptr is pointing at  n is returned as -(ESC_REF + n); ESC_REF is the highest ESC_xxx macro. When
753  the \. On exit, it is on the final character of the escape sequence.  UTF-8 is enabled, a positive value greater than 255 may be returned. On entry,
754    ptr is pointing at the \. On exit, it is on the final character of the escape
755    sequence.
756    
757  Arguments:  Arguments:
758    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
# Line 355  Arguments: Line 763  Arguments:
763    
764  Returns:         zero or positive => a data character  Returns:         zero or positive => a data character
765                   negative => a special escape sequence                   negative => a special escape sequence
766                   on error, errorptr is set                   on error, errorcodeptr is set
767  */  */
768    
769  static int  static int
770  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const pcre_uchar **ptrptr, int *errorcodeptr, int bracount,
771    int options, BOOL isclass)    int options, BOOL isclass)
772  {  {
773  const uschar *ptr = *ptrptr;  /* PCRE_UTF16 has the same value as PCRE_UTF8. */
774  int c, i;  BOOL utf = (options & PCRE_UTF8) != 0;
775    const pcre_uchar *ptr = *ptrptr + 1;
776    pcre_int32 c;
777    int i;
778    
779    GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
780    ptr--;                            /* Set pointer back to the last byte */
781    
782  /* 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. */
783    
 c = *(++ptr);  
784  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
785    
786  /* 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
787  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.
788  Otherwise further processing may be required. */  Otherwise further processing may be required. */
789    
790  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
791  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  /* Not alphanumeric */
792  else if ((i = escapes[c - '0']) != 0) c = i;  else if (c < CHAR_0 || c > CHAR_z) {}
793    else if ((i = escapes[c - CHAR_0]) != 0) c = i;
794  #else          /* EBCDIC coding */  
795  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  #else           /* EBCDIC coding */
796    /* Not alphanumeric */
797    else if (c < CHAR_a || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}
798  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
799  #endif  #endif
800    
# Line 387  else if ((i = escapes[c - 0x48]) != 0) Line 802  else if ((i = escapes[c - 0x48]) != 0)
802    
803  else  else
804    {    {
805    const uschar *oldptr;    const pcre_uchar *oldptr;
806      BOOL braced, negated;
807    
808    switch (c)    switch (c)
809      {      {
810      /* 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
811      error. */      error. */
812    
813      case 'l':      case CHAR_l:
814      case 'L':      case CHAR_L:
     case 'N':  
     case 'u':  
     case 'U':  
815      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
816      break;      break;
817    
818        case CHAR_u:
819        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
820          {
821          /* In JavaScript, \u must be followed by four hexadecimal numbers.
822          Otherwise it is a lowercase u letter. */
823          if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
824            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0
825            && MAX_255(ptr[3]) && (digitab[ptr[3]] & ctype_xdigit) != 0
826            && MAX_255(ptr[4]) && (digitab[ptr[4]] & ctype_xdigit) != 0)
827            {
828            c = 0;
829            for (i = 0; i < 4; ++i)
830              {
831              register int cc = *(++ptr);
832    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
833              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
834              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
835    #else           /* EBCDIC coding */
836              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
837              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
838    #endif
839              }
840    
841    #ifdef COMPILE_PCRE8
842            if (c > (utf ? 0x10ffff : 0xff))
843    #else
844    #ifdef COMPILE_PCRE16
845            if (c > (utf ? 0x10ffff : 0xffff))
846    #endif
847    #endif
848              {
849              *errorcodeptr = ERR76;
850              }
851            else if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
852            }
853          }
854        else
855          *errorcodeptr = ERR37;
856        break;
857    
858        case CHAR_U:
859        /* In JavaScript, \U is an uppercase U letter. */
860        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
861        break;
862    
863        /* In a character class, \g is just a literal "g". Outside a character
864        class, \g must be followed by one of a number of specific things:
865    
866        (1) A number, either plain or braced. If positive, it is an absolute
867        backreference. If negative, it is a relative backreference. This is a Perl
868        5.10 feature.
869    
870        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
871        is part of Perl's movement towards a unified syntax for back references. As
872        this is synonymous with \k{name}, we fudge it up by pretending it really
873        was \k.
874    
875        (3) For Oniguruma compatibility we also support \g followed by a name or a
876        number either in angle brackets or in single quotes. However, these are
877        (possibly recursive) subroutine calls, _not_ backreferences. Just return
878        the -ESC_g code (cf \k). */
879    
880        case CHAR_g:
881        if (isclass) break;
882        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
883          {
884          c = -ESC_g;
885          break;
886          }
887    
888        /* Handle the Perl-compatible cases */
889    
890        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
891          {
892          const pcre_uchar *p;
893          for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
894            if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
895          if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
896            {
897            c = -ESC_k;
898            break;
899            }
900          braced = TRUE;
901          ptr++;
902          }
903        else braced = FALSE;
904    
905        if (ptr[1] == CHAR_MINUS)
906          {
907          negated = TRUE;
908          ptr++;
909          }
910        else negated = FALSE;
911    
912        /* The integer range is limited by the machine's int representation. */
913        c = 0;
914        while (IS_DIGIT(ptr[1]))
915          {
916          if (((unsigned int)c) > INT_MAX / 10) /* Integer overflow */
917            {
918            c = -1;
919            break;
920            }
921          c = c * 10 + *(++ptr) - CHAR_0;
922          }
923        if (((unsigned int)c) > INT_MAX) /* Integer overflow */
924          {
925          while (IS_DIGIT(ptr[1]))
926            ptr++;
927          *errorcodeptr = ERR61;
928          break;
929          }
930    
931        if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
932          {
933          *errorcodeptr = ERR57;
934          break;
935          }
936    
937        if (c == 0)
938          {
939          *errorcodeptr = ERR58;
940          break;
941          }
942    
943        if (negated)
944          {
945          if (c > bracount)
946            {
947            *errorcodeptr = ERR15;
948            break;
949            }
950          c = bracount - (c - 1);
951          }
952    
953        c = -(ESC_REF + c);
954        break;
955    
956      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
957      starting with one that is not zero is not straightforward. By experiment,      starting with one that is not zero is not straightforward. By experiment,
958      the way Perl works seems to be as follows:      the way Perl works seems to be as follows:
# Line 413  else Line 965  else
965      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
966      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
967    
968      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:
969      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
970    
971      if (!isclass)      if (!isclass)
972        {        {
973        oldptr = ptr;        oldptr = ptr;
974        c -= '0';        /* The integer range is limited by the machine's int representation. */
975        while ((digitab[ptr[1]] & ctype_digit) != 0)        c -= CHAR_0;
976          c = c * 10 + *(++ptr) - '0';        while (IS_DIGIT(ptr[1]))
977            {
978            if (((unsigned int)c) > INT_MAX / 10) /* Integer overflow */
979              {
980              c = -1;
981              break;
982              }
983            c = c * 10 + *(++ptr) - CHAR_0;
984            }
985          if (((unsigned int)c) > INT_MAX) /* Integer overflow */
986            {
987            while (IS_DIGIT(ptr[1]))
988              ptr++;
989            *errorcodeptr = ERR61;
990            break;
991            }
992        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
993          {          {
994          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 434  else Line 1001  else
1001      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.
1002      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
1003    
1004      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
1005        {        {
1006        ptr--;        ptr--;
1007        c = 0;        c = 0;
# Line 442  else Line 1009  else
1009        }        }
1010    
1011      /* \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
1012      larger first octal digit. */      larger first octal digit. The original code used just to take the least
1013        significant 8 bits of octal numbers (I think this is what early Perls used
1014      case '0':      to do). Nowadays we allow for larger numbers in UTF-8 mode and 16-bit mode,
1015      c -= '0';      but no more than 3 octal digits. */
1016      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')  
1017          c = c * 8 + *(++ptr) - '0';      case CHAR_0:
1018      c &= 255;     /* Take least significant 8 bits */      c -= CHAR_0;
1019        while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
1020            c = c * 8 + *(++ptr) - CHAR_0;
1021    #ifdef COMPILE_PCRE8
1022        if (!utf && c > 0xff) *errorcodeptr = ERR51;
1023    #endif
1024      break;      break;
1025    
1026      /* \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
1027      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.
1028        If not, { is treated as a data character. */
1029    
1030        case CHAR_x:
1031        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1032          {
1033          /* In JavaScript, \x must be followed by two hexadecimal numbers.
1034          Otherwise it is a lowercase x letter. */
1035          if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1036            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1037            {
1038            c = 0;
1039            for (i = 0; i < 2; ++i)
1040              {
1041              register int cc = *(++ptr);
1042    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1043              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1044              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1045    #else           /* EBCDIC coding */
1046              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1047              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1048    #endif
1049              }
1050            }
1051          break;
1052          }
1053    
1054      case 'x':      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
 #ifdef SUPPORT_UTF8  
     if (ptr[1] == '{' && (options & PCRE_UTF8) != 0)  
1055        {        {
1056        const uschar *pt = ptr + 2;        const pcre_uchar *pt = ptr + 2;
1057        register int count = 0;  
1058        c = 0;        c = 0;
1059        while ((digitab[*pt] & ctype_xdigit) != 0)        while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0)
1060          {          {
1061          int cc = *pt++;          register int cc = *pt++;
1062          count++;          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1063  #if !EBCDIC    /* ASCII coding */  
1064          if (cc >= 'a') cc -= 32;               /* Convert to upper case */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1065          c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1066  #else          /* EBCDIC coding */          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1067          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */  #else           /* EBCDIC coding */
1068          c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1069            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1070    #endif
1071    
1072    #ifdef COMPILE_PCRE8
1073            if (c > (utf ? 0x10ffff : 0xff)) { c = -1; break; }
1074    #else
1075    #ifdef COMPILE_PCRE16
1076            if (c > (utf ? 0x10ffff : 0xffff)) { c = -1; break; }
1077  #endif  #endif
1078    #endif
1079            }
1080    
1081          if (c < 0)
1082            {
1083            while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0) pt++;
1084            *errorcodeptr = ERR34;
1085          }          }
1086        if (*pt == '}')  
1087          if (*pt == CHAR_RIGHT_CURLY_BRACKET)
1088          {          {
1089          if (c < 0 || count > 8) *errorcodeptr = ERR34;          if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1090          ptr = pt;          ptr = pt;
1091          break;          break;
1092          }          }
1093    
1094        /* 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
1095        recognize this construct; fall through to the normal \x handling. */        recognize this construct; fall through to the normal \x handling. */
1096        }        }
 #endif  
1097    
1098      /* Read just a single hex char */      /* Read just a single-byte hex-defined char */
1099    
1100      c = 0;      c = 0;
1101      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1102        {        {
1103        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
1104        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
1105  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1106        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1107        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1108  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
1109        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1110        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1111  #endif  #endif
1112        }        }
1113      break;      break;
1114    
1115      /* Other special escapes not starting with a digit are straightforward */      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
1116        An error is given if the byte following \c is not an ASCII character. This
1117        coding is ASCII-specific, but then the whole concept of \cx is
1118        ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
1119    
1120      case 'c':      case CHAR_c:
1121      c = *(++ptr);      c = *(++ptr);
1122      if (c == 0)      if (c == 0)
1123        {        {
1124        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
1125        return 0;        break;
1126        }        }
1127    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
1128      /* A letter is upper-cased; then the 0x40 bit is flipped. This coding      if (c > 127)  /* Excludes all non-ASCII in either mode */
1129      is ASCII-specific, but then the whole concept of \cx is ASCII-specific.        {
1130      (However, an EBCDIC equivalent has now been added.) */        *errorcodeptr = ERR68;
1131          break;
1132  #if !EBCDIC    /* ASCII coding */        }
1133      if (c >= 'a' && c <= 'z') c -= 32;      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
1134      c ^= 0x40;      c ^= 0x40;
1135  #else          /* EBCDIC coding */  #else             /* EBCDIC coding */
1136      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
1137      c ^= 0xC0;      c ^= 0xC0;
1138  #endif  #endif
1139      break;      break;
1140    
1141      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
1142      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
1143      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
1144      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
1145      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
1146    
1147      default:      default:
1148      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 541  else Line 1155  else
1155      }      }
1156    }    }
1157    
1158    /* Perl supports \N{name} for character names, as well as plain \N for "not
1159    newline". PCRE does not support \N{name}. However, it does support
1160    quantification such as \N{2,3}. */
1161    
1162    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1163         !is_counted_repeat(ptr+2))
1164      *errorcodeptr = ERR37;
1165    
1166    /* If PCRE_UCP is set, we change the values for \d etc. */
1167    
1168    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
1169      c -= (ESC_DU - ESC_D);
1170    
1171    /* Set the pointer to the final character before returning. */
1172    
1173  *ptrptr = ptr;  *ptrptr = ptr;
1174  return c;  return c;
