/[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 85 by nigel, Sat Feb 24 21:41:13 2007 UTC revision 1219 by ph10, Sun Nov 11 18:04:37 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 DEBUG is defined, we need the pcre_printint() function, which is also  /* When PCRE_DEBUG is defined, we need the pcre(16|32)_printint() function, which
57  used by pcretest. DEBUG is not defined when building a production library. */  is also used by pcretest. PCRE_DEBUG is not defined when building a production
58    library. We do not need to select pcre16_printint.c specially, because the
59    COMPILE_PCREx macro will already be appropriately set. */
60    
61  #ifdef DEBUG  #ifdef PCRE_DEBUG
62  #include "pcre_printint.src"  /* pcre_printint.c should not include any headers */
63    #define PCRE_INCLUDED
64    #include "pcre_printint.c"
65    #undef PCRE_INCLUDED
66  #endif  #endif
67    
68    
69    /* Macro for setting individual bits in class bitmaps. */
70    
71    #define SETBIT(a,b) a[(b)/8] |= (1 << ((b)&7))
72    
73    /* Maximum length value to check against when making sure that the integer that
74    holds the compiled pattern length does not overflow. We make it a bit less than
75    INT_MAX to allow for adding in group terminating bytes, so that we don't have
76    to check them every time. */
77    
78    #define OFLOW_MAX (INT_MAX - 20)
79    
80    /* Definitions to allow mutual recursion */
81    
82    static int
83      add_list_to_class(pcre_uint8 *, pcre_uchar **, int, compile_data *,
84        const pcre_uint32 *, unsigned int);
85    
86    static BOOL
87      compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL, int, int,
88        pcre_uint32 *, pcre_int32 *, pcre_uint32 *, pcre_int32 *, branch_chain *,
89        compile_data *, int *);
90    
91    
92    
93  /*************************************************  /*************************************************
94  *      Code parameters and static tables         *  *      Code parameters and static tables         *
95  *************************************************/  *************************************************/
96    
97  /* 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
98  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
99  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
100  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
101  compile time. */  does, however, check for an overrun. The largest amount I've seen used is 218,
102    so this number is very generous.
103    
104    The same workspace is used during the second, actual compile phase for
105    remembering forward references to groups so that they can be filled in at the
106    end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE
107    is 4 there is plenty of room for most patterns. However, the memory can get
108    filled up by repetitions of forward references, for example patterns like
109    /(?1){0,1999}(b)/, and one user did hit the limit. The code has been changed so
110    that the workspace is expanded using malloc() in this situation. The value
111    below is therefore a minimum, and we put a maximum on it for safety. The
112    minimum is now also defined in terms of LINK_SIZE so that the use of malloc()
113    kicks in at the same number of forward references in all cases. */
114    
115    #define COMPILE_WORK_SIZE (2048*LINK_SIZE)
116    #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)
117    
118    /* The overrun tests check for a slightly smaller size so that they detect the
119    overrun before it actually does run off the end of the data block. */
120    
121    #define WORK_SIZE_SAFETY_MARGIN (100)
122    
123    /* Private flags added to firstchar and reqchar. */
124    
125    #define REQ_CASELESS    (1 << 0)        /* Indicates caselessness */
126    #define REQ_VARY        (1 << 1)        /* Reqchar followed non-literal item */
127    /* Negative values for the firstchar and reqchar flags */
128    #define REQ_UNSET       (-2)
129    #define REQ_NONE        (-1)
130    
131  #define BRASTACK_SIZE 200  /* Repeated character flags. */
132    
133    #define UTF_LENGTH     0x10000000l      /* The char contains its length. */
134    
135  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
136  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
137  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
138  is invalid. */  is invalid. */
139    
140  #if !EBCDIC   /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC
141    
142    /* This is the "normal" table for ASCII systems or for EBCDIC systems running
143    in UTF-8 mode. */
144    
145  static const short int escapes[] = {  static const short int escapes[] = {
146       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,                       0,
147       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,                       0,
148     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */       0,                       0,
149       0,      0,      0,      0,      0,      0,      0,      0,   /* H - O */       0,                       0,
150  -ESC_P, -ESC_Q,      0, -ESC_S,      0,      0,      0, -ESC_W,   /* P - W */       0,                       0,
151  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */       CHAR_COLON,              CHAR_SEMICOLON,
152     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
153       0,      0,      0,      0,      0,      0,  ESC_n,      0,   /* h - o */       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
154  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0,      0, -ESC_w,   /* p - w */       CHAR_COMMERCIAL_AT,      -ESC_A,
155       0,      0, -ESC_z                                            /* x - z */       -ESC_B,                  -ESC_C,
156         -ESC_D,                  -ESC_E,
157         0,                       -ESC_G,
158         -ESC_H,                  0,
159         0,                       -ESC_K,
160         0,                       0,
161         -ESC_N,                  0,
162         -ESC_P,                  -ESC_Q,
163         -ESC_R,                  -ESC_S,
164         0,                       0,
165         -ESC_V,                  -ESC_W,
166         -ESC_X,                  0,
167         -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,
168         CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,
169         CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,
170         CHAR_GRAVE_ACCENT,       7,
171         -ESC_b,                  0,
172         -ESC_d,                  ESC_e,
173         ESC_f,                   0,
174         -ESC_h,                  0,
175         0,                       -ESC_k,
176         0,                       0,
177         ESC_n,                   0,
178         -ESC_p,                  0,
179         ESC_r,                   -ESC_s,
180         ESC_tee,                 0,
181         -ESC_v,                  -ESC_w,
182         0,                       0,
183         -ESC_z
184  };  };
185    
186  #else         /* This is the "abnormal" table for EBCDIC systems */  #else
187    
188    /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
189    
190  static const short int escapes[] = {  static const short int escapes[] = {
191  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
192  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 96  static const short int escapes[] = { Line 196  static const short int escapes[] = {
196  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,
197  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',
198  /*  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,
199  /*  88 */     0,     0,      0,     '{',      0,     0,      0,      0,  /*  88 */-ESC_h,     0,      0,     '{',      0,     0,      0,      0,
200  /*  90 */     0,     0,      0,     'l',      0, ESC_n,      0, -ESC_p,  /*  90 */     0,     0, -ESC_k,     'l',      0, ESC_n,      0, -ESC_p,
201  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,
202  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,     0, -ESC_w,      0,  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,-ESC_v, -ESC_w,      0,
203  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,
204  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,
205  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
206  /*  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,
207  /*  C8 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
208  /*  D0 */   '}',     0,      0,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,-ESC_N,      0, -ESC_P,
209  /*  D8 */-ESC_Q,     0,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
210  /*  E0 */  '\\',     0, -ESC_S,       0,      0,     0, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
211  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
212  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,
213  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0
# Line 115  static const short int escapes[] = { Line 215  static const short int escapes[] = {
215  #endif  #endif
216    
217    
218  /* 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
219  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
220  as this is assumed for handling case independence. */  the number of relocations when a shared library is dynamically linked. The
221    string is built from string macros so that it works in UTF-8 mode on EBCDIC
222  static const char *const posix_names[] = {  platforms. */
223    "alpha", "lower", "upper",  
224    "alnum", "ascii", "blank", "cntrl", "digit", "graph",  typedef struct verbitem {
225    "print", "punct", "space", "word",  "xdigit" };    int   len;                 /* Length of verb name */
226      int   op;                  /* Op when no arg, or -1 if arg mandatory */
227      int   op_arg;              /* Op when arg present, or -1 if not allowed */
228    } verbitem;
229    
230    static const char verbnames[] =
231      "\0"                       /* Empty name is a shorthand for MARK */
232      STRING_MARK0
233      STRING_ACCEPT0
234      STRING_COMMIT0
235      STRING_F0
236      STRING_FAIL0
237      STRING_PRUNE0
238      STRING_SKIP0
239      STRING_THEN;
240    
241    static const verbitem verbs[] = {
242      { 0, -1,        OP_MARK },
243      { 4, -1,        OP_MARK },
244      { 6, OP_ACCEPT, -1 },
245      { 6, OP_COMMIT, -1 },
246      { 1, OP_FAIL,   -1 },
247      { 4, OP_FAIL,   -1 },
248      { 5, OP_PRUNE,  OP_PRUNE_ARG },
249      { 4, OP_SKIP,   OP_SKIP_ARG  },
250      { 4, OP_THEN,   OP_THEN_ARG  }
251    };
252    
253    static const int verbcount = sizeof(verbs)/sizeof(verbitem);
254    
255    
256    /* Tables of names of POSIX character classes and their lengths. The names are
257    now all in a single string, to reduce the number of relocations when a shared
258    library is dynamically loaded. The list of lengths is terminated by a zero
259    length entry. The first three must be alpha, lower, upper, as this is assumed
260    for handling case independence. */
261    
262    static const char posix_names[] =
263      STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
264      STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
265      STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
266      STRING_word0  STRING_xdigit;
267    
268  static const uschar posix_name_lengths[] = {  static const pcre_uint8 posix_name_lengths[] = {
269    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 };
270    
271  /* 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
272  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
273  the vertical space characters. */  classes, there is some additional tweaking: for [:blank:] the vertical space
274    characters are removed, and for [:alpha:] and [:alnum:] the underscore
275    character is removed. The triples in the table consist of the base map offset,
276    second map offset or -1 if no second map, and a non-negative value for map
277    addition or a negative value for map subtraction (if there are two maps). The
278    absolute value of the third field has these meanings: 0 => no tweaking, 1 =>
279    remove vertical space characters, 2 => remove underscore. */
280    
281  static const int posix_class_maps[] = {  static const int posix_class_maps[] = {
282    cbit_lower, cbit_upper, -1,             /* alpha */    cbit_word,  cbit_digit, -2,             /* alpha */
283    cbit_lower, -1,         -1,             /* lower */    cbit_lower, -1,          0,             /* lower */
284    cbit_upper, -1,         -1,             /* upper */    cbit_upper, -1,          0,             /* upper */
285    cbit_digit, cbit_lower, cbit_upper,     /* alnum */    cbit_word,  -1,          2,             /* alnum - word without underscore */
286    cbit_print, cbit_cntrl, -1,             /* ascii */    cbit_print, cbit_cntrl,  0,             /* ascii */
287    cbit_space, -1,         -1,             /* blank - a GNU extension */    cbit_space, -1,          1,             /* blank - a GNU extension */
288    cbit_cntrl, -1,         -1,             /* cntrl */    cbit_cntrl, -1,          0,             /* cntrl */
289    cbit_digit, -1,         -1,             /* digit */    cbit_digit, -1,          0,             /* digit */
290    cbit_graph, -1,         -1,             /* graph */    cbit_graph, -1,          0,             /* graph */
291    cbit_print, -1,         -1,             /* print */    cbit_print, -1,          0,             /* print */
292    cbit_punct, -1,         -1,             /* punct */    cbit_punct, -1,          0,             /* punct */
293    cbit_space, -1,         -1,             /* space */    cbit_space, -1,          0,             /* space */
294    cbit_word,  -1,         -1,             /* word - a Perl extension */    cbit_word,  -1,          0,             /* word - a Perl extension */
295    cbit_xdigit,-1,         -1              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
296  };  };
297    
298    /* Table of substitutes for \d etc when PCRE_UCP is set. The POSIX class
299    substitutes must be in the order of the names, defined above, and there are
300    both positive and negative cases. NULL means no substitute. */
301    
302  /* The texts of compile-time error messages. These are "char *" because they  #ifdef SUPPORT_UCP
303  are passed to the outside world. */  static const pcre_uchar string_PNd[]  = {
304      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
305      CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
306    static const pcre_uchar string_pNd[]  = {
307      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
308      CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
309    static const pcre_uchar string_PXsp[] = {
310      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
311      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
312    static const pcre_uchar string_pXsp[] = {
313      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
314      CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
315    static const pcre_uchar string_PXwd[] = {
316      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
317      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
318    static const pcre_uchar string_pXwd[] = {
319      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
320      CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
321    
322    static const pcre_uchar *substitutes[] = {
323      string_PNd,           /* \D */
324      string_pNd,           /* \d */
325      string_PXsp,          /* \S */       /* NOTE: Xsp is Perl space */
326      string_pXsp,          /* \s */
327      string_PXwd,          /* \W */
328      string_pXwd           /* \w */
329    };
330    
331    static const pcre_uchar string_pL[] =   {
332      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
333      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
334    static const pcre_uchar string_pLl[] =  {
335      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
336      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
337    static const pcre_uchar string_pLu[] =  {
338      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
339      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
340    static const pcre_uchar string_pXan[] = {
341      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
342      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
343    static const pcre_uchar string_h[] =    {
344      CHAR_BACKSLASH, CHAR_h, '\0' };
345    static const pcre_uchar string_pXps[] = {
346      CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
347      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
348    static const pcre_uchar string_PL[] =   {
349      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
350      CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
351    static const pcre_uchar string_PLl[] =  {
352      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
353      CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
354    static const pcre_uchar string_PLu[] =  {
355      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
356      CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
357    static const pcre_uchar string_PXan[] = {
358      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
359      CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
360    static const pcre_uchar string_H[] =    {
361      CHAR_BACKSLASH, CHAR_H, '\0' };
362    static const pcre_uchar string_PXps[] = {
363      CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
364      CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
365    
366    static const pcre_uchar *posix_substitutes[] = {
367      string_pL,            /* alpha */
368      string_pLl,           /* lower */
369      string_pLu,           /* upper */
370      string_pXan,          /* alnum */
371      NULL,                 /* ascii */
372      string_h,             /* blank */
373      NULL,                 /* cntrl */
374      string_pNd,           /* digit */
375      NULL,                 /* graph */
376      NULL,                 /* print */
377      NULL,                 /* punct */
378      string_pXps,          /* space */    /* NOTE: Xps is POSIX space */
379      string_pXwd,          /* word */
380      NULL,                 /* xdigit */
381      /* Negated cases */
382      string_PL,            /* ^alpha */
383      string_PLl,           /* ^lower */
384      string_PLu,           /* ^upper */
385      string_PXan,          /* ^alnum */
386      NULL,                 /* ^ascii */
387      string_H,             /* ^blank */
388      NULL,                 /* ^cntrl */
389      string_PNd,           /* ^digit */
390      NULL,                 /* ^graph */
391      NULL,                 /* ^print */
392      NULL,                 /* ^punct */
393      string_PXps,          /* ^space */   /* NOTE: Xps is POSIX space */
394      string_PXwd,          /* ^word */
395      NULL                  /* ^xdigit */
396    };
397    #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))
398    #endif
399    
400  static const char *error_texts[] = {  #define STRING(a)  # a
401    "no error",  #define XSTRING(s) STRING(s)
402    "\\ at end of pattern",  
403    "\\c at end of pattern",  /* The texts of compile-time error messages. These are "char *" because they
404    "unrecognized character follows \\",  are passed to the outside world. Do not ever re-use any error number, because
405    "numbers out of order in {} quantifier",  they are documented. Always add a new error instead. Messages marked DEAD below
406    are no longer used. This used to be a table of strings, but in order to reduce
407    the number of relocations needed when a shared library is loaded dynamically,
408    it is now one long string. We cannot use a table of offsets, because the
409    lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
410    simply count through to the one we want - this isn't a performance issue
411    because these strings are used only when there is a compilation error.
