/[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 574 by ph10, Sat Nov 20 17:47:27 2010 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) 1997-2010 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_printint() function, which is
57  used by pcretest. DEBUG is not defined when building a production library. */  also used by pcretest. PCRE_DEBUG is not defined when building a production
58    library. */
59    
60  #ifdef DEBUG  #ifdef PCRE_DEBUG
61  #include "pcre_printint.src"  #include "pcre_printint.src"
62  #endif  #endif
63    
64    
65    /* Macro for setting individual bits in class bitmaps. */
66    
67    #define SETBIT(a,b) a[b/8] |= (1 << (b%8))
68    
69    /* Maximum length value to check against when making sure that the integer that
70    holds the compiled pattern length does not overflow. We make it a bit less than
71    INT_MAX to allow for adding in group terminating bytes, so that we don't have
72    to check them every time. */
73    
74    #define OFLOW_MAX (INT_MAX - 20)
75    
76    
77  /*************************************************  /*************************************************
78  *      Code parameters and static tables         *  *      Code parameters and static tables         *
79  *************************************************/  *************************************************/
80    
81  /* 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
82  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
83  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
84  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
85  compile time. */  does, however, check for an overrun. The largest amount I've seen used is 218,
86    so this number is very generous.
87    
88    The same workspace is used during the second, actual compile phase for
89    remembering forward references to groups so that they can be filled in at the
90    end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE
91    is 4 there is plenty of room. */
92    
93    #define COMPILE_WORK_SIZE (4096)
94    
95  #define BRASTACK_SIZE 200  /* The overrun tests check for a slightly smaller size so that they detect the
96    overrun before it actually does run off the end of the data block. */
97    
98    #define WORK_SIZE_CHECK (COMPILE_WORK_SIZE - 100)
99    
100    
101  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
# Line 72  are simple data values; negative values Line 103  are simple data values; negative values
103  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
104  is invalid. */  is invalid. */
105    
106  #if !EBCDIC   /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC
107    
108    /* This is the "normal" table for ASCII systems or for EBCDIC systems running
109    in UTF-8 mode. */
110    
111  static const short int escapes[] = {  static const short int escapes[] = {
112       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,                       0,
113       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,                       0,
114     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */       0,                       0,
115       0,      0,      0,      0,      0,      0,      0,      0,   /* H - O */       0,                       0,
116  -ESC_P, -ESC_Q,      0, -ESC_S,      0,      0,      0, -ESC_W,   /* P - W */       0,                       0,
117  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */       CHAR_COLON,              CHAR_SEMICOLON,
118     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
119       0,      0,      0,      0,      0,      0,  ESC_n,      0,   /* h - o */       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
120  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0,      0, -ESC_w,   /* p - w */       CHAR_COMMERCIAL_AT,      -ESC_A,
121       0,      0, -ESC_z                                            /* x - z */       -ESC_B,                  -ESC_C,
122         -ESC_D,                  -ESC_E,
123         0,                       -ESC_G,
124         -ESC_H,                  0,
125         0,                       -ESC_K,
126         0,                       0,
127         -ESC_N,                  0,
128         -ESC_P,                  -ESC_Q,
129         -ESC_R,                  -ESC_S,
130         0,                       0,
131         -ESC_V,                  -ESC_W,
132         -ESC_X,                  0,
133         -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,
134         CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,
135         CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,
136         CHAR_GRAVE_ACCENT,       7,
137         -ESC_b,                  0,
138         -ESC_d,                  ESC_e,
139         ESC_f,                   0,
140         -ESC_h,                  0,
141         0,                       -ESC_k,
142         0,                       0,
143         ESC_n,                   0,
144         -ESC_p,                  0,
145         ESC_r,                   -ESC_s,
146         ESC_tee,                 0,
147         -ESC_v,                  -ESC_w,
148         0,                       0,
149         -ESC_z
150  };  };
151    
152  #else         /* This is the "abnormal" table for EBCDIC systems */  #else
153    
154    /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
155    
156  static const short int escapes[] = {  static const short int escapes[] = {
157  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
158  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 96  static const short int escapes[] = { Line 162  static const short int escapes[] = {
162  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,
163  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',
164  /*  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,
165  /*  88 */     0,     0,      0,     '{',      0,     0,      0,      0,  /*  88 */-ESC_h,     0,      0,     '{',      0,     0,      0,      0,
166  /*  90 */     0,     0,      0,     'l',      0, ESC_n,      0, -ESC_p,  /*  90 */     0,     0, -ESC_k,     'l',      0, ESC_n,      0, -ESC_p,
167  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,
168  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,     0, -ESC_w,      0,  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,-ESC_v, -ESC_w,      0,
169  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,
170  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,
171  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
172  /*  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,
173  /*  C8 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
174  /*  D0 */   '}',     0,      0,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,-ESC_N,      0, -ESC_P,
175  /*  D8 */-ESC_Q,     0,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
176  /*  E0 */  '\\',     0, -ESC_S,       0,      0,     0, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
177  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
178  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,
179  /*  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 181  static const short int escapes[] = {
181  #endif  #endif
182    
183    
184  /* 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
185  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
186  as this is assumed for handling case independence. */  the number of relocations when a shared library is dynamically linked. The
187    string is built from string macros so that it works in UTF-8 mode on EBCDIC
188  static const char *const posix_names[] = {  platforms. */
189    "alpha", "lower", "upper",  
190    "alnum", "ascii", "blank", "cntrl", "digit", "graph",  typedef struct verbitem {
191    "print", "punct", "space", "word",  "xdigit" };    int   len;                 /* Length of verb name */
192      int   op;                  /* Op when no arg, or -1 if arg mandatory */
193      int   op_arg;              /* Op when arg present, or -1 if not allowed */
194    } verbitem;
195    
196    static const char verbnames[] =
197      "\0"                       /* Empty name is a shorthand for MARK */
198      STRING_MARK0
199      STRING_ACCEPT0
200      STRING_COMMIT0
201      STRING_F0
202      STRING_FAIL0
203      STRING_PRUNE0
204      STRING_SKIP0
205      STRING_THEN;
206    
207    static const verbitem verbs[] = {
208      { 0, -1,        OP_MARK },
209      { 4, -1,        OP_MARK },
210      { 6, OP_ACCEPT, -1 },
211      { 6, OP_COMMIT, -1 },
212      { 1, OP_FAIL,   -1 },
213      { 4, OP_FAIL,   -1 },
214      { 5, OP_PRUNE,  OP_PRUNE_ARG },
215      { 4, OP_SKIP,   OP_SKIP_ARG  },
216      { 4, OP_THEN,   OP_THEN_ARG  }
217    };
218    
219    static const int verbcount = sizeof(verbs)/sizeof(verbitem);
220    
221    
222    /* Tables of names of POSIX character classes and their lengths. The names are
223    now all in a single string, to reduce the number of relocations when a shared
224    library is dynamically loaded. The list of lengths is terminated by a zero
225    length entry. The first three must be alpha, lower, upper, as this is assumed
226    for handling case independence. */
227    
228    static const char posix_names[] =
229      STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
230      STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
231      STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
232      STRING_word0  STRING_xdigit;
233    
234  static const uschar posix_name_lengths[] = {  static const uschar posix_name_lengths[] = {
235    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 };
236    
237  /* 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
238  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
239  the vertical space characters. */  classes, there is some additional tweaking: for [:blank:] the vertical space
240    characters are removed, and for [:alpha:] and [:alnum:] the underscore
241    character is removed. The triples in the table consist of the base map offset,
242    second map offset or -1 if no second map, and a non-negative value for map
243    addition or a negative value for map subtraction (if there are two maps). The
244    absolute value of the third field has these meanings: 0 => no tweaking, 1 =>
245    remove vertical space characters, 2 => remove underscore. */
246    
247  static const int posix_class_maps[] = {  static const int posix_class_maps[] = {
248    cbit_lower, cbit_upper, -1,             /* alpha */    cbit_word,  cbit_digit, -2,             /* alpha */
249    cbit_lower, -1,         -1,             /* lower */    cbit_lower, -1,          0,             /* lower */
250    cbit_upper, -1,         -1,             /* upper */    cbit_upper, -1,          0,             /* upper */
251    cbit_digit, cbit_lower, cbit_upper,     /* alnum */    cbit_word,  -1,          2,             /* alnum - word without underscore */
252    cbit_print, cbit_cntrl, -1,             /* ascii */    cbit_print, cbit_cntrl,  0,             /* ascii */
253    cbit_space, -1,         -1,             /* blank - a GNU extension */    cbit_space, -1,          1,             /* blank - a GNU extension */
254    cbit_cntrl, -1,         -1,             /* cntrl */    cbit_cntrl, -1,          0,             /* cntrl */
255    cbit_digit, -1,         -1,             /* digit */    cbit_digit, -1,          0,             /* digit */
256    cbit_graph, -1,         -1,             /* graph */    cbit_graph, -1,          0,             /* graph */
257    cbit_print, -1,         -1,             /* print */    cbit_print, -1,          0,             /* print */
258    cbit_punct, -1,         -1,             /* punct */    cbit_punct, -1,          0,             /* punct */
259    cbit_space, -1,         -1,             /* space */    cbit_space, -1,          0,             /* space */
260    cbit_word,  -1,         -1,             /* word - a Perl extension */    cbit_word,  -1,          0,             /* word - a Perl extension */
261    cbit_xdigit,-1,         -1              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
262  };  };
263    
264    /* Table of substitutes for \d etc when PCRE_UCP is set. The POSIX class
265    substitutes must be in the order of the names, defined above, and there are
266    both positive and negative cases. NULL means no substitute. */
267    
268  /* The texts of compile-time error messages. These are "char *" because they  #ifdef SUPPORT_UCP
269  are passed to the outside world. */  static const uschar *substitutes[] = {
270      (uschar *)"\\P{Nd}",    /* \D */
271      (uschar *)"\\p{Nd}",    /* \d */
272      (uschar *)"\\P{Xsp}",   /* \S */       /* NOTE: Xsp is Perl space */
273      (uschar *)"\\p{Xsp}",   /* \s */
274      (uschar *)"\\P{Xwd}",   /* \W */
275      (uschar *)"\\p{Xwd}"    /* \w */
276    };
277    
278    static const uschar *posix_substitutes[] = {
279      (uschar *)"\\p{L}",     /* alpha */
280      (uschar *)"\\p{Ll}",    /* lower */
281      (uschar *)"\\p{Lu}",    /* upper */
282      (uschar *)"\\p{Xan}",   /* alnum */
283      NULL,                   /* ascii */
284      (uschar *)"\\h",        /* blank */
285      NULL,                   /* cntrl */
286      (uschar *)"\\p{Nd}",    /* digit */
287      NULL,                   /* graph */
288      NULL,                   /* print */
289      NULL,                   /* punct */
290      (uschar *)"\\p{Xps}",   /* space */    /* NOTE: Xps is POSIX space */
291      (uschar *)"\\p{Xwd}",   /* word */
292      NULL,                   /* xdigit */
293      /* Negated cases */
294      (uschar *)"\\P{L}",     /* ^alpha */
295      (uschar *)"\\P{Ll}",    /* ^lower */
296      (uschar *)"\\P{Lu}",    /* ^upper */
297      (uschar *)"\\P{Xan}",   /* ^alnum */
298      NULL,                   /* ^ascii */
299      (uschar *)"\\H",        /* ^blank */
300      NULL,                   /* ^cntrl */
301      (uschar *)"\\P{Nd}",    /* ^digit */
302      NULL,                   /* ^graph */
303      NULL,                   /* ^print */
304      NULL,                   /* ^punct */
305      (uschar *)"\\P{Xps}",   /* ^space */   /* NOTE: Xps is POSIX space */
306      (uschar *)"\\P{Xwd}",   /* ^word */
307      NULL                    /* ^xdigit */
308    };
309    #define POSIX_SUBSIZE (sizeof(posix_substitutes)/sizeof(uschar *))
310    #endif
311    
312  static const char *error_texts[] = {  #define STRING(a)  # a
313    "no error",  #define XSTRING(s) STRING(s)
314    "\\ at end of pattern",  
315    "\\c at end of pattern",  /* The texts of compile-time error messages. These are "char *" because they
316    "unrecognized character follows \\",  are passed to the outside world. Do not ever re-use any error number, because
317    "numbers out of order in {} quantifier",  they are documented. Always add a new error instead. Messages marked DEAD below
318    are no longer used. This used to be a table of strings, but in order to reduce
319    the number of relocations needed when a shared library is loaded dynamically,
320    it is now one long string. We cannot use a table of offsets, because the
321    lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
322    simply count through to the one we want - this isn't a performance issue
323    because these strings are used only when there is a compilation error.