1175  }  }
# Line 560  escape sequence. Line 1189  escape sequence.
1189  Argument:  Argument:
1190    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
1191    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
1192      dptr           points to an int that is set to the detailed property value
1193    errorcodeptr   points to the error code variable    errorcodeptr   points to the error code variable
1194    
1195  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
1196  */  */
1197    
1198  static int  static int
1199  get_ucp(const uschar **ptrptr, BOOL *negptr, int *errorcodeptr)  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)
1200  {  {
1201  int c, i, bot, top;  int c, i, bot, top;
1202  const uschar *ptr = *ptrptr;  const pcre_uchar *ptr = *ptrptr;
1203  char name[4];  pcre_uchar name[32];
1204    
1205  c = *(++ptr);  c = *(++ptr);
1206  if (c == 0) goto ERROR_RETURN;  if (c == 0) goto ERROR_RETURN;
1207    
1208  *negptr = FALSE;  *negptr = FALSE;
1209    
1210  /* \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
1211  preceded by ^ for negation. */  negation. */
1212    
1213  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
1214    {    {
1215    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1216      {      {
1217      *negptr = TRUE;      *negptr = TRUE;
1218      ptr++;      ptr++;
1219      }      }
1220    for (i = 0; i <= 2; i++)    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1221      {      {
1222      c = *(++ptr);      c = *(++ptr);
1223      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
1224      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1225      name[i] = c;      name[i] = c;
1226      }      }
1227    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;  
     }  
1228    name[i] = 0;    name[i] = 0;
1229    }    }
1230    
# Line 615  else Line 1241  else
1241  /* Search for a recognized property name using binary chop */  /* Search for a recognized property name using binary chop */
1242    
1243  bot = 0;  bot = 0;
1244  top = _pcre_utt_size;  top = PRIV(utt_size);
1245    
1246  while (bot < top)  while (bot < top)
1247    {    {
1248    i = (bot + top)/2;    i = (bot + top) >> 1;
1249    c = strcmp(name, _pcre_utt[i].name);    c = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1250    if (c == 0) return _pcre_utt[i].value;    if (c == 0)
1251        {
1252        *dptr = PRIV(utt)[i].value;
1253        return PRIV(utt)[i].type;
1254        }
1255    if (c > 0) bot = i + 1; else top = i;    if (c > 0) bot = i + 1; else top = i;
1256    }    }
1257    
 UNKNOWN_RETURN:  
1258  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
1259  *ptrptr = ptr;  *ptrptr = ptr;
1260  return -1;  return -1;
# Line 641  return -1; Line 1270  return -1;
1270    
1271    
1272  /*************************************************  /*************************************************
 *            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 == '}');  
 }  
   
   
   
 /*************************************************  
1273  *         Read repeat counts                     *  *         Read repeat counts                     *
1274  *************************************************/  *************************************************/
1275    
# Line 692  Returns:         pointer to '}' on succe Line 1288  Returns:         pointer to '}' on succe
1288                   current ptr on error, with errorcodeptr set non-zero                   current ptr on error, with errorcodeptr set non-zero
1289  */  */
1290    
1291  static const uschar *  static const pcre_uchar *
1292  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)
1293  {  {
1294  int min = 0;  int min = 0;
1295  int max = -1;  int max = -1;
1296    
1297  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  /* Read the minimum value and do a paranoid check: a negative value indicates
1298    an integer overflow. */
1299    
1300    while (IS_DIGIT(*p)) min = min * 10 + *p++ - CHAR_0;
1301    if (min < 0 || min > 65535)
1302      {
1303      *errorcodeptr = ERR5;
1304      return p;
1305      }
1306    
1307    /* Read the maximum value if there is one, and again do a paranoid on its size.
1308    Also, max must not be less than min. */
1309    
1310  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1311    {    {
1312    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1313      {      {
1314      max = 0;      max = 0;
1315      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while(IS_DIGIT(*p)) max = max * 10 + *p++ - CHAR_0;
1316        if (max < 0 || max > 65535)
1317          {
1318          *errorcodeptr = ERR5;
1319          return p;
1320          }
1321      if (max < min)      if (max < min)
1322        {        {
1323        *errorcodeptr = ERR4;        *errorcodeptr = ERR4;
# Line 714  if (*p == '}') max = min; else Line 1326  if (*p == '}') max = min; else
1326      }      }
1327    }    }
1328    
1329  /* 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
1330  pointer to the terminating '}'. */  '}'. */
1331    
1332  if (min > 65535 || max > 65535)  *minp = min;
1333    *errorcodeptr = ERR5;  *maxp = max;
 else  
   {  
   *minp = min;  
   *maxp = max;  
   }  
1334  return p;  return p;
1335  }  }
1336    
1337    
1338    
1339  /*************************************************  /*************************************************
1340  *      Find first significant op code            *  *  Subroutine for finding forward reference      *
1341  *************************************************/  *************************************************/
1342    
1343  /* This is called by several functions that scan a compiled expression looking  /* This recursive function is called only from find_parens() below. The
1344  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
1345  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
1346  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
1347  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
1348    returns when it reaches a given numbered subpattern. Recursion is used to keep
1349    track of subpatterns that reset the capturing group numbers - the (?| feature.
1350    
1351    This function was originally called only from the second pass, in which we know
1352    that if (?< or (?' or (?P< is encountered, the name will be correctly
1353    terminated because that is checked in the first pass. There is now one call to
1354    this function in the first pass, to check for a recursive back reference by
1355    name (so that we can make the whole group atomic). In this case, we need check
1356    only up to the current position in the pattern, and that is still OK because
1357    and previous occurrences will have been checked. To make this work, the test
1358    for "end of pattern" is a check against cd->end_pattern in the main loop,
1359    instead of looking for a binary zero. This means that the special first-pass
1360    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1361    processing items within the loop are OK, because afterwards the main loop will
1362    terminate.)
1363    
1364  Arguments:  Arguments:
1365    code         pointer to the start of the group    ptrptr       address of the current character pointer (updated)
1366    options      pointer to external options    cd           compile background data
1367    optbit       the option bit whose changing is significant, or    name         name to seek, or NULL if seeking a numbered subpattern
1368                   zero if none are    lorn         name length, or subpattern number if name is NULL
1369    skipassert   TRUE if certain assertions are to be skipped    xmode        TRUE if we are in /x mode
1370      utf          TRUE if we are in UTF-8 / UTF-16 mode
1371      count        pointer to the current capturing subpattern number (updated)
1372    
1373  Returns:       pointer to the first significant opcode  Returns:       the number of the named subpattern, or -1 if not found
1374  */  */
1375    
1376  static const uschar*  static int
1377  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,
1378    BOOL skipassert)    BOOL xmode, BOOL utf, int *count)
1379  {  {
1380  for (;;)  pcre_uchar *ptr = *ptrptr;
1381    int start_count = *count;
1382    int hwm_count = start_count;
1383    BOOL dup_parens = FALSE;
1384    
1385    /* If the first character is a parenthesis, check on the type of group we are
1386    dealing with. The very first call may not start with a parenthesis. */
1387    
1388    if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1389    {    {
1390    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;  
1391    
1392      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;  
1393    
1394      case OP_WORD_BOUNDARY:    /* Handle a normal, unnamed capturing parenthesis. */
1395      case OP_NOT_WORD_BOUNDARY:  
1396      if (!skipassert) return code;    else if (ptr[1] != CHAR_QUESTION_MARK)
1397        {
1398        *count += 1;
1399        if (name == NULL && *count == lorn) return *count;
1400        ptr++;
1401        }
1402    
1403      /* All cases now have (? at the start. Remember when we are in a group
1404      where the parenthesis numbers are duplicated. */
1405    
1406      else if (ptr[2] == CHAR_VERTICAL_LINE)
1407        {
1408        ptr += 3;
1409        dup_parens = TRUE;
1410        }
1411    
1412      /* Handle comments; all characters are allowed until a ket is reached. */
1413    
1414      else if (ptr[2] == CHAR_NUMBER_SIGN)
1415        {
1416        for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1417        goto FAIL_EXIT;
1418        }
1419    
1420      /* Handle a condition. If it is an assertion, just carry on so that it
1421      is processed as normal. If not, skip to the closing parenthesis of the
1422      condition (there can't be any nested parens). */
1423    
1424      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1425        {
1426        ptr += 2;
1427        if (ptr[1] != CHAR_QUESTION_MARK)
1428          {
1429          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1430          if (*ptr != 0) ptr++;
1431          }
1432        }
1433    
1434      /* Start with (? but not a condition. */
1435    
1436      else
1437        {
1438        ptr += 2;
1439        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1440    
1441        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1442    
1443        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1444            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1445          {
1446          int term;
1447          const pcre_uchar *thisname;
1448          *count += 1;
1449          if (name == NULL && *count == lorn) return *count;
1450          term = *ptr++;
1451          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1452          thisname = ptr;
1453          while (*ptr != term) ptr++;
1454          if (name != NULL && lorn == ptr - thisname &&
1455              STRNCMP_UC_UC(name, thisname, lorn) == 0)
1456            return *count;
1457          term++;
1458          }
1459        }
1460      }
1461    
1462    /* Past any initial parenthesis handling, scan for parentheses or vertical
1463    bars. Stop if we get to cd->end_pattern. Note that this is important for the
1464    first-pass call when this value is temporarily adjusted to stop at the current