412    
413    Each substring ends with \0 to insert a null character. This includes the final
414    substring, so that the whole string ends with \0\0, which can be detected when
415    counting through. */
416    
417    static const char error_texts[] =
418      "no error\0"
419      "\\ at end of pattern\0"
420      "\\c at end of pattern\0"
421      "unrecognized character follows \\\0"
422      "numbers out of order in {} quantifier\0"
423    /* 5 */    /* 5 */
424    "number too big in {} quantifier",    "number too big in {} quantifier\0"
425    "missing terminating ] for character class",    "missing terminating ] for character class\0"
426    "invalid escape sequence in character class",    "invalid escape sequence in character class\0"
427    "range out of order in character class",    "range out of order in character class\0"
428    "nothing to repeat",    "nothing to repeat\0"
429    /* 10 */    /* 10 */
430    "operand of unlimited repeat could match the empty string",    "operand of unlimited repeat could match the empty string\0"  /** DEAD **/
431    "internal error: unexpected repeat",    "internal error: unexpected repeat\0"
432    "unrecognized character after (?",    "unrecognized character after (? or (?-\0"
433    "POSIX named classes are supported only within a class",    "POSIX named classes are supported only within a class\0"
434    "missing )",    "missing )\0"
435    /* 15 */    /* 15 */
436    "reference to non-existent subpattern",    "reference to non-existent subpattern\0"
437    "erroffset passed as NULL",    "erroffset passed as NULL\0"
438    "unknown option bit(s) set",    "unknown option bit(s) set\0"
439    "missing ) after comment",    "missing ) after comment\0"
440    "parentheses nested too deeply",    "parentheses nested too deeply\0"  /** DEAD **/
441    /* 20 */    /* 20 */
442    "regular expression too large",    "regular expression is too large\0"
443    "failed to get memory",    "failed to get memory\0"
444    "unmatched parentheses",    "unmatched parentheses\0"
445    "internal error: code overflow",    "internal error: code overflow\0"
446    "unrecognized character after (?<",    "unrecognized character after (?<\0"
447    /* 25 */    /* 25 */
448    "lookbehind assertion is not fixed length",    "lookbehind assertion is not fixed length\0"
449    "malformed number after (?(",    "malformed number or name after (?(\0"
450    "conditional group contains more than two branches",    "conditional group contains more than two branches\0"
451    "assertion expected after (?(",    "assertion expected after (?(\0"
452    "(?R or (?digits must be followed by )",    "(?R or (?[+-]digits must be followed by )\0"
453    /* 30 */    /* 30 */
454    "unknown POSIX class name",    "unknown POSIX class name\0"
455    "POSIX collating elements are not supported",    "POSIX collating elements are not supported\0"
456    "this version of PCRE is not compiled with PCRE_UTF8 support",    "this version of PCRE is compiled without UTF support\0"
457    "spare error",    "spare error\0"  /** DEAD **/
458    "character value in \\x{...} sequence is too large",    "character value in \\x{...} sequence is too large\0"
459    /* 35 */    /* 35 */
460    "invalid condition (?(0)",    "invalid condition (?(0)\0"
461    "\\C not allowed in lookbehind assertion",    "\\C not allowed in lookbehind assertion\0"
462    "PCRE does not support \\L, \\l, \\N, \\U, or \\u",    "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
463    "number after (?C is > 255",    "number after (?C is > 255\0"
464    "closing ) for (?C expected",    "closing ) for (?C expected\0"
465    /* 40 */    /* 40 */
466    "recursive call could loop indefinitely",    "recursive call could loop indefinitely\0"
467    "unrecognized character after (?P",    "unrecognized character after (?P\0"
468    "syntax error after (?P",    "syntax error in subpattern name (missing terminator)\0"
469    "two named groups have the same name",    "two named subpatterns have the same name\0"
470    "invalid UTF-8 string",    "invalid UTF-8 string\0"
471    /* 45 */    /* 45 */
472    "support for \\P, \\p, and \\X has not been compiled",    "support for \\P, \\p, and \\X has not been compiled\0"
473    "malformed \\P or \\p sequence",    "malformed \\P or \\p sequence\0"
474    "unknown property name after \\P or \\p"    "unknown property name after \\P or \\p\0"
475  };    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
476      "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
477      /* 50 */
478      "repeated subpattern is too long\0"    /** DEAD **/
479      "octal value is greater than \\377 in 8-bit non-UTF-8 mode\0"
480      "internal error: overran compiling workspace\0"
481      "internal error: previously-checked referenced subpattern not found\0"
482      "DEFINE group contains more than one branch\0"
483      /* 55 */
484      "repeating a DEFINE group is not allowed\0"  /** DEAD **/
485      "inconsistent NEWLINE options\0"
486      "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
487      "a numbered reference must not be zero\0"
488      "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
489      /* 60 */
490      "(*VERB) not recognized\0"
491      "number is too big\0"
492      "subpattern name expected\0"
493      "digit expected after (?+\0"
494      "] is an invalid data character in JavaScript compatibility mode\0"
495      /* 65 */
496      "different names for subpatterns of the same number are not allowed\0"
497      "(*MARK) must have an argument\0"
498      "this version of PCRE is not compiled with Unicode property support\0"
499      "\\c must be followed by an ASCII character\0"
500      "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
501      /* 70 */
502      "internal error: unknown opcode in find_fixedlength()\0"
503      "\\N is not supported in a class\0"
504      "too many forward references\0"
505      "disallowed Unicode code point (>= 0xd800 && <= 0xdfff)\0"
506      "invalid UTF-16 string\0"
507      /* 75 */
508      "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"
509      "character value in \\u.... sequence is too large\0"
510      "invalid UTF-32 string\0"
511      ;
512    
513  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
514  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 229  For convenience, we use the same bit def Line 526  For convenience, we use the same bit def
526    
527  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
528    
529  #if !EBCDIC    /* This is the "normal" case, for ASCII systems */  /* Using a simple comparison for decimal numbers rather than a memory read
530  static const unsigned char digitab[] =  is much faster, and the resulting code is simpler (the compiler turns it
531    into a subtraction and unsigned comparison). */
532    
533    #define IS_DIGIT(x) ((x) >= CHAR_0 && (x) <= CHAR_9)
534    
535    #ifndef EBCDIC
536    
537    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
538    UTF-8 mode. */
539    
540    static const pcre_uint8 digitab[] =
541    {    {
542    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
543    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15 */
# Line 265  static const unsigned char digitab[] = Line 572  static const unsigned char digitab[] =
572    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
573    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
574    
575  #else          /* This is the "abnormal" case, for EBCDIC systems */  #else
576  static const unsigned char digitab[] =  
577    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
578    
579    static const pcre_uint8 digitab[] =
580    {    {
581    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
582    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15    */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   8- 15    */
# Line 279  static const unsigned char digitab[] = Line 589  static const unsigned char digitab[] =
589    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */
590    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */
591    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */
592    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88-     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88- 95    */
593    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */
594    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */
595    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */
# Line 301  static const unsigned char digitab[] = Line 611  static const unsigned char digitab[] =
611    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */    0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /*  0 - 7  F0 */
612    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */    0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/*  8 -255    */
613    
614  static const unsigned char ebcdic_chartab[] = { /* chartable partial dup */  static const pcre_uint8 ebcdic_chartab[] = { /* chartable partial dup */
615    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */    0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*   0-  7 */
616    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */    0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /*   8- 15 */
617    0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*  16- 23 */    0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /*  16- 23 */
# Line 313  static const unsigned char ebcdic_charta Line 623  static const unsigned char ebcdic_charta
623    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */
624    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */
625    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */
626    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88-  */    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88- 95 */
627    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */
628    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */
629    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */
# Line 337  static const unsigned char ebcdic_charta Line 647  static const unsigned char ebcdic_charta
647  #endif  #endif
648    
649    
650  /* Definition to allow mutual recursion */  
651    /*************************************************
652    *            Find an error text                  *
653    *************************************************/
654    
655    /* The error texts are now all in one long string, to save on relocations. As
656    some of the text is of unknown length, we can't use a table of offsets.
657    Instead, just count through the strings. This is not a performance issue
658    because it happens only when there has been a compilation error.
659    
660    Argument:   the error number
661    Returns:    pointer to the error string
662    */
663    
664    static const char *
665    find_error_text(int n)
666    {
667    const char *s = error_texts;
668    for (; n > 0; n--)
669      {
670      while (*s++ != CHAR_NULL) {};
671      if (*s == CHAR_NULL) return "Error text not found (please report)";
672      }
673    return s;
674    }
675    
676    
677    /*************************************************
678    *           Expand the workspace                 *
679    *************************************************/
680    
681    /* This function is called during the second compiling phase, if the number of
682    forward references fills the existing workspace, which is originally a block on
683    the stack. A larger block is obtained from malloc() unless the ultimate limit
684    has been reached or the increase will be rather small.
685    
686    Argument: pointer to the compile data block
687    Returns:  0 if all went well, else an error number
688    */
689    
690    static int
691    expand_workspace(compile_data *cd)
692    {
693    pcre_uchar *newspace;
694    int newsize = cd->workspace_size * 2;
695    
696    if (newsize > COMPILE_WORK_SIZE_MAX) newsize = COMPILE_WORK_SIZE_MAX;
697    if (cd->workspace_size >= COMPILE_WORK_SIZE_MAX ||
698        newsize - cd->workspace_size < WORK_SIZE_SAFETY_MARGIN)
699     return ERR72;
700    
701    newspace = (PUBL(malloc))(IN_UCHARS(newsize));
702    if (newspace == NULL) return ERR21;
703    memcpy(newspace, cd->start_workspace, cd->workspace_size * sizeof(pcre_uchar));
704    cd->hwm = (pcre_uchar *)newspace + (cd->hwm - cd->start_workspace);
705    if (cd->workspace_size > COMPILE_WORK_SIZE)
706      (PUBL(free))((void *)cd->start_workspace);
707    cd->start_workspace = newspace;
708    cd->workspace_size = newsize;
709    return 0;
710    }
711    
712    
713    
714    /*************************************************
715    *            Check for counted repeat            *
716    *************************************************/
717    
718    /* This function is called when a '{' is encountered in a place where it might
719    start a quantifier. It looks ahead to see if it really is a quantifier or not.
720    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
721    where the ddds are digits.
722    
723    Arguments:
724      p         pointer to the first char after '{'
725    
726    Returns:    TRUE or FALSE
727    */
728    
729  static BOOL  static BOOL
730    compile_regex(int, int, int *, uschar **, const uschar **, int *, BOOL, int,  is_counted_repeat(const pcre_uchar *p)
731      int *, int *, branch_chain *, compile_data *);  {
732    if (!IS_DIGIT(*p)) return FALSE;
733    p++;
734    while (IS_DIGIT(*p)) p++;
735    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
736    
737    if (*p++ != CHAR_COMMA) return FALSE;
738    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
739    
740    if (!IS_DIGIT(*p)) return FALSE;
741    p++;
742    while (IS_DIGIT(*p)) p++;
743    
744    return (*p == CHAR_RIGHT_CURLY_BRACKET);
745    }
746    
747    
748    
# Line 350  static BOOL Line 751  static BOOL
751  *************************************************/  *************************************************/
752    
753  /* 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
754  positive value for a simple escape such as \n, or a negative value which  positive value for a simple escape such as \n, or 0 for a data character
755  encodes one of the more complicated things such as \d. When UTF-8 is enabled,  which will be placed in chptr. A backreference to group n is returned as
756  a positive value greater than 255 may be returned. On entry, ptr is pointing at  negative n. When UTF-8 is enabled, a positive value greater than 255 may
757  the \. On exit, it is on the final character of the escape sequence.  be returned in chptr.
758    On entry,ptr is pointing at the \. On exit, it is on the final character of the
759    escape sequence.
760    
761  Arguments:  Arguments:
762    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
763      chptr          points to the data character
764    errorcodeptr   points to the errorcode variable    errorcodeptr   points to the errorcode variable
765    bracount       number of previous extracting brackets    bracount       number of previous extracting brackets
766    options        the options bits    options        the options bits
767    isclass        TRUE if inside a character class    isclass        TRUE if inside a character class
768    
769  Returns:         zero or positive => a data character  Returns:         zero => a data character
770                   negative => a special escape sequence                   positive => a special escape sequence
771                   on error, errorptr is set                   negative => a back reference
772                     on error, errorcodeptr is set
773  */  */
774    
775  static int  static int
776  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,
777    int options, BOOL isclass)    int bracount, int options, BOOL isclass)
778  {  {
779  const uschar *ptr = *ptrptr;  /* PCRE_UTF16 has the same value as PCRE_UTF8. */
780  int c, i;  BOOL utf = (options & PCRE_UTF8) != 0;
781    const pcre_uchar *ptr = *ptrptr + 1;
782    pcre_uint32 c;
783    int escape = 0;
784    int i;
785    
786    GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
787    ptr--;                            /* Set pointer back to the last byte */
788    
789  /* 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. */
790    
791  c = *(++ptr);  if (c == CHAR_NULL) *errorcodeptr = ERR1;
 if (c == 0) *errorcodeptr = ERR1;  
792    
793  /* 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
794  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.
795  Otherwise further processing may be required. */  Otherwise further processing may be required. */
796    
797  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
798  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  /* Not alphanumeric */
799  else if ((i = escapes[c - '0']) != 0) c = i;  else if (c < CHAR_0 || c > CHAR_z) {}
800    else if ((i = escapes[c - CHAR_0]) != 0) { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
801  #else          /* EBCDIC coding */  
802  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  #else           /* EBCDIC coding */
803  else if ((i = escapes[c - 0x48]) != 0)  c = i;  /* Not alphanumeric */
804    else if (c < CHAR_a || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}
805    else if ((i = escapes[c - 0x48]) != 0)  { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
806  #endif  #endif
807    
808  /* Escapes that need further processing, or are illegal. */  /* Escapes that need further processing, or are illegal. */
809    
810  else  else
811    {    {
812    const uschar *oldptr;    const pcre_uchar *oldptr;
813      BOOL braced, negated, overflow;
814      int s;
815    
816    switch (c)    switch (c)
817      {      {
818      /* 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
819      error. */      error. */
820    
821      case 'l':      case CHAR_l:
822      case 'L':      case CHAR_L:
     case 'N':  
     case 'u':  
     case 'U':  
823      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
824      break;      break;
825    
826        case CHAR_u:
827        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
828          {
829          /* In JavaScript, \u must be followed by four hexadecimal numbers.
830          Otherwise it is a lowercase u letter. */
831          if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
832            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0
833            && MAX_255(ptr[3]) && (digitab[ptr[3]] & ctype_xdigit) != 0
834            && MAX_255(ptr[4]) && (digitab[ptr[4]] & ctype_xdigit) != 0)
835            {
836            c = 0;
837            for (i = 0; i < 4; ++i)
838              {
839              register pcre_uint32 cc = *(++ptr);
840    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
841              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
842              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
843    #else           /* EBCDIC coding */
844              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
845              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
846    #endif
847              }
848    
849    #if defined COMPILE_PCRE8
850            if (c > (utf ? 0x10ffff : 0xff))
851    #elif defined COMPILE_PCRE16
852            if (c > (utf ? 0x10ffff : 0xffff))
853    #elif defined COMPILE_PCRE32
854            if (utf && c > 0x10ffff)
855    #endif
856              {
857              *errorcodeptr = ERR76;
858              }
859            else if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
860            }
861          }
862        else
863          *errorcodeptr = ERR37;
864        break;
865    
866        case CHAR_U:
867        /* In JavaScript, \U is an uppercase U letter. */
868        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
869        break;
870    
871        /* In a character class, \g is just a literal "g". Outside a character
872        class, \g must be followed by one of a number of specific things:
873    
874        (1) A number, either plain or braced. If positive, it is an absolute
875        backreference. If negative, it is a relative backreference. This is a Perl
876        5.10 feature.
877    
878        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
879        is part of Perl's movement towards a unified syntax for back references. As
880        this is synonymous with \k{name}, we fudge it up by pretending it really
881        was \k.
882    
883        (3) For Oniguruma compatibility we also support \g followed by a name or a
884        number either in angle brackets or in single quotes. However, these are
885        (possibly recursive) subroutine calls, _not_ backreferences. Just return
886        the ESC_g code (cf \k). */
887    
888        case CHAR_g:
889        if (isclass) break;
890        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
891          {
892          escape = ESC_g;
893          break;
894          }
895    
896        /* Handle the Perl-compatible cases */
897    
898        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
899          {
900          const pcre_uchar *p;
901          for (p = ptr+2; *p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
902            if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
903          if (*p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET)
904            {
905            escape = ESC_k;
906            break;
907            }
908          braced = TRUE;
909          ptr++;
910          }
911        else braced = FALSE;
912    
913        if (ptr[1] == CHAR_MINUS)
914          {
915          negated = TRUE;
916          ptr++;
917          }
918        else negated = FALSE;
919    
920        /* The integer range is limited by the machine's int representation. */
921        s = 0;
922        overflow = FALSE;
923        while (IS_DIGIT(ptr[1]))
924          {
925          if (s > INT_MAX / 10 - 1) /* Integer overflow */
926            {
927            overflow = TRUE;
928            break;
929            }
930          s = s * 10 + (int)(*(++ptr) - CHAR_0);
931          }
932        if (overflow) /* Integer overflow */
933          {
934          while (IS_DIGIT(ptr[1]))
935            ptr++;
936          *errorcodeptr = ERR61;
937          break;
938          }
939    
940        if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
941          {
942          *errorcodeptr = ERR57;
943          break;
944          }
945    
946        if (s == 0)
947          {
948          *errorcodeptr = ERR58;
949          break;
950          }
951    
952        if (negated)
953          {
954          if (s > bracount)
955            {
956            *errorcodeptr = ERR15;
957            break;
958            }
959          s = bracount - (s - 1);
960          }
961    
962        escape = -s;
963        break;
964    
965      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
966      starting with one that is not zero is not straightforward. By experiment,      starting with one that is not zero is not straightforward. By experiment,
967      the way Perl works seems to be as follows:      the way Perl works seems to be as follows:
# Line 422  else Line 974  else
974      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
975      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
976    
977      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:
978      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
979    
980      if (!isclass)      if (!isclass)
981        {        {
982        oldptr = ptr;        oldptr = ptr;
983        c -= '0';        /* The integer range is limited by the machine's int representation. */
984        while ((digitab[ptr[1]] & ctype_digit) != 0)        s = (int)(c -CHAR_0);
985          c = c * 10 + *(++ptr) - '0';        overflow = FALSE;
986        if (c < 10 || c <= bracount)        while (IS_DIGIT(ptr[1]))
987            {
988            if (s > INT_MAX / 10 - 1) /* Integer overflow */
989              {
990              overflow = TRUE;
991              break;
992              }
993            s = s * 10 + (int)(*(++ptr) - CHAR_0);
994            }
995          if (overflow) /* Integer overflow */
996            {
997            while (IS_DIGIT(ptr[1]))
998              ptr++;
999            *errorcodeptr = ERR61;
1000            break;
1001            }
1002          if (s < 10 || s <= bracount)
1003          {          {
1004          c = -(ESC_REF + c);          escape = -s;
1005          break;          break;
1006          }          }
1007        ptr = oldptr;      /* Put the pointer back and fall through */        ptr = oldptr;      /* Put the pointer back and fall through */
# Line 443  else Line 1011  else
1011      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.