324    
325    Each substring ends with \0 to insert a null character. This includes the final
326    substring, so that the whole string ends with \0\0, which can be detected when
327    counting through. */
328    
329    static const char error_texts[] =
330      "no error\0"
331      "\\ at end of pattern\0"
332      "\\c at end of pattern\0"
333      "unrecognized character follows \\\0"
334      "numbers out of order in {} quantifier\0"
335    /* 5 */    /* 5 */
336    "number too big in {} quantifier",    "number too big in {} quantifier\0"
337    "missing terminating ] for character class",    "missing terminating ] for character class\0"
338    "invalid escape sequence in character class",    "invalid escape sequence in character class\0"
339    "range out of order in character class",    "range out of order in character class\0"
340    "nothing to repeat",    "nothing to repeat\0"
341    /* 10 */    /* 10 */
342    "operand of unlimited repeat could match the empty string",    "operand of unlimited repeat could match the empty string\0"  /** DEAD **/
343    "internal error: unexpected repeat",    "internal error: unexpected repeat\0"
344    "unrecognized character after (?",    "unrecognized character after (? or (?-\0"
345    "POSIX named classes are supported only within a class",    "POSIX named classes are supported only within a class\0"
346    "missing )",    "missing )\0"
347    /* 15 */    /* 15 */
348    "reference to non-existent subpattern",    "reference to non-existent subpattern\0"
349    "erroffset passed as NULL",    "erroffset passed as NULL\0"
350    "unknown option bit(s) set",    "unknown option bit(s) set\0"
351    "missing ) after comment",    "missing ) after comment\0"
352    "parentheses nested too deeply",    "parentheses nested too deeply\0"  /** DEAD **/
353    /* 20 */    /* 20 */
354    "regular expression too large",    "regular expression is too large\0"
355    "failed to get memory",    "failed to get memory\0"
356    "unmatched parentheses",    "unmatched parentheses\0"
357    "internal error: code overflow",    "internal error: code overflow\0"
358    "unrecognized character after (?<",    "unrecognized character after (?<\0"
359    /* 25 */    /* 25 */
360    "lookbehind assertion is not fixed length",    "lookbehind assertion is not fixed length\0"
361    "malformed number after (?(",    "malformed number or name after (?(\0"
362    "conditional group contains more than two branches",    "conditional group contains more than two branches\0"
363    "assertion expected after (?(",    "assertion expected after (?(\0"
364    "(?R or (?digits must be followed by )",    "(?R or (?[+-]digits must be followed by )\0"
365    /* 30 */    /* 30 */
366    "unknown POSIX class name",    "unknown POSIX class name\0"
367    "POSIX collating elements are not supported",    "POSIX collating elements are not supported\0"
368    "this version of PCRE is not compiled with PCRE_UTF8 support",    "this version of PCRE is not compiled with PCRE_UTF8 support\0"
369    "spare error",    "spare error\0"  /** DEAD **/
370    "character value in \\x{...} sequence is too large",    "character value in \\x{...} sequence is too large\0"
371    /* 35 */    /* 35 */
372    "invalid condition (?(0)",    "invalid condition (?(0)\0"
373    "\\C not allowed in lookbehind assertion",    "\\C not allowed in lookbehind assertion\0"
374    "PCRE does not support \\L, \\l, \\N, \\U, or \\u",    "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
375    "number after (?C is > 255",    "number after (?C is > 255\0"
376    "closing ) for (?C expected",    "closing ) for (?C expected\0"
377    /* 40 */    /* 40 */
378    "recursive call could loop indefinitely",    "recursive call could loop indefinitely\0"
379    "unrecognized character after (?P",    "unrecognized character after (?P\0"
380    "syntax error after (?P",    "syntax error in subpattern name (missing terminator)\0"
381    "two named groups have the same name",    "two named subpatterns have the same name\0"
382    "invalid UTF-8 string",    "invalid UTF-8 string\0"
383    /* 45 */    /* 45 */
384    "support for \\P, \\p, and \\X has not been compiled",    "support for \\P, \\p, and \\X has not been compiled\0"
385    "malformed \\P or \\p sequence",    "malformed \\P or \\p sequence\0"
386    "unknown property name after \\P or \\p"    "unknown property name after \\P or \\p\0"
387  };    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
388      "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
389      /* 50 */
390      "repeated subpattern is too long\0"    /** DEAD **/
391      "octal value is greater than \\377 (not in UTF-8 mode)\0"
392      "internal error: overran compiling workspace\0"
393      "internal error: previously-checked referenced subpattern not found\0"
394      "DEFINE group contains more than one branch\0"
395      /* 55 */
396      "repeating a DEFINE group is not allowed\0"
397      "inconsistent NEWLINE options\0"
398      "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
399      "a numbered reference must not be zero\0"
400      "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
401      /* 60 */
402      "(*VERB) not recognized\0"
403      "number is too big\0"
404      "subpattern name expected\0"
405      "digit expected after (?+\0"
406      "] is an invalid data character in JavaScript compatibility mode\0"
407      /* 65 */
408      "different names for subpatterns of the same number are not allowed\0"
409      "(*MARK) must have an argument\0"
410      "this version of PCRE is not compiled with PCRE_UCP support\0"
411      "\\c must be followed by an ASCII character\0"
412      ;
413    
414  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
415  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 427  For convenience, we use the same bit def
427    
428  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
429    
430  #if !EBCDIC    /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC
431    
432    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
433    UTF-8 mode. */
434    
435  static const unsigned char digitab[] =  static const unsigned char digitab[] =
436    {    {
437    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
# Line 265  static const unsigned char digitab[] = Line 467  static const unsigned char digitab[] =
467    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
468    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
469    
470  #else          /* This is the "abnormal" case, for EBCDIC systems */  #else
471    
472    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
473    
474  static const unsigned char digitab[] =  static const unsigned char digitab[] =
475    {    {
476    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
# Line 279  static const unsigned char digitab[] = Line 484  static const unsigned char digitab[] =
484    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */
485    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */
486    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */
487    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88-     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88- 95    */
488    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */
489    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */
490    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */
# Line 313  static const unsigned char ebcdic_charta Line 518  static const unsigned char ebcdic_charta
518    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */
519    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */
520    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */
521    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88-  */    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88- 95 */
522    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */
523    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */
524    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */
# Line 340  static const unsigned char ebcdic_charta Line 545  static const unsigned char ebcdic_charta
545  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
546    
547  static BOOL  static BOOL
548    compile_regex(int, int, int *, uschar **, const uschar **, int *, BOOL, int,    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,
549      int *, int *, branch_chain *, compile_data *);      int *, int *, branch_chain *, compile_data *, int *);
550    
551    
552    
553    /*************************************************
554    *            Find an error text                  *
555    *************************************************/
556    
557    /* The error texts are now all in one long string, to save on relocations. As
558    some of the text is of unknown length, we can't use a table of offsets.
559    Instead, just count through the strings. This is not a performance issue
560    because it happens only when there has been a compilation error.
561    
562    Argument:   the error number
563    Returns:    pointer to the error string
564    */
565    
566    static const char *
567    find_error_text(int n)
568    {
569    const char *s = error_texts;
570    for (; n > 0; n--)
571      {
572      while (*s++ != 0) {};
573      if (*s == 0) return "Error text not found (please report)";
574      }
575    return s;
576    }
577    
578    
579  /*************************************************  /*************************************************
# Line 351  static BOOL Line 582  static BOOL
582    
583  /* 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
584  positive value for a simple escape such as \n, or a negative value which  positive value for a simple escape such as \n, or a negative value which
585  encodes one of the more complicated things such as \d. When UTF-8 is enabled,  encodes one of the more complicated things such as \d. A backreference to group
586  a positive value greater than 255 may be returned. On entry, ptr is pointing at  n is returned as -(ESC_REF + n); ESC_REF is the highest ESC_xxx macro. When
587  the \. On exit, it is on the final character of the escape sequence.  UTF-8 is enabled, a positive value greater than 255 may be returned. On entry,
588    ptr is pointing at the \. On exit, it is on the final character of the escape
589    sequence.
590    
591  Arguments:  Arguments:
592    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
# Line 364  Arguments: Line 597  Arguments:
597    
598  Returns:         zero or positive => a data character  Returns:         zero or positive => a data character
599                   negative => a special escape sequence                   negative => a special escape sequence
600                   on error, errorptr is set                   on error, errorcodeptr is set
601  */  */
602    
603  static int  static int
604  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,
605    int options, BOOL isclass)    int options, BOOL isclass)
606  {  {
607  const uschar *ptr = *ptrptr;  BOOL utf8 = (options & PCRE_UTF8) != 0;
608    const uschar *ptr = *ptrptr + 1;
609  int c, i;  int c, i;
610    
611    GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
612    ptr--;                            /* Set pointer back to the last byte */
613    
614  /* 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. */
615    
 c = *(++ptr);  
616  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
617    
618  /* 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
619  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.
620  Otherwise further processing may be required. */  Otherwise further processing may be required. */
621    
622  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
623  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */
624  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0) c = i;
625    
626  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
627  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */
628  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
629  #endif  #endif
630    
# Line 397  else if ((i = escapes[c - 0x48]) != 0) Line 633  else if ((i = escapes[c - 0x48]) != 0)
633  else  else
634    {    {
635    const uschar *oldptr;    const uschar *oldptr;
636      BOOL braced, negated;
637    
638    switch (c)    switch (c)
639      {      {
640      /* 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
641      error. */      error. */
642    
643      case 'l':      case CHAR_l:
644      case 'L':      case CHAR_L:
645      case 'N':      case CHAR_u:
646      case 'u':      case CHAR_U:
     case 'U':  
647      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
648      break;      break;
649    
650        /* \g must be followed by one of a number of specific things:
651    
652        (1) A number, either plain or braced. If positive, it is an absolute
653        backreference. If negative, it is a relative backreference. This is a Perl
654        5.10 feature.
655    
656        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
657        is part of Perl's movement towards a unified syntax for back references. As
658        this is synonymous with \k{name}, we fudge it up by pretending it really
659        was \k.
660    
661        (3) For Oniguruma compatibility we also support \g followed by a name or a
662        number either in angle brackets or in single quotes. However, these are
663        (possibly recursive) subroutine calls, _not_ backreferences. Just return
664        the -ESC_g code (cf \k). */
665    
666        case CHAR_g:
667        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
668          {
669          c = -ESC_g;
670          break;
671          }
672    
673        /* Handle the Perl-compatible cases */
674    
675        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
676          {
677          const uschar *p;
678          for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
679            if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
680          if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
681            {
682            c = -ESC_k;
683            break;
684            }
685          braced = TRUE;
686          ptr++;
687          }
688        else braced = FALSE;
689    
690        if (ptr[1] == CHAR_MINUS)
691          {
692          negated = TRUE;
693          ptr++;
694          }
695        else negated = FALSE;
696    
697        c = 0;
698        while ((digitab[ptr[1]] & ctype_digit) != 0)
699          c = c * 10 + *(++ptr) - CHAR_0;
700    
701        if (c < 0)   /* Integer overflow */
702          {
703          *errorcodeptr = ERR61;
704          break;
705          }
706    
707        if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
708          {
709          *errorcodeptr = ERR57;
710          break;
711          }
712    
713        if (c == 0)
714          {
715          *errorcodeptr = ERR58;
716          break;
717          }
718    
719        if (negated)
720          {
721          if (c > bracount)
722            {
723            *errorcodeptr = ERR15;
724            break;
725            }
726          c = bracount - (c - 1);
727          }
728    
729        c = -(ESC_REF + c);
730        break;
731    
732      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
733      starting with one that is not zero is not straightforward. By experiment,      starting with one that is not zero is not straightforward. By experiment,
734      the way Perl works seems to be as follows:      the way Perl works seems to be as follows:
# Line 422  else Line 741  else
741      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
742      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
743    
744      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:
745      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
746    
747      if (!isclass)      if (!isclass)
748        {        {
749        oldptr = ptr;        oldptr = ptr;
750        c -= '0';        c -= CHAR_0;
751        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
752          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - CHAR_0;
753          if (c < 0)    /* Integer overflow */
754            {
755            *errorcodeptr = ERR61;
756            break;
757            }
758        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
759          {          {
760          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 443  else Line 767  else
767      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.