1465    position. So DO NOT change this to a test for binary zero. */
1466    
1467    for (; ptr < cd->end_pattern; ptr++)
1468      {
1469      /* Skip over backslashed characters and also entire \Q...\E */
1470    
1471      if (*ptr == CHAR_BACKSLASH)
1472        {
1473        if (*(++ptr) == 0) goto FAIL_EXIT;
1474        if (*ptr == CHAR_Q) for (;;)
1475          {
1476          while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1477          if (*ptr == 0) goto FAIL_EXIT;
1478          if (*(++ptr) == CHAR_E) break;
1479          }
1480        continue;
1481        }
1482    
1483      /* Skip over character classes; this logic must be similar to the way they
1484      are handled for real. If the first character is '^', skip it. Also, if the
1485      first few characters (either before or after ^) are \Q\E or \E we skip them
1486      too. This makes for compatibility with Perl. Note the use of STR macros to
1487      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1488    
1489      if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1490        {
1491        BOOL negate_class = FALSE;
1492        for (;;)
1493          {
1494          if (ptr[1] == CHAR_BACKSLASH)
1495            {
1496            if (ptr[2] == CHAR_E)
1497              ptr+= 2;
1498            else if (STRNCMP_UC_C8(ptr + 2,
1499                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1500              ptr += 4;
1501            else
1502              break;
1503            }
1504          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1505            {
1506            negate_class = TRUE;
1507            ptr++;
1508            }
1509          else break;
1510          }
1511    
1512        /* If the next character is ']', it is a data character that must be
1513        skipped, except in JavaScript compatibility mode. */
1514    
1515        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1516            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1517          ptr++;
1518    
1519        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1520          {
1521          if (*ptr == 0) return -1;
1522          if (*ptr == CHAR_BACKSLASH)
1523            {
1524            if (*(++ptr) == 0) goto FAIL_EXIT;
1525            if (*ptr == CHAR_Q) for (;;)
1526              {
1527              while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1528              if (*ptr == 0) goto FAIL_EXIT;
1529              if (*(++ptr) == CHAR_E) break;
1530              }
1531            continue;
1532            }
1533          }
1534        continue;
1535        }
1536    
1537      /* Skip comments in /x mode */
1538    
1539      if (xmode && *ptr == CHAR_NUMBER_SIGN)
1540        {
1541        ptr++;
1542        while (*ptr != 0)
1543          {
1544          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1545          ptr++;
1546    #ifdef SUPPORT_UTF
1547          if (utf) FORWARDCHAR(ptr);
1548    #endif
1549          }
1550        if (*ptr == 0) goto FAIL_EXIT;
1551        continue;
1552        }
1553    
1554      /* Check for the special metacharacters */
1555    
1556      if (*ptr == CHAR_LEFT_PARENTHESIS)
1557        {
1558        int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, count);
1559        if (rc > 0) return rc;
1560        if (*ptr == 0) goto FAIL_EXIT;
1561        }
1562    
1563      else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1564        {
1565        if (dup_parens && *count < hwm_count) *count = hwm_count;
1566        goto FAIL_EXIT;
1567        }
1568    
1569      else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1570        {
1571        if (*count > hwm_count) hwm_count = *count;
1572        *count = start_count;
1573        }
1574      }
1575    
1576    FAIL_EXIT:
1577    *ptrptr = ptr;
1578    return -1;
1579    }
1580    
1581    
1582    
1583    
1584    /*************************************************
1585    *       Find forward referenced subpattern       *
1586    *************************************************/
1587    
1588    /* This function scans along a pattern's text looking for capturing
1589    subpatterns, and counting them. If it finds a named pattern that matches the
1590    name it is given, it returns its number. Alternatively, if the name is NULL, it
1591    returns when it reaches a given numbered subpattern. This is used for forward
1592    references to subpatterns. We used to be able to start this scan from the
1593    current compiling point, using the current count value from cd->bracount, and
1594    do it all in a single loop, but the addition of the possibility of duplicate
1595    subpattern numbers means that we have to scan from the very start, in order to
1596    take account of such duplicates, and to use a recursive function to keep track
1597    of the different types of group.
1598    
1599    Arguments:
1600      cd           compile background data
1601      name         name to seek, or NULL if seeking a numbered subpattern
1602      lorn         name length, or subpattern number if name is NULL
1603      xmode        TRUE if we are in /x mode
1604      utf          TRUE if we are in UTF-8 / UTF-16 mode
1605    
1606    Returns:       the number of the found subpattern, or -1 if not found
1607    */
1608    
1609    static int
1610    find_parens(compile_data *cd, const pcre_uchar *name, int lorn, BOOL xmode,
1611      BOOL utf)
1612    {
1613    pcre_uchar *ptr = (pcre_uchar *)cd->start_pattern;
1614    int count = 0;
1615    int rc;
1616    
1617    /* If the pattern does not start with an opening parenthesis, the first call
1618    to find_parens_sub() will scan right to the end (if necessary). However, if it
1619    does start with a parenthesis, find_parens_sub() will return when it hits the
1620    matching closing parens. That is why we have to have a loop. */
1621    
1622    for (;;)
1623      {
1624      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, &count);
1625      if (rc > 0 || *ptr++ == 0) break;
1626      }
1627    
1628    return rc;
1629    }
1630    
1631    
1632    
1633    
1634    /*************************************************
1635    *      Find first significant op code            *
1636    *************************************************/
1637    
1638    /* This is called by several functions that scan a compiled expression looking
1639    for a fixed first character, or an anchoring op code etc. It skips over things
1640    that do not influence this. For some calls, it makes sense to skip negative
1641    forward and all backward assertions, and also the \b assertion; for others it
1642    does not.
1643    
1644    Arguments:
1645      code         pointer to the start of the group
1646      skipassert   TRUE if certain assertions are to be skipped
1647    
1648    Returns:       pointer to the first significant opcode
1649    */
1650    
1651    static const pcre_uchar*
1652    first_significant_code(const pcre_uchar *code, BOOL skipassert)
1653    {
1654    for (;;)
1655      {
1656      switch ((int)*code)
1657        {
1658        case OP_ASSERT_NOT:
1659        case OP_ASSERTBACK:
1660        case OP_ASSERTBACK_NOT:
1661        if (!skipassert) return code;
1662        do code += GET(code, 1); while (*code == OP_ALT);
1663        code += PRIV(OP_lengths)[*code];
1664        break;
1665    
1666        case OP_WORD_BOUNDARY:
1667        case OP_NOT_WORD_BOUNDARY:
1668        if (!skipassert) return code;
1669      /* Fall through */      /* Fall through */
1670    
1671      case OP_CALLOUT:      case OP_CALLOUT:
1672      case OP_CREF:      case OP_CREF:
1673      case OP_BRANUMBER:      case OP_NCREF:
1674      code += _pcre_OP_lengths[*code];      case OP_RREF:
1675        case OP_NRREF:
1676        case OP_DEF:
1677        code += PRIV(OP_lengths)[*code];
1678      break;      break;
1679    
1680      default:      default:
# Line 793  for (;;) Line 1688  for (;;)
1688    
1689    
1690  /*************************************************  /*************************************************
1691  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1692  *************************************************/  *************************************************/
1693    
1694  /* 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,
1695  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.
1696  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
1697    temporarily terminated with OP_END when this function is called.
1698    
1699    This function is called when a backward assertion is encountered, so that if it
1700    fails, the error message can point to the correct place in the pattern.
1701    However, we cannot do this when the assertion contains subroutine calls,
1702    because they can be forward references. We solve this by remembering this case
1703    and doing the check at the end; a flag specifies which mode we are running in.
1704    
1705  Arguments:  Arguments:
1706    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1707    options  the compiling options    utf      TRUE in UTF-8 / UTF-16 mode
1708      atend    TRUE if called when the pattern is complete
1709  Returns:   the fixed length, or -1 if there is no fixed length,    cd       the "compile data" structure
1710               or -2 if \C was encountered  
1711    Returns:   the fixed length,
1712                 or -1 if there is no fixed length,
1713                 or -2 if \C was encountered (in UTF-8 mode only)
1714                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1715                 or -4 if an unknown opcode was encountered (internal error)
1716  */  */
1717    
1718  static int  static int
1719  find_fixedlength(uschar *code, int options)  find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd)
1720  {  {
1721  int length = -1;  int length = -1;
1722    
1723  register int branchlength = 0;  register int branchlength = 0;
1724  register uschar *cc = code + 1 + LINK_SIZE;  register pcre_uchar *cc = code + 1 + LINK_SIZE;
1725    
1726  /* 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
1727  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 1729  branch, check the length against that of
1729  for (;;)  for (;;)
1730    {    {
1731    int d;    int d;
1732      pcre_uchar *ce, *cs;
1733    register int op = *cc;    register int op = *cc;
   if (op >= OP_BRA) op = OP_BRA;  
1734    
1735    switch (op)    switch (op)
1736      {      {
1737        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1738        OP_BRA (normal non-capturing bracket) because the other variants of these
1739        opcodes are all concerned with unlimited repeated groups, which of course
1740        are not of fixed length. */
1741    
1742        case OP_CBRA:
1743      case OP_BRA:      case OP_BRA:
1744      case OP_ONCE:      case OP_ONCE:
1745        case OP_ONCE_NC:
1746      case OP_COND:      case OP_COND:
1747      d = find_fixedlength(cc, options);      d = find_fixedlength(cc + ((op == OP_CBRA)? IMM2_SIZE : 0), utf, atend, cd);
1748      if (d < 0) return d;      if (d < 0) return d;
1749      branchlength += d;      branchlength += d;
1750      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1751      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1752      break;      break;
1753    
1754      /* 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.