1012      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
1013    
1014      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
1015        {        {
1016        ptr--;        ptr--;
1017        c = 0;        c = 0;
# Line 451  else Line 1019  else
1019        }        }
1020    
1021      /* \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
1022      larger first octal digit. */      larger first octal digit. The original code used just to take the least
1023        significant 8 bits of octal numbers (I think this is what early Perls used
1024      case '0':      to do). Nowadays we allow for larger numbers in UTF-8 mode and 16-bit mode,
1025      c -= '0';      but no more than 3 octal digits. */
1026      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')  
1027          c = c * 8 + *(++ptr) - '0';      case CHAR_0:
1028      c &= 255;     /* Take least significant 8 bits */      c -= CHAR_0;
1029        while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
1030            c = c * 8 + *(++ptr) - CHAR_0;
1031    #ifdef COMPILE_PCRE8
1032        if (!utf && c > 0xff) *errorcodeptr = ERR51;
1033    #endif
1034      break;      break;
1035    
1036      /* \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
1037      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.
1038        If not, { is treated as a data character. */
1039    
1040        case CHAR_x:
1041        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1042          {
1043          /* In JavaScript, \x must be followed by two hexadecimal numbers.
1044          Otherwise it is a lowercase x letter. */
1045          if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1046            && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1047            {
1048            c = 0;
1049            for (i = 0; i < 2; ++i)
1050              {
1051              register pcre_uint32 cc = *(++ptr);
1052    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1053              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1054              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1055    #else           /* EBCDIC coding */
1056              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1057              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1058    #endif
1059              }
1060            }
1061          break;
1062          }
1063    
1064      case 'x':      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
 #ifdef SUPPORT_UTF8  
     if (ptr[1] == '{' && (options & PCRE_UTF8) != 0)  
1065        {        {
1066        const uschar *pt = ptr + 2;        const pcre_uchar *pt = ptr + 2;
1067        register int count = 0;  
1068        c = 0;        c = 0;
1069        while ((digitab[*pt] & ctype_xdigit) != 0)        overflow = FALSE;
1070          while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0)
1071          {          {
1072          int cc = *pt++;          register pcre_uint32 cc = *pt++;
1073          count++;          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1074  #if !EBCDIC    /* ASCII coding */  
1075          if (cc >= 'a') cc -= 32;               /* Convert to upper case */  #ifdef COMPILE_PCRE32
1076          c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));          if (c >= 0x10000000l) { overflow = TRUE; break; }
1077  #else          /* EBCDIC coding */  #endif
1078          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */  
1079          c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1080            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1081            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1082    #else           /* EBCDIC coding */
1083            if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1084            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1085    #endif
1086    
1087    #if defined COMPILE_PCRE8
1088            if (c > (utf ? 0x10ffff : 0xff)) { overflow = TRUE; break; }
1089    #elif defined COMPILE_PCRE16
1090            if (c > (utf ? 0x10ffff : 0xffff)) { overflow = TRUE; break; }
1091    #elif defined COMPILE_PCRE32
1092            if (utf && c > 0x10ffff) { overflow = TRUE; break; }
1093  #endif  #endif
1094          }          }
1095        if (*pt == '}')  
1096          if (overflow)
1097            {
1098            while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0) pt++;
1099            *errorcodeptr = ERR34;
1100            }
1101    
1102          if (*pt == CHAR_RIGHT_CURLY_BRACKET)
1103          {          {
1104          if (c < 0 || count > 8) *errorcodeptr = ERR34;          if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1105          ptr = pt;          ptr = pt;
1106          break;          break;
1107          }          }
1108    
1109        /* 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
1110        recognize this construct; fall through to the normal \x handling. */        recognize this construct; fall through to the normal \x handling. */
1111        }        }
 #endif  
1112    
1113      /* Read just a single hex char */      /* Read just a single-byte hex-defined char */
1114    
1115      c = 0;      c = 0;
1116      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1117        {        {
1118        int cc;                               /* Some compilers don't like ++ */        pcre_uint32 cc;                          /* Some compilers don't like */
1119        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
1120  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1121        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1122        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1123  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
1124        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1125        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1126  #endif  #endif
1127        }        }
1128      break;      break;
1129    
1130      /* Other special escapes not starting with a digit are straightforward */      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
1131        An error is given if the byte following \c is not an ASCII character. This
1132        coding is ASCII-specific, but then the whole concept of \cx is
1133        ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
1134    
1135      case 'c':      case CHAR_c:
1136      c = *(++ptr);      c = *(++ptr);
1137      if (c == 0)      if (c == CHAR_NULL)
1138        {        {
1139        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
1140        return 0;        break;
1141        }        }
1142    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
1143      /* A letter is upper-cased; then the 0x40 bit is flipped. This coding      if (c > 127)  /* Excludes all non-ASCII in either mode */
1144      is ASCII-specific, but then the whole concept of \cx is ASCII-specific.        {
1145      (However, an EBCDIC equivalent has now been added.) */        *errorcodeptr = ERR68;
1146          break;
1147  #if !EBCDIC    /* ASCII coding */        }
1148      if (c >= 'a' && c <= 'z') c -= 32;      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
1149      c ^= 0x40;      c ^= 0x40;
1150  #else          /* EBCDIC coding */  #else             /* EBCDIC coding */
1151      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
1152      c ^= 0xC0;      c ^= 0xC0;
1153  #endif  #endif
1154      break;      break;
1155    
1156      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
1157      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
1158      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
1159      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
1160      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
1161    
1162      default:      default:
1163      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 550  else Line 1170  else
1170      }      }
1171    }    }
1172    
1173  *ptrptr = ptr;  /* Perl supports \N{name} for character names, as well as plain \N for "not
1174  return c;  newline". PCRE does not support \N{name}. However, it does support
1175  }  quantification such as \N{2,3}. */
1176    
1177    if (escape == ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1178         !is_counted_repeat(ptr+2))
1179      *errorcodeptr = ERR37;
1180    
1181    /* If PCRE_UCP is set, we change the values for \d etc. */
1182    
1183    if ((options & PCRE_UCP) != 0 && escape >= ESC_D && escape <= ESC_w)
1184      escape += (ESC_DU - ESC_D);
1185    
1186    /* Set the pointer to the final character before returning. */
1187    
1188    *ptrptr = ptr;
1189    *chptr = c;
1190    return escape;
1191    }
1192    
1193  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
1194  /*************************************************  /*************************************************
# Line 569  escape sequence. Line 1203  escape sequence.
1203  Argument:  Argument:
1204    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
1205    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
1206      ptypeptr       points to an unsigned int that is set to the type value
1207      pdataptr       points to an unsigned int that is set to the detailed property value
1208    errorcodeptr   points to the error code variable    errorcodeptr   points to the error code variable
1209    
1210  Returns:     value from ucp_type_table, or -1 for an invalid type  Returns:         TRUE if the type value was found, or FALSE for an invalid type
1211  */  */
1212    
1213  static int  static BOOL
1214  get_ucp(const uschar **ptrptr, BOOL *negptr, int *errorcodeptr)  get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, unsigned int *ptypeptr,
1215      unsigned int *pdataptr, int *errorcodeptr)
1216  {  {
1217  int c, i, bot, top;  pcre_uchar c;
1218  const uschar *ptr = *ptrptr;  int i, bot, top;
1219  char name[4];  const pcre_uchar *ptr = *ptrptr;
1220    pcre_uchar name[32];
1221    
1222  c = *(++ptr);  c = *(++ptr);
1223  if (c == 0) goto ERROR_RETURN;  if (c == CHAR_NULL) goto ERROR_RETURN;
1224    
1225  *negptr = FALSE;  *negptr = FALSE;
1226    
1227  /* \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
1228  preceded by ^ for negation. */  negation. */
1229    
1230  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
1231    {    {
1232    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1233      {      {
1234      *negptr = TRUE;      *negptr = TRUE;
1235      ptr++;      ptr++;
1236      }      }
1237    for (i = 0; i <= 2; i++)    for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1238      {      {
1239      c = *(++ptr);      c = *(++ptr);
1240      if (c == 0) goto ERROR_RETURN;      if (c == CHAR_NULL) goto ERROR_RETURN;
1241      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1242      name[i] = c;      name[i] = c;
1243      }      }
1244    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;  
     }  
1245    name[i] = 0;    name[i] = 0;
1246    }    }
1247    
# Line 624  else Line 1258  else
1258  /* Search for a recognized property name using binary chop */  /* Search for a recognized property name using binary chop */
1259    
1260  bot = 0;  bot = 0;
1261  top = _pcre_utt_size;  top = PRIV(utt_size);
1262    
1263  while (bot < top)  while (bot < top)
1264    {    {
1265    i = (bot + top)/2;    int r;
1266    c = strcmp(name, _pcre_utt[i].name);    i = (bot + top) >> 1;
1267    if (c == 0) return _pcre_utt[i].value;    r = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1268    if (c > 0) bot = i + 1; else top = i;    if (r == 0)
1269        {
1270        *ptypeptr = PRIV(utt)[i].type;
1271        *pdataptr = PRIV(utt)[i].value;
1272        return TRUE;
1273        }
1274      if (r > 0) bot = i + 1; else top = i;
1275    }    }
1276    
 UNKNOWN_RETURN:  
1277  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
1278  *ptrptr = ptr;  *ptrptr = ptr;
1279  return -1;  return FALSE;
1280    
1281  ERROR_RETURN:  ERROR_RETURN:
1282  *errorcodeptr = ERR46;  *errorcodeptr = ERR46;
1283  *ptrptr = ptr;  *ptrptr = ptr;
1284  return -1;  return FALSE;
1285  }  }
1286  #endif  #endif
1287    
# Line 650  return -1; Line 1289  return -1;
1289    
1290    
1291  /*************************************************  /*************************************************
 *            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 == '}');  
 }  
   
   
   
 /*************************************************  
1292  *         Read repeat counts                     *  *         Read repeat counts                     *
1293  *************************************************/  *************************************************/
1294    
# Line 701  Returns:         pointer to '}' on succe Line 1307  Returns:         pointer to '}' on succe
1307                   current ptr on error, with errorcodeptr set non-zero                   current ptr on error, with errorcodeptr set non-zero
1308  */  */
1309    
1310  static const uschar *  static const pcre_uchar *
1311  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)
1312  {  {
1313  int min = 0;  int min = 0;
1314  int max = -1;  int max = -1;
# Line 710  int max = -1; Line 1316  int max = -1;
1316  /* Read the minimum value and do a paranoid check: a negative value indicates  /* Read the minimum value and do a paranoid check: a negative value indicates
1317  an integer overflow. */  an integer overflow. */
1318    
1319  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while (IS_DIGIT(*p)) min = min * 10 + (int)(*p++ - CHAR_0);
1320  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1321    {    {
1322    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 720  if (min < 0 || min > 65535) Line 1326  if (min < 0 || min > 65535)
1326  /* Read the maximum value if there is one, and again do a paranoid on its size.  /* Read the maximum value if there is one, and again do a paranoid on its size.
1327  Also, max must not be less than min. */  Also, max must not be less than min. */
1328    
1329  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1330    {    {
1331    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1332      {      {
1333      max = 0;      max = 0;
1334      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while(IS_DIGIT(*p)) max = max * 10 + (int)(*p++ - CHAR_0);
1335      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1336        {        {
1337        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 750  return p; Line 1356  return p;
1356    
1357    
1358  /*************************************************  /*************************************************
1359  *      Find first significant op code            *  *  Subroutine for finding forward reference      *
1360  *************************************************/  *************************************************/
1361    
1362  /* This is called by several functions that scan a compiled expression looking  /* This recursive function is called only from find_parens() below. The
1363  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
1364  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
1365  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
1366  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
1367    returns when it reaches a given numbered subpattern. Recursion is used to keep
1368    track of subpatterns that reset the capturing group numbers - the (?| feature.
1369    
1370    This function was originally called only from the second pass, in which we know
1371    that if (?< or (?' or (?P< is encountered, the name will be correctly
1372    terminated because that is checked in the first pass. There is now one call to
1373    this function in the first pass, to check for a recursive back reference by
1374    name (so that we can make the whole group atomic). In this case, we need check
1375    only up to the current position in the pattern, and that is still OK because
1376    and previous occurrences will have been checked. To make this work, the test
1377    for "end of pattern" is a check against cd->end_pattern in the main loop,
1378    instead of looking for a binary zero. This means that the special first-pass
1379    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1380    processing items within the loop are OK, because afterwards the main loop will
1381    terminate.)