768      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
769    
770      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
771        {        {
772        ptr--;        ptr--;
773        c = 0;        c = 0;
# Line 451  else Line 775  else
775        }        }
776    
777      /* \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
778      larger first octal digit. */      larger first octal digit. The original code used just to take the least
779        significant 8 bits of octal numbers (I think this is what early Perls used
780      case '0':      to do). Nowadays we allow for larger numbers in UTF-8 mode, but no more
781      c -= '0';      than 3 octal digits. */
782      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')  
783          c = c * 8 + *(++ptr) - '0';      case CHAR_0:
784      c &= 255;     /* Take least significant 8 bits */      c -= CHAR_0;
785        while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
786            c = c * 8 + *(++ptr) - CHAR_0;
787        if (!utf8 && c > 255) *errorcodeptr = ERR51;
788      break;      break;
789    
790      /* \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
791      which can be greater than 0xff, but only if the ddd are hex digits. */      than 0xff in utf8 mode, but only if the ddd are hex digits. If not, { is
792        treated as a data character. */
793    
794      case 'x':      case CHAR_x:
795  #ifdef SUPPORT_UTF8      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
     if (ptr[1] == '{' && (options & PCRE_UTF8) != 0)  
796        {        {
797        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
798        register int count = 0;        int count = 0;
799    
800        c = 0;        c = 0;
801        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
802          {          {
803          int cc = *pt++;          register int cc = *pt++;
804            if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
805          count++;          count++;
806  #if !EBCDIC    /* ASCII coding */  
807          if (cc >= 'a') cc -= 32;               /* Convert to upper case */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
808          c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
809  #else          /* EBCDIC coding */          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
810          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */  #else           /* EBCDIC coding */
811          c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
812            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
813  #endif  #endif
814          }          }
815        if (*pt == '}')  
816          if (*pt == CHAR_RIGHT_CURLY_BRACKET)
817          {          {
818          if (c < 0 || count > 8) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
819          ptr = pt;          ptr = pt;
820          break;          break;
821          }          }
822    
823        /* 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
824        recognize this construct; fall through to the normal \x handling. */        recognize this construct; fall through to the normal \x handling. */
825        }        }
 #endif  
826    
827      /* Read just a single hex char */      /* Read just a single-byte hex-defined char */
828    
829      c = 0;      c = 0;
830      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
831        {        {
832        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
833        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
834  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
835        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
836        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
837  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
838        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
839        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
840  #endif  #endif
841        }        }
842      break;      break;
843    
844      /* Other special escapes not starting with a digit are straightforward */      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
845        An error is given if the byte following \c is not an ASCII character. This
846        coding is ASCII-specific, but then the whole concept of \cx is
847        ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
848    
849      case 'c':      case CHAR_c:
850      c = *(++ptr);      c = *(++ptr);
851      if (c == 0)      if (c == 0)
852        {        {
853        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
854        return 0;        break;
855        }        }
856    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
857      /* A letter is upper-cased; then the 0x40 bit is flipped. This coding      if (c > 127)  /* Excludes all non-ASCII in either mode */
858      is ASCII-specific, but then the whole concept of \cx is ASCII-specific.        {
859      (However, an EBCDIC equivalent has now been added.) */        *errorcodeptr = ERR68;
860          break;
861  #if !EBCDIC    /* ASCII coding */        }
862      if (c >= 'a' && c <= 'z') c -= 32;      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
863      c ^= 0x40;      c ^= 0x40;
864  #else          /* EBCDIC coding */  #else             /* EBCDIC coding */
865      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
866      c ^= 0xC0;      c ^= 0xC0;
867  #endif  #endif
868      break;      break;
869    
870      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
871      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
872      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
873      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
874      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
875    
876      default:      default:
877      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 550  else Line 884  else
884      }      }
885    }    }
886    
887    /* Perl supports \N{name} for character names, as well as plain \N for "not
888    newline". PCRE does not support \N{name}. */
889    
890    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET)
891      *errorcodeptr = ERR37;
892    
893    /* If PCRE_UCP is set, we change the values for \d etc. */
894    
895    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
896      c -= (ESC_DU - ESC_D);
897    
898    /* Set the pointer to the final character before returning. */
899    
900  *ptrptr = ptr;  *ptrptr = ptr;
901  return c;  return c;
902  }  }
# Line 569  escape sequence. Line 916  escape sequence.
916  Argument:  Argument:
917    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
918    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
919      dptr           points to an int that is set to the detailed property value
920    errorcodeptr   points to the error code variable    errorcodeptr   points to the error code variable
921    
922  Returns:     value from ucp_type_table, or -1 for an invalid type  Returns:         type value from ucp_type_table, or -1 for an invalid type
923  */  */
924    
925  static int  static int
926  get_ucp(const uschar **ptrptr, BOOL *negptr, int *errorcodeptr)  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)
927  {  {
928  int c, i, bot, top;  int c, i, bot, top;
929  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
930  char name[4];  char name[32];
931    
932  c = *(++ptr);  c = *(++ptr);
933  if (c == 0) goto ERROR_RETURN;  if (c == 0) goto ERROR_RETURN;
934    
935  *negptr = FALSE;  *negptr = FALSE;
936    
937  /* \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
938  preceded by ^ for negation. */  negation. */
939    
940  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
941    {    {
942    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
943      {      {
944      *negptr = TRUE;      *negptr = TRUE;
945      ptr++;      ptr++;
946      }      }
947    for (i = 0; i <= 2; i++)    for (i = 0; i < (int)sizeof(name) - 1; i++)
948      {      {
949      c = *(++ptr);      c = *(++ptr);
950      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
951      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
952      name[i] = c;      name[i] = c;
953      }      }
954    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;  
     }  
955    name[i] = 0;    name[i] = 0;
956    }    }
957    
# Line 628  top = _pcre_utt_size; Line 972  top = _pcre_utt_size;
972    
973  while (bot < top)  while (bot < top)
974    {    {
975    i = (bot + top)/2;    i = (bot + top) >> 1;
976    c = strcmp(name, _pcre_utt[i].name);    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);
977    if (c == 0) return _pcre_utt[i].value;    if (c == 0)
978        {
979        *dptr = _pcre_utt[i].value;
980        return _pcre_utt[i].type;
981        }
982    if (c > 0) bot = i + 1; else top = i;    if (c > 0) bot = i + 1; else top = i;
983    }    }
984    
 UNKNOWN_RETURN:  
985  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
986  *ptrptr = ptr;  *ptrptr = ptr;
987  return -1;  return -1;
# Line 669  is_counted_repeat(const uschar *p) Line 1016  is_counted_repeat(const uschar *p)
1016  {  {
1017  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
1018  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
1019  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
1020    
1021  if (*p++ != ',') return FALSE;  if (*p++ != CHAR_COMMA) return FALSE;
1022  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
1023    
1024  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
1025  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
1026    
1027  return (*p == '}');  return (*p == CHAR_RIGHT_CURLY_BRACKET);
1028  }  }
1029    
1030    
# Line 710  int max = -1; Line 1057  int max = -1;
1057  /* 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
1058  an integer overflow. */  an integer overflow. */
1059    
1060  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;
1061  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1062    {    {
1063    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 720  if (min < 0 || min > 65535) Line 1067  if (min < 0 || min > 65535)
1067  /* 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.
1068  Also, max must not be less than min. */  Also, max must not be less than min. */
1069    
1070  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1071    {    {
1072    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1073      {      {
1074      max = 0;      max = 0;
1075      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;
1076      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1077        {        {
1078        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 750  return p; Line 1097  return p;
1097    
1098    
1099  /*************************************************  /*************************************************
1100    *  Subroutine for finding forward reference      *
1101    *************************************************/
1102    
1103    /* This recursive function is called only from find_parens() below. The
1104    top-level call starts at the beginning of the pattern. All other calls must
1105    start at a parenthesis. It scans along a pattern's text looking for capturing
1106    subpatterns, and counting them. If it finds a named pattern that matches the
1107    name it is given, it returns its number. Alternatively, if the name is NULL, it
1108    returns when it reaches a given numbered subpattern. We know that if (?P< is
1109    encountered, the name will be terminated by '>' because that is checked in the
1110    first pass. Recursion is used to keep track of subpatterns that reset the
1111    capturing group numbers - the (?| feature.