1755      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
1756      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
1757        the same code. Note that we must not include the OP_KETRxxx opcodes here,
1758        because they all imply an unlimited repeat. */
1759    
1760      case OP_ALT:      case OP_ALT:
1761      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1762      case OP_END:      case OP_END:
1763        case OP_ACCEPT:
1764        case OP_ASSERT_ACCEPT:
1765      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1766        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
1767      if (*cc != OP_ALT) return length;      if (*cc != OP_ALT) return length;
# Line 853  for (;;) Line 1769  for (;;)
1769      branchlength = 0;      branchlength = 0;
1770      break;      break;
1771    
1772        /* A true recursion implies not fixed length, but a subroutine call may
1773        be OK. If the subroutine is a forward reference, we can't deal with
1774        it until the end of the pattern, so return -3. */
1775    
1776        case OP_RECURSE:
1777        if (!atend) return -3;
1778        cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1779        do ce += GET(ce, 1); while (*ce == OP_ALT);           /* End subpattern */
1780        if (cc > cs && cc < ce) return -1;                    /* Recursion */
1781        d = find_fixedlength(cs + IMM2_SIZE, utf, atend, cd);
1782        if (d < 0) return d;
1783        branchlength += d;
1784        cc += 1 + LINK_SIZE;
1785        break;
1786    
1787      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1788    
1789      case OP_ASSERT:      case OP_ASSERT:
# Line 860  for (;;) Line 1791  for (;;)
1791      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1792      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1793      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1794      /* Fall through */      cc += PRIV(OP_lengths)[*cc];
1795        break;
1796    
1797      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1798    
1799      case OP_REVERSE:      case OP_MARK:
1800      case OP_BRANUMBER:      case OP_PRUNE_ARG:
1801      case OP_CREF:      case OP_SKIP_ARG:
1802      case OP_OPT:      case OP_THEN_ARG:
1803        cc += cc[1] + PRIV(OP_lengths)[*cc];
1804        break;
1805    
1806      case OP_CALLOUT:      case OP_CALLOUT:
     case OP_SOD:  
     case OP_SOM:  
     case OP_EOD:  
     case OP_EODN:  
1807      case OP_CIRC:      case OP_CIRC:
1808        case OP_CIRCM:
1809        case OP_CLOSE:
1810        case OP_COMMIT:
1811        case OP_CREF:
1812        case OP_DEF:
1813      case OP_DOLL:      case OP_DOLL:
1814        case OP_DOLLM:
1815        case OP_EOD:
1816        case OP_EODN:
1817        case OP_FAIL:
1818        case OP_NCREF:
1819        case OP_NRREF:
1820      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1821        case OP_PRUNE:
1822        case OP_REVERSE:
1823        case OP_RREF:
1824        case OP_SET_SOM:
1825        case OP_SKIP:
1826        case OP_SOD:
1827        case OP_SOM:
1828        case OP_THEN:
1829      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1830      cc += _pcre_OP_lengths[*cc];      cc += PRIV(OP_lengths)[*cc];
1831      break;      break;
1832    
1833      /* Handle literal characters */      /* Handle literal characters */
1834    
1835      case OP_CHAR:      case OP_CHAR:
1836      case OP_CHARNC:      case OP_CHARI:
1837        case OP_NOT:
1838        case OP_NOTI:
1839      branchlength++;      branchlength++;
1840      cc += 2;      cc += 2;
1841  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1842      if ((options & PCRE_UTF8) != 0)      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
       {  
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1843  #endif  #endif
1844      break;      break;
1845    
# Line 898  for (;;) Line 1847  for (;;)
1847      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1848    
1849      case OP_EXACT:      case OP_EXACT:
1850        case OP_EXACTI:
1851        case OP_NOTEXACT:
1852        case OP_NOTEXACTI:
1853      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1854      cc += 4;      cc += 2 + IMM2_SIZE;
1855  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1856      if ((options & PCRE_UTF8) != 0)      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
       {  
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1857  #endif  #endif
1858      break;      break;
1859    
1860      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1861      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1862      cc += 4;      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)
1863          cc += 2;
1864        cc += 1 + IMM2_SIZE + 1;
1865      break;      break;
1866    
1867      /* Handle single-char matchers */      /* Handle single-char matchers */
1868    
1869      case OP_PROP:      case OP_PROP:
1870      case OP_NOTPROP:      case OP_NOTPROP:
1871      cc++;      cc += 2;
1872      /* Fall through */      /* Fall through */
1873    
1874        case OP_HSPACE:
1875        case OP_VSPACE:
1876        case OP_NOT_HSPACE:
1877        case OP_NOT_VSPACE:
1878      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
1879      case OP_DIGIT:      case OP_DIGIT:
1880      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
# Line 927  for (;;) Line 1882  for (;;)
1882      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1883      case OP_WORDCHAR:      case OP_WORDCHAR:
1884      case OP_ANY:      case OP_ANY:
1885        case OP_ALLANY:
1886      branchlength++;      branchlength++;
1887      cc++;      cc++;
1888      break;      break;
1889    
1890      /* The single-byte matcher isn't allowed */      /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1891        otherwise \C is coded as OP_ALLANY. */
1892    
1893      case OP_ANYBYTE:      case OP_ANYBYTE:
1894      return -2;      return -2;
1895    
1896      /* Check a class for variable quantification */      /* Check a class for variable quantification */
1897    
1898  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || defined COMPILE_PCRE16
1899      case OP_XCLASS:      case OP_XCLASS:
1900      cc += GET(cc, 1) - 33;      cc += GET(cc, 1) - PRIV(OP_lengths)[OP_CLASS];
1901      /* Fall through */      /* Fall through */
1902  #endif  #endif
1903    
1904      case OP_CLASS:      case OP_CLASS:
1905      case OP_NCLASS:      case OP_NCLASS:
1906      cc += 33;      cc += PRIV(OP_lengths)[OP_CLASS];
1907    
1908      switch (*cc)      switch (*cc)
1909        {        {
1910          case OP_CRPLUS:
1911          case OP_CRMINPLUS:
1912        case OP_CRSTAR:        case OP_CRSTAR:
1913        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1914        case OP_CRQUERY:        case OP_CRQUERY:
# Line 958  for (;;) Line 1917  for (;;)
1917    
1918        case OP_CRRANGE:        case OP_CRRANGE:
1919        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1920        if (GET2(cc,1) != GET2(cc,3)) return -1;        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1921        branchlength += GET2(cc,1);        branchlength += GET2(cc,1);
1922        cc += 5;        cc += 1 + 2 * IMM2_SIZE;
1923        break;        break;
1924    
1925        default:        default:
# Line 970  for (;;) Line 1929  for (;;)
1929    
1930      /* Anything else is variable length */      /* Anything else is variable length */
1931    
1932      default:      case OP_ANYNL:
1933        case OP_BRAMINZERO:
1934        case OP_BRAPOS:
1935        case OP_BRAPOSZERO:
1936        case OP_BRAZERO:
1937        case OP_CBRAPOS:
1938        case OP_EXTUNI:
1939        case OP_KETRMAX:
1940        case OP_KETRMIN:
1941        case OP_KETRPOS:
1942        case OP_MINPLUS:
1943        case OP_MINPLUSI:
1944        case OP_MINQUERY:
1945        case OP_MINQUERYI:
1946        case OP_MINSTAR:
1947        case OP_MINSTARI:
1948        case OP_MINUPTO:
1949        case OP_MINUPTOI:
1950        case OP_NOTMINPLUS:
1951        case OP_NOTMINPLUSI:
1952        case OP_NOTMINQUERY:
1953        case OP_NOTMINQUERYI:
1954        case OP_NOTMINSTAR:
1955        case OP_NOTMINSTARI:
1956        case OP_NOTMINUPTO:
1957        case OP_NOTMINUPTOI:
1958        case OP_NOTPLUS:
1959        case OP_NOTPLUSI:
1960        case OP_NOTPOSPLUS:
1961        case OP_NOTPOSPLUSI:
1962        case OP_NOTPOSQUERY:
1963        case OP_NOTPOSQUERYI:
1964        case OP_NOTPOSSTAR:
1965        case OP_NOTPOSSTARI:
1966        case OP_NOTPOSUPTO:
1967        case OP_NOTPOSUPTOI:
1968        case OP_NOTQUERY:
1969        case OP_NOTQUERYI:
1970        case OP_NOTSTAR:
1971        case OP_NOTSTARI:
1972        case OP_NOTUPTO:
1973        case OP_NOTUPTOI:
1974        case OP_PLUS:
1975        case OP_PLUSI:
1976        case OP_POSPLUS:
1977        case OP_POSPLUSI:
1978        case OP_POSQUERY:
1979        case OP_POSQUERYI:
1980        case OP_POSSTAR:
1981        case OP_POSSTARI:
1982        case OP_POSUPTO:
1983        case OP_POSUPTOI:
1984        case OP_QUERY:
1985        case OP_QUERYI:
1986        case OP_REF:
1987        case OP_REFI:
1988        case OP_SBRA:
1989        case OP_SBRAPOS:
1990        case OP_SCBRA:
1991        case OP_SCBRAPOS:
1992        case OP_SCOND:
1993        case OP_SKIPZERO:
1994        case OP_STAR:
1995        case OP_STARI:
1996        case OP_TYPEMINPLUS:
1997        case OP_TYPEMINQUERY:
1998        case OP_TYPEMINSTAR:
1999        case OP_TYPEMINUPTO:
2000        case OP_TYPEPLUS:
2001        case OP_TYPEPOSPLUS:
2002        case OP_TYPEPOSQUERY:
2003        case OP_TYPEPOSSTAR:
2004        case OP_TYPEPOSUPTO:
2005        case OP_TYPEQUERY:
2006        case OP_TYPESTAR:
2007        case OP_TYPEUPTO:
2008        case OP_UPTO:
2009        case OP_UPTOI:
2010      return -1;      return -1;
2011    
2012        /* Catch unrecognized opcodes so that when new ones are added they
2013        are not forgotten, as has happened in the past. */
2014    
2015        default:
2016        return -4;
2017      }      }
2018    }    }
2019  /* Control never gets here */  /* Control never gets here */
# Line 981  for (;;) Line 2023  for (;;)
2023    
2024    
2025  /*************************************************  /*************************************************
2026  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
2027  *************************************************/  *************************************************/
2028    
2029  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
2030  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
2031    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
2032    so that it can be called from pcre_study() when finding the minimum matching
2033    length.