1382    
1383  Arguments:  Arguments:
1384    code         pointer to the start of the group    ptrptr       address of the current character pointer (updated)
1385    options      pointer to external options    cd           compile background data
1386    optbit       the option bit whose changing is significant, or    name         name to seek, or NULL if seeking a numbered subpattern
1387                   zero if none are    lorn         name length, or subpattern number if name is NULL
1388    skipassert   TRUE if certain assertions are to be skipped    xmode        TRUE if we are in /x mode
1389      utf          TRUE if we are in UTF-8 / UTF-16 / UTF-32 mode
1390      count        pointer to the current capturing subpattern number (updated)
1391    
1392  Returns:       pointer to the first significant opcode  Returns:       the number of the named subpattern, or -1 if not found
1393  */  */
1394    
1395  static const uschar*  static int
1396  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,
1397    BOOL skipassert)    BOOL xmode, BOOL utf, int *count)
1398  {  {
1399  for (;;)  pcre_uchar *ptr = *ptrptr;
1400    {  int start_count = *count;
1401    switch ((int)*code)  int hwm_count = start_count;
1402      {  BOOL dup_parens = FALSE;
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
1403    
1404      case OP_ASSERT_NOT:  /* If the first character is a parenthesis, check on the type of group we are
1405      case OP_ASSERTBACK:  dealing with. The very first call may not start with a parenthesis. */
     case OP_ASSERTBACK_NOT:  
     if (!skipassert) return code;  
     do code += GET(code, 1); while (*code == OP_ALT);  
     code += _pcre_OP_lengths[*code];  
     break;  
1406    
1407      case OP_WORD_BOUNDARY:  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1408      case OP_NOT_WORD_BOUNDARY:    {
1409      if (!skipassert) return code;    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
     /* Fall through */  
1410    
1411      case OP_CALLOUT:    if (ptr[1] == CHAR_ASTERISK) ptr += 2;
     case OP_CREF:  
     case OP_BRANUMBER:  
     code += _pcre_OP_lengths[*code];  
     break;  
1412    
1413      default:    /* Handle a normal, unnamed capturing parenthesis. */
1414      return code;  
1415      else if (ptr[1] != CHAR_QUESTION_MARK)
1416        {
1417        *count += 1;
1418        if (name == NULL && *count == lorn) return *count;
1419        ptr++;
1420      }      }
1421    }  
1422  /* Control never reaches here */    /* All cases now have (? at the start. Remember when we are in a group
1423  }    where the parenthesis numbers are duplicated. */
1424    
1425      else if (ptr[2] == CHAR_VERTICAL_LINE)
1426        {
1427        ptr += 3;
1428        dup_parens = TRUE;
1429        }
1430    
1431      /* Handle comments; all characters are allowed until a ket is reached. */
1432    
1433      else if (ptr[2] == CHAR_NUMBER_SIGN)
1434        {
1435        for (ptr += 3; *ptr != CHAR_NULL; ptr++)
1436          if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1437        goto FAIL_EXIT;
1438        }
1439    
1440      /* Handle a condition. If it is an assertion, just carry on so that it
1441      is processed as normal. If not, skip to the closing parenthesis of the
1442      condition (there can't be any nested parens). */
1443    
1444      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1445        {
1446        ptr += 2;
1447        if (ptr[1] != CHAR_QUESTION_MARK)
1448          {
1449          while (*ptr != CHAR_NULL && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1450          if (*ptr != CHAR_NULL) ptr++;
1451          }
1452        }
1453    
1454      /* Start with (? but not a condition. */
1455    
1456      else
1457        {
1458        ptr += 2;
1459        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1460    
1461        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1462    
1463        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1464            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1465          {
1466          pcre_uchar term;
1467          const pcre_uchar *thisname;
1468          *count += 1;
1469          if (name == NULL && *count == lorn) return *count;
1470          term = *ptr++;
1471          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1472          thisname = ptr;
1473          while (*ptr != term) ptr++;
1474          if (name != NULL && lorn == (int)(ptr - thisname) &&
1475              STRNCMP_UC_UC(name, thisname, (unsigned int)lorn) == 0)
1476            return *count;
1477          term++;
1478          }
1479        }
1480      }
1481    
1482    /* Past any initial parenthesis handling, scan for parentheses or vertical
1483    bars. Stop if we get to cd->end_pattern. Note that this is important for the
1484    first-pass call when this value is temporarily adjusted to stop at the current
1485    position. So DO NOT change this to a test for binary zero. */
1486    
1487    for (; ptr < cd->end_pattern; ptr++)
1488      {
1489      /* Skip over backslashed characters and also entire \Q...\E */
1490    
1491      if (*ptr == CHAR_BACKSLASH)
1492        {
1493        if (*(++ptr) == CHAR_NULL) goto FAIL_EXIT;
1494        if (*ptr == CHAR_Q) for (;;)
1495          {
1496          while (*(++ptr) != CHAR_NULL && *ptr != CHAR_BACKSLASH) {};
1497          if (*ptr == CHAR_NULL) goto FAIL_EXIT;
1498          if (*(++ptr) == CHAR_E) break;
1499          }
1500        continue;
1501        }
1502    
1503      /* Skip over character classes; this logic must be similar to the way they
1504      are handled for real. If the first character is '^', skip it. Also, if the
1505      first few characters (either before or after ^) are \Q\E or \E we skip them
1506      too. This makes for compatibility with Perl. Note the use of STR macros to
1507      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1508    
1509      if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1510        {
1511        BOOL negate_class = FALSE;
1512        for (;;)
1513          {
1514          if (ptr[1] == CHAR_BACKSLASH)
1515            {
1516            if (ptr[2] == CHAR_E)
1517              ptr+= 2;
1518            else if (STRNCMP_UC_C8(ptr + 2,
1519                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1520              ptr += 4;
1521            else
1522              break;
1523            }
1524          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1525            {
1526            negate_class = TRUE;
1527            ptr++;
1528            }
1529          else break;
1530          }
1531    
1532        /* If the next character is ']', it is a data character that must be
1533        skipped, except in JavaScript compatibility mode. */
1534    
1535        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1536            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1537          ptr++;
1538    
1539        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1540          {
1541          if (*ptr == CHAR_NULL) return -1;
1542          if (*ptr == CHAR_BACKSLASH)
1543            {
1544            if (*(++ptr) == CHAR_NULL) goto FAIL_EXIT;
1545            if (*ptr == CHAR_Q) for (;;)
1546              {
1547              while (*(++ptr) != CHAR_NULL && *ptr != CHAR_BACKSLASH) {};
1548              if (*ptr == CHAR_NULL) goto FAIL_EXIT;
1549              if (*(++ptr) == CHAR_E) break;
1550              }
1551            continue;
1552            }
1553          }
1554        continue;
1555        }
1556    
1557      /* Skip comments in /x mode */
1558    
1559      if (xmode && *ptr == CHAR_NUMBER_SIGN)
1560        {
1561        ptr++;
1562        while (*ptr != CHAR_NULL)
1563          {
1564          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1565          ptr++;
1566    #ifdef SUPPORT_UTF
1567          if (utf) FORWARDCHAR(ptr);
1568    #endif
1569          }
1570        if (*ptr == CHAR_NULL) goto FAIL_EXIT;
1571        continue;
1572        }
1573    
1574      /* Check for the special metacharacters */
1575    
1576      if (*ptr == CHAR_LEFT_PARENTHESIS)
1577        {
1578        int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, count);
1579        if (rc > 0) return rc;
1580        if (*ptr == CHAR_NULL) goto FAIL_EXIT;
1581        }
1582    
1583      else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1584        {
1585        if (dup_parens && *count < hwm_count) *count = hwm_count;
1586        goto FAIL_EXIT;
1587        }
1588    
1589      else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1590        {
1591        if (*count > hwm_count) hwm_count = *count;
1592        *count = start_count;
1593        }
1594      }
1595    
1596    FAIL_EXIT:
1597    *ptrptr = ptr;
1598    return -1;
1599    }
1600    
1601    
1602    
1603    
1604    /*************************************************
1605    *       Find forward referenced subpattern       *
1606    *************************************************/
1607    
1608    /* This function scans along a pattern's text looking for capturing
1609    subpatterns, and counting them. If it finds a named pattern that matches the
1610    name it is given, it returns its number. Alternatively, if the name is NULL, it
1611    returns when it reaches a given numbered subpattern. This is used for forward
1612    references to subpatterns. We used to be able to start this scan from the
1613    current compiling point, using the current count value from cd->bracount, and
1614    do it all in a single loop, but the addition of the possibility of duplicate
1615    subpattern numbers means that we have to scan from the very start, in order to
1616    take account of such duplicates, and to use a recursive function to keep track
1617    of the different types of group.
1618    
1619    Arguments:
1620      cd           compile background data
1621      name         name to seek, or NULL if seeking a numbered subpattern
1622      lorn         name length, or subpattern number if name is NULL
1623      xmode        TRUE if we are in /x mode
1624      utf          TRUE if we are in UTF-8 / UTF-16 / UTF-32 mode
1625    
1626    Returns:       the number of the found subpattern, or -1 if not found
1627    */
1628    
1629    static int
1630    find_parens(compile_data *cd, const pcre_uchar *name, int lorn, BOOL xmode,
1631      BOOL utf)
1632    {
1633    pcre_uchar *ptr = (pcre_uchar *)cd->start_pattern;
1634    int count = 0;
1635    int rc;
1636    
1637    /* If the pattern does not start with an opening parenthesis, the first call
1638    to find_parens_sub() will scan right to the end (if necessary). However, if it
1639    does start with a parenthesis, find_parens_sub() will return when it hits the
1640    matching closing parens. That is why we have to have a loop. */
1641    
1642    for (;;)
1643      {
1644      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf, &count);
1645      if (rc > 0 || *ptr++ == CHAR_NULL) break;
1646      }
1647    
1648    return rc;
1649    }
1650    
1651    
1652    
1653    
1654    /*************************************************
1655    *      Find first significant op code            *
1656    *************************************************/
1657    
1658    /* This is called by several functions that scan a compiled expression looking
1659    for a fixed first character, or an anchoring op code etc. It skips over things
1660    that do not influence this. For some calls, it makes sense to skip negative
1661    forward and all backward assertions, and also the \b assertion; for others it
1662    does not.
1663    
1664    Arguments:
1665      code         pointer to the start of the group
1666      skipassert   TRUE if certain assertions are to be skipped
1667    
1668    Returns:       pointer to the first significant opcode
1669    */
1670    
1671    static const pcre_uchar*
1672    first_significant_code(const pcre_uchar *code, BOOL skipassert)
1673    {
1674    for (;;)
1675      {
1676      switch ((int)*code)
1677        {
1678        case OP_ASSERT_NOT:
1679        case OP_ASSERTBACK:
1680        case OP_ASSERTBACK_NOT:
1681        if (!skipassert) return code;
1682        do code += GET(code, 1); while (*code == OP_ALT);
1683        code += PRIV(OP_lengths)[*code];
1684        break;
1685    
1686        case OP_WORD_BOUNDARY:
1687        case OP_NOT_WORD_BOUNDARY:
1688        if (!skipassert) return code;
1689        /* Fall through */
1690    
1691        case OP_CALLOUT:
1692        case OP_CREF:
1693        case OP_NCREF:
1694        case OP_RREF:
1695        case OP_NRREF:
1696        case OP_DEF:
1697        code += PRIV(OP_lengths)[*code];
1698        break;
1699    
1700        default:
1701        return code;
1702        }
1703      }
1704    /* Control never reaches here */
1705    }
1706    
1707    
1708    
1709    
1710  /*************************************************  /*************************************************
1711  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1712  *************************************************/  *************************************************/
1713    
1714  /* 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,
1715  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.
1716  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
1717    temporarily terminated with OP_END when this function is called.
1718    
1719    This function is called when a backward assertion is encountered, so that if it
1720    fails, the error message can point to the correct place in the pattern.
1721    However, we cannot do this when the assertion contains subroutine calls,
1722    because they can be forward references. We solve this by remembering this case
1723    and doing the check at the end; a flag specifies which mode we are running in.
1724    
1725  Arguments:  Arguments:
1726    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1727    options  the compiling options    utf      TRUE in UTF-8 / UTF-16 / UTF-32 mode
1728      atend    TRUE if called when the pattern is complete
1729  Returns:   the fixed length, or -1 if there is no fixed length,    cd       the "compile data" structure
1730               or -2 if \C was encountered  
1731    Returns:   the fixed length,
1732                 or -1 if there is no fixed length,
1733                 or -2 if \C was encountered (in UTF-8 mode only)
1734                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1735                 or -4 if an unknown opcode was encountered (internal error)
1736  */  */
1737    
1738  static int  static int
1739  find_fixedlength(uschar *code, int options)  find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd)
1740  {  {
1741  int length = -1;  int length = -1;
1742    
1743  register int branchlength = 0;  register int branchlength = 0;
1744  register uschar *cc = code + 1 + LINK_SIZE;  register pcre_uchar *cc = code + 1 + LINK_SIZE;
1745    
1746  /* 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
1747  branch, check the length against that of the other branches. */  branch, check the length against that of the other branches. */
# Line 842  branch, check the length against that of Line 1749  branch, check the length against that of
1749  for (;;)  for (;;)
1750    {    {
1751    int d;    int d;
1752    register int op = *cc;    pcre_uchar *ce, *cs;
1753    if (op >= OP_BRA) op = OP_BRA;    register pcre_uchar op = *cc;
1754    
1755    switch (op)    switch (op)
1756      {      {
1757        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1758        OP_BRA (normal non-capturing bracket) because the other variants of these
1759        opcodes are all concerned with unlimited repeated groups, which of course
1760        are not of fixed length. */
1761    
1762        case OP_CBRA:
1763      case OP_BRA:      case OP_BRA:
1764      case OP_ONCE:      case OP_ONCE:
1765        case OP_ONCE_NC:
1766      case OP_COND:      case OP_COND:
1767      d = find_fixedlength(cc, options);      d = find_fixedlength(cc + ((op == OP_CBRA)? IMM2_SIZE : 0), utf, atend, cd);
1768      if (d < 0) return d;      if (d < 0) return d;
1769      branchlength += d;      branchlength += d;
1770      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1771      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1772      break;      break;
1773    
1774      /* 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.
1775      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
1776      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
1777        the same code. Note that we must not include the OP_KETRxxx opcodes here,
1778        because they all imply an unlimited repeat. */
1779    
1780      case OP_ALT:      case OP_ALT:
1781      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1782      case OP_END:      case OP_END:
1783        case OP_ACCEPT:
1784        case OP_ASSERT_ACCEPT:
1785      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1786        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
1787      if (*cc != OP_ALT) return length;      if (*cc != OP_ALT) return length;
# Line 873  for (;;) Line 1789  for (;;)
1789      branchlength = 0;      branchlength = 0;
1790      break;      break;
1791    
1792        /* A true recursion implies not fixed length, but a subroutine call may
1793        be OK. If the subroutine is a forward reference, we can't deal with
1794        it until the end of the pattern, so return -3. */
1795    
1796        case OP_RECURSE:
1797        if (!atend) return -3;
1798        cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1799        do ce += GET(ce, 1); while (*ce == OP_ALT);           /* End subpattern */
1800        if (cc > cs && cc < ce) return -1;                    /* Recursion */
1801        d = find_fixedlength(cs + IMM2_SIZE, utf, atend, cd);
1802        if (d < 0) return d;
1803        branchlength += d;
1804        cc += 1 + LINK_SIZE;
1805        break;
1806    
1807      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1808    
1809      case OP_ASSERT:      case OP_ASSERT:
# Line 880  for (;;) Line 1811  for (;;)
1811      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1812      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1813      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1814      /* Fall through */      cc += PRIV(OP_lengths)[*cc];
1815        break;
1816    
1817      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1818    
1819      case OP_REVERSE:      case OP_MARK:
1820      case OP_BRANUMBER:      case OP_PRUNE_ARG:
1821      case OP_CREF:      case OP_SKIP_ARG:
1822      case OP_OPT:      case OP_THEN_ARG:
1823        cc += cc[1] + PRIV(OP_lengths)[*cc];
1824        break;
1825    
1826      case OP_CALLOUT:      case OP_CALLOUT:
     case OP_SOD:  
     case OP_SOM:  
     case OP_EOD:  
     case OP_EODN:  
1827      case OP_CIRC:      case OP_CIRC:
1828        case OP_CIRCM:
1829        case OP_CLOSE:
1830        case OP_COMMIT:
1831        case OP_CREF:
1832        case OP_DEF:
1833      case OP_DOLL:      case OP_DOLL:
1834        case OP_DOLLM:
1835        case OP_EOD:
1836        case OP_EODN:
1837        case OP_FAIL:
1838        case OP_NCREF:
1839        case OP_NRREF:
1840      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1841        case OP_PRUNE:
1842        case OP_REVERSE:
1843        case OP_RREF:
1844        case OP_SET_SOM:
1845        case OP_SKIP:
1846        case OP_SOD:
1847        case OP_SOM:
1848        case OP_THEN:
1849      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1850      cc += _pcre_OP_lengths[*cc];      cc += PRIV(OP_lengths)[*cc];
1851      break;      break;
1852    
1853      /* Handle literal characters */      /* Handle literal characters */
1854    
1855      case OP_CHAR:      case OP_CHAR:
1856      case OP_CHARNC:      case OP_CHARI:
1857        case OP_NOT:
1858        case OP_NOTI:
1859      branchlength++;      branchlength++;
1860      cc += 2;      cc += 2;
1861  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
1862      if ((options & PCRE_UTF8) != 0)      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
       {  
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1863  #endif  #endif
1864      break;      break;
1865    
# Line 918  for (;;) Line 1867  for (;;)
1867      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1868    
1869      case OP_EXACT:      case OP_EXACT:
1870      branchlength += GET2(cc,1);      case OP_EXACTI:
1871      cc += 4;      case OP_NOTEXACT:
1872  #ifdef SUPPORT_UTF8      case OP_NOTEXACTI:
1873      if ((options & PCRE_UTF8) != 0)      branchlength += (int)GET2(cc,1);
1874        {      cc += 2 + IMM2_SIZE;
1875        while((*cc & 0x80) == 0x80) cc++;  #ifdef SUPPORT_UTF
1876        }      if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1877  #endif  #endif
1878      break;      break;
1879    
1880      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1881      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1882      cc += 4;      if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)
1883          cc += 2;
1884        cc += 1 + IMM2_SIZE + 1;
1885      break;      break;
1886    
1887      /* Handle single-char matchers */      /* Handle single-char matchers */
1888    
1889      case OP_PROP:      case OP_PROP:
1890      case OP_NOTPROP:      case OP_NOTPROP:
1891      cc++;      cc += 2;
1892      /* Fall through */      /* Fall through */
1893    
1894        case OP_HSPACE:
1895        case OP_VSPACE:
1896        case OP_NOT_HSPACE:
1897        case OP_NOT_VSPACE:
1898      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
1899      case OP_DIGIT:      case OP_DIGIT:
1900      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
# Line 947  for (;;) Line 1902  for (;;)
1902      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1903      case OP_WORDCHAR:      case OP_WORDCHAR:
1904      case OP_ANY:      case OP_ANY:
1905        case OP_ALLANY:
1906      branchlength++;      branchlength++;
1907      cc++;      cc++;
1908      break;      break;
1909    
1910      /* The single-byte matcher isn't allowed */      /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1911        otherwise \C is coded as OP_ALLANY. */
1912    
1913      case OP_ANYBYTE:      case OP_ANYBYTE:
1914      return -2;      return -2;
1915    
1916      /* Check a class for variable quantification */      /* Check a class for variable quantification */
1917    
 #ifdef SUPPORT_UTF8  
     case OP_XCLASS:  
     cc += GET(cc, 1) - 33;  
     /* Fall through */  
 #endif  
   
1918      case OP_CLASS:      case OP_CLASS:
1919      case OP_NCLASS:      case OP_NCLASS:
1920      cc += 33;  #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1921        case OP_XCLASS:
1922        /* The original code caused an unsigned overflow in 64 bit systems,
1923        so now we use a conditional statement. */
1924        if (op == OP_XCLASS)
1925          cc += GET(cc, 1);
1926        else
1927          cc += PRIV(OP_lengths)[OP_CLASS];
1928    #else
1929        cc += PRIV(OP_lengths)[OP_CLASS];
1930    #endif
1931    
1932      switch (*cc)      switch (*cc)
1933        {        {
1934          case OP_CRPLUS:
1935          case OP_CRMINPLUS:
1936        case OP_CRSTAR:        case OP_CRSTAR:
1937        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1938        case OP_CRQUERY:        case OP_CRQUERY:
# Line 978  for (;;) Line 1941  for (;;)
1941    
1942        case OP_CRRANGE:        case OP_CRRANGE:
1943        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1944        if (GET2(cc,1) != GET2(cc,3)) return -1;        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1945        branchlength += GET2(cc,1);        branchlength += (int)GET2(cc,1);
1946        cc += 5;        cc += 1 + 2 * IMM2_SIZE;
1947        break;        break;
1948    
1949        default:        default:
# Line 990  for (;;) Line 1953  for (;;)
1953    
1954      /* Anything else is variable length */      /* Anything else is variable length */
1955    
1956      default:      case OP_ANYNL:
1957        case OP_BRAMINZERO:
1958        case OP_BRAPOS:
1959        case OP_BRAPOSZERO:
1960        case OP_BRAZERO:
1961        case OP_CBRAPOS:
1962        case OP_EXTUNI:
1963        case OP_KETRMAX:
1964        case OP_KETRMIN:
1965        case OP_KETRPOS:
1966        case OP_MINPLUS:
1967        case OP_MINPLUSI:
1968        case OP_MINQUERY:
1969        case OP_MINQUERYI:
1970        case OP_MINSTAR:
1971        case OP_MINSTARI:
1972        case OP_MINUPTO:
1973        case OP_MINUPTOI:
1974        case OP_NOTMINPLUS:
1975        case OP_NOTMINPLUSI:
1976        case OP_NOTMINQUERY:
1977        case OP_NOTMINQUERYI:
1978        case OP_NOTMINSTAR:
1979        case OP_NOTMINSTARI:
1980        case OP_NOTMINUPTO:
1981        case OP_NOTMINUPTOI:
1982        case OP_NOTPLUS:
1983        case OP_NOTPLUSI:
1984        case OP_NOTPOSPLUS:
1985        case OP_NOTPOSPLUSI:
1986        case OP_NOTPOSQUERY:
1987        case OP_NOTPOSQUERYI:
1988        case OP_NOTPOSSTAR:
1989        case OP_NOTPOSSTARI:
1990        case OP_NOTPOSUPTO:
1991        case OP_NOTPOSUPTOI:
1992        case OP_NOTQUERY:
1993        case OP_NOTQUERYI:
1994        case OP_NOTSTAR:
1995        case OP_NOTSTARI:
1996        case OP_NOTUPTO:
1997        case OP_NOTUPTOI:
1998        case OP_PLUS:
1999        case OP_PLUSI:
2000        case OP_POSPLUS:
2001        case OP_POSPLUSI:
2002        case OP_POSQUERY:
2003        case OP_POSQUERYI:
2004        case OP_POSSTAR:
2005        case OP_POSSTARI:
2006        case OP_POSUPTO:
2007        case OP_POSUPTOI:
2008        case OP_QUERY:
2009        case OP_QUERYI:
2010        case OP_REF:
2011        case OP_REFI:
2012        case OP_SBRA:
2013        case OP_SBRAPOS:
2014        case OP_SCBRA:
2015        case OP_SCBRAPOS:
2016        case OP_SCOND:
2017        case OP_SKIPZERO:
2018        case OP_STAR:
2019        case OP_STARI:
2020        case OP_TYPEMINPLUS:
2021        case OP_TYPEMINQUERY:
2022        case OP_TYPEMINSTAR:
2023        case OP_TYPEMINUPTO:
2024        case OP_TYPEPLUS:
2025        case OP_TYPEPOSPLUS:
2026        case OP_TYPEPOSQUERY:
2027        case OP_TYPEPOSSTAR:
2028        case OP_TYPEPOSUPTO:
2029        case OP_TYPEQUERY:
2030        case OP_TYPESTAR:
2031        case OP_TYPEUPTO:
2032        case OP_UPTO:
2033        case OP_UPTOI:
2034      return -1;      return -1;
2035    
2036        /* Catch unrecognized opcodes so that when new ones are added they
2037        are not forgotten, as has happened in the past. */
2038    
2039        default:
2040        return -4;
2041      }      }
2042    }    }
2043  /* Control never gets here */  /* Control never gets here */
# Line 1001  for (;;) Line 2047  for (;;)
2047    
2048    
2049  /*************************************************  /*************************************************
2050  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
2051  *************************************************/  *************************************************/
2052    
2053  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
2054  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
2055    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
2056    so that it can be called from pcre_study() when finding the minimum matching
2057    length.