1112    
1113    Arguments:
1114      ptrptr       address of the current character pointer (updated)
1115      cd           compile background data
1116      name         name to seek, or NULL if seeking a numbered subpattern
1117      lorn         name length, or subpattern number if name is NULL
1118      xmode        TRUE if we are in /x mode
1119      utf8         TRUE if we are in UTF-8 mode
1120      count        pointer to the current capturing subpattern number (updated)
1121    
1122    Returns:       the number of the named subpattern, or -1 if not found
1123    */
1124    
1125    static int
1126    find_parens_sub(uschar **ptrptr, compile_data *cd, const uschar *name, int lorn,
1127      BOOL xmode, BOOL utf8, int *count)
1128    {
1129    uschar *ptr = *ptrptr;
1130    int start_count = *count;
1131    int hwm_count = start_count;
1132    BOOL dup_parens = FALSE;
1133    
1134    /* If the first character is a parenthesis, check on the type of group we are
1135    dealing with. The very first call may not start with a parenthesis. */
1136    
1137    if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1138      {
1139      /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1140    
1141      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1142    
1143      /* Handle a normal, unnamed capturing parenthesis. */
1144    
1145      else if (ptr[1] != CHAR_QUESTION_MARK)
1146        {
1147        *count += 1;
1148        if (name == NULL && *count == lorn) return *count;
1149        ptr++;
1150        }
1151    
1152      /* All cases now have (? at the start. Remember when we are in a group
1153      where the parenthesis numbers are duplicated. */
1154    
1155      else if (ptr[2] == CHAR_VERTICAL_LINE)
1156        {
1157        ptr += 3;
1158        dup_parens = TRUE;
1159        }
1160    
1161      /* Handle comments; all characters are allowed until a ket is reached. */
1162    
1163      else if (ptr[2] == CHAR_NUMBER_SIGN)
1164        {
1165        for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1166        goto FAIL_EXIT;
1167        }
1168    
1169      /* Handle a condition. If it is an assertion, just carry on so that it
1170      is processed as normal. If not, skip to the closing parenthesis of the
1171      condition (there can't be any nested parens). */
1172    
1173      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1174        {
1175        ptr += 2;
1176        if (ptr[1] != CHAR_QUESTION_MARK)
1177          {
1178          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1179          if (*ptr != 0) ptr++;
1180          }
1181        }
1182    
1183      /* Start with (? but not a condition. */
1184    
1185      else
1186        {
1187        ptr += 2;
1188        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1189    
1190        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1191    
1192        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1193            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1194          {
1195          int term;
1196          const uschar *thisname;
1197          *count += 1;
1198          if (name == NULL && *count == lorn) return *count;
1199          term = *ptr++;
1200          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1201          thisname = ptr;
1202          while (*ptr != term) ptr++;
1203          if (name != NULL && lorn == ptr - thisname &&
1204              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1205            return *count;
1206          term++;
1207          }
1208        }
1209      }
1210    
1211    /* Past any initial parenthesis handling, scan for parentheses or vertical
1212    bars. */
1213    
1214    for (; *ptr != 0; ptr++)
1215      {
1216      /* Skip over backslashed characters and also entire \Q...\E */
1217    
1218      if (*ptr == CHAR_BACKSLASH)
1219        {
1220        if (*(++ptr) == 0) goto FAIL_EXIT;
1221        if (*ptr == CHAR_Q) for (;;)
1222          {
1223          while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1224          if (*ptr == 0) goto FAIL_EXIT;
1225          if (*(++ptr) == CHAR_E) break;
1226          }
1227        continue;
1228        }
1229    
1230      /* Skip over character classes; this logic must be similar to the way they
1231      are handled for real. If the first character is '^', skip it. Also, if the
1232      first few characters (either before or after ^) are \Q\E or \E we skip them
1233      too. This makes for compatibility with Perl. Note the use of STR macros to
1234      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1235    
1236      if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1237        {
1238        BOOL negate_class = FALSE;
1239        for (;;)
1240          {
1241          if (ptr[1] == CHAR_BACKSLASH)
1242            {
1243            if (ptr[2] == CHAR_E)
1244              ptr+= 2;
1245            else if (strncmp((const char *)ptr+2,
1246                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1247              ptr += 4;
1248            else
1249              break;
1250            }
1251          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1252            {
1253            negate_class = TRUE;
1254            ptr++;
1255            }
1256          else break;
1257          }
1258    
1259        /* If the next character is ']', it is a data character that must be
1260        skipped, except in JavaScript compatibility mode. */
1261    
1262        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1263            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1264          ptr++;
1265    
1266        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1267          {
1268          if (*ptr == 0) return -1;
1269          if (*ptr == CHAR_BACKSLASH)
1270            {
1271            if (*(++ptr) == 0) goto FAIL_EXIT;
1272            if (*ptr == CHAR_Q) for (;;)
1273              {
1274              while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1275              if (*ptr == 0) goto FAIL_EXIT;
1276              if (*(++ptr) == CHAR_E) break;
1277              }
1278            continue;
1279            }
1280          }
1281        continue;
1282        }
1283    
1284      /* Skip comments in /x mode */
1285    
1286      if (xmode && *ptr == CHAR_NUMBER_SIGN)
1287        {
1288        ptr++;
1289        while (*ptr != 0)
1290          {
1291          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1292          ptr++;
1293    #ifdef SUPPORT_UTF8
1294          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
1295    #endif
1296          }
1297        if (*ptr == 0) goto FAIL_EXIT;
1298        continue;
1299        }
1300    
1301      /* Check for the special metacharacters */
1302    
1303      if (*ptr == CHAR_LEFT_PARENTHESIS)
1304        {
1305        int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, count);
1306        if (rc > 0) return rc;
1307        if (*ptr == 0) goto FAIL_EXIT;
1308        }
1309    
1310      else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1311        {
1312        if (dup_parens && *count < hwm_count) *count = hwm_count;
1313        goto FAIL_EXIT;
1314        }
1315    
1316      else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1317        {
1318        if (*count > hwm_count) hwm_count = *count;
1319        *count = start_count;
1320        }
1321      }
1322    
1323    FAIL_EXIT:
1324    *ptrptr = ptr;
1325    return -1;
1326    }
1327    
1328    
1329    
1330    
1331    /*************************************************
1332    *       Find forward referenced subpattern       *
1333    *************************************************/
1334    
1335    /* This function scans along a pattern's text looking for capturing
1336    subpatterns, and counting them. If it finds a named pattern that matches the
1337    name it is given, it returns its number. Alternatively, if the name is NULL, it
1338    returns when it reaches a given numbered subpattern. This is used for forward
1339    references to subpatterns. We used to be able to start this scan from the
1340    current compiling point, using the current count value from cd->bracount, and
1341    do it all in a single loop, but the addition of the possibility of duplicate
1342    subpattern numbers means that we have to scan from the very start, in order to
1343    take account of such duplicates, and to use a recursive function to keep track
1344    of the different types of group.
1345    
1346    Arguments:
1347      cd           compile background data
1348      name         name to seek, or NULL if seeking a numbered subpattern
1349      lorn         name length, or subpattern number if name is NULL
1350      xmode        TRUE if we are in /x mode
1351      utf8         TRUE if we are in UTF-8 mode
1352    
1353    Returns:       the number of the found subpattern, or -1 if not found
1354    */
1355    
1356    static int
1357    find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode,
1358      BOOL utf8)
1359    {
1360    uschar *ptr = (uschar *)cd->start_pattern;
1361    int count = 0;
1362    int rc;
1363    
1364    /* If the pattern does not start with an opening parenthesis, the first call
1365    to find_parens_sub() will scan right to the end (if necessary). However, if it
1366    does start with a parenthesis, find_parens_sub() will return when it hits the
1367    matching closing parens. That is why we have to have a loop. */
1368    
1369    for (;;)
1370      {
1371      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, &count);
1372      if (rc > 0 || *ptr++ == 0) break;
1373      }
1374    
1375    return rc;
1376    }
1377    
1378    
1379    
1380    
1381    /*************************************************
1382  *      Find first significant op code            *  *      Find first significant op code            *
1383  *************************************************/  *************************************************/
1384    
# Line 798  for (;;) Line 1427  for (;;)
1427    
1428      case OP_CALLOUT:      case OP_CALLOUT:
1429      case OP_CREF:      case OP_CREF:
1430      case OP_BRANUMBER:      case OP_NCREF:
1431        case OP_RREF:
1432        case OP_NRREF:
1433        case OP_DEF:
1434      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1435      break;      break;
1436    
# Line 813  for (;;) Line 1445  for (;;)
1445    
1446    
1447  /*************************************************  /*************************************************
1448  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1449  *************************************************/  *************************************************/
1450    
1451  /* 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,
1452  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.
1453  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
1454    temporarily terminated with OP_END when this function is called.
1455    
1456    This function is called when a backward assertion is encountered, so that if it
1457    fails, the error message can point to the correct place in the pattern.
1458    However, we cannot do this when the assertion contains subroutine calls,
1459    because they can be forward references. We solve this by remembering this case
1460    and doing the check at the end; a flag specifies which mode we are running in.
1461    
1462  Arguments:  Arguments:
1463    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1464    options  the compiling options    options  the compiling options
1465      atend    TRUE if called when the pattern is complete
1466      cd       the "compile data" structure
1467    
1468  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1469                 or -1 if there is no fixed length,
1470               or -2 if \C was encountered               or -2 if \C was encountered
1471                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1472  */  */
1473    
1474  static int  static int
1475  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)
1476  {  {
1477  int length = -1;  int length = -1;
1478    
# Line 842  branch, check the length against that of Line 1485  branch, check the length against that of
1485  for (;;)  for (;;)
1486    {    {
1487    int d;    int d;
1488      uschar *ce, *cs;
1489    register int op = *cc;    register int op = *cc;
   if (op >= OP_BRA) op = OP_BRA;  
   
1490    switch (op)    switch (op)
1491      {      {
1492        case OP_CBRA:
1493      case OP_BRA:      case OP_BRA:
1494      case OP_ONCE:      case OP_ONCE:
1495      case OP_COND:      case OP_COND:
1496      d = find_fixedlength(cc, options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);
1497      if (d < 0) return d;      if (d < 0) return d;
1498      branchlength += d;      branchlength += d;
1499      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 873  for (;;) Line 1516  for (;;)
1516      branchlength = 0;      branchlength = 0;
1517      break;      break;
1518    
1519        /* A true recursion implies not fixed length, but a subroutine call may
1520        be OK. If the subroutine is a forward reference, we can't deal with
1521        it until the end of the pattern, so return -3. */
1522    
1523        case OP_RECURSE:
1524        if (!atend) return -3;
1525        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1526        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1527        if (cc > cs && cc < ce) return -1;                /* Recursion */
1528        d = find_fixedlength(cs + 2, options, atend, cd);
1529        if (d < 0) return d;
1530        branchlength += d;
1531        cc += 1 + LINK_SIZE;
1532        break;
1533    
1534      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1535    
1536      case OP_ASSERT:      case OP_ASSERT:
# Line 885  for (;;) Line 1543  for (;;)
1543      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1544    
1545      case OP_REVERSE:      case OP_REVERSE:
     case OP_BRANUMBER:  
1546      case OP_CREF:      case OP_CREF:
1547        case OP_NCREF:
1548        case OP_RREF:
1549        case OP_NRREF:
1550        case OP_DEF:
1551      case OP_OPT:      case OP_OPT:
1552      case OP_CALLOUT:      case OP_CALLOUT:
1553      case OP_SOD:      case OP_SOD:
1554      case OP_SOM:      case OP_SOM:
1555        case OP_SET_SOM:
1556      case OP_EOD:      case OP_EOD:
1557      case OP_EODN:      case OP_EODN:
1558      case OP_CIRC:      case OP_CIRC:
# Line 904  for (;;) Line 1566  for (;;)
1566    
1567      case OP_CHAR:      case OP_CHAR:
1568      case OP_CHARNC:      case OP_CHARNC:
1569        case OP_NOT:
1570      branchlength++;      branchlength++;
1571      cc += 2;      cc += 2;
1572  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1573      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1574        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1575  #endif  #endif
1576      break;      break;
1577    
# Line 921  for (;;) Line 1582  for (;;)
1582      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1583      cc += 4;      cc += 4;
1584  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1585      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1586        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1587  #endif  #endif
1588      break;      break;
1589    
1590      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1591      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1592        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1593      cc += 4;      cc += 4;
1594      break;      break;
1595    
# Line 937  for (;;) Line 1597  for (;;)
1597    
1598      case OP_PROP:      case OP_PROP:
1599      case OP_NOTPROP:      case OP_NOTPROP:
1600      cc++;      cc += 2;
1601      /* Fall through */      /* Fall through */
1602    
1603      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
# Line 947  for (;;) Line 1607  for (;;)
1607      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1608      case OP_WORDCHAR:      case OP_WORDCHAR:
1609      case OP_ANY:      case OP_ANY:
1610        case OP_ALLANY:
1611      branchlength++;      branchlength++;
1612      cc++;      cc++;
1613      break;      break;
# Line 1001  for (;;) Line 1662  for (;;)
1662    
1663    
1664  /*************************************************  /*************************************************
1665  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1666  *************************************************/  *************************************************/
1667    
1668  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1669  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1670    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1671    so that it can be called from pcre_study() when finding the minimum matching
1672    length.