2034    
2035  Arguments:  Arguments:
2036    code        points to start of expression    code        points to start of expression
2037    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 mode
2038    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
2039    
2040  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
2041  */  */
2042    
2043  static const uschar *  const pcre_uchar *
2044  find_bracket(const uschar *code, BOOL utf8, int number)  PRIV(find_bracket)(const pcre_uchar *code, BOOL utf, int number)
2045  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
2046  for (;;)  for (;;)
2047    {    {
2048    register int c = *code;    register int c = *code;
2049    
2050    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2051    else if (c > OP_BRA)  
2052      /* XCLASS is used for classes that cannot be represented just by a bit
2053      map. This includes negated single high-valued characters. The length in
2054      the table is zero; the actual length is stored in the compiled code. */
2055    
2056      if (c == OP_XCLASS) code += GET(code, 1);
2057    
2058      /* Handle recursion */
2059    
2060      else if (c == OP_REVERSE)
2061        {
2062        if (number < 0) return (pcre_uchar *)code;
2063        code += PRIV(OP_lengths)[c];
2064        }
2065    
2066      /* Handle capturing bracket */
2067    
2068      else if (c == OP_CBRA || c == OP_SCBRA ||
2069               c == OP_CBRAPOS || c == OP_SCBRAPOS)
2070      {      {
2071      int n = c - OP_BRA;      int n = GET2(code, 1+LINK_SIZE);
2072      if (n > EXTRACT_BASIC_MAX) n = GET2(code, 2+LINK_SIZE);      if (n == number) return (pcre_uchar *)code;
2073      if (n == number) return (uschar *)code;      code += PRIV(OP_lengths)[c];
     code += _pcre_OP_lengths[OP_BRA];  
2074      }      }
2075    
2076      /* Otherwise, we can get the item's length from the table, except that for
2077      repeated character types, we have to test for \p and \P, which have an extra
2078      two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2079      must add in its length. */
2080    
2081    else    else
2082      {      {
2083      code += _pcre_OP_lengths[c];      switch(c)
2084          {
2085          case OP_TYPESTAR:
2086          case OP_TYPEMINSTAR:
2087          case OP_TYPEPLUS:
2088          case OP_TYPEMINPLUS:
2089          case OP_TYPEQUERY:
2090          case OP_TYPEMINQUERY:
2091          case OP_TYPEPOSSTAR:
2092          case OP_TYPEPOSPLUS:
2093          case OP_TYPEPOSQUERY:
2094          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2095          break;
2096    
2097  #ifdef SUPPORT_UTF8        case OP_TYPEUPTO:
2098          case OP_TYPEMINUPTO:
2099          case OP_TYPEEXACT:
2100          case OP_TYPEPOSUPTO:
2101          if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2102            code += 2;
2103          break;
2104    
2105      /* In UTF-8 mode, opcodes that are followed by a character may be followed        case OP_MARK:
2106      by a multi-byte character. The length in the table is a minimum, so we have        case OP_PRUNE_ARG:
2107      to scan along to skip the extra bytes. All opcodes are less than 128, so we        case OP_SKIP_ARG:
2108      can use relatively efficient code. */        code += code[1];
2109          break;
2110    
2111          case OP_THEN_ARG:
2112          code += code[1];
2113          break;
2114          }
2115    
2116        /* Add in the fixed length from the table */
2117    
2118        code += PRIV(OP_lengths)[c];
2119    
2120      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
2121      a multi-byte character. The length in the table is a minimum, so we have to
2122      arrange to skip the extra bytes. */
2123    
2124      if (utf8) switch(c)  #ifdef SUPPORT_UTF
2125        if (utf) switch(c)
2126        {        {
2127        case OP_CHAR:        case OP_CHAR:
2128        case OP_CHARNC:        case OP_CHARI:
2129        case OP_EXACT:        case OP_EXACT:
2130          case OP_EXACTI:
2131        case OP_UPTO:        case OP_UPTO:
2132          case OP_UPTOI:
2133        case OP_MINUPTO:        case OP_MINUPTO:
2134          case OP_MINUPTOI:
2135          case OP_POSUPTO:
2136          case OP_POSUPTOI:
2137        case OP_STAR:        case OP_STAR:
2138          case OP_STARI:
2139        case OP_MINSTAR:        case OP_MINSTAR:
2140          case OP_MINSTARI:
2141          case OP_POSSTAR:
2142          case OP_POSSTARI:
2143        case OP_PLUS:        case OP_PLUS:
2144          case OP_PLUSI:
2145        case OP_MINPLUS:        case OP_MINPLUS:
2146          case OP_MINPLUSI:
2147          case OP_POSPLUS:
2148          case OP_POSPLUSI:
2149        case OP_QUERY:        case OP_QUERY:
2150          case OP_QUERYI:
2151        case OP_MINQUERY:        case OP_MINQUERY:
2152        while ((*code & 0xc0) == 0x80) code++;        case OP_MINQUERYI:
2153        break;        case OP_POSQUERY:
2154          case OP_POSQUERYI:
2155        /* 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;  
2156        break;        break;
2157        }        }
2158    #else
2159        (void)(utf);  /* Keep compiler happy by referencing function argument */
2160  #endif  #endif
2161      }      }
2162    }    }
# Line 1064  instance of OP_RECURSE. Line 2173  instance of OP_RECURSE.
2173    
2174  Arguments:  Arguments:
2175    code        points to start of expression    code        points to start of expression
2176    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 mode
2177    
2178  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
2179  */  */
2180    
2181  static const uschar *  static const pcre_uchar *
2182  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const pcre_uchar *code, BOOL utf)
2183  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
2184  for (;;)  for (;;)
2185    {    {
2186    register int c = *code;    register int c = *code;
2187    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2188    else if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
2189    else if (c > OP_BRA)  
2190      {    /* XCLASS is used for classes that cannot be represented just by a bit
2191      code += _pcre_OP_lengths[OP_BRA];    map. This includes negated single high-valued characters. The length in
2192      }    the table is zero; the actual length is stored in the compiled code. */
2193    
2194      if (c == OP_XCLASS) code += GET(code, 1);
2195    
2196      /* Otherwise, we can get the item's length from the table, except that for
2197      repeated character types, we have to test for \p and \P, which have an extra
2198      two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2199      must add in its length. */
2200    
2201    else    else
2202      {      {
2203      code += _pcre_OP_lengths[c];      switch(c)
2204          {
2205          case OP_TYPESTAR:
2206          case OP_TYPEMINSTAR:
2207          case OP_TYPEPLUS:
2208          case OP_TYPEMINPLUS:
2209          case OP_TYPEQUERY:
2210          case OP_TYPEMINQUERY:
2211          case OP_TYPEPOSSTAR:
2212          case OP_TYPEPOSPLUS:
2213          case OP_TYPEPOSQUERY:
2214          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2215          break;
2216    
2217          case OP_TYPEPOSUPTO:
2218          case OP_TYPEUPTO:
2219          case OP_TYPEMINUPTO:
2220          case OP_TYPEEXACT:
2221          if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2222            code += 2;
2223          break;
2224    
2225          case OP_MARK:
2226          case OP_PRUNE_ARG:
2227          case OP_SKIP_ARG:
2228          code += code[1];
2229          break;
2230    
2231          case OP_THEN_ARG:
2232          code += code[1];
2233          break;
2234          }
2235    
2236        /* Add in the fixed length from the table */
2237    
2238  #ifdef SUPPORT_UTF8      code += PRIV(OP_lengths)[c];
2239    
2240      /* 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
2241      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
2242      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. */  
2243    
2244      if (utf8) switch(c)  #ifdef SUPPORT_UTF
2245        if (utf) switch(c)
2246        {        {
2247        case OP_CHAR:        case OP_CHAR:
2248        case OP_CHARNC:        case OP_CHARI:
2249          case OP_NOT:
2250          case OP_NOTI:
2251        case OP_EXACT:        case OP_EXACT:
2252          case OP_EXACTI:
2253          case OP_NOTEXACT:
2254          case OP_NOTEXACTI:
2255        case OP_UPTO:        case OP_UPTO:
2256          case OP_UPTOI:
2257          case OP_NOTUPTO:
2258          case OP_NOTUPTOI:
2259        case OP_MINUPTO:        case OP_MINUPTO:
2260          case OP_MINUPTOI:
2261          case OP_NOTMINUPTO:
2262          case OP_NOTMINUPTOI:
2263          case OP_POSUPTO:
2264          case OP_POSUPTOI:
2265          case OP_NOTPOSUPTO:
2266          case OP_NOTPOSUPTOI:
2267        case OP_STAR:        case OP_STAR:
2268          case OP_STARI:
2269          case OP_NOTSTAR:
2270          case OP_NOTSTARI:
2271        case OP_MINSTAR:        case OP_MINSTAR:
2272          case OP_MINSTARI:
2273          case OP_NOTMINSTAR:
2274          case OP_NOTMINSTARI:
2275          case OP_POSSTAR:
2276          case OP_POSSTARI:
2277          case OP_NOTPOSSTAR:
2278          case OP_NOTPOSSTARI:
2279        case OP_PLUS:        case OP_PLUS:
2280          case OP_PLUSI:
2281          case OP_NOTPLUS:
2282          case OP_NOTPLUSI:
2283        case OP_MINPLUS:        case OP_MINPLUS:
2284          case OP_MINPLUSI:
2285          case OP_NOTMINPLUS:
2286          case OP_NOTMINPLUSI:
2287          case OP_POSPLUS:
2288          case OP_POSPLUSI:
2289          case OP_NOTPOSPLUS:
2290          case OP_NOTPOSPLUSI:
2291        case OP_QUERY:        case OP_QUERY:
2292          case OP_QUERYI:
2293          case OP_NOTQUERY:
2294          case OP_NOTQUERYI:
2295        case OP_MINQUERY:        case OP_MINQUERY:
2296        while ((*code & 0xc0) == 0x80) code++;        case OP_MINQUERYI:
2297        break;        case OP_NOTMINQUERY:
2298          case OP_NOTMINQUERYI:
2299        /* XCLASS is used for classes that cannot be represented just by a bit        case OP_POSQUERY:
2300        map. This includes negated single high-valued characters. The length in        case OP_POSQUERYI:
2301        the table is zero; the actual length is stored in the compiled code. */        case OP_NOTPOSQUERY:
2302          case OP_NOTPOSQUERYI:
2303        case OP_XCLASS:        if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
       code += GET(code, 1) + 1;  
2304        break;        break;
2305        }        }
2306    #else
2307        (void)(utf);  /* Keep compiler happy by referencing function argument */
2308  #endif  #endif
2309      }      }
2310    }    }
# Line 1132  for (;;) Line 2317  for (;;)
2317  *************************************************/  *************************************************/
2318    
2319  /* 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
2320  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()
2321  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
2322  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
2323  whose current branch will already have been scanned.  backward and negative forward assertions when its final argument is TRUE. If we
2324    hit an unclosed bracket, we return "empty" - this means we've struck an inner
2325    bracket whose current branch will already have been scanned.
2326    
2327  Arguments:  Arguments:
2328    code        points to start of search    code        points to start of search
2329    endcode     points to where to stop    endcode     points to where to stop
2330    utf8        TRUE if in UTF8 mode    utf         TRUE if in UTF-8 / UTF-16 mode
2331      cd          contains pointers to tables etc.
2332    
2333  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2334  */  */
2335    
2336  static BOOL  static BOOL
2337  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2338      BOOL utf, compile_data *cd)
2339  {  {
2340  register int c;  register int c;
2341  for (code = first_significant_code(code + 1 + LINK_SIZE, NULL, 0, TRUE);  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2342       code < endcode;       code < endcode;
2343       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
2344    {    {
2345    const uschar *ccode;    const pcre_uchar *ccode;
2346    
2347    c = *code;    c = *code;
2348    
2349    if (c >= OP_BRA)    /* Skip over forward assertions; the other assertions are skipped by
2350      first_significant_code() with a TRUE final argument. */
2351    
2352      if (c == OP_ASSERT)
2353        {
2354        do code += GET(code, 1); while (*code == OP_ALT);
2355        c = *code;
2356        continue;
2357        }
2358    
2359      /* For a recursion/subroutine call, if its end has been reached, which
2360      implies a backward reference subroutine call, we can scan it. If it's a
2361      forward reference subroutine call, we can't. To detect forward reference
2362      we have to scan up the list that is kept in the workspace. This function is
2363      called only when doing the real compile, not during the pre-compile that
2364      measures the size of the compiled pattern. */
2365    
2366      if (c == OP_RECURSE)
2367      {      {
2368        const pcre_uchar *scode;
2369      BOOL empty_branch;      BOOL empty_branch;
     if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */  
2370    
2371      /* Scan a closed bracket */      /* Test for forward reference */
2372    
2373        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
2374          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
2375    
2376        /* Not a forward reference, test for completed backward reference */
2377    
2378      empty_branch = FALSE;      empty_branch = FALSE;
2379      do      scode = cd->start_code + GET(code, 1);
2380        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2381    
2382        /* Completed backwards reference */
2383    
2384        do
2385        {        {
2386        if (!empty_branch && could_be_empty_branch(code, endcode, utf8))        if (could_be_empty_branch(scode, endcode, utf, cd))
2387            {
2388          empty_branch = TRUE;          empty_branch = TRUE;
2389            break;
2390            }
2391          scode += GET(scode, 1);
2392          }
2393        while (*scode == OP_ALT);
2394    
2395        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2396        continue;
2397        }
2398    
2399      /* Groups with zero repeats can of course be empty; skip them. */
2400    
2401      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2402          c == OP_BRAPOSZERO)
2403        {
2404        code += PRIV(OP_lengths)[c];
2405        do code += GET(code, 1); while (*code == OP_ALT);
2406        c = *code;
2407        continue;
2408        }
2409    
2410      /* A nested group that is already marked as "could be empty" can just be
2411      skipped. */
2412    
2413      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2414          c == OP_SCBRA || c == OP_SCBRAPOS)
2415        {
2416        do code += GET(code, 1); while (*code == OP_ALT);
2417        c = *code;
2418        continue;
2419        }
2420    
2421      /* For other groups, scan the branches. */
2422    
2423      if (c == OP_BRA  || c == OP_BRAPOS ||
2424          c == OP_CBRA || c == OP_CBRAPOS ||
2425          c == OP_ONCE || c == OP_ONCE_NC ||
2426          c == OP_COND)
2427        {
2428        BOOL empty_branch;
2429        if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
2430    
2431        /* If a conditional group has only one branch, there is a second, implied,
2432        empty branch, so just skip over the conditional, because it could be empty.