2058    
2059  Arguments:  Arguments:
2060    code        points to start of expression    code        points to start of expression
2061    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2062    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
2063    
2064  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
2065  */  */
2066    
2067  static const uschar *  const pcre_uchar *
2068  find_bracket(const uschar *code, BOOL utf8, int number)  PRIV(find_bracket)(const pcre_uchar *code, BOOL utf, int number)
2069  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
2070  for (;;)  for (;;)
2071    {    {
2072    register int c = *code;    register pcre_uchar c = *code;
2073    
2074    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2075    else if (c > OP_BRA)  
2076      /* XCLASS is used for classes that cannot be represented just by a bit
2077      map. This includes negated single high-valued characters. The length in
2078      the table is zero; the actual length is stored in the compiled code. */
2079    
2080      if (c == OP_XCLASS) code += GET(code, 1);
2081    
2082      /* Handle recursion */
2083    
2084      else if (c == OP_REVERSE)
2085        {
2086        if (number < 0) return (pcre_uchar *)code;
2087        code += PRIV(OP_lengths)[c];
2088        }
2089    
2090      /* Handle capturing bracket */
2091    
2092      else if (c == OP_CBRA || c == OP_SCBRA ||
2093               c == OP_CBRAPOS || c == OP_SCBRAPOS)
2094      {      {
2095      int n = c - OP_BRA;      int n = (int)GET2(code, 1+LINK_SIZE);
2096      if (n > EXTRACT_BASIC_MAX) n = GET2(code, 2+LINK_SIZE);      if (n == number) return (pcre_uchar *)code;
2097      if (n == number) return (uschar *)code;      code += PRIV(OP_lengths)[c];
     code += _pcre_OP_lengths[OP_BRA];  
2098      }      }
2099    
2100      /* Otherwise, we can get the item's length from the table, except that for
2101      repeated character types, we have to test for \p and \P, which have an extra
2102      two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2103      must add in its length. */
2104    
2105    else    else
2106      {      {
2107      code += _pcre_OP_lengths[c];      switch(c)
2108          {
2109          case OP_TYPESTAR:
2110          case OP_TYPEMINSTAR:
2111          case OP_TYPEPLUS:
2112          case OP_TYPEMINPLUS:
2113          case OP_TYPEQUERY:
2114          case OP_TYPEMINQUERY:
2115          case OP_TYPEPOSSTAR:
2116          case OP_TYPEPOSPLUS:
2117          case OP_TYPEPOSQUERY:
2118          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2119          break;
2120    
2121          case OP_TYPEUPTO:
2122          case OP_TYPEMINUPTO:
2123          case OP_TYPEEXACT:
2124          case OP_TYPEPOSUPTO:
2125          if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2126            code += 2;
2127          break;
2128    
2129  #ifdef SUPPORT_UTF8        case OP_MARK:
2130          case OP_PRUNE_ARG:
2131          case OP_SKIP_ARG:
2132          code += code[1];
2133          break;
2134    
2135      /* In UTF-8 mode, opcodes that are followed by a character may be followed        case OP_THEN_ARG:
2136      by a multi-byte character. The length in the table is a minimum, so we have        code += code[1];
2137      to scan along to skip the extra bytes. All opcodes are less than 128, so we        break;
2138      can use relatively efficient code. */        }
2139    
2140        /* Add in the fixed length from the table */
2141    
2142        code += PRIV(OP_lengths)[c];
2143    
2144      if (utf8) switch(c)    /* In UTF-8 mode, opcodes that are followed by a character may be followed by
2145      a multi-byte character. The length in the table is a minimum, so we have to
2146      arrange to skip the extra bytes. */
2147    
2148    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2149        if (utf) switch(c)
2150        {        {
2151        case OP_CHAR:        case OP_CHAR:
2152        case OP_CHARNC:        case OP_CHARI:
2153        case OP_EXACT:        case OP_EXACT:
2154          case OP_EXACTI:
2155        case OP_UPTO:        case OP_UPTO:
2156          case OP_UPTOI:
2157        case OP_MINUPTO:        case OP_MINUPTO:
2158          case OP_MINUPTOI:
2159          case OP_POSUPTO:
2160          case OP_POSUPTOI:
2161        case OP_STAR:        case OP_STAR:
2162          case OP_STARI:
2163        case OP_MINSTAR:        case OP_MINSTAR:
2164          case OP_MINSTARI:
2165          case OP_POSSTAR:
2166          case OP_POSSTARI:
2167        case OP_PLUS:        case OP_PLUS:
2168          case OP_PLUSI:
2169        case OP_MINPLUS:        case OP_MINPLUS:
2170          case OP_MINPLUSI:
2171          case OP_POSPLUS:
2172          case OP_POSPLUSI:
2173        case OP_QUERY:        case OP_QUERY:
2174          case OP_QUERYI:
2175        case OP_MINQUERY:        case OP_MINQUERY:
2176        while ((*code & 0xc0) == 0x80) code++;        case OP_MINQUERYI:
2177        break;        case OP_POSQUERY:
2178          case OP_POSQUERYI:
2179        /* 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;  
2180        break;        break;
2181        }        }
2182    #else
2183        (void)(utf);  /* Keep compiler happy by referencing function argument */
2184  #endif  #endif
2185      }      }
2186    }    }
# Line 1084  instance of OP_RECURSE. Line 2197  instance of OP_RECURSE.
2197    
2198  Arguments:  Arguments:
2199    code        points to start of expression    code        points to start of expression
2200    utf8        TRUE in UTF-8 mode    utf         TRUE in UTF-8 / UTF-16 / UTF-32 mode
2201    
2202  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
2203  */  */
2204    
2205  static const uschar *  static const pcre_uchar *
2206  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const pcre_uchar *code, BOOL utf)
2207  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
2208  for (;;)  for (;;)
2209    {    {
2210    register int c = *code;    register pcre_uchar c = *code;
2211    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2212    else if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
2213    else if (c > OP_BRA)  
2214      {    /* XCLASS is used for classes that cannot be represented just by a bit
2215      code += _pcre_OP_lengths[OP_BRA];    map. This includes negated single high-valued characters. The length in
2216      }    the table is zero; the actual length is stored in the compiled code. */
2217    
2218      if (c == OP_XCLASS) code += GET(code, 1);
2219    
2220      /* Otherwise, we can get the item's length from the table, except that for
2221      repeated character types, we have to test for \p and \P, which have an extra
2222      two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2223      must add in its length. */
2224    
2225    else    else
2226      {      {
2227      code += _pcre_OP_lengths[c];      switch(c)
2228          {
2229          case OP_TYPESTAR:
2230          case OP_TYPEMINSTAR:
2231          case OP_TYPEPLUS:
2232          case OP_TYPEMINPLUS:
2233          case OP_TYPEQUERY:
2234          case OP_TYPEMINQUERY:
2235          case OP_TYPEPOSSTAR:
2236          case OP_TYPEPOSPLUS:
2237          case OP_TYPEPOSQUERY:
2238          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2239          break;
2240    
2241  #ifdef SUPPORT_UTF8        case OP_TYPEPOSUPTO:
2242          case OP_TYPEUPTO:
2243          case OP_TYPEMINUPTO:
2244          case OP_TYPEEXACT:
2245          if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2246            code += 2;
2247          break;
2248    
2249      /* In UTF-8 mode, opcodes that are followed by a character may be followed        case OP_MARK:
2250      by a multi-byte character. The length in the table is a minimum, so we have        case OP_PRUNE_ARG:
2251      to scan along to skip the extra bytes. All opcodes are less than 128, so we        case OP_SKIP_ARG:
2252      can use relatively efficient code. */        code += code[1];
2253          break;
2254    
2255      if (utf8) switch(c)        case OP_THEN_ARG:
2256        {        code += code[1];
2257        case OP_CHAR:        break;
2258        case OP_CHARNC:        }
2259        case OP_EXACT:  
2260        /* Add in the fixed length from the table */
2261    
2262        code += PRIV(OP_lengths)[c];
2263    
2264        /* In UTF-8 mode, opcodes that are followed by a character may be followed
2265        by a multi-byte character. The length in the table is a minimum, so we have
2266        to arrange to skip the extra bytes. */
2267    
2268    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2269        if (utf) switch(c)
2270          {
2271          case OP_CHAR:
2272          case OP_CHARI:
2273          case OP_NOT:
2274          case OP_NOTI:
2275          case OP_EXACT:
2276          case OP_EXACTI:
2277          case OP_NOTEXACT:
2278          case OP_NOTEXACTI:
2279        case OP_UPTO:        case OP_UPTO:
2280          case OP_UPTOI:
2281          case OP_NOTUPTO:
2282          case OP_NOTUPTOI:
2283        case OP_MINUPTO:        case OP_MINUPTO:
2284          case OP_MINUPTOI:
2285          case OP_NOTMINUPTO:
2286          case OP_NOTMINUPTOI:
2287          case OP_POSUPTO:
2288          case OP_POSUPTOI:
2289          case OP_NOTPOSUPTO:
2290          case OP_NOTPOSUPTOI:
2291        case OP_STAR:        case OP_STAR:
2292          case OP_STARI:
2293          case OP_NOTSTAR:
2294          case OP_NOTSTARI:
2295        case OP_MINSTAR:        case OP_MINSTAR:
2296          case OP_MINSTARI:
2297          case OP_NOTMINSTAR:
2298          case OP_NOTMINSTARI:
2299          case OP_POSSTAR:
2300          case OP_POSSTARI:
2301          case OP_NOTPOSSTAR:
2302          case OP_NOTPOSSTARI:
2303        case OP_PLUS:        case OP_PLUS:
2304          case OP_PLUSI:
2305          case OP_NOTPLUS:
2306          case OP_NOTPLUSI:
2307        case OP_MINPLUS:        case OP_MINPLUS:
2308          case OP_MINPLUSI:
2309          case OP_NOTMINPLUS:
2310          case OP_NOTMINPLUSI:
2311          case OP_POSPLUS:
2312          case OP_POSPLUSI:
2313          case OP_NOTPOSPLUS:
2314          case OP_NOTPOSPLUSI:
2315        case OP_QUERY:        case OP_QUERY:
2316          case OP_QUERYI:
2317          case OP_NOTQUERY:
2318          case OP_NOTQUERYI:
2319        case OP_MINQUERY:        case OP_MINQUERY:
2320        while ((*code & 0xc0) == 0x80) code++;        case OP_MINQUERYI:
2321        break;        case OP_NOTMINQUERY:
2322          case OP_NOTMINQUERYI:
2323        /* XCLASS is used for classes that cannot be represented just by a bit        case OP_POSQUERY:
2324        map. This includes negated single high-valued characters. The length in        case OP_POSQUERYI:
2325        the table is zero; the actual length is stored in the compiled code. */        case OP_NOTPOSQUERY:
2326          case OP_NOTPOSQUERYI:
2327        case OP_XCLASS:        if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
       code += GET(code, 1) + 1;  
2328        break;        break;
2329        }        }
2330    #else
2331        (void)(utf);  /* Keep compiler happy by referencing function argument */
2332  #endif  #endif
2333      }      }
2334    }    }
# Line 1152  for (;;) Line 2341  for (;;)
2341  *************************************************/  *************************************************/
2342    
2343  /* 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
2344  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()
2345  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
2346  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
2347  whose current branch will already have been scanned.  backward and negative forward assertions when its final argument is TRUE. If we
2348    hit an unclosed bracket, we return "empty" - this means we've struck an inner
2349    bracket whose current branch will already have been scanned.
2350    
2351  Arguments:  Arguments:
2352    code        points to start of search    code        points to start of search
2353    endcode     points to where to stop    endcode     points to where to stop
2354    utf8        TRUE if in UTF8 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2355      cd          contains pointers to tables etc.