1673    
1674  Arguments:  Arguments:
1675    code        points to start of expression    code        points to start of expression
1676    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1677    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1678    
1679  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
1680  */  */
1681    
1682  static const uschar *  const uschar *
1683  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1684  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
1685  for (;;)  for (;;)
1686    {    {
1687    register int c = *code;    register int c = *code;
1688    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1689    else if (c > OP_BRA)  
1690      /* XCLASS is used for classes that cannot be represented just by a bit
1691      map. This includes negated single high-valued characters. The length in
1692      the table is zero; the actual length is stored in the compiled code. */
1693    
1694      if (c == OP_XCLASS) code += GET(code, 1);
1695    
1696      /* Handle recursion */
1697    
1698      else if (c == OP_REVERSE)
1699        {
1700        if (number < 0) return (uschar *)code;
1701        code += _pcre_OP_lengths[c];
1702        }
1703    
1704      /* Handle capturing bracket */
1705    
1706      else if (c == OP_CBRA)
1707      {      {
1708      int n = c - OP_BRA;      int n = GET2(code, 1+LINK_SIZE);
     if (n > EXTRACT_BASIC_MAX) n = GET2(code, 2+LINK_SIZE);  
1709      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
1710      code += _pcre_OP_lengths[OP_BRA];      code += _pcre_OP_lengths[c];
1711      }      }
1712    
1713      /* Otherwise, we can get the item's length from the table, except that for
1714      repeated character types, we have to test for \p and \P, which have an extra
1715      two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1716      must add in its length. */
1717    
1718    else    else
1719      {      {
1720      code += _pcre_OP_lengths[c];      switch(c)
1721          {
1722          case OP_TYPESTAR:
1723          case OP_TYPEMINSTAR:
1724          case OP_TYPEPLUS:
1725          case OP_TYPEMINPLUS:
1726          case OP_TYPEQUERY:
1727          case OP_TYPEMINQUERY:
1728          case OP_TYPEPOSSTAR:
1729          case OP_TYPEPOSPLUS:
1730          case OP_TYPEPOSQUERY:
1731          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1732          break;
1733    
1734  #ifdef SUPPORT_UTF8        case OP_TYPEUPTO:
1735          case OP_TYPEMINUPTO:
1736          case OP_TYPEEXACT:
1737          case OP_TYPEPOSUPTO:
1738          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1739          break;
1740    
1741      /* In UTF-8 mode, opcodes that are followed by a character may be followed        case OP_MARK:
1742      by a multi-byte character. The length in the table is a minimum, so we have        case OP_PRUNE_ARG:
1743      to scan along to skip the extra bytes. All opcodes are less than 128, so we        case OP_SKIP_ARG:
1744      can use relatively efficient code. */        code += code[1];
1745          break;
1746    
1747          case OP_THEN_ARG:
1748          code += code[1+LINK_SIZE];
1749          break;
1750          }
1751    
1752        /* Add in the fixed length from the table */
1753    
1754        code += _pcre_OP_lengths[c];
1755    
1756      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1757      a multi-byte character. The length in the table is a minimum, so we have to
1758      arrange to skip the extra bytes. */
1759    
1760    #ifdef SUPPORT_UTF8
1761      if (utf8) switch(c)      if (utf8) switch(c)
1762        {        {
1763        case OP_CHAR:        case OP_CHAR:
# Line 1051  for (;;) Line 1765  for (;;)
1765        case OP_EXACT:        case OP_EXACT:
1766        case OP_UPTO:        case OP_UPTO:
1767        case OP_MINUPTO:        case OP_MINUPTO:
1768          case OP_POSUPTO:
1769        case OP_STAR:        case OP_STAR:
1770        case OP_MINSTAR:        case OP_MINSTAR:
1771          case OP_POSSTAR:
1772        case OP_PLUS:        case OP_PLUS:
1773        case OP_MINPLUS:        case OP_MINPLUS:
1774          case OP_POSPLUS:
1775        case OP_QUERY:        case OP_QUERY:
1776        case OP_MINQUERY:        case OP_MINQUERY:
1777        while ((*code & 0xc0) == 0x80) code++;        case OP_POSQUERY:
1778        break;        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
   
       /* XCLASS is used for classes that cannot be represented just by a bit  
       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;  
1779        break;        break;
1780        }        }
1781    #else
1782        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1783  #endif  #endif
1784      }      }
1785    }    }
# Line 1092  Returns:      pointer to the opcode for Line 1804  Returns:      pointer to the opcode for
1804  static const uschar *  static const uschar *
1805  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const uschar *code, BOOL utf8)
1806  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
1807  for (;;)  for (;;)
1808    {    {
1809    register int c = *code;    register int c = *code;
1810    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1811    else if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
1812    else if (c > OP_BRA)  
1813      {    /* XCLASS is used for classes that cannot be represented just by a bit
1814      code += _pcre_OP_lengths[OP_BRA];    map. This includes negated single high-valued characters. The length in
1815      }    the table is zero; the actual length is stored in the compiled code. */
1816    
1817      if (c == OP_XCLASS) code += GET(code, 1);
1818    
1819      /* Otherwise, we can get the item's length from the table, except that for
1820      repeated character types, we have to test for \p and \P, which have an extra
1821      two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1822      must add in its length. */
1823    
1824    else    else
1825      {      {
1826      code += _pcre_OP_lengths[c];      switch(c)
1827          {
1828          case OP_TYPESTAR:
1829          case OP_TYPEMINSTAR:
1830          case OP_TYPEPLUS:
1831          case OP_TYPEMINPLUS:
1832          case OP_TYPEQUERY:
1833          case OP_TYPEMINQUERY:
1834          case OP_TYPEPOSSTAR:
1835          case OP_TYPEPOSPLUS:
1836          case OP_TYPEPOSQUERY:
1837          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1838          break;
1839    
1840  #ifdef SUPPORT_UTF8        case OP_TYPEPOSUPTO:
1841          case OP_TYPEUPTO:
1842          case OP_TYPEMINUPTO:
1843          case OP_TYPEEXACT:
1844          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1845          break;
1846    
1847          case OP_MARK:
1848          case OP_PRUNE_ARG:
1849          case OP_SKIP_ARG:
1850          code += code[1];
1851          break;
1852    
1853          case OP_THEN_ARG:
1854          code += code[1+LINK_SIZE];
1855          break;
1856          }
1857    
1858        /* Add in the fixed length from the table */
1859    
1860        code += _pcre_OP_lengths[c];
1861    
1862      /* In UTF-8 mode, opcodes that are followed by a character may be followed      /* In UTF-8 mode, opcodes that are followed by a character may be followed
1863      by a multi-byte character. The length in the table is a minimum, so we have      by a multi-byte character. The length in the table is a minimum, so we have
1864      to scan along to skip the extra bytes. All opcodes are less than 128, so we      to arrange to skip the extra bytes. */
     can use relatively efficient code. */  
1865    
1866    #ifdef SUPPORT_UTF8
1867      if (utf8) switch(c)      if (utf8) switch(c)
1868        {        {
1869        case OP_CHAR:        case OP_CHAR:
# Line 1123  for (;;) Line 1871  for (;;)
1871        case OP_EXACT:        case OP_EXACT:
1872        case OP_UPTO:        case OP_UPTO:
1873        case OP_MINUPTO:        case OP_MINUPTO:
1874          case OP_POSUPTO:
1875        case OP_STAR:        case OP_STAR:
1876        case OP_MINSTAR:        case OP_MINSTAR:
1877          case OP_POSSTAR:
1878        case OP_PLUS:        case OP_PLUS:
1879        case OP_MINPLUS:        case OP_MINPLUS:
1880          case OP_POSPLUS:
1881        case OP_QUERY:        case OP_QUERY:
1882        case OP_MINQUERY:        case OP_MINQUERY:
1883        while ((*code & 0xc0) == 0x80) code++;        case OP_POSQUERY:
1884        break;        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
   
       /* XCLASS is used for classes that cannot be represented just by a bit  
       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;  
1885        break;        break;
1886        }        }
1887    #else
1888        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1889  #endif  #endif
1890      }      }
1891    }    }
# Line 1152  for (;;) Line 1898  for (;;)
1898  *************************************************/  *************************************************/
1899    
1900  /* 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
1901  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()
1902  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
1903  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
1904  whose current branch will already have been scanned.  backward and negative forward assertions when its final argument is TRUE. If we
1905    hit an unclosed bracket, we return "empty" - this means we've struck an inner
1906    bracket whose current branch will already have been scanned.
1907    
1908  Arguments:  Arguments:
1909    code        points to start of search    code        points to start of search
1910    endcode     points to where to stop    endcode     points to where to stop
1911    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
1912      cd          contains pointers to tables etc.
1913    
1914  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
1915  */  */
1916    
1917  static BOOL  static BOOL
1918  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
1919      compile_data *cd)
1920  {  {
1921  register int c;  register int c;
1922  for (code = first_significant_code(code + 1 + LINK_SIZE, NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);
1923       code < endcode;       code < endcode;
1924       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))
1925    {    {
# Line 1177  for (code = first_significant_code(code Line 1927  for (code = first_significant_code(code
1927    
1928    c = *code;    c = *code;
1929    
1930    if (c >= OP_BRA)    /* Skip over forward assertions; the other assertions are skipped by
1931      first_significant_code() with a TRUE final argument. */
1932    
1933      if (c == OP_ASSERT)
1934      {      {
1935      BOOL empty_branch;      do code += GET(code, 1); while (*code == OP_ALT);
1936      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      c = *code;
1937        continue;
1938        }
1939    
1940      /* Groups with zero repeats can of course be empty; skip them. */
1941    
1942      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)
1943        {
1944        code += _pcre_OP_lengths[c];
1945        do code += GET(code, 1); while (*code == OP_ALT);
1946        c = *code;
1947        continue;
1948        }
1949    
1950      /* Scan a closed bracket */    /* For a recursion/subroutine call, if its end has been reached, which
1951      implies a subroutine call, we can scan it. */
1952    
1953      empty_branch = FALSE;    if (c == OP_RECURSE)
1954        {
1955        BOOL empty_branch = FALSE;
1956        const uschar *scode = cd->start_code + GET(code, 1);
1957        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
1958      do      do
1959        {        {
1960        if (!empty_branch && could_be_empty_branch(code, endcode, utf8))        if (could_be_empty_branch(scode, endcode, utf8, cd))
1961            {
1962          empty_branch = TRUE;          empty_branch = TRUE;
1963            break;
1964            }
1965          scode += GET(scode, 1);
1966          }
1967        while (*scode == OP_ALT);
1968        if (!empty_branch) return FALSE;  /* All branches are non-empty */
1969        continue;
1970        }
1971    
1972      /* For other groups, scan the branches. */
1973    
1974      if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)
1975        {
1976        BOOL empty_branch;
1977        if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
1978    
1979        /* If a conditional group has only one branch, there is a second, implied,
1980        empty branch, so just skip over the conditional, because it could be empty.
1981        Otherwise, scan the individual branches of the group. */
1982    
1983        if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
1984        code += GET(code, 1);        code += GET(code, 1);
1985        else
1986          {
1987          empty_branch = FALSE;
1988          do
1989            {
1990            if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
1991              empty_branch = TRUE;
1992            code += GET(code, 1);
1993            }
1994          while (*code == OP_ALT);
1995          if (!empty_branch) return FALSE;   /* All branches are non-empty */
1996        }        }
1997      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
     code += 1 + LINK_SIZE;  
1998      c = *code;      c = *code;
1999        continue;
2000      }      }
2001    
2002    else switch (c)    /* Handle the other opcodes */
2003    
2004      switch (c)
2005      {      {
2006      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
2007        cannot be represented just by a bit map. This includes negated single
2008        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
2009        actual length is stored in the compiled code, so we must update "code"
2010        here. */
2011    
2012  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2013      case OP_XCLASS:      case OP_XCLASS:
2014      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
2015      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
2016  #endif  #endif
2017    
# Line 1247  for (code = first_significant_code(code Line 2055  for (code = first_significant_code(code
2055      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2056      case OP_WORDCHAR:      case OP_WORDCHAR:
2057      case OP_ANY:      case OP_ANY:
2058        case OP_ALLANY:
2059      case OP_ANYBYTE:      case OP_ANYBYTE:
2060      case OP_CHAR:      case OP_CHAR:
2061      case OP_CHARNC:      case OP_CHARNC:
2062      case OP_NOT:      case OP_NOT:
2063      case OP_PLUS:      case OP_PLUS:
2064      case OP_MINPLUS:      case OP_MINPLUS:
2065        case OP_POSPLUS:
2066      case OP_EXACT:      case OP_EXACT:
2067      case OP_NOTPLUS:      case OP_NOTPLUS:
2068      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
2069        case OP_NOTPOSPLUS:
2070      case OP_NOTEXACT:      case OP_NOTEXACT:
2071      case OP_TYPEPLUS:      case OP_TYPEPLUS:
2072      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
2073        case OP_TYPEPOSPLUS:
2074      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2075      return FALSE;      return FALSE;
2076    
2077        /* These are going to continue, as they may be empty, but we have to
2078        fudge the length for the \p and \P cases. */
2079    
2080        case OP_TYPESTAR:
2081        case OP_TYPEMINSTAR:
2082        case OP_TYPEPOSSTAR:
2083        case OP_TYPEQUERY:
2084        case OP_TYPEMINQUERY:
2085        case OP_TYPEPOSQUERY:
2086        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2087        break;
2088    
2089        /* Same for these */
2090    
2091        case OP_TYPEUPTO:
2092        case OP_TYPEMINUPTO:
2093        case OP_TYPEPOSUPTO:
2094        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
2095        break;
2096    
2097      /* End of branch */      /* End of branch */
2098    
2099      case OP_KET:      case OP_KET:
# Line 1270  for (code = first_significant_code(code Line 2102  for (code = first_significant_code(code
2102      case OP_ALT:      case OP_ALT:
2103      return TRUE;      return TRUE;
2104    
2105      /* In UTF-8 mode, STAR, MINSTAR, QUERY, MINQUERY, UPTO, and MINUPTO  may be      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2106      followed by a multibyte character */      MINUPTO, and POSUPTO may be followed by a multibyte character */
2107    
2108  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2109      case OP_STAR:      case OP_STAR:
2110      case OP_MINSTAR:      case OP_MINSTAR:
2111        case OP_POSSTAR:
2112      case OP_QUERY:      case OP_QUERY:
2113      case OP_MINQUERY:      case OP_MINQUERY:
2114        case OP_POSQUERY:
2115        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2116        break;
2117    
2118      case OP_UPTO:      case OP_UPTO:
2119      case OP_MINUPTO:      case OP_MINUPTO:
2120      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_POSUPTO:
2121        if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2122      break;      break;
2123  #endif  #endif
2124    
2125        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2126        string. */
2127    
2128        case OP_MARK:
2129        case OP_PRUNE_ARG:
2130        case OP_SKIP_ARG:
2131        code += code[1];
2132        break;
2133    
2134        case OP_THEN_ARG:
2135        code += code[1+LINK_SIZE];
2136        break;
2137    
2138        /* None of the remaining opcodes are required to match a character. */
2139    
2140        default:
2141        break;
2142      }      }
2143    }    }
2144    
# Line 1305  Arguments: Line 2161  Arguments:
2161    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2162    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2163    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2164      cd          pointers to tables etc
2165    
2166  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2167  */  */
2168    
2169  static BOOL  static BOOL
2170  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2171    BOOL utf8)    BOOL utf8, compile_data *cd)
2172  {  {
2173  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2174    {    {
2175    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2176        return FALSE;
2177    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2178    }    }
2179  return TRUE;  return TRUE;
# Line 1328  return TRUE; Line 2186  return TRUE;
2186  *************************************************/  *************************************************/
2187    
2188  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2189  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
2190  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2191  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2192    
2193    Originally, this function only recognized a sequence of letters between the
2194    terminators, but it seems that Perl recognizes any sequence of characters,
2195    though of course unknown POSIX names are subsequently rejected. Perl gives an
2196    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2197    didn't consider this to be a POSIX class. Likewise for [:1234:].