2433        Otherwise, scan the individual branches of the group. */
2434    
2435        if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
2436        code += GET(code, 1);        code += GET(code, 1);
2437        else
2438          {
2439          empty_branch = FALSE;
2440          do
2441            {
2442            if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd))
2443              empty_branch = TRUE;
2444            code += GET(code, 1);
2445            }
2446          while (*code == OP_ALT);
2447          if (!empty_branch) return FALSE;   /* All branches are non-empty */
2448        }        }
2449      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
     code += 1 + LINK_SIZE;  
2450      c = *code;      c = *code;
2451        continue;
2452      }      }
2453    
2454    else switch (c)    /* Handle the other opcodes */
2455    
2456      switch (c)
2457      {      {
2458      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
2459        cannot be represented just by a bit map. This includes negated single
2460        high-valued characters. The length in PRIV(OP_lengths)[] is zero; the
2461        actual length is stored in the compiled code, so we must update "code"
2462        here. */
2463    
2464  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2465      case OP_XCLASS:      case OP_XCLASS:
2466      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
2467      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
2468  #endif  #endif
2469    
2470      case OP_CLASS:      case OP_CLASS:
2471      case OP_NCLASS:      case OP_NCLASS:
2472      ccode = code + 33;      ccode = code + PRIV(OP_lengths)[OP_CLASS];
2473    
2474  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2475      CHECK_CLASS_REPEAT:      CHECK_CLASS_REPEAT:
2476  #endif  #endif
2477    
# Line 1227  for (code = first_significant_code(code Line 2507  for (code = first_significant_code(code
2507      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2508      case OP_WORDCHAR:      case OP_WORDCHAR:
2509      case OP_ANY:      case OP_ANY:
2510        case OP_ALLANY:
2511      case OP_ANYBYTE:      case OP_ANYBYTE:
2512      case OP_CHAR:      case OP_CHAR:
2513      case OP_CHARNC:      case OP_CHARI:
2514      case OP_NOT:      case OP_NOT:
2515        case OP_NOTI:
2516      case OP_PLUS:      case OP_PLUS:
2517      case OP_MINPLUS:      case OP_MINPLUS:
2518        case OP_POSPLUS:
2519      case OP_EXACT:      case OP_EXACT:
2520      case OP_NOTPLUS:      case OP_NOTPLUS:
2521      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
2522        case OP_NOTPOSPLUS:
2523      case OP_NOTEXACT:      case OP_NOTEXACT:
2524      case OP_TYPEPLUS:      case OP_TYPEPLUS:
2525      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
2526        case OP_TYPEPOSPLUS:
2527      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2528      return FALSE;      return FALSE;
2529    
2530        /* These are going to continue, as they may be empty, but we have to
2531        fudge the length for the \p and \P cases. */
2532    
2533        case OP_TYPESTAR:
2534        case OP_TYPEMINSTAR:
2535        case OP_TYPEPOSSTAR:
2536        case OP_TYPEQUERY:
2537        case OP_TYPEMINQUERY:
2538        case OP_TYPEPOSQUERY:
2539        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2540        break;
2541    
2542        /* Same for these */
2543    
2544        case OP_TYPEUPTO:
2545        case OP_TYPEMINUPTO:
2546        case OP_TYPEPOSUPTO:
2547        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2548          code += 2;
2549        break;
2550    
2551      /* End of branch */      /* End of branch */
2552    
2553      case OP_KET:      case OP_KET:
2554      case OP_KETRMAX:      case OP_KETRMAX:
2555      case OP_KETRMIN:      case OP_KETRMIN:
2556        case OP_KETRPOS:
2557      case OP_ALT:      case OP_ALT:
2558      return TRUE;      return TRUE;
2559    
2560      /* In UTF-8 mode, STAR, MINSTAR, QUERY, MINQUERY, UPTO, and MINUPTO  may be      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2561      followed by a multibyte character */      MINUPTO, and POSUPTO may be followed by a multibyte character */
2562    
2563  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2564      case OP_STAR:      case OP_STAR:
2565        case OP_STARI:
2566      case OP_MINSTAR:      case OP_MINSTAR:
2567        case OP_MINSTARI:
2568        case OP_POSSTAR:
2569        case OP_POSSTARI:
2570      case OP_QUERY:      case OP_QUERY:
2571        case OP_QUERYI:
2572      case OP_MINQUERY:      case OP_MINQUERY:
2573        case OP_MINQUERYI:
2574        case OP_POSQUERY:
2575        case OP_POSQUERYI:
2576        if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2577        break;
2578    
2579      case OP_UPTO:      case OP_UPTO:
2580        case OP_UPTOI:
2581      case OP_MINUPTO:      case OP_MINUPTO:
2582      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_MINUPTOI:
2583        case OP_POSUPTO:
2584        case OP_POSUPTOI:
2585        if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2586      break;      break;
2587  #endif  #endif
2588    
2589        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2590        string. */
2591    
2592        case OP_MARK:
2593        case OP_PRUNE_ARG:
2594        case OP_SKIP_ARG:
2595        code += code[1];
2596        break;
2597    
2598        case OP_THEN_ARG:
2599        code += code[1];
2600        break;
2601    
2602        /* None of the remaining opcodes are required to match a character. */
2603    
2604        default:
2605        break;
2606      }      }
2607    }    }
2608    
# Line 1279  return TRUE; Line 2619  return TRUE;
2619  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
2620  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,
2621  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.
2622    This function is called only during the real compile, not during the
2623    pre-compile.
2624    
2625  Arguments:  Arguments:
2626    code        points to start of the recursion    code        points to start of the recursion
2627    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2628    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2629    utf8        TRUE if in UTF-8 mode    utf         TRUE if in UTF-8 / UTF-16 mode
2630      cd          pointers to tables etc
2631    
2632  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2633  */  */
2634    
2635  static BOOL  static BOOL
2636  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
2637    BOOL utf8)    branch_chain *bcptr, BOOL utf, compile_data *cd)
2638  {  {
2639  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2640    {    {
2641    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd))
2642        return FALSE;
2643    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2644    }    }
2645  return TRUE;  return TRUE;
# Line 1308  return TRUE; Line 2652  return TRUE;
2652  *************************************************/  *************************************************/
2653    
2654  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2655  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
2656  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2657  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2658    
2659    Originally, this function only recognized a sequence of letters between the
2660    terminators, but it seems that Perl recognizes any sequence of characters,
2661    though of course unknown POSIX names are subsequently rejected. Perl gives an
2662    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2663    didn't consider this to be a POSIX class. Likewise for [:1234:].
2664    
2665    The problem in trying to be exactly like Perl is in the handling of escapes. We
2666    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2667    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2668    below handles the special case of \], but does not try to do any other escape
2669    processing. This makes it different from Perl for cases such as [:l\ower:]
2670    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2671    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2672    I think.
2673    
2674    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2675    It seems that the appearance of a nested POSIX class supersedes an apparent
2676    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2677    a digit.
2678    
2679    In Perl, unescaped square brackets may also appear as part of class names. For
2680    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2681    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2682    seem right at all. PCRE does not allow closing square brackets in POSIX class
2683    names.
2684    
2685  Argument:  Arguments:
2686    ptr      pointer to the initial [    ptr      pointer to the initial [
2687    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2688    
2689  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2690  */  */
2691    
2692  static BOOL  static BOOL
2693  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
2694  {  {
2695  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2696  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2697  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2698    {    {
2699    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2700    return TRUE;      ptr++;
2701      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2702      else
2703        {
2704        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2705          {
2706          *endptr = ptr;
2707          return TRUE;
2708          }
2709        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2710             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2711              ptr[1] == CHAR_EQUALS_SIGN) &&
2712            check_posix_syntax(ptr, endptr))
2713          return FALSE;
2714        }
2715    }    }
2716  return FALSE;  return FALSE;
2717  }  }
# Line 1353  Returns:     a value representing the na Line 2734  Returns:     a value representing the na
2734  */  */
2735    
2736  static int  static int
2737  check_posix_name(const uschar *ptr, int len)  check_posix_name(const pcre_uchar *ptr, int len)
2738  {  {
2739    const char *pn = posix_names;
2740  register int yield = 0;  register int yield = 0;
2741  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2742    {    {
2743    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2744      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      STRNCMP_UC_C8(ptr, pn, len) == 0) return yield;
2745      pn += posix_name_lengths[yield] + 1;
2746    yield++;    yield++;
2747    }    }
2748  return -1;  return -1;
# Line 1374  return -1; Line 2757  return -1;
2757  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2758  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2759  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
2760  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
2761  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
2762  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
2763  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
2764  partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2765    OP_END.
2766    
2767    This function has been extended with the possibility of forward references for
2768    recursions and subroutine calls. It must also check the list of such references
2769    for the group we are dealing with. If it finds that one of the recursions in
2770    the current group is on this list, it adjusts the offset in the list, not the
2771    value in the reference (which is a group number).