2356    
2357  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2358  */  */
2359    
2360  static BOOL  static BOOL
2361  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2362      BOOL utf, compile_data *cd)
2363  {  {
2364  register int c;  register pcre_uchar c;
2365  for (code = first_significant_code(code + 1 + LINK_SIZE, NULL, 0, TRUE);  for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2366       code < endcode;       code < endcode;
2367       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
2368    {    {
2369    const uschar *ccode;    const pcre_uchar *ccode;
2370    
2371    c = *code;    c = *code;
2372    
2373    if (c >= OP_BRA)    /* Skip over forward assertions; the other assertions are skipped by
2374      first_significant_code() with a TRUE final argument. */
2375    
2376      if (c == OP_ASSERT)
2377      {      {
2378        do code += GET(code, 1); while (*code == OP_ALT);
2379        c = *code;
2380        continue;
2381        }
2382    
2383      /* For a recursion/subroutine call, if its end has been reached, which
2384      implies a backward reference subroutine call, we can scan it. If it's a
2385      forward reference subroutine call, we can't. To detect forward reference
2386      we have to scan up the list that is kept in the workspace. This function is
2387      called only when doing the real compile, not during the pre-compile that
2388      measures the size of the compiled pattern. */
2389    
2390      if (c == OP_RECURSE)
2391        {
2392        const pcre_uchar *scode;
2393      BOOL empty_branch;      BOOL empty_branch;
     if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */  
2394    
2395      /* Scan a closed bracket */      /* Test for forward reference */
2396    
2397        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
2398          if ((int)GET(scode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;
2399    
2400        /* Not a forward reference, test for completed backward reference */
2401    
2402      empty_branch = FALSE;      empty_branch = FALSE;
2403        scode = cd->start_code + GET(code, 1);
2404        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2405    
2406        /* Completed backwards reference */
2407    
2408      do      do
2409        {        {
2410        if (!empty_branch && could_be_empty_branch(code, endcode, utf8))        if (could_be_empty_branch(scode, endcode, utf, cd))
2411            {
2412          empty_branch = TRUE;          empty_branch = TRUE;
2413            break;
2414            }
2415          scode += GET(scode, 1);
2416          }
2417        while (*scode == OP_ALT);
2418    
2419        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2420        continue;
2421        }
2422    
2423      /* Groups with zero repeats can of course be empty; skip them. */
2424    
2425      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2426          c == OP_BRAPOSZERO)
2427        {
2428        code += PRIV(OP_lengths)[c];
2429        do code += GET(code, 1); while (*code == OP_ALT);
2430        c = *code;
2431        continue;
2432        }
2433    
2434      /* A nested group that is already marked as "could be empty" can just be
2435      skipped. */
2436    
2437      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2438          c == OP_SCBRA || c == OP_SCBRAPOS)
2439        {
2440        do code += GET(code, 1); while (*code == OP_ALT);
2441        c = *code;
2442        continue;
2443        }
2444    
2445      /* For other groups, scan the branches. */
2446    
2447      if (c == OP_BRA  || c == OP_BRAPOS ||
2448          c == OP_CBRA || c == OP_CBRAPOS ||
2449          c == OP_ONCE || c == OP_ONCE_NC ||
2450          c == OP_COND)
2451        {
2452        BOOL empty_branch;
2453        if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
2454    
2455        /* If a conditional group has only one branch, there is a second, implied,
2456        empty branch, so just skip over the conditional, because it could be empty.
2457        Otherwise, scan the individual branches of the group. */
2458    
2459        if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
2460        code += GET(code, 1);        code += GET(code, 1);
2461        else
2462          {
2463          empty_branch = FALSE;
2464          do
2465            {
2466            if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd))
2467              empty_branch = TRUE;
2468            code += GET(code, 1);
2469            }
2470          while (*code == OP_ALT);
2471          if (!empty_branch) return FALSE;   /* All branches are non-empty */
2472        }        }
2473      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
     code += 1 + LINK_SIZE;  
2474      c = *code;      c = *code;
2475        continue;
2476      }      }
2477    
2478    else switch (c)    /* Handle the other opcodes */
2479    
2480      switch (c)
2481      {      {
2482      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
2483        cannot be represented just by a bit map. This includes negated single
2484        high-valued characters. The length in PRIV(OP_lengths)[] is zero; the
2485        actual length is stored in the compiled code, so we must update "code"
2486        here. */
2487    
2488  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2489      case OP_XCLASS:      case OP_XCLASS:
2490      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
2491      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
2492  #endif  #endif
2493    
2494      case OP_CLASS:      case OP_CLASS:
2495      case OP_NCLASS:      case OP_NCLASS:
2496      ccode = code + 33;      ccode = code + PRIV(OP_lengths)[OP_CLASS];
2497    
2498  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2499      CHECK_CLASS_REPEAT:      CHECK_CLASS_REPEAT:
2500  #endif  #endif
2501    
# Line 1247  for (code = first_significant_code(code Line 2531  for (code = first_significant_code(code
2531      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2532      case OP_WORDCHAR:      case OP_WORDCHAR:
2533      case OP_ANY:      case OP_ANY:
2534        case OP_ALLANY:
2535      case OP_ANYBYTE:      case OP_ANYBYTE:
2536      case OP_CHAR:      case OP_CHAR:
2537      case OP_CHARNC:      case OP_CHARI:
2538      case OP_NOT:      case OP_NOT:
2539        case OP_NOTI:
2540      case OP_PLUS:      case OP_PLUS:
2541      case OP_MINPLUS:      case OP_MINPLUS:
2542        case OP_POSPLUS:
2543      case OP_EXACT:      case OP_EXACT:
2544      case OP_NOTPLUS:      case OP_NOTPLUS:
2545      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
2546        case OP_NOTPOSPLUS:
2547      case OP_NOTEXACT:      case OP_NOTEXACT:
2548      case OP_TYPEPLUS:      case OP_TYPEPLUS:
2549      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
2550        case OP_TYPEPOSPLUS:
2551      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2552      return FALSE;      return FALSE;
2553    
2554        /* These are going to continue, as they may be empty, but we have to
2555        fudge the length for the \p and \P cases. */
2556    
2557        case OP_TYPESTAR:
2558        case OP_TYPEMINSTAR:
2559        case OP_TYPEPOSSTAR:
2560        case OP_TYPEQUERY:
2561        case OP_TYPEMINQUERY:
2562        case OP_TYPEPOSQUERY:
2563        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2564        break;
2565    
2566        /* Same for these */
2567    
2568        case OP_TYPEUPTO:
2569        case OP_TYPEMINUPTO:
2570        case OP_TYPEPOSUPTO:
2571        if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2572          code += 2;
2573        break;
2574    
2575      /* End of branch */      /* End of branch */
2576    
2577      case OP_KET:      case OP_KET:
2578      case OP_KETRMAX:      case OP_KETRMAX:
2579      case OP_KETRMIN:      case OP_KETRMIN:
2580        case OP_KETRPOS:
2581      case OP_ALT:      case OP_ALT:
2582      return TRUE;      return TRUE;
2583    
2584      /* In UTF-8 mode, STAR, MINSTAR, QUERY, MINQUERY, UPTO, and MINUPTO  may be      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2585      followed by a multibyte character */      MINUPTO, and POSUPTO may be followed by a multibyte character */
2586    
2587  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2588      case OP_STAR:      case OP_STAR:
2589        case OP_STARI:
2590      case OP_MINSTAR:      case OP_MINSTAR:
2591        case OP_MINSTARI:
2592        case OP_POSSTAR:
2593        case OP_POSSTARI:
2594      case OP_QUERY:      case OP_QUERY:
2595        case OP_QUERYI:
2596      case OP_MINQUERY:      case OP_MINQUERY:
2597        case OP_MINQUERYI:
2598        case OP_POSQUERY:
2599        case OP_POSQUERYI:
2600        if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2601        break;
2602    
2603      case OP_UPTO:      case OP_UPTO:
2604        case OP_UPTOI:
2605      case OP_MINUPTO:      case OP_MINUPTO:
2606      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_MINUPTOI:
2607        case OP_POSUPTO:
2608        case OP_POSUPTOI:
2609        if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2610      break;      break;
2611  #endif  #endif
2612    
2613        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2614        string. */
2615    
2616        case OP_MARK:
2617        case OP_PRUNE_ARG:
2618        case OP_SKIP_ARG:
2619        code += code[1];
2620        break;
2621    
2622        case OP_THEN_ARG:
2623        code += code[1];
2624        break;
2625    
2626        /* None of the remaining opcodes are required to match a character. */
2627    
2628        default:
2629        break;
2630      }      }
2631    }    }
2632    
# Line 1299  return TRUE; Line 2643  return TRUE;
2643  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
2644  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,
2645  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.
2646    This function is called only during the real compile, not during the
2647    pre-compile.
2648    
2649  Arguments:  Arguments:
2650    code        points to start of the recursion    code        points to start of the recursion
2651    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2652    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2653    utf8        TRUE if in UTF-8 mode    utf         TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2654      cd          pointers to tables etc
2655    
2656  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2657  */  */
2658    
2659  static BOOL  static BOOL
2660  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
2661    BOOL utf8)    branch_chain *bcptr, BOOL utf, compile_data *cd)
2662  {  {
2663  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2664    {    {
2665    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd))
2666        return FALSE;
2667    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2668    }    }
2669  return TRUE;  return TRUE;
# Line 1328  return TRUE; Line 2676  return TRUE;
2676  *************************************************/  *************************************************/
2677    
2678  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2679  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
2680  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2681  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2682    
2683    Originally, this function only recognized a sequence of letters between the
2684    terminators, but it seems that Perl recognizes any sequence of characters,
2685    though of course unknown POSIX names are subsequently rejected. Perl gives an
2686    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2687    didn't consider this to be a POSIX class. Likewise for [:1234:].
2688    
2689    The problem in trying to be exactly like Perl is in the handling of escapes. We
2690    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2691    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2692    below handles the special case of \], but does not try to do any other escape
2693    processing. This makes it different from Perl for cases such as [:l\ower:]
2694    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2695    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2696    I think.
2697    
2698    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2699    It seems that the appearance of a nested POSIX class supersedes an apparent
2700    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2701    a digit.
2702    
2703    In Perl, unescaped square brackets may also appear as part of class names. For
2704    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2705    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2706    seem right at all. PCRE does not allow closing square brackets in POSIX class
2707    names.
2708    
2709  Argument:  Arguments:
2710    ptr      pointer to the initial [    ptr      pointer to the initial [
2711    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2712    
2713  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2714  */  */
2715    
2716  static BOOL  static BOOL
2717  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
2718  {  {
2719  int terminator;          /* Don't combine these lines; the Solaris cc */  pcre_uchar terminator;          /* Don't combine these lines; the Solaris cc */
2720  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2721  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != CHAR_NULL; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2722    {    {
2723    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2724    return TRUE;      ptr++;
2725      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2726      else
2727        {
2728        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2729          {
2730          *endptr = ptr;
2731          return TRUE;
2732          }
2733        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2734             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2735              ptr[1] == CHAR_EQUALS_SIGN) &&
2736            check_posix_syntax(ptr, endptr))
2737          return FALSE;
2738        }
2739    }    }
2740  return FALSE;  return FALSE;
2741  }  }
# Line 1373  Returns:     a value representing the na Line 2758  Returns:     a value representing the na
2758  */  */
2759    
2760  static int  static int
2761  check_posix_name(const uschar *ptr, int len)  check_posix_name(const pcre_uchar *ptr, int len)
2762  {  {
2763    const char *pn = posix_names;
2764  register int yield = 0;  register int yield = 0;
2765  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2766    {    {
2767    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2768      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      STRNCMP_UC_C8(ptr, pn, (unsigned int)len) == 0) return yield;
2769      pn += posix_name_lengths[yield] + 1;
2770    yield++;    yield++;
2771    }    }
2772  return -1;  return -1;
# Line 1394  return -1; Line 2781  return -1;
2781  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2782  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2783  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
2784  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
2785  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
2786  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
2787  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
2788  partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2789    OP_END.
2790    
2791    This function has been extended with the possibility of forward references for
2792    recursions and subroutine calls. It must also check the list of such references
2793    for the group we are dealing with. If it finds that one of the recursions in
2794    the current group is on this list, it adjusts the offset in the list, not the
2795    value in the reference (which is a group number).
2796    
2797  Arguments:  Arguments:
2798    group      points to the start of the group    group      points to the start of the group
2799    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
2800    utf8       TRUE in UTF-8 mode    utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode
2801    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
2802      save_hwm   the hwm forward reference pointer at the start of the group
2803    
2804  Returns:     nothing  Returns:     nothing
2805  */  */
2806    
2807  static void  static void
2808  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd)  adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
2809      pcre_uchar *save_hwm)
2810  {  {
2811  uschar *ptr = group;  pcre_uchar *ptr = group;
2812  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  
2813    while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
2814    {    {
2815    int offset = GET(ptr, 1);    int offset;
2816    if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);    pcre_uchar *hc;
2817    
2818      /* See if this recursion is on the forward reference list. If so, adjust the
2819      reference. */
2820    
2821      for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
2822        {
2823        offset = (int)GET(hc, 0);
2824        if (cd->start_code + offset == ptr + 1)
2825          {
2826          PUT(hc, 0, offset + adjust);
2827          break;
2828          }
2829        }
2830    
2831      /* Otherwise, adjust the recursion offset if it's after the start of this
2832      group. */
2833    
2834      if (hc >= cd->hwm)
2835        {
2836        offset = (int)GET(ptr, 1);
2837        if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
2838        }
2839    
2840    ptr += 1 + LINK_SIZE;    ptr += 1 + LINK_SIZE;
2841    }    }
2842  }  }
# Line 1438  Arguments: Line 2858  Arguments:
2858  Returns:         new code pointer  Returns:         new code pointer
2859  */  */
2860    
2861  static uschar *  static pcre_uchar *
2862  auto_callout(uschar *code, const uschar *ptr, compile_data *cd)  auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
2863  {  {
2864  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2865  *code++ = 255;  *code++ = 255;
2866  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2867  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2868  return code + 2*LINK_SIZE;  return code + 2 * LINK_SIZE;
2869  }  }
2870    
2871    
# Line 1467  Returns:             nothing Line 2887  Returns:             nothing
2887  */  */
2888    
2889  static void  static void
2890  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
2891  {  {
2892  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2893  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2894  }  }
2895    
# Line 1481  PUT(previous_callout, 2 + LINK_SIZE, len Line 2901  PUT(previous_callout, 2 + LINK_SIZE, len
2901  *************************************************/  *************************************************/
2902    
2903  /* 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
2904  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
2905  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
2906  start address.  start address. A character with multiple other cases is returned on its own
2907    with a special return value.
2908    
2909  Arguments:  Arguments:
2910    cptr        points to starting character value; updated    cptr        points to starting character value; updated
# Line 1491  Arguments: Line 2912  Arguments:
2912    ocptr       where to put start of othercase range    ocptr       where to put start of othercase range
2913    odptr       where to put end of othercase range    odptr       where to put end of othercase range
2914    
2915  Yield:        TRUE when range returned; FALSE when no more  Yield:        -1 when no more
2916                   0 when a range is returned
2917                  >0 the CASESET offset for char with multiple other cases
2918                    in this case, ocptr contains the original
2919  */  */
2920    
2921  static BOOL  static int
2922  get_othercase_range(int *cptr, int d, int *ocptr, int *odptr)  get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,
2923      pcre_uint32 *odptr)
2924  {  {
2925  int c, chartype, othercase, next;  pcre_uint32 c, othercase, next;
2926    unsigned int co;
2927    
2928    /* Find the first character that has an other case. If it has multiple other
2929    cases, return its case offset value. */
2930    
2931  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2932    {    {
2933    if (_pcre_ucp_findchar(c, &chartype, &othercase) == ucp_L && othercase != 0)    if ((co = UCD_CASESET(c)) != 0)
2934      break;      {
2935        *ocptr = c++;   /* Character that has the set */
2936        *cptr = c;      /* Rest of input range */
2937        return (int)co;
2938        }
2939      if ((othercase = UCD_OTHERCASE(c)) != c) break;
2940    }    }
2941    
2942  if (c > d) return FALSE;  if (c > d) return -1;  /* Reached end of range */
2943    
2944  *ocptr = othercase;  *ocptr = othercase;
2945  next = othercase + 1;  next = othercase + 1;
2946    
2947  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2948    {    {
2949    if (_pcre_ucp_findchar(c, &chartype, &othercase) != ucp_L ||    if (UCD_OTHERCASE(c) != next) break;
         othercase != next)  
     break;  
2950    next++;    next++;
2951    }    }
2952    
2953  *odptr = next - 1;  *odptr = next - 1;     /* End of othercase range */
2954  *cptr = c;  *cptr = c;             /* Rest of input range */
2955    return 0;
 return TRUE;  
2956  }  }
2957  #endif  /* SUPPORT_UCP */  
2958    
2959    
2960  /*************************************************  /*************************************************
2961  *           Compile one branch                   *  *        Check a character and a property        *
2962  *************************************************/  *************************************************/
2963    
2964  /* 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
2965  changed during the branch, the pointer is used to change the external options  is adjacent to a fixed character.
 bits.  
2966    
2967  Arguments:  Arguments:
2968    optionsptr     pointer to the option bits    c            the character
2969    brackets       points to number of extracting brackets used    ptype        the property type
2970    codeptr        points to the pointer to the current code point    pdata        the data for the type
2971    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.  