2198    
2199    The problem in trying to be exactly like Perl is in the handling of escapes. We
2200    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2201    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2202    below handles the special case of \], but does not try to do any other escape
2203    processing. This makes it different from Perl for cases such as [:l\ower:]
2204    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2205    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2206    I think.
2207    
2208  Argument:  Arguments:
2209    ptr      pointer to the initial [    ptr      pointer to the initial [
2210    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2211    
2212  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2213  */  */
2214    
2215  static BOOL  static BOOL
2216  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
2217  {  {
2218  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2219  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2220  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2221    {    {
2222    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else
2223    return TRUE;      {
2224        if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2225        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2226          {
2227          *endptr = ptr;
2228          return TRUE;
2229          }
2230        }
2231    }    }
2232  return FALSE;  return FALSE;
2233  }  }
# Line 1375  Returns:     a value representing the na Line 2252  Returns:     a value representing the na
2252  static int  static int
2253  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
2254  {  {
2255    const char *pn = posix_names;
2256  register int yield = 0;  register int yield = 0;
2257  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2258    {    {
2259    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2260      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
2261      pn += posix_name_lengths[yield] + 1;
2262    yield++;    yield++;
2263    }    }
2264  return -1;  return -1;
# Line 1394  return -1; Line 2273  return -1;
2273  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2274  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2275  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
2276  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
2277  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
2278  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
2279  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
2280  partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2281    OP_END.
2282    
2283    This function has been extended with the possibility of forward references for
2284    recursions and subroutine calls. It must also check the list of such references
2285    for the group we are dealing with. If it finds that one of the recursions in
2286    the current group is on this list, it adjusts the offset in the list, not the
2287    value in the reference (which is a group number).
2288    
2289  Arguments:  Arguments:
2290    group      points to the start of the group    group      points to the start of the group
2291    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
2292    utf8       TRUE in UTF-8 mode    utf8       TRUE in UTF-8 mode
2293    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
2294      save_hwm   the hwm forward reference pointer at the start of the group
2295    
2296  Returns:     nothing  Returns:     nothing
2297  */  */
2298    
2299  static void  static void
2300  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd)  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,
2301      uschar *save_hwm)
2302  {  {
2303  uschar *ptr = group;  uschar *ptr = group;
2304    
2305  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
2306    {    {
2307    int offset = GET(ptr, 1);    int offset;
2308    if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);    uschar *hc;
2309    
2310      /* See if this recursion is on the forward reference list. If so, adjust the
2311      reference. */
2312    
2313      for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
2314        {
2315        offset = GET(hc, 0);
2316        if (cd->start_code + offset == ptr + 1)
2317          {
2318          PUT(hc, 0, offset + adjust);
2319          break;
2320          }
2321        }
2322    
2323      /* Otherwise, adjust the recursion offset if it's after the start of this
2324      group. */
2325    
2326      if (hc >= cd->hwm)
2327        {
2328        offset = GET(ptr, 1);
2329        if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
2330        }
2331    
2332    ptr += 1 + LINK_SIZE;    ptr += 1 + LINK_SIZE;
2333    }    }
2334  }  }
# Line 1443  auto_callout(uschar *code, const uschar Line 2355  auto_callout(uschar *code, const uschar
2355  {  {
2356  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2357  *code++ = 255;  *code++ = 255;
2358  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2359  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2360  return code + 2*LINK_SIZE;  return code + 2*LINK_SIZE;
2361  }  }
2362    
# Line 1469  Returns:             nothing Line 2381  Returns:             nothing
2381  static void  static void
2382  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)
2383  {  {
2384  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2385  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2386  }  }
2387    
# Line 1495  Yield:        TRUE when range returned; Line 2407  Yield:        TRUE when range returned;
2407  */  */
2408    
2409  static BOOL  static BOOL
2410  get_othercase_range(int *cptr, int d, int *ocptr, int *odptr)  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,
2411      unsigned int *odptr)
2412  {  {
2413  int c, chartype, othercase, next;  unsigned int c, othercase, next;
2414    
2415  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2416    {    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
   if (_pcre_ucp_findchar(c, &chartype, &othercase) == ucp_L && othercase != 0)  
     break;  
   }  
2417    
2418  if (c > d) return FALSE;  if (c > d) return FALSE;
2419    
# Line 1512  next = othercase + 1; Line 2422  next = othercase + 1;
2422    
2423  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2424    {    {
2425    if (_pcre_ucp_findchar(c, &chartype, &othercase) != ucp_L ||    if (UCD_OTHERCASE(c) != next) break;
         othercase != next)  
     break;  
2426    next++;    next++;
2427    }    }
2428    
# Line 1523  for (++c; c <= d; c++) Line 2431  for (++c; c <= d; c++)
2431    
2432  return TRUE;  return TRUE;
2433  }  }
2434  #endif  /* SUPPORT_UCP */  
2435    
2436    
2437  /*************************************************  /*************************************************
2438  *           Compile one branch                   *  *        Check a character and a property        *
2439  *************************************************/  *************************************************/
2440    
2441  /* 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
2442  changed during the branch, the pointer is used to change the external options  is adjacent to a fixed character.
 bits.  
2443    
2444  Arguments:  Arguments:
2445    optionsptr     pointer to the option bits    c            the character
2446    brackets       points to number of extracting brackets used    ptype        the property type
2447    codeptr        points to the pointer to the current code point    pdata        the data for the type
2448    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.  
2449    
2450  Returns:         TRUE on success  Returns:       TRUE if auto-possessifying is OK
                  FALSE, with *errorcodeptr set non-zero on error  
2451  */  */
2452    
2453  static BOOL  static BOOL
2454  compile_branch(int *optionsptr, int *brackets, uschar **codeptr,  check_char_prop(int c, int ptype, int pdata, BOOL negated)
   const uschar **ptrptr, int *errorcodeptr, int *firstbyteptr,  
   int *reqbyteptr, branch_chain *bcptr, compile_data *cd)  
2455  {  {
2456  int repeat_type, op_type;  const ucd_record *prop = GET_UCD(c);
2457  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  switch(ptype)
2458  int bravalue = 0;    {
2459  int greedy_default, greedy_non_default;    case PT_LAMP:
2460  int firstbyte, reqbyte;    return (prop->chartype == ucp_Lu ||
2461  int zeroreqbyte, zerofirstbyte;            prop->chartype == ucp_Ll ||
2462  int req_caseopt, reqvary, tempreqvary;            prop->chartype == ucp_Lt) == negated;
2463  int condcount = 0;  
2464  int options = *optionsptr;    case PT_GC:
2465  int after_manual_callout = 0;    return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;
2466  register int c;  
2467  register uschar *code = *codeptr;    case PT_PC:
2468  uschar *tempcode;    return (pdata == prop->chartype) == negated;
2469  BOOL inescq = FALSE;  
2470  BOOL groupsetfirstbyte = FALSE;    case PT_SC:
2471  const uschar *ptr = *ptrptr;    return (pdata == prop->script) == negated;
2472  const uschar *tempptr;  
2473  uschar *previous = NULL;    /* These are specials */
2474  uschar *previous_callout = NULL;  
2475  uschar classbits[32];    case PT_ALNUM:
2476      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2477  #ifdef SUPPORT_UTF8            _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;
2478  BOOL class_utf8;  
2479  BOOL utf8 = (options & PCRE_UTF8) != 0;    case PT_SPACE:    /* Perl space */
2480  uschar *class_utf8data;    return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2481  uschar utf8_char[6];            c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2482  #else            == negated;
2483  BOOL utf8 = FALSE;  
2484  #endif    case PT_PXSPACE:  /* POSIX space */
2485      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2486              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2487              c == CHAR_FF || c == CHAR_CR)
2488              == negated;
2489    
2490      case PT_WORD:
2491      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2492              _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2493              c == CHAR_UNDERSCORE) == negated;
2494      }
2495    return FALSE;
2496    }
2497    #endif  /* SUPPORT_UCP */
2498    
 /* Set up the default and non-default settings for greediness */  
2499    
 greedy_default = ((options & PCRE_UNGREEDY) != 0);  
 greedy_non_default = greedy_default ^ 1;  
2500    
2501  /* Initialize no first byte, no required byte. REQ_UNSET means "no char  /*************************************************
2502  matching encountered yet". It gets changed to REQ_NONE if we hit something that  *     Check if auto-possessifying is possible    *
2503  matches a non-fixed char first char; reqbyte just remains unset if we never  *************************************************/
 find one.  
2504    
2505  When we hit a repeat whose minimum is zero, we may have to adjust these values  /* This function is called for unlimited repeats of certain items, to see
2506  to take the zero repeat into account. This is implemented by setting them to  whether the next thing could possibly match the repeated item. If not, it makes
2507  zerofirstbyte and zeroreqbyte when such a repeat is encountered. The individual  sense to automatically possessify the repeated item.
 item types that can be repeated set these backoff variables appropriately. */  
2508    
2509  firstbyte = reqbyte = zerofirstbyte = zeroreqbyte = REQ_UNSET;  Arguments:
2510      previous      pointer to the repeated opcode
2511      utf8          TRUE in UTF-8 mode
2512      ptr           next character in pattern
2513      options       options bits
2514      cd            contains pointers to tables etc.