2772    
2773  Arguments:  Arguments:
2774    group      points to the start of the group    group      points to the start of the group
2775    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
2776    utf8       TRUE in UTF-8 mode    utf        TRUE in UTF-8 / UTF-16 mode
2777    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
2778      save_hwm   the hwm forward reference pointer at the start of the group
2779    
2780  Returns:     nothing  Returns:     nothing
2781  */  */
2782    
2783  static void  static void
2784  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd)  adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
2785      pcre_uchar *save_hwm)
2786  {  {
2787  uschar *ptr = group;  pcre_uchar *ptr = group;
2788  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  
2789    while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
2790    {    {
2791    int offset = GET(ptr, 1);    int offset;
2792    if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);    pcre_uchar *hc;
2793    
2794      /* See if this recursion is on the forward reference list. If so, adjust the
2795      reference. */
2796    
2797      for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
2798        {
2799        offset = GET(hc, 0);
2800        if (cd->start_code + offset == ptr + 1)
2801          {
2802          PUT(hc, 0, offset + adjust);
2803          break;
2804          }
2805        }
2806    
2807      /* Otherwise, adjust the recursion offset if it's after the start of this
2808      group. */
2809    
2810      if (hc >= cd->hwm)
2811        {
2812        offset = GET(ptr, 1);
2813        if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
2814        }
2815    
2816    ptr += 1 + LINK_SIZE;    ptr += 1 + LINK_SIZE;
2817    }    }
2818  }  }
# Line 1418  Arguments: Line 2834  Arguments:
2834  Returns:         new code pointer  Returns:         new code pointer
2835  */  */
2836    
2837  static uschar *  static pcre_uchar *
2838  auto_callout(uschar *code, const uschar *ptr, compile_data *cd)  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
2839  {  {
2840  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2841  *code++ = 255;  *code++ = 255;
2842  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2843  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2844  return code + 2*LINK_SIZE;  return code + 2 * LINK_SIZE;
2845  }  }
2846    
2847    
# Line 1447  Returns:             nothing Line 2863  Returns:             nothing
2863  */  */
2864    
2865  static void  static void
2866  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
2867  {  {
2868  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2869  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2870  }  }
2871    
# Line 1461  PUT(previous_callout, 2 + LINK_SIZE, len Line 2877  PUT(previous_callout, 2 + LINK_SIZE, len
2877  *************************************************/  *************************************************/
2878    
2879  /* 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
2880  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
2881  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
2882  start address.  start address. A character with multiple other cases is returned on its own
2883    with a special return value.
2884    
2885  Arguments:  Arguments:
2886    cptr        points to starting character value; updated    cptr        points to starting character value; updated
# Line 1471  Arguments: Line 2888  Arguments:
2888    ocptr       where to put start of othercase range    ocptr       where to put start of othercase range
2889    odptr       where to put end of othercase range    odptr       where to put end of othercase range
2890    
2891  Yield:        TRUE when range returned; FALSE when no more  Yield:        -1 when no more
2892                   0 when a range is returned
2893                  >0 the CASESET offset for char with multiple other cases
2894                    in this case, ocptr contains the original
2895  */  */
2896    
2897  static BOOL  static int
2898  get_othercase_range(int *cptr, int d, int *ocptr, int *odptr)  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,
2899      unsigned int *odptr)
2900  {  {
2901  int c, chartype, othercase, next;  unsigned int c, othercase, next;
2902    int co;
2903    
2904    /* Find the first character that has an other case. If it has multiple other
2905    cases, return its case offset value. */
2906    
2907  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2908    {    {
2909    if (_pcre_ucp_findchar(c, &chartype, &othercase) == ucp_L && othercase != 0)    if ((co = UCD_CASESET(c)) != 0)
2910      break;      {
2911        *ocptr = c++;   /* Character that has the set */
2912        *cptr = c;      /* Rest of input range */
2913        return co;
2914        }
2915      if ((othercase = UCD_OTHERCASE(c)) != c) break;
2916    }    }
2917    
2918  if (c > d) return FALSE;  if (c > d) return -1;  /* Reached end of range */
2919    
2920  *ocptr = othercase;  *ocptr = othercase;
2921  next = othercase + 1;  next = othercase + 1;
2922    
2923  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2924    {    {
2925    if (_pcre_ucp_findchar(c, &chartype, &othercase) != ucp_L ||    if (UCD_OTHERCASE(c) != next) break;
         othercase != next)  
     break;  
2926    next++;    next++;
2927    }    }
2928    
2929  *odptr = next - 1;  *odptr = next - 1;     /* End of othercase range */
2930  *cptr = c;  *cptr = c;             /* Rest of input range */
2931    return 0;
 return TRUE;  
2932  }  }
2933  #endif  /* SUPPORT_UCP */  
2934    
2935    
2936  /*************************************************  /*************************************************
2937  *           Compile one branch                   *  *        Check a character and a property        *
2938  *************************************************/  *************************************************/
2939    
2940  /* 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
2941  changed during the branch, the pointer is used to change the external options  is adjacent to a fixed character.
 bits.  
2942    
2943  Arguments:  Arguments:
2944    optionsptr     pointer to the option bits    c            the character
2945    brackets       points to number of extracting brackets used    ptype        the property type
2946    codeptr        points to the pointer to the current code point    pdata        the data for the type
2947    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.  
2948    
2949  Returns:         TRUE on success  Returns:       TRUE if auto-possessifying is OK
                  FALSE, with *errorcodeptr set non-zero on error  
2950  */  */
2951    
2952  static BOOL  static BOOL
2953  compile_branch(int *optionsptr, int *brackets, uschar **codeptr,  check_char_prop(int c, int ptype, int pdata, BOOL negated)
   const uschar **ptrptr, int *errorcodeptr, int *firstbyteptr,  
   int *reqbyteptr, branch_chain *bcptr, compile_data *cd)  
2954  {  {
2955  int repeat_type, op_type;  #ifdef SUPPORT_UCP
2956  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;  
2957  #endif  #endif
2958    
2959  /* 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;  
   
 /* Initialize no first byte, no required byte. REQ_UNSET means "no char  
 matching encountered yet". It gets changed to REQ_NONE if we hit something that  
 matches a non-fixed char first char; reqbyte just remains unset if we never  
 find one.  
2960    
2961  When we hit a repeat whose minimum is zero, we may have to adjust these values  switch(ptype)
2962  to take the zero repeat into account. This is implemented by setting them to    {
2963  zerofirstbyte and zeroreqbyte when such a repeat is encountered. The individual    case PT_LAMP:
2964  item types that can be repeated set these backoff variables appropriately. */    return (prop->chartype == ucp_Lu ||
2965              prop->chartype == ucp_Ll ||
2966              prop->chartype == ucp_Lt) == negated;
2967    
2968      case PT_GC:
2969      return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2970    
2971      case PT_PC:
2972      return (pdata == prop->chartype) == negated;
2973    
2974      case PT_SC:
2975      return (pdata == prop->script) == negated;
2976    
2977      /* These are specials */
2978    
2979      case PT_ALNUM:
2980      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2981              PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2982    
2983      case PT_SPACE:    /* Perl space */
2984      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2985              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2986              == negated;
2987    
2988      case PT_PXSPACE:  /* POSIX space */
2989      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2990              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2991              c == CHAR_FF || c == CHAR_CR)
2992              == negated;
2993    
2994      case PT_WORD:
2995      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2996              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2997              c == CHAR_UNDERSCORE) == negated;
2998    
2999    #ifdef SUPPORT_UCP
3000      case PT_CLIST:
3001      p = PRIV(ucd_caseless_sets) + prop->caseset;
3002      for (;;)
3003        {
3004        if ((unsigned int)c < *p) return !negated;
3005        if ((unsigned int)c == *p++) return negated;
3006        }
3007      break;  /* Control never reaches here */
3008    #endif
3009      }
3010    
3011    return FALSE;
3012    }
3013    #endif  /* SUPPORT_UCP */
3014    
 firstbyte = reqbyte = zerofirstbyte = zeroreqbyte = REQ_UNSET;  
3015    
 /* 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. */  
3016    
3017  req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;  /*************************************************
3018    *     Check if auto-possessifying is possible    *
3019    *************************************************/
3020    
3021  /* Switch on next character until the end of the branch */  /* This function is called for unlimited repeats of certain items, to see
3022    whether the next thing could possibly match the repeated item. If not, it makes
3023    sense to automatically possessify the repeated item.
3024    
3025  for (;; ptr++)  Arguments:
3026    {    previous      pointer to the repeated opcode
3027    BOOL negate_class;    utf           TRUE in UTF-8 / UTF-16 mode
3028    BOOL possessive_quantifier;    ptr           next character in pattern
3029    BOOL is_quantifier;    options       options bits
3030    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];  
3031    
3032    /* Next byte in the pattern */  Returns:        TRUE if possessifying is wanted
3033    */
3034    
3035    c = *ptr;  static BOOL
3036    check_auto_possessive(const pcre_uchar *previous, BOOL utf,
3037      const pcre_uchar *ptr, int options, compile_data *cd)
3038    {
3039    pcre_int32 c, next;
3040    int op_code = *previous++;
3041    
3042    /* If in \Q...\E, check for the end; if not, we have a literal */  /* Skip whitespace and comments in extended mode */
3043    
3044    if (inescq && c != 0)  if ((options & PCRE_EXTENDED) != 0)
3045      {
3046      for (;;)
3047      {      {
3048      if (c == '\\' && ptr[1] == 'E')      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
3049        if (*ptr == CHAR_NUMBER_SIGN)
3050        {        {
       inescq = FALSE;  
3051        ptr++;        ptr++;
3052        continue;        while (*ptr != 0)
       }  
     else  
       {  
       if (previous_callout != NULL)  
         {  
         complete_callout(previous_callout, ptr, cd);  
         previous_callout = NULL;  
         }  
       if ((options & PCRE_AUTO_CALLOUT) != 0)  
3053          {          {
3054          previous_callout = code;          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
3055          code = auto_callout(code, ptr, cd);          ptr++;
3056    #ifdef SUPPORT_UTF
3057            if (utf) FORWARDCHAR(ptr);
3058    #endif
3059          }          }
       goto NORMAL_CHAR;  
3060        }        }
3061        else break;
3062      }      }
3063      }
3064    
3065    /* 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
3066    a quantifier. */  value is a character, a negative value is an escape value. */
   
   is_quantifier = c == '*' || c == '+' || c == '?' ||  
     (c == '{' && is_counted_repeat(ptr+1));  
3067    
3068    if (!is_quantifier && previous_callout != NULL &&  if (*ptr == CHAR_BACKSLASH)
3069         after_manual_callout-- <= 0)    {
3070      {    int temperrorcode = 0;
3071      complete_callout(previous_callout, ptr, cd);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
3072      previous_callout = NULL;    if (temperrorcode != 0) return FALSE;
3073      }    ptr++;    /* Point after the escape sequence */
3074      }
3075    else if (!MAX_255(*ptr) || (cd->ctypes[*ptr] & ctype_meta) == 0)