2972    
2973  Returns:         TRUE on success  Returns:       TRUE if auto-possessifying is OK
                  FALSE, with *errorcodeptr set non-zero on error  
2974  */  */
2975    
2976  static BOOL  static BOOL
2977  compile_branch(int *optionsptr, int *brackets, uschar **codeptr,  check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata, BOOL negated)
   const uschar **ptrptr, int *errorcodeptr, int *firstbyteptr,  
   int *reqbyteptr, branch_chain *bcptr, compile_data *cd)  
2978  {  {
2979  int repeat_type, op_type;  #ifdef SUPPORT_UCP
2980  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;  
2981  #endif  #endif
2982    
2983  /* 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;  
2984    
2985  /* Initialize no first byte, no required byte. REQ_UNSET means "no char  switch(ptype)
2986  matching encountered yet". It gets changed to REQ_NONE if we hit something that    {
2987  matches a non-fixed char first char; reqbyte just remains unset if we never    case PT_LAMP:
2988  find one.    return (prop->chartype == ucp_Lu ||
2989              prop->chartype == ucp_Ll ||
2990              prop->chartype == ucp_Lt) == negated;
2991    
2992      case PT_GC:
2993      return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2994    
2995      case PT_PC:
2996      return (pdata == prop->chartype) == negated;
2997    
2998      case PT_SC:
2999      return (pdata == prop->script) == negated;
3000    
3001      /* These are specials */
3002    
3003      case PT_ALNUM:
3004      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
3005              PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
3006    
3007      case PT_SPACE:    /* Perl space */
3008      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
3009              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
3010              == negated;
3011    
3012      case PT_PXSPACE:  /* POSIX space */
3013      return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
3014              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
3015              c == CHAR_FF || c == CHAR_CR)
3016              == negated;
3017    
3018      case PT_WORD:
3019      return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
3020              PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
3021              c == CHAR_UNDERSCORE) == negated;
3022    
3023    #ifdef SUPPORT_UCP
3024      case PT_CLIST:
3025      p = PRIV(ucd_caseless_sets) + prop->caseset;
3026      for (;;)
3027        {
3028        if (c < *p) return !negated;
3029        if (c == *p++) return negated;
3030        }
3031      break;  /* Control never reaches here */
3032    #endif
3033      }
3034    
3035  When we hit a repeat whose minimum is zero, we may have to adjust these values  return FALSE;
3036  to take the zero repeat into account. This is implemented by setting them to  }
3037  zerofirstbyte and zeroreqbyte when such a repeat is encountered. The individual  #endif  /* SUPPORT_UCP */
 item types that can be repeated set these backoff variables appropriately. */  
3038    
 firstbyte = reqbyte = zerofirstbyte = zeroreqbyte = REQ_UNSET;  
3039    
 /* 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. */  
3040    
3041  req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;  /*************************************************
3042    *     Check if auto-possessifying is possible    *
3043    *************************************************/
3044    
3045  /* Switch on next character until the end of the branch */  /* This function is called for unlimited repeats of certain items, to see
3046    whether the next thing could possibly match the repeated item. If not, it makes
3047    sense to automatically possessify the repeated item.
3048    
3049  for (;; ptr++)  Arguments:
3050    {    previous      pointer to the repeated opcode
3051    BOOL negate_class;    utf           TRUE in UTF-8 / UTF-16 / UTF-32 mode
3052    BOOL possessive_quantifier;    ptr           next character in pattern
3053    BOOL is_quantifier;    options       options bits
3054    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];  
3055    
3056    /* Next byte in the pattern */  Returns:        TRUE if possessifying is wanted
3057    */
3058    
3059    c = *ptr;  static BOOL
3060    check_auto_possessive(const pcre_uchar *previous, BOOL utf,
3061      const pcre_uchar *ptr, int options, compile_data *cd)
3062    {
3063    pcre_uint32 c = NOTACHAR;
3064    pcre_uint32 next;
3065    int escape;
3066    pcre_uchar op_code = *previous++;
3067    
3068    /* If in \Q...\E, check for the end; if not, we have a literal */  /* Skip whitespace and comments in extended mode */
3069    
3070    if (inescq && c != 0)  if ((options & PCRE_EXTENDED) != 0)
3071      {
3072      for (;;)
3073      {      {
3074      if (c == '\\' && ptr[1] == 'E')      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
3075        if (*ptr == CHAR_NUMBER_SIGN)
3076        {        {
       inescq = FALSE;  
3077        ptr++;        ptr++;
3078        continue;        while (*ptr != CHAR_NULL)
       }  
     else  
       {  
       if (previous_callout != NULL)  
         {  
         complete_callout(previous_callout, ptr, cd);  
         previous_callout = NULL;  
         }  
       if ((options & PCRE_AUTO_CALLOUT) != 0)  
3079          {          {
3080          previous_callout = code;          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
3081          code = auto_callout(code, ptr, cd);          ptr++;
3082    #ifdef SUPPORT_UTF
3083            if (utf) FORWARDCHAR(ptr);
3084    #endif
3085          }          }
       goto NORMAL_CHAR;  
3086        }        }
3087        else break;
3088      }      }
3089      }
3090    
3091    /* 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
3092    a quantifier. */  value is a character, a negative value is an escape value. */
   
   is_quantifier = c == '*' || c == '+' || c == '?' ||  
     (c == '{' && is_counted_repeat(ptr+1));  
3093    
3094    if (!is_quantifier && previous_callout != NULL &&  if (*ptr == CHAR_BACKSLASH)
3095         after_manual_callout-- <= 0)    {
3096      {    int temperrorcode = 0;
3097      complete_callout(previous_callout, ptr, cd);    escape = check_escape(&ptr, &next, &temperrorcode, cd->bracount, options, FALSE);
3098      previous_callout = NULL;    if (temperrorcode != 0) return FALSE;
3099      }    ptr++;    /* Point after the escape sequence */
3100      }
3101    else if (!MAX_255(*ptr) || (cd->ctypes[*ptr] & ctype_meta) == 0)
3102      {
3103      escape = 0;
3104    #ifdef SUPPORT_UTF
3105      if (utf) { GETCHARINC(next, ptr); } else
3106    #endif
3107      next = *ptr++;
3108      }
3109    else return FALSE;
3110    
3111    /* In extended mode, skip white space and comments */  /* Skip whitespace and comments in extended mode */
3112    
3113    if ((options & PCRE_EXTENDED) != 0)  if ((options & PCRE_EXTENDED) != 0)
3114      {
3115      for (;;)
3116      {      {
3117      if ((cd->ctypes[c] & ctype_space) != 0) continue;      while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
3118      if (c == '#')      if (*ptr == CHAR_NUMBER_SIGN)
3119        {        {
3120        /* The space before the ; is to avoid a warning on a silly compiler        ptr++;
3121        on the Macintosh. */        while (*ptr != CHAR_NULL)
3122        while ((c = *(++ptr)) != 0 && c != NEWLINE) ;          {
3123        if (c != 0) continue;   /* Else fall through to handle end of string */          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
3124            ptr++;
3125    #ifdef SUPPORT_UTF
3126            if (utf) FORWARDCHAR(ptr);
3127    #endif
3128            }
3129        }        }
3130        else break;
3131      }      }
3132      }
3133    
3134    /* No auto callout for quantifiers. */  /* If the next thing is itself optional, we have to give up. */
3135    
3136    if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3137      STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3138        return FALSE;
3139    
3140    /* If the previous item is a character, get its value. */
3141    
3142    if (op_code == OP_CHAR || op_code == OP_CHARI ||
3143        op_code == OP_NOT || op_code == OP_NOTI)
3144      {
3145    #ifdef SUPPORT_UTF
3146      GETCHARTEST(c, previous);
3147    #else
3148      c = *previous;
3149    #endif
3150      }
3151    
3152    /* Now compare the next item with the previous opcode. First, handle cases when
3153    the next item is a character. */
3154    
3155    if (escape == 0)
3156      {
3157      /* For a caseless UTF match, the next character may have more than one other
3158      case, which maps to the special PT_CLIST property. Check this first. */
3159    
3160    #ifdef SUPPORT_UCP
3161      if (utf && c != NOTACHAR && (options & PCRE_CASELESS) != 0)
3162      {      {
3163      previous_callout = code;      unsigned int ocs = UCD_CASESET(next);
3164      code = auto_callout(code, ptr, cd);      if (ocs > 0) return check_char_prop(c, PT_CLIST, ocs, op_code >= OP_NOT);
3165      }      }
3166    #endif
3167    
3168    switch(c)    switch(op_code)
3169      {      {
3170      /* The branch terminates at end of string, |, or ). */      case OP_CHAR:
3171        return c != next;
     case 0:  
     case '|':  
     case ')':  
     *firstbyteptr = firstbyte;  
     *reqbyteptr = reqbyte;  
     *codeptr = code;  
     *ptrptr = ptr;  
     return TRUE;  
   
     /* Handle single-character metacharacters. In multiline mode, ^ disables  
     the setting of any following char as a first character. */  
3172    
3173      case '^':      /* For CHARI (caseless character) we must check the other case. If we have
3174      if ((options & PCRE_MULTILINE) != 0)      Unicode property support, we can use it to test the other case of
3175        high-valued characters. We know that next can have only one other case,
3176        because multi-other-case characters are dealt with above. */
3177    
3178        case OP_CHARI:
3179        if (c == next) return FALSE;
3180    #ifdef SUPPORT_UTF
3181        if (utf)
3182        {        {
3183        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        pcre_uint32 othercase;
3184          if (next < 128) othercase = cd->fcc[next]; else
3185    #ifdef SUPPORT_UCP
3186          othercase = UCD_OTHERCASE(next);
3187    #else
3188          othercase = NOTACHAR;
3189    #endif
3190          return c != othercase;
3191        }        }
3192      previous = NULL;      else
3193      *code++ = OP_CIRC;  #endif  /* SUPPORT_UTF */
3194      break;      return (c != TABLE_GET(next, cd->fcc, next));  /* Not UTF */
3195    
3196        case OP_NOT:
3197        return c == next;
3198    
3199        case OP_NOTI:
3200        if (c == next) return TRUE;
3201    #ifdef SUPPORT_UTF
3202        if (utf)
3203          {
3204          pcre_uint32 othercase;
3205          if (next < 128) othercase = cd->fcc[next]; else
3206    #ifdef SUPPORT_UCP
3207          othercase = UCD_OTHERCASE(next);
3208    #else
3209          othercase = NOTACHAR;
3210    #endif
3211          return c == othercase;
3212          }
3213        else
3214    #endif  /* SUPPORT_UTF */
3215        return (c == TABLE_GET(next, cd->fcc, next));  /* Not UTF */
3216    
3217      case '$':      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
3218      previous = NULL;      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
     *code++ = OP_DOLL;  
     break;  
3219    
3220      /* There can never be a first char if '.' is first, whatever happens about      case OP_DIGIT:
3221      repeats. The value of reqbyte doesn't change either. */      return next > 255 || (cd->ctypes[next] & ctype_digit) == 0;
3222    
3223      case '.':      case OP_NOT_DIGIT:
3224      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      return next <= 255 && (cd->ctypes[next] & ctype_digit) != 0;
     zerofirstbyte = firstbyte;  
     zeroreqbyte = reqbyte;  
     previous = code;  
     *code++ = OP_ANY;  
     break;  
3225    
3226      /* Character classes. If the included characters are all < 255 in value, we      case OP_WHITESPACE:
3227      build a 32-byte bitmap of the permitted characters, except in the special      return next > 255 || (cd->ctypes[next] & ctype_space) == 0;
     case where there is only one such character. For negated classes, we build  
     the map as usual, then invert it at the end. However, we use a different  
     opcode so that data characters > 255 can be handled correctly.  
3228    
3229      If the class contains characters outside the 0-255 range, a different      case OP_NOT_WHITESPACE:
3230      opcode is compiled. It may optionally have a bit map for characters < 256,      return next <= 255 && (cd->ctypes[next] & ctype_space) != 0;
     but those above are are explicitly listed afterwards. A flag byte tells  
     whether the bitmap is present, and whether this is a negated class or not.  
     */  
3231    
3232      case '[':      case OP_WORDCHAR:
3233      previous = code;      return next > 255 || (cd->ctypes[next] & ctype_word) == 0;
3234    
3235      /* PCRE supports POSIX class stuff inside a class. Perl gives an error if      case OP_NOT_WORDCHAR:
3236      they are encountered at the top level, so we'll do that too. */      return next <= 255 && (cd->ctypes[next] & ctype_word) != 0;
3237    
3238      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      case OP_HSPACE:
3239          check_posix_syntax(ptr, &tempptr, cd))      case OP_NOT_HSPACE:
3240        switch(next)
3241        {        {
3242        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        HSPACE_CASES:
3243        goto FAILED;        return op_code == OP_NOT_HSPACE;
       }  
3244    
3245      /* If the first character is '^', set the negation flag and skip it. */        default:
3246          return op_code != OP_NOT_HSPACE;
     if ((c = *(++ptr)) == '^')  
       {  
       negate_class = TRUE;  
       c = *(++ptr);  
3247        }        }
3248      else  
3249        case OP_ANYNL:
3250        case OP_VSPACE:
3251        case OP_NOT_VSPACE:
3252        switch(next)
3253        {        {
3254        negate_class = FALSE;        VSPACE_CASES:
3255        }        return op_code == OP_NOT_VSPACE;
3256    
3257      /* Keep a count of chars with values < 256 so that we can optimize the case        default:
3258      of just a single character (as long as it's < 256). For higher valued UTF-8        return op_code != OP_NOT_VSPACE;
3259      characters, we don't yet do any optimization. */        }
3260    
3261      class_charcount = 0;  #ifdef SUPPORT_UCP
3262      class_lastchar = -1;      case OP_PROP:
3263        return check_char_prop(next, previous[0], previous[1], FALSE);
3264    
3265  #ifdef SUPPORT_UTF8      case OP_NOTPROP:
3266      class_utf8 = FALSE;                       /* No chars >= 256 */      return check_char_prop(next, previous[0], previous[1], TRUE);
     class_utf8data = code + LINK_SIZE + 34;   /* For UTF-8 items */  
3267  #endif  #endif
3268    
3269      /* Initialize the 32-char bit map to all zeros. We have to build the      default:
3270      map in a temporary bit of store, in case the class contains only 1      return FALSE;
3271      character (< 256), because in that case the compiled code doesn't use the      }
3272      bit map. */    }
   
     memset(classbits, 0, 32 * sizeof(uschar));  
   
     /* Process characters until ] is reached. By writing this as a "do" it  
     means that an initial ] is taken as a data character. The first pass  
     through the regex checked the overall syntax, so we don't need to be very  
     strict here. At the start of the loop, c contains the first byte of the  
     character. */  
3273    
3274      do  /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP
3275        {  is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
3276  #ifdef SUPPORT_UTF8  generated only when PCRE_UCP is *not* set, that is, when only ASCII
3277        if (utf8 && c > 127)  characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
3278          {                           /* Braces are required because the */  replaced by OP_PROP codes when PCRE_UCP is set. */
         GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */  
         }  
 #endif  
3279    
3280        /* Inside \Q...\E everything is literal except \E */  switch(op_code)
3281      {
3282      case OP_CHAR:
3283      case OP_CHARI:
3284      switch(escape)
3285        {
3286        case ESC_d:
3287        return c > 255 || (cd->ctypes[c] & ctype_digit) == 0;
3288    
3289        if (inescq)      case ESC_D:
3290          {      return c <= 255 && (cd->ctypes[c] & ctype_digit) != 0;
         if (c == '\\' && ptr[1] == 'E')  
           {  
           inescq = FALSE;  
           ptr++;  
           continue;  
           }  
         else goto LONE_SINGLE_CHARACTER;  
         }  
3291    
3292        /* Handle POSIX class names. Perl allows a negation extension of the      case ESC_s:
3293        form [:^name:]. A square bracket that doesn't match the syntax is      return c > 255 || (cd->ctypes[c] & ctype_space) == 0;
       treated as a literal. We also recognize the POSIX constructions  
       [.ch.] and [=ch=] ("collating elements") and fault them, as Perl  
       5.6 and 5.8 do. */  
3294    
3295        if (c == '[' &&      case ESC_S:
3296            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      return c <= 255 && (cd->ctypes[c] & ctype_space) != 0;
           check_posix_syntax(ptr, &tempptr, cd))  
         {  
         BOOL local_negate = FALSE;  
         int posix_class, i;  
         register const uschar *cbits = cd->cbits;  
3297    
3298          if (ptr[1] != ':')      case ESC_w:
3299            {      return c > 255 || (cd->ctypes[c] & ctype_word) == 0;
           *errorcodeptr = ERR31;  
           goto FAILED;  
           }  
3300    
3301          ptr += 2;      case ESC_W:
3302          if (*ptr == '^')      return c <= 255 && (cd->ctypes[c] & ctype_word) != 0;
           {  
           local_negate = TRUE;  
           ptr++;  
           }  
3303    
3304          posix_class = check_posix_name(ptr, tempptr - ptr);      case ESC_h:
3305          if (posix_class < 0)      case ESC_H:
3306            {      switch(c)
3307            *errorcodeptr = ERR30;        {
3308            goto FAILED;        HSPACE_CASES:
3309            }        return escape != ESC_h;
3310    
3311          default:
3312          return escape == ESC_h;
3313          }
3314    
3315          /* If matching is caseless, upper and lower are converted to      case ESC_v:
3316          alpha. This relies on the fact that the class table starts with      case ESC_V:
3317          alpha, lower, upper as the first 3 entries. */      switch(c)
3318          {
3319          VSPACE_CASES:
3320          return escape != ESC_v;
3321    
3322          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)        default:
3323            posix_class = 0;        return escape == ESC_v;
3324          }
3325    
3326          /* Or into the map we are building up to 3 of the static class      /* When PCRE_UCP is set, these values get generated for \d etc. Find
3327          tables, or their negations. The [:blank:] class sets up the same      their substitutions and process them. The result will always be either