3076      {
3077    #ifdef SUPPORT_UTF
3078      if (utf) { GETCHARINC(next, ptr); } else
3079    #endif
3080      next = *ptr++;
3081      }
3082    else return FALSE;
3083    
3084    /* In extended mode, skip white space and comments */  /* Skip whitespace and comments in extended mode */
3085    
3086    if ((options & PCRE_EXTENDED) != 0)  if ((options & PCRE_EXTENDED) != 0)
3087      {
3088      for (;;)
3089      {      {
3090      if ((cd->ctypes[c] & ctype_space) != 0) continue;      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
3091      if (c == '#')      if (*ptr == CHAR_NUMBER_SIGN)
3092        {        {
3093        /* The space before the ; is to avoid a warning on a silly compiler        ptr++;
3094        on the Macintosh. */        while (*ptr != 0)
3095        while ((c = *(++ptr)) != 0 && c != NEWLINE) ;          {
3096        if (c != 0) continue;   /* Else fall through to handle end of string */          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
3097            ptr++;
3098    #ifdef SUPPORT_UTF
3099            if (utf) FORWARDCHAR(ptr);
3100    #endif
3101            }
3102        }        }
3103        else break;
3104      }      }
3105      }
3106    
3107    /* No auto callout for quantifiers. */  /* If the next thing is itself optional, we have to give up. */
3108    
3109    if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3110      {    STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3111      previous_callout = code;      return FALSE;
3112      code = auto_callout(code, ptr, cd);  
3113      }  /* If the previous item is a character, get its value. */
3114    
3115    switch(c)  if (op_code == OP_CHAR || op_code == OP_CHARI ||
3116      {      op_code == OP_NOT || op_code == OP_NOTI)
3117      /* The branch terminates at end of string, |, or ). */    {
3118    #ifdef SUPPORT_UTF
3119      GETCHARTEST(c, previous);
3120    #else
3121      c = *previous;
3122    #endif
3123      }
3124    
3125      case 0:  /* Now compare the next item with the previous opcode. First, handle cases when
3126      case '|':  the next item is a character. For a caseless UTF match, the next character may
3127      case ')':  have more than one other case; convert this to a special property. */
     *firstbyteptr = firstbyte;  
     *reqbyteptr = reqbyte;  
     *codeptr = code;  
     *ptrptr = ptr;  
     return TRUE;  
3128    
3129      /* Handle single-character metacharacters. In multiline mode, ^ disables  if (next >= 0)
3130      the setting of any following char as a first character. */    {
3131    #ifdef SUPPORT_UCP
3132      if (utf && (options & PCRE_CASELESS) != 0)
3133        {
3134        int ocs = UCD_CASESET(next);
3135        if (ocs > 0) return check_char_prop(c, PT_CLIST, ocs, FALSE);
3136        }
3137    #endif
3138    
3139      case '^':    switch(op_code)
3140      if ((options & PCRE_MULTILINE) != 0)      {
3141        case OP_CHAR:
3142        return c != next;
3143    
3144        /* For CHARI (caseless character) we must check the other case. If we have
3145        Unicode property support, we can use it to test the other case of
3146        high-valued characters. We know that next can have only one other case,
3147        because multi-other-case characters are dealt with above. */
3148    
3149        case OP_CHARI:
3150        if (c == next) return FALSE;
3151    #ifdef SUPPORT_UTF
3152        if (utf)
3153        {        {
3154        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        unsigned int othercase;
3155          if (next < 128) othercase = cd->fcc[next]; else
3156    #ifdef SUPPORT_UCP
3157          othercase = UCD_OTHERCASE((unsigned int)next);
3158    #else
3159          othercase = NOTACHAR;
3160    #endif
3161          return (unsigned int)c != othercase;
3162        }        }
3163      previous = NULL;      else
3164      *code++ = OP_CIRC;  #endif  /* SUPPORT_UTF */
3165      break;      return (c != TABLE_GET((unsigned int)next, cd->fcc, next));  /* Not UTF */
3166    
3167        case OP_NOT:
3168        return c == next;
3169    
3170        case OP_NOTI:
3171        if (c == next) return TRUE;
3172    #ifdef SUPPORT_UTF
3173        if (utf)
3174          {
3175          unsigned int othercase;
3176          if (next < 128) othercase = cd->fcc[next]; else
3177    #ifdef SUPPORT_UCP
3178          othercase = UCD_OTHERCASE((unsigned int)next);
3179    #else
3180          othercase = NOTACHAR;
3181    #endif
3182          return (unsigned int)c == othercase;
3183          }
3184        else
3185    #endif  /* SUPPORT_UTF */
3186        return (c == TABLE_GET((unsigned int)next, cd->fcc, next));  /* Not UTF */
3187    
3188        /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
3189        When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3190    
3191        case OP_DIGIT:
3192        return next > 255 || (cd->ctypes[next] & ctype_digit) == 0;
3193    
3194        case OP_NOT_DIGIT:
3195        return next <= 255 && (cd->ctypes[next] & ctype_digit) != 0;
3196    
3197        case OP_WHITESPACE:
3198        return next > 255 || (cd->ctypes[next] & ctype_space) == 0;
3199    
3200        case OP_NOT_WHITESPACE:
3201        return next <= 255 && (cd->ctypes[next] & ctype_space) != 0;
3202    
3203        case OP_WORDCHAR:
3204        return next > 255 || (cd->ctypes[next] & ctype_word) == 0;
3205    
3206        case OP_NOT_WORDCHAR:
3207        return next <= 255 && (cd->ctypes[next] & ctype_word) != 0;
3208    
3209        case OP_HSPACE:
3210        case OP_NOT_HSPACE:
3211        switch(next)
3212          {
3213          HSPACE_CASES:
3214          return op_code == OP_NOT_HSPACE;
3215    
3216          default:
3217          return op_code != OP_NOT_HSPACE;
3218          }
3219    
3220        case OP_ANYNL:
3221        case OP_VSPACE:
3222        case OP_NOT_VSPACE:
3223        switch(next)
3224          {
3225          VSPACE_CASES:
3226          return op_code == OP_NOT_VSPACE;
3227    
3228          default:
3229          return op_code != OP_NOT_VSPACE;
3230          }
3231    
3232    #ifdef SUPPORT_UCP
3233        case OP_PROP:
3234        return check_char_prop(next, previous[0], previous[1], FALSE);
3235    
3236        case OP_NOTPROP:
3237        return check_char_prop(next, previous[0], previous[1], TRUE);
3238    #endif
3239    
3240        default:
3241        return FALSE;
3242        }
3243      }
3244    
3245      case '$':  /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP
3246      previous = NULL;  is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
3247      *code++ = OP_DOLL;  generated only when PCRE_UCP is *not* set, that is, when only ASCII
3248      break;  characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
3249    replaced by OP_PROP codes when PCRE_UCP is set. */
3250    
3251      /* There can never be a first char if '.' is first, whatever happens about  switch(op_code)
3252      repeats. The value of reqbyte doesn't change either. */    {
3253      case OP_CHAR:
3254      case OP_CHARI:
3255      switch(-next)
3256        {
3257        case ESC_d:
3258        return c > 255 || (cd->ctypes[c] & ctype_digit) == 0;
3259    
3260      case '.':      case ESC_D:
3261      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      return c <= 255 && (cd->ctypes[c] & ctype_digit) != 0;
     zerofirstbyte = firstbyte;  
     zeroreqbyte = reqbyte;  
     previous = code;  
     *code++ = OP_ANY;  
     break;  
3262    
3263      /* Character classes. If the included characters are all < 255 in value, we      case ESC_s:
3264      build a 32-byte bitmap of the permitted characters, except in the special      return c > 255 || (cd->ctypes[c] & 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.  
3265    
3266      If the class contains characters outside the 0-255 range, a different      case ESC_S:
3267      opcode is compiled. It may optionally have a bit map for characters < 256,      return c <= 255 && (cd->ctypes[c] & 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.  
     */  
3268    
3269      case '[':      case ESC_w:
3270      previous = code;      return c > 255 || (cd->ctypes[c] & ctype_word) == 0;
3271    
3272      /* PCRE supports POSIX class stuff inside a class. Perl gives an error if      case ESC_W:
3273      they are encountered at the top level, so we'll do that too. */      return c <= 255 && (cd->ctypes[c] & ctype_word) != 0;
3274    
3275      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      case ESC_h:
3276          check_posix_syntax(ptr, &tempptr, cd))      case ESC_H:
3277        switch(c)
3278        {        {
3279        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        HSPACE_CASES:
3280        goto FAILED;        return -next != ESC_h;
3281    
3282          default:
3283          return -next == ESC_h;
3284        }        }
3285    
3286      /* If the first character is '^', set the negation flag and skip it. */      case ESC_v:
3287        case ESC_V:
3288      if ((c = *(++ptr)) == '^')      switch(c)
3289        {        {
3290        negate_class = TRUE;        VSPACE_CASES:
3291        c = *(++ptr);        return -next != ESC_v;
3292        }  
3293      else        default:
3294        {        return -next == ESC_v;
       negate_class = FALSE;  
3295        }        }
3296    
3297      /* Keep a count of chars with values < 256 so that we can optimize the case      /* When PCRE_UCP is set, these values get generated for \d etc. Find
3298      of just a single character (as long as it's < 256). For higher valued UTF-8      their substitutions and process them. The result will always be either
3299      characters, we don't yet do any optimization. */      -ESC_p or -ESC_P. Then fall through to process those values. */
   
     class_charcount = 0;  
     class_lastchar = -1;  
3300    
3301  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UCP
3302      class_utf8 = FALSE;                       /* No chars >= 256 */      case ESC_du:
3303      class_utf8data = code + LINK_SIZE + 34;   /* For UTF-8 items */      case ESC_DU:
3304  #endif      case ESC_wu:
3305        case ESC_WU:
3306        case ESC_su:
3307        case ESC_SU:
3308          {
3309          int temperrorcode = 0;
3310          ptr = substitutes[-next - ESC_DU];
3311          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
3312          if (temperrorcode != 0) return FALSE;
3313          ptr++;    /* For compatibility */
3314          }
3315        /* Fall through */
3316    
3317      /* Initialize the 32-char bit map to all zeros. We have to build the      case ESC_p:
3318      map in a temporary bit of store, in case the class contains only 1      case ESC_P:
3319      character (< 256), because in that case the compiled code doesn't use the        {
3320      bit map. */        int ptype, pdata, errorcodeptr;
3321          BOOL negated;
3322    
3323      memset(classbits, 0, 32 * sizeof(uschar));        ptr--;      /* Make ptr point at the p or P */
3324          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
3325          if (ptype < 0) return FALSE;
3326          ptr++;      /* Point past the final curly ket */
3327    
3328          /* If the property item is optional, we have to give up. (When generated
3329          from \d etc by PCRE_UCP, this test will have been applied much earlier,
3330          to the original \d etc. At this point, ptr will point to a zero byte. */
3331    
3332          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3333            STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3334              return FALSE;
3335    
3336      /* Process characters until ] is reached. By writing this as a "do" it        /* Do the property check. */
     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. */  
3337    
3338      do        return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
3339        {        }
 #ifdef SUPPORT_UTF8  
       if (utf8 && c > 127)  
         {                           /* Braces are required because the */  
         GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */  
         }  
3340  #endif  #endif
3341    
3342        /* Inside \Q...\E everything is literal except \E */      default:
3343        return FALSE;
3344        }
3345    
3346        if (inescq)    /* In principle, support for Unicode properties should be integrated here as
3347          {    well. It means re-organizing the above code so as to get hold of the property
3348          if (c == '\\' && ptr[1] == 'E')    values before switching on the op-code. However, I wonder how many patterns
3349            {    combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
3350            inescq = FALSE;    these op-codes are never generated.) */
           ptr++;  
           continue;  
           }  
         else goto LONE_SINGLE_CHARACTER;  
         }  
3351    
3352        /* Handle POSIX class names. Perl allows a negation extension of the &