3328          chars as the [:space:] class (all white space). We remove the vertical      ESC_p or ESC_P. Then fall through to process those values. */
         white space chars afterwards. */  
3329    
3330          posix_class *= 3;  #ifdef SUPPORT_UCP
3331          for (i = 0; i < 3; i++)      case ESC_du:
3332            {      case ESC_DU:
3333            BOOL blankclass = strncmp((char *)ptr, "blank", 5) == 0;      case ESC_wu:
3334            int taboffset = posix_class_maps[posix_class + i];      case ESC_WU:
3335            if (taboffset < 0) break;      case ESC_su:
3336            if (local_negate)      case ESC_SU:
3337              {        {
3338              if (i == 0)        int temperrorcode = 0;
3339                for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+taboffset];        ptr = substitutes[escape - ESC_DU];
3340              else        escape = check_escape(&ptr, &next, &temperrorcode, 0, options, FALSE);
3341                for (c = 0; c < 32; c++) classbits[c] &= ~cbits[c+taboffset];        if (temperrorcode != 0) return FALSE;
3342              if (blankclass) classbits[1] |= 0x3c;        ptr++;    /* For compatibility */
3343              }        }
3344            else      /* Fall through */
             {  
             for (c = 0; c < 32; c++) classbits[c] |= cbits[c+taboffset];  
             if (blankclass) classbits[1] &= ~0x3c;  
             }  
           }  
3345    
3346          ptr = tempptr + 1;      case ESC_p:
3347          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */      case ESC_P:
3348          continue;    /* End of POSIX syntax handling */        {
3349          }        unsigned int ptype = 0, pdata = 0;
3350          int errorcodeptr;
3351          BOOL negated;
3352    
3353        /* Backslash may introduce a single character, or it may introduce one        ptr--;      /* Make ptr point at the p or P */
3354        of the specials, which just set a flag. Escaped items are checked for        if (!get_ucp(&ptr, &negated, &ptype, &pdata, &errorcodeptr))
3355        validity in the pre-compiling pass. The sequence \b is a special case.          return FALSE;
3356        Inside a class (and only there) it is treated as backspace. Elsewhere        ptr++;      /* Point past the final curly ket */
3357        it marks a word boundary. Other escapes have preset maps ready to  
3358        or into the one we are building. We assume they have more than one        /* If the property item is optional, we have to give up. (When generated
3359        character in them, so set class_charcount bigger than one. */        from \d etc by PCRE_UCP, this test will have been applied much earlier,
3360          to the original \d etc. At this point, ptr will point to a zero byte. */
3361        if (c == '\\')  
3362          {        if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3363          c = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);          STRNCMP_UC_C8(ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3364              return FALSE;
         if (-c == ESC_b) c = '\b';       /* \b is backslash in a class */  
         else if (-c == ESC_X) c = 'X';   /* \X is literal X in a class */  
         else if (-c == ESC_Q)            /* Handle start of quoted string */  
           {  
           if (ptr[1] == '\\' && ptr[2] == 'E')  
             {  
             ptr += 2; /* avoid empty string */  
             }  
           else inescq = TRUE;  
           continue;  
           }  
3365    
3366          if (c < 0)        /* Do the property check. */
           {  
           register const uschar *cbits = cd->cbits;  
           class_charcount += 2;     /* Greater than 1 is what matters */  
           switch (-c)  
             {  
             case ESC_d:  
             for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];  
             continue;  
3367    
3368              case ESC_D:        return check_char_prop(c, ptype, pdata, (escape == ESC_P) != negated);
3369              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];        }
3370              continue;  #endif
3371    
3372              case ESC_w:      default:
3373              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_word];      return FALSE;
3374              continue;      }
3375    
3376              case ESC_W:    /* In principle, support for Unicode properties should be integrated here as
3377              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];    well. It means re-organizing the above code so as to get hold of the property
3378              continue;    values before switching on the op-code. However, I wonder how many patterns
3379      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
3380      these op-codes are never generated.) */
3381    
3382              case ESC_s:    case OP_DIGIT:
3383              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];    return escape == ESC_D || escape == ESC_s || escape == ESC_W ||
3384              classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */           escape == ESC_h || escape == ESC_v || escape == ESC_R;
             continue;  
3385    
3386              case ESC_S:    case OP_NOT_DIGIT:
3387              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];    return escape == ESC_d;
             classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */  
             continue;  
3388    
3389  #ifdef SUPPORT_UCP    case OP_WHITESPACE:
3390              case ESC_p:    return escape == ESC_S || escape == ESC_d || escape == ESC_w;
             case ESC_P:  
               {  
               BOOL negated;  
               int property = get_ucp(&ptr, &negated, errorcodeptr);  
               if (property < 0) goto FAILED;  
               class_utf8 = TRUE;  
               *class_utf8data++ = ((-c == ESC_p) != negated)?  
                 XCL_PROP : XCL_NOTPROP;  
               *class_utf8data++ = property;  
               class_charcount -= 2;   /* Not a < 256 character */  
               }  
             continue;  
 #endif  
3391    
3392              /* Unrecognized escapes are faulted if PCRE is running in its    case OP_NOT_WHITESPACE:
3393              strict mode. By default, for compatibility with Perl, they are    return escape == ESC_s || escape == ESC_h || escape == ESC_v || escape == ESC_R;
             treated as literals. */  
3394    
3395              default:    case OP_HSPACE:
3396              if ((options & PCRE_EXTRA) != 0)    return escape == ESC_S || escape == ESC_H || escape == ESC_d ||
3397                {           escape == ESC_w || escape == ESC_v || escape == ESC_R;
               *errorcodeptr = ERR7;  
               goto FAILED;  
               }  
             c = *ptr;              /* The final character */  
             class_charcount -= 2;  /* Undo the default count from above */  
             }  
           }  
3398    
3399          /* Fall through if we have a single character (c >= 0). This may be    case OP_NOT_HSPACE:
3400          > 256 in UTF-8 mode. */    return escape == ESC_h;
3401    
3402          }   /* End of backslash handling */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
3403      case OP_ANYNL:
3404      case OP_VSPACE:
3405      return escape == ESC_V || escape == ESC_d || escape == ESC_w;
3406    
3407        /* A single character may be followed by '-' to form a range. However,    case OP_NOT_VSPACE:
3408        Perl does not permit ']' to be the end of the range. A '-' character    return escape == ESC_v || escape == ESC_R;
       here is treated as a literal. */  
3409    
3410        if (ptr[1] == '-' && ptr[2] != ']')    case OP_WORDCHAR:
3411          {    return escape == ESC_W || escape == ESC_s || escape == ESC_h ||
3412          int d;           escape == ESC_v || escape == ESC_R;
         ptr += 2;  
3413    
3414  #ifdef SUPPORT_UTF8    case OP_NOT_WORDCHAR:
3415          if (utf8)    return escape == ESC_w || escape == ESC_d;
           {                           /* Braces are required because the */  
           GETCHARLEN(d, ptr, ptr);    /* macro generates multiple statements */  
           }  
         else  
 #endif  
         d = *ptr;  /* Not UTF-8 mode */  
3416    
3417          /* The second part of a range can be a single-character escape, but    default:
3418          not any of the other escapes. Perl 5.6 treats a hyphen as a literal    return FALSE;
3419          in such circumstances. */    }
3420    
3421          if (d == '\\')  /* Control does not reach here */
3422            {  }
           const uschar *oldptr = ptr;  
           d = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);  
3423    
           /* \b is backslash; \X is literal X; any other special means the '-'  
           was literal */  
3424    
           if (d < 0)  
             {  
             if (d == -ESC_b) d = '\b';  
             else if (d == -ESC_X) d = 'X'; else  
               {  
               ptr = oldptr - 2;  
               goto LONE_SINGLE_CHARACTER;  /* A few lines below */  
               }  
             }  
           }  
3425    
3426          /* The check that the two values are in the correct order happens in  /*************************************************
3427          the pre-pass. Optimize one-character ranges */  *        Add a character or range to a class     *
3428    *************************************************/
3429    
3430          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */  /* This function packages up the logic of adding a character or range of
3431    characters to a class. The character values in the arguments will be within the
3432    valid values for the current mode (8-bit, 16-bit, UTF, etc). This function is
3433    mutually recursive with the function immediately below.
3434    
3435          /* In UTF-8 mode, if the upper limit is > 255, or > 127 for caseless  Arguments:
3436          matching, we have to use an XCLASS with extra data items. Caseless    classbits     the bit map for characters < 256
3437          matching for characters > 127 is available only if UCP support is    uchardptr     points to the pointer for extra data
3438          available. */    options       the options word
3439      cd            contains pointers to tables etc.
3440      start         start of range character
3441      end           end of range character
3442    
3443    Returns:        the number of < 256 characters added
3444                    the pointer to extra data is updated
3445    */
3446    
3447  #ifdef SUPPORT_UTF8  static int
3448          if (utf8 && (d > 255 || ((options & PCRE_CASELESS) != 0 && d > 127)))  add_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
3449            {    compile_data *cd, pcre_uint32 start, pcre_uint32 end)
3450            class_utf8 = TRUE;  {
3451    pcre_uint32 c;
3452    int n8 = 0;
3453    
3454            /* With UCP support, we can find the other case equivalents of  /* If caseless matching is required, scan the range and process alternate
3455            the relevant characters. There may be several ranges. Optimize how  cases. In Unicode, there are 8-bit characters that have alternate cases that
3456            they fit with the basic range. */  are greater than 255 and vice-versa. Sometimes we can just extend the original
3457    range. */
3458    
3459    if ((options & PCRE_CASELESS) != 0)
3460      {
3461  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3462            if ((options & PCRE_CASELESS) != 0)    if ((options & PCRE_UTF8) != 0)
3463              {      {
3464              int occ, ocd;      int rc;
3465              int cc = c;      pcre_uint32 oc, od;
3466              int origd = d;  
3467              while (get_othercase_range(&cc, origd, &occ, &ocd))      options &= ~PCRE_CASELESS;   /* Remove for recursive calls */
3468                {      c = start;
3469                if (occ >= c && ocd <= d) continue;  /* Skip embedded ranges */  
3470        while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)
3471                if (occ < c  && ocd >= c - 1)        /* Extend the basic range */        {
3472                  {                                  /* if there is overlap,   */        /* Handle a single character that has more than one other case. */
3473                  c = occ;                           /* noting that if occ < c */  
3474                  continue;                          /* we can't have ocd > d  */        if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,
3475                  }                                  /* because a subrange is  */          PRIV(ucd_caseless_sets) + rc, oc);
3476                if (ocd > d && occ <= d + 1)         /* always shorter than    */  
3477                  {                                  /* the basic range.       */        /* Do nothing if the other case range is within the original range. */
3478                  d = ocd;  
3479                  continue;        else if (oc >= start && od <= end) continue;
3480                  }  
3481          /* Extend the original range if there is overlap, noting that if oc < c, we
3482                if (occ == ocd)        can't have od > end because a subrange is always shorter than the basic
3483                  {        range. Otherwise, use a recursive call to add the additional range. */
3484                  *class_utf8data++ = XCL_SINGLE;  
3485                  }        else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */
3486                else        else if (od > end && oc <= end + 1) end = od;       /* Extend upwards */
3487                  {        else n8 += add_to_class(classbits, uchardptr, options, cd, oc, od);
3488                  *class_utf8data++ = XCL_RANGE;        }
3489                  class_utf8data += _pcre_ord2utf8(occ, class_utf8data);      }
3490                  }    else
               class_utf8data += _pcre_ord2utf8(ocd, class_utf8data);  
               }  
             }  
3491  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3492    
3493            /* Now record the original range, possibly modified for UCP caseless    /* Not UTF-mode, or no UCP */
3494            overlapping ranges. */  
3495      for (c = start; c <= end && c < 256; c++)
3496        {
3497        SETBIT(classbits, cd->fcc[c]);
3498        n8++;
3499        }
3500      }
3501    
3502    /* Now handle the original range. Adjust the final value according to the bit
3503    length - this means that the same lists of (e.g.) horizontal spaces can be used
3504    in all cases. */
3505    
3506    #if defined COMPILE_PCRE8
3507    #ifdef SUPPORT_UTF
3508      if ((options & PCRE_UTF8) == 0)
3509    #endif
3510      if (end > 0xff) end = 0xff;
3511    
3512    #elif defined COMPILE_PCRE16
3513    #ifdef SUPPORT_UTF
3514      if ((options & PCRE_UTF16) == 0)
3515    #endif
3516      if (end > 0xffff) end = 0xffff;
3517    
3518            *class_utf8data++ = XCL_RANGE;  #endif /* COMPILE_PCRE[8|16] */
           class_utf8data += _pcre_ord2utf8(c, class_utf8data);  
           class_utf8data += _pcre_ord2utf8(d, class_utf8data);  
   
           /* With UCP support, we are done. Without UCP support, there is no  
           caseless matching for UTF-8 characters > 127; we can use the bit map  
           for the smaller ones. */  
3519    
3520  #ifdef SUPPORT_UCP  /* If all characters are less than 256, use the bit map. Otherwise use extra
3521            continue;    /* With next character in the class */  data. */
 #else  
           if ((options & PCRE_CASELESS) == 0 || c > 127) continue;  
3522    
3523            /* Adjust upper limit and fall through to set up the map */  if (end < 0x100)
3524      {
3525      for (c = start; c <= end; c++)
3526        {
3527        n8++;
3528        SETBIT(classbits, c);
3529        }
3530      }
3531    
3532    else
3533      {
3534      pcre_uchar *uchardata = *uchardptr;
3535    
3536    #ifdef SUPPORT_UTF
3537      if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */
3538        {
3539        if (start < end)
3540          {
3541          *uchardata++ = XCL_RANGE;
3542          uchardata += PRIV(ord2utf)(start, uchardata);
3543          uchardata += PRIV(ord2utf)(end, uchardata);
3544          }
3545        else if (start == end)
3546          {
3547          *uchardata++ = XCL_SINGLE;
3548          uchardata += PRIV(ord2utf)(start, uchardata);
3549          }
3550        }
3551      else
3552    #endif  /* SUPPORT_UTF */
3553    
3554      /* Without UTF support, character values are constrained by the bit length,
3555      and can only be > 256 for 16-bit and 32-bit libraries. */
3556    
3557    #ifdef COMPILE_PCRE8
3558        {}
3559    #else
3560      if (start < end)
3561        {
3562        *uchardata++ = XCL_RANGE;
3563        *uchardata++ = start;
3564        *uchardata++ = end;
3565        }
3566      else if (start == end)
3567        {
3568        *uchardata++ = XCL_SINGLE;
3569        *uchardata++ = start;
3570        }
3571    #endif
3572    
3573            d = 127;    *uchardptr = uchardata;   /* Updata extra data pointer */
3574      }
3575    
3576    return n8;    /* Number of 8-bit characters */
3577    }
3578    
3579    
3580    
 #endif  /* SUPPORT_UCP */  
           }  
 #endif  /* SUPPORT_UTF8 */  
3581    
3582          /* We use the bit map for all cases when not in UTF-8 mode; else  /*************************************************
3583          ranges that lie entirely within 0-127 when there is UCP support; else  *        Add a list of characters to a class     *
3584          for partial ranges without UCP support. */  *************************************************/
3585    
3586          for (; c <= d; c++)  /* This function is used for adding a list of case-equivalent characters to a
3587            {  class, and also for adding a list of horizontal or vertical whitespace. If the
3588            classbits[c/8] |= (1 << (c&7));  list is in order (which it should be), ranges of characters are detected and
3589            if ((options & PCRE_CASELESS) != 0)  handled appropriately. This function is mutually recursive with the function
3590              {  above.
             int uc = cd->fcc[c];           /* flip case */  
             classbits[uc/8] |= (1 << (uc&7));  
             }  
           class_charcount++;                /* in case a one-char range */  
           class_lastchar = c;  
           }  
3591    
3592          continue;   /* Go get the next char in the class */  Arguments:
3593          }    classbits     the bit map for characters < 256
3594      uchardptr     points to the pointer for extra data
3595      options       the options word
3596      cd            contains pointers to tables etc.
3597      p             points to row of 32-bit values, terminated by NOTACHAR
3598      except        character to omit; this is used when adding lists of
3599                      case-equivalent characters to avoid including the one we
3600                      already know about
3601    
3602    Returns:        the number of < 256 characters added
3603                    the pointer to extra data is updated
3604    */
3605    
3606        /* Handle a lone single character - we can get here for a normal