2515    
2516  /* The variable req_caseopt contains either the REQ_CASELESS value or zero,  Returns:        TRUE if possessifying is wanted
2517  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. */  
2518    
2519  req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;  static BOOL
2520    check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
2521      int options, compile_data *cd)
2522    {
2523    int c, next;
2524    int op_code = *previous++;
2525    
2526  /* Switch on next character until the end of the branch */  /* Skip whitespace and comments in extended mode */
2527    
2528  for (;; ptr++)  if ((options & PCRE_EXTENDED) != 0)
2529    {    {
2530    BOOL negate_class;    for (;;)
   BOOL possessive_quantifier;  
   BOOL is_quantifier;  
   int class_charcount;  
   int class_lastchar;  
   int newoptions;  
   int recno;  
   int skipbytes;  
   int subreqbyte;  
   int subfirstbyte;  
   int mclength;  
   uschar mcbuffer[8];  
   
   /* Next byte in the pattern */  
   
   c = *ptr;  
   
   /* If in \Q...\E, check for the end; if not, we have a literal */  
   
   if (inescq && c != 0)  
2531      {      {
2532      if (c == '\\' && ptr[1] == 'E')      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2533        {      if (*ptr == CHAR_NUMBER_SIGN)
       inescq = FALSE;  
       ptr++;  
       continue;  
       }  
     else  
2534        {        {
2535        if (previous_callout != NULL)        ptr++;
2536          {        while (*ptr != 0)
         complete_callout(previous_callout, ptr, cd);  
         previous_callout = NULL;  
         }  
       if ((options & PCRE_AUTO_CALLOUT) != 0)  
2537          {          {
2538          previous_callout = code;          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2539          code = auto_callout(code, ptr, cd);          ptr++;
2540    #ifdef SUPPORT_UTF8
2541            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2542    #endif
2543          }          }
       goto NORMAL_CHAR;  
2544        }        }
2545        else break;
2546      }      }
2547      }
2548    
2549    /* 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
2550    a quantifier. */  value is a character, a negative value is an escape value. */
   
   is_quantifier = c == '*' || c == '+' || c == '?' ||  
     (c == '{' && is_counted_repeat(ptr+1));  
2551    
2552    if (!is_quantifier && previous_callout != NULL &&  if (*ptr == CHAR_BACKSLASH)
2553         after_manual_callout-- <= 0)    {
2554      {    int temperrorcode = 0;
2555      complete_callout(previous_callout, ptr, cd);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
2556      previous_callout = NULL;    if (temperrorcode != 0) return FALSE;
2557      }    ptr++;    /* Point after the escape sequence */
2558      }
2559    
2560    /* In extended mode, skip white space and comments */  else if ((cd->ctypes[*ptr] & ctype_meta) == 0)
2561      {
2562    #ifdef SUPPORT_UTF8
2563      if (utf8) { GETCHARINC(next, ptr); } else
2564    #endif
2565      next = *ptr++;
2566      }
2567    
2568    if ((options & PCRE_EXTENDED) != 0)  else return FALSE;
2569    
2570    /* Skip whitespace and comments in extended mode */
2571    
2572    if ((options & PCRE_EXTENDED) != 0)
2573      {
2574      for (;;)
2575      {      {
2576      if ((cd->ctypes[c] & ctype_space) != 0) continue;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2577      if (c == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2578        {        {
2579        /* The space before the ; is to avoid a warning on a silly compiler        ptr++;
2580        on the Macintosh. */        while (*ptr != 0)
2581        while ((c = *(++ptr)) != 0 && c != NEWLINE) ;          {
2582        if (c != 0) continue;   /* Else fall through to handle end of string */          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2583            ptr++;
2584    #ifdef SUPPORT_UTF8
2585            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2586    #endif
2587            }
2588        }        }
2589        else break;
2590      }      }
2591      }
2592    
2593    /* No auto callout for quantifiers. */  /* If the next thing is itself optional, we have to give up. */
2594    
2595    if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2596      {    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2597      previous_callout = code;      return FALSE;
     code = auto_callout(code, ptr, cd);  
     }  
2598    
2599    switch(c)  /* Now compare the next item with the previous opcode. First, handle cases when
2600      {  the next item is a character. */
     /* The branch terminates at end of string, |, or ). */  
2601    
2602      case 0:  if (next >= 0) switch(op_code)
2603      case '|':    {
2604      case ')':    case OP_CHAR:
2605      *firstbyteptr = firstbyte;  #ifdef SUPPORT_UTF8
2606      *reqbyteptr = reqbyte;    GETCHARTEST(c, previous);
2607      *codeptr = code;  #else
2608      *ptrptr = ptr;    c = *previous;
2609      return TRUE;  #endif
2610      return c != next;
2611    
2612      /* Handle single-character metacharacters. In multiline mode, ^ disables    /* For CHARNC (caseless character) we must check the other case. If we have
2613      the setting of any following char as a first character. */    Unicode property support, we can use it to test the other case of
2614      high-valued characters. */
2615    
2616      case '^':    case OP_CHARNC:
2617      if ((options & PCRE_MULTILINE) != 0)  #ifdef SUPPORT_UTF8
2618        {    GETCHARTEST(c, previous);
2619        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;  #else
2620        }    c = *previous;
2621      previous = NULL;  #endif
2622      *code++ = OP_CIRC;    if (c == next) return FALSE;
2623      break;  #ifdef SUPPORT_UTF8
2624      if (utf8)
2625        {
2626        unsigned int othercase;
2627        if (next < 128) othercase = cd->fcc[next]; else
2628    #ifdef SUPPORT_UCP
2629        othercase = UCD_OTHERCASE((unsigned int)next);
2630    #else
2631        othercase = NOTACHAR;
2632    #endif
2633        return (unsigned int)c != othercase;
2634        }
2635      else
2636    #endif  /* SUPPORT_UTF8 */
2637      return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2638    
2639      case '$':    /* For OP_NOT, its data is always a single-byte character. */
     previous = NULL;  
     *code++ = OP_DOLL;  
     break;  
2640    
2641      /* There can never be a first char if '.' is first, whatever happens about    case OP_NOT:
2642      repeats. The value of reqbyte doesn't change either. */    if ((c = *previous) == next) return TRUE;
2643      if ((options & PCRE_CASELESS) == 0) return FALSE;
2644    #ifdef SUPPORT_UTF8
2645      if (utf8)
2646        {
2647        unsigned int othercase;
2648        if (next < 128) othercase = cd->fcc[next]; else
2649    #ifdef SUPPORT_UCP
2650        othercase = UCD_OTHERCASE(next);
2651    #else
2652        othercase = NOTACHAR;
2653    #endif
2654        return (unsigned int)c == othercase;
2655        }
2656      else
2657    #endif  /* SUPPORT_UTF8 */
2658      return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2659    
2660      case '.':    /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2661      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;    When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
     zerofirstbyte = firstbyte;  
     zeroreqbyte = reqbyte;  
     previous = code;  
     *code++ = OP_ANY;  
     break;  
2662    
2663      /* Character classes. If the included characters are all < 255 in value, we    case OP_DIGIT:
2664      build a 32-byte bitmap of the permitted characters, except in the special    return next > 127 || (cd->ctypes[next] & ctype_digit) == 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.  
2665    
2666      If the class contains characters outside the 0-255 range, a different    case OP_NOT_DIGIT:
2667      opcode is compiled. It may optionally have a bit map for characters < 256,    return next <= 127 && (cd->ctypes[next] & ctype_digit) != 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.  
     */  
2668    
2669      case '[':    case OP_WHITESPACE:
2670      previous = code;    return next > 127 || (cd->ctypes[next] & ctype_space) == 0;
2671    
2672      /* PCRE supports POSIX class stuff inside a class. Perl gives an error if    case OP_NOT_WHITESPACE:
2673      they are encountered at the top level, so we'll do that too. */    return next <= 127 && (cd->ctypes[next] & ctype_space) != 0;
2674    
2675      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&    case OP_WORDCHAR:
2676          check_posix_syntax(ptr, &tempptr, cd))    return next > 127 || (cd->ctypes[next] & ctype_word) == 0;
       {  
       *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;  
       goto FAILED;  
       }  
2677    
2678      /* If the first character is '^', set the negation flag and skip it. */    case OP_NOT_WORDCHAR:
2679      return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;
2680    
2681      if ((c = *(++ptr)) == '^')    case OP_HSPACE:
2682        {    case OP_NOT_HSPACE:
2683        negate_class = TRUE;    switch(next)
2684        c = *(++ptr);      {
2685        }      case 0x09:
2686      else      case 0x20:
2687        {      case 0xa0:
2688        negate_class = FALSE;      case 0x1680:
2689        }      case 0x180e:
2690        case 0x2000:
2691        case 0x2001:
2692        case 0x2002:
2693        case 0x2003:
2694        case 0x2004:
2695        case 0x2005:
2696        case 0x2006:
2697        case 0x2007:
2698        case 0x2008:
2699        case 0x2009:
2700        case 0x200A:
2701        case 0x202f:
2702        case 0x205f:
2703        case 0x3000:
2704        return op_code == OP_NOT_HSPACE;
2705        default:
2706        return op_code != OP_NOT_HSPACE;
2707        }
2708    
2709      /* Keep a count of chars with values < 256 so that we can optimize the case    case OP_ANYNL:
2710      of just a single character (as long as it's < 256). For higher valued UTF-8    case OP_VSPACE:
2711      characters, we don't yet do any optimization. */    case OP_NOT_VSPACE:
2712      switch(next)
2713        {
2714        case 0x0a:
2715        case 0x0b:
2716        case 0x0c:
2717        case 0x0d:
2718        case 0x85:
2719        case 0x2028:
2720        case 0x2029:
2721        return op_code == OP_NOT_VSPACE;
2722        default:
2723        return op_code != OP_NOT_VSPACE;
2724        }
2725    
2726      class_charcount = 0;  #ifdef SUPPORT_UCP
2727      class_lastchar = -1;    case OP_PROP:
2728      return check_char_prop(next, previous[0], previous[1], FALSE);
2729    
2730  #ifdef SUPPORT_UTF8    case OP_NOTPROP:
2731      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 */  
2732  #endif  #endif
2733    
2734      /* Initialize the 32-char bit map to all zeros. We have to build the    default:
2735      map in a temporary bit of store, in case the class contains only 1    return FALSE;
2736      character (< 256), because in that case the compiled code doesn't use the    }
     bit map. */  
2737    
     memset(classbits, 0, 32 * sizeof(uschar));  
2738    
2739      /* Process characters until ] is reached. By writing this as a "do" it  /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP
2740      means that an initial ] is taken as a data character. The first pass  is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
2741      through the regex checked the overall syntax, so we don't need to be very  generated only when PCRE_UCP is *not* set, that is, when only ASCII
2742      strict here. At the start of the loop, c contains the first byte of the  characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
2743      character. */  replaced by OP_PROP codes when PCRE_UCP is set. */
2744    
2745      do  switch(op_code)
2746        {    {
2747      case OP_CHAR:
2748      case OP_CHARNC:
2749  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2750        if (utf8 && c > 127)    GETCHARTEST(c, previous);
2751          {                           /* Braces are required because the */  #else
2752          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */    c = *previous;
         }  
2753  #endif  #endif
2754      switch(-next)
2755        {
2756        case ESC_d:
2757        return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
2758    
2759        /* Inside \Q...\E everything is literal except \E */      case ESC_D:
2760        return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
       if (inescq)  
         {  
         if (c == '\\' && ptr[1] == 'E')  
           {  
           inescq = FALSE;  
           ptr++;  
           continue;  
           }  
         else goto LONE_SINGLE_CHARACTER;  
         }  
   
       /* Handle POSIX class names. Perl allows a negation extension of the  
       form [:^name:]. A square bracket that doesn't match the syntax is  
       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. */  
2761    
2762        if (c == '[' &&      case ESC_s:
2763            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      return c > 127 || (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;  
2764    
2765          if (ptr[1] != ':')      case ESC_S:
2766            {      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
2767            *errorcodeptr = ERR31;  
2768            goto FAILED;      case ESC_w:
2769            }      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
2770    
2771        case ESC_W:
2772        return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
2773    
2774        case ESC_h:
2775        case ESC_H:
2776        switch(c)
2777          {
2778          case 0x09:
2779          case 0x20:
2780          case 0xa0:
2781          case 0x1680:
2782          case 0x180e:
2783          case 0x2000:
2784          case 0x2001:
2785          case 0x2002:
2786          case 0x2003:
2787          case 0x2004:
2788          case 0x2005:
2789          case 0x2006:
2790          case 0x2007:
2791          case 0x2008:
2792          case 0x2009:
2793          case 0x200A:
2794          case 0x202f:
2795          case 0x205f:
2796          case 0x3000:
2797          return -next != ESC_h;
2798          default:
2799          return -next == ESC_h;
2800          }
2801    
2802          ptr += 2;      case ESC_v:
2803          if (*ptr == '^')      case ESC_V:
2804            {      switch(c)
2805            local_negate = TRUE;        {
2806            ptr++;        case 0x0a:
2807            }        case 0x0b:
2808          case 0x0c:
2809          case 0x0d:
2810          case 0x85:
2811          case 0x2028:
2812          case 0x2029:
2813          return -next != ESC_v;
2814          default:
2815          return -next == ESC_v;
2816          }
2817    
2818          posix_class = check_posix_name(ptr, tempptr - ptr);      /* When PCRE_UCP is set, these values get generated for \d etc. Find
2819          if (posix_class < 0)      their substitutions and process them. The result will always be either
2820            {      -ESC_p or -ESC_P. Then fall through to process those values. */
           *errorcodeptr = ERR30;  
           goto FAILED;  
           }  
2821    
2822          /* If matching is caseless, upper and lower are converted to  #ifdef SUPPORT_UCP
2823          alpha. This relies on the fact that the class table starts with      case ESC_du:
2824          alpha, lower, upper as the first 3 entries. */      case ESC_DU:
2825        case ESC_wu:
2826        case ESC_WU:
2827        case ESC_su:
2828        case ESC_SU:
2829          {
2830          int temperrorcode = 0;
2831          ptr = substitutes[-next - ESC_DU];
2832          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
2833          if (temperrorcode != 0) return FALSE;
2834          ptr++;    /* For compatibility */
2835          }
2836        /* Fall through */
2837    
2838          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)      case ESC_p:
2839            posix_class = 0;      case ESC_P:
2840          {
2841          int ptype, pdata, errorcodeptr;
2842          BOOL negated;
2843    
2844          /* Or into the map we are building up to 3 of the static class        ptr--;      /* Make ptr point at the p or P */
2845          tables, or their negations. The [:blank:] class sets up the same        ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
2846          chars as the [:space:] class (all white space). We remove the vertical        if (ptype < 0) return FALSE;
2847          white space chars afterwards. */        ptr++;      /* Point past the final curly ket */
2848    
2849          /* If the property item is optional, we have to give up. (When generated
2850          from \d etc by PCRE_UCP, this test will have been applied much earlier,
2851          to the original \d etc. At this point, ptr will point to a zero byte. */
2852    
2853          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2854            strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2855              return FALSE;
2856    
2857          posix_class *= 3;        /* Do the property check. */
         for (i = 0; i < 3; i++)  
           {  
           BOOL blankclass = strncmp((char *)ptr, "blank", 5) == 0;  
           int taboffset = posix_class_maps[posix_class + i];  
           if (taboffset < 0) break;  
           if (local_negate)  
             {  
             if (i == 0)  
               for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+taboffset];  
             else  
               for (c = 0; c < 32; c++) classbits[c] &= ~cbits[c+taboffset];  
             if (blankclass) classbits[1] |= 0x3c;  
             }  
           else  
             {  
             for (c = 0; c < 32; c++) classbits[c] |= cbits[c+taboffset];  
             if (blankclass) classbits[1] &= ~0x3c;  
             }  
           }  
2858    
2859          ptr = tempptr + 1;        return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
2860          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */        }
2861          continue;    /* End of POSIX syntax handling */  #endif
         }  
2862    
2863        /* Backslash may introduce a single character, or it may introduce one      default:
2864        of the specials, which just set a flag. Escaped items are checked for      return FALSE;
2865        validity in the pre-compiling pass. The sequence \b is a special case.      }
       Inside a class (and only there) it is treated as backspace. Elsewhere  
       it marks a word boundary. Other escapes have preset maps ready to  
       or into the one we are building. We assume they have more than one  
       character in them, so set class_charcount bigger than one. */  
2866    
2867        if (c == '\\')    /* In principle, support for Unicode properties should be integrated here as
2868          {    well. It means re-organizing the above code so as to get hold of the property
2869          c = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);    values before switching on the op-code. However, I wonder how many patterns
2870      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
2871      these op-codes are never generated.) */
2872    
2873          if (-c == ESC_b) c = '\b';       /* \b is backslash in a class */    case OP_DIGIT:
2874          else if (-c == ESC_X) c = 'X';   /* \X is literal X in a class */    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2875          else if (-c == ESC_Q)            /* Handle start of quoted string */           next == -ESC_h || next == -ESC_v || next == -ESC_R;
           {  
           if (ptr[1] == '\\' && ptr[2] == 'E')  
             {  
             ptr += 2; /* avoid empty string */  
             }  
           else inescq = TRUE;  
           continue;  
           }  
2876    
2877          if (c < 0)    case OP_NOT_DIGIT:
2878            {    return next == -ESC_d;
           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;  
2879    
2880              case ESC_D:    case OP_WHITESPACE:
2881              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
             continue;  
2882    
2883              case ESC_w:    case OP_NOT_WHITESPACE:
2884              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_word];    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
             continue;  
2885    
2886              case ESC_W:    case OP_HSPACE:
2887              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
2888              continue;           next == -ESC_w || next == -ESC_v || next == -ESC_R;
2889    
2890              case ESC_s:    case OP_NOT_HSPACE:
2891              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];    return next == -ESC_h;
             classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */  
             continue;  
2892    
2893              case ESC_S:    /* Can't have \S in here because VT matches \S (Perl anomaly) */
2894              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];    case OP_ANYNL:
2895              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */    case OP_VSPACE:
2896              continue;    return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2897    
2898  #ifdef SUPPORT_UCP    case OP_NOT_VSPACE:
2899              case ESC_p:    return next == -ESC_v || next == -ESC_R;
             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  
2900    
2901              /* Unrecognized escapes are faulted if PCRE is running in its    case OP_WORDCHAR:
2902              strict mode. By default, for compatibility with Perl, they are    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
2903              treated as literals. */           next == -ESC_v || next == -ESC_R;
2904    
2905              default:    case OP_NOT_WORDCHAR:
2906              if ((options & PCRE_EXTRA) != 0)    return next == -ESC_w || next == -ESC_d;
               {  
               *errorcodeptr = ERR7;  
               goto FAILED;  
               }  
             c = *ptr;              /* The final character */  
             class_charcount -= 2;  /* Undo the default count from above */  
             }  
           }  
2907    
2908          /* Fall through if we have a single character (c >= 0). This may be    default:
2909          > 256 in UTF-8 mode. */    return FALSE;
2910      }
2911    
2912          }   /* End of backslash handling */  /* Control does not reach here */
2913    }
2914    
       /* A single character may be followed by '-' to form a range. However,  
       Perl does not permit ']' to be the end of the range. A '-' character  
       here is treated as a literal. */  
2915    
       if (ptr[1] == '-' && ptr[2] != ']')  
         {  
         int d;  
         ptr += 2;  
2916    
2917  #ifdef SUPPORT_UTF8  /*************************************************
2918          if (utf8)  *           Compile one branch                   *
2919            {                           /* Braces are required because the */  *************************************************/
           GETCHARLEN(d, ptr, ptr);    /* macro generates multiple statements */  
           }  
         else  
 #endif  
         d = *ptr;  /* Not UTF-8 mode */  
2920    
2921          /* The second part of a range can be a single-character escape, but  /* Scan the pattern, compiling it into the a vector. If the options are
2922          not any of the other escapes. Perl 5.6 treats a hyphen as a literal  changed during the branch, the pointer is used to change the external options
2923          in such circumstances. */  bits. This function is used during the pre-compile phase when we are trying
2924    to find out the amount of memory needed, as well as during the real compile
2925    phase. The value of lengthptr distinguishes the two phases.
2926    
2927          if (d == '\\')  Arguments:
2928      optionsptr     pointer to the option bits
2929      codeptr        points to the pointer to the current code point
2930      ptrptr         points to the current pattern pointer
2931      errorcodeptr   points to error code variable
2932      firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
2933      reqbyteptr     set to the last literal character required, else < 0
2934      bcptr          points to current branch chain
2935      cd             contains pointers to tables etc.
2936      lengthptr      NULL during the real compile phase
2937                     points to length accumulator during pre-compile phase
2938    
2939    Returns:         TRUE on success
2940                     FALSE, with *errorcodeptr set non-zero on error
2941    */
2942    
2943    static BOOL
2944    compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
2945      int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
2946      compile_data *cd, int *lengthptr)
2947    {
2948    int repeat_type, op_type;
2949    int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
2950    int bravalue = 0;
2951    int greedy_default, greedy_non_default;
2952    int firstbyte, reqbyte;
2953    int zeroreqbyte, zerofirstbyte;
2954    int req_caseopt, reqvary, tempreqvary;
2955    int options = *optionsptr;
2956    int after_manual_callout = 0;
2957    int length_prevgroup = 0;
2958    register int c;
2959    register uschar *code = *codeptr;
2960    uschar *last_code = code;
2961    uschar *orig_code = code;
2962    uschar *tempcode;
2963    BOOL inescq = FALSE;
2964    BOOL groupsetfirstbyte = FALSE;
2965    const uschar *ptr = *ptrptr;
2966    const uschar *tempptr;
2967    const uschar *nestptr = NULL;
2968    uschar *previous = NULL;
2969    uschar *previous_callout = NULL;
2970    uschar *save_hwm = NULL;
2971    uschar classbits[32];
2972    
2973    #ifdef SUPPORT_UTF8
2974    BOOL class_utf8;
2975    BOOL utf8 = (options & PCRE_UTF8) != 0;
2976    uschar *class_utf8data;
2977    uschar *class_utf8data_base;
2978    uschar utf8_char[6];
2979    #else
2980    BOOL utf8 = FALSE;
2981    uschar *utf8_char = NULL;
2982    #endif
2983    
2984    #ifdef PCRE_DEBUG
2985    if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2986    #endif
2987    
2988    /* Set up the default and non-default settings for greediness */
2989    
2990    greedy_default = ((options & PCRE_UNGREEDY) != 0);
2991    greedy_non_default = greedy_default ^ 1;
2992    
2993    /* Initialize no first byte, no required byte. REQ_UNSET means "no char
2994    matching encountered yet". It gets changed to REQ_NONE if we hit something that
2995    matches a non-fixed char first char; reqbyte just remains unset if we never
2996    find one.
2997    
2998    When we hit a repeat whose minimum is zero, we may have to adjust these values
2999    to take the zero repeat into account. This is implemented by setting them to
3000    zerofirstbyte and zeroreqbyte when such a repeat is encountered. The individual
3001    item types that can be repeated set these backoff variables appropriately. */
3002    
3003    firstbyte = reqbyte = zerofirstbyte = zeroreqbyte = REQ_UNSET;
3004    
3005    /* The variable req_caseopt contains either the REQ_CASELESS value or zero,
3006    according to the current setting of the caseless flag. REQ_CASELESS is a bit
3007    value > 255. It is added into the firstbyte or reqbyte variables to record the
3008    case status of the value. This is used only for ASCII characters. */
3009    
3010    req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
3011    
3012    /* Switch on next character until the end of the branch */
3013    
3014    for (;; ptr++)
3015      {
3016      BOOL negate_class;
3017      BOOL should_flip_negation;
3018      BOOL possessive_quantifier;
3019      BOOL is_quantifier;
3020      BOOL is_recurse;
3021      BOOL reset_bracount;
3022      int class_charcount;
3023      int class_lastchar;
3024      int newoptions;
3025      int recno;
3026      int refsign;
3027      int skipbytes;
3028      int subreqbyte;
3029      int subfirstbyte;
3030      int terminator;
3031      int mclength;
3032      uschar mcbuffer[8];
3033    
3034      /* Get next byte in the pattern */
3035    
3036      c = *ptr;
3037    
3038      /* If we are at the end of a nested substitution, revert to the outer level
3039      string. Nesting only happens one level deep. */
3040    
3041      if (c == 0 && nestptr != NULL)
3042        {
3043        ptr = nestptr;
3044        nestptr = NULL;
3045        c = *ptr;
3046        }
3047    
3048      /* If we are in the pre-compile phase, accumulate the length used for the
3049      previous cycle of this loop. */
3050    
3051      if (lengthptr != NULL)
3052        {
3053    #ifdef PCRE_DEBUG
3054        if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3055    #endif
3056        if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
3057          {
3058          *errorcodeptr = ERR52;
3059          goto FAILED;
3060          }
3061    
3062        /* There is at least one situation where code goes backwards: this is the
3063        case of a zero quantifier after a class (e.g. [ab]{0}). At compile time,
3064        the class is simply eliminated. However, it is created first, so we have to
3065        allow memory for it. Therefore, don't ever reduce the length at this point.
3066        */
3067    
3068        if (code < last_code) code = last_code;
3069    
3070        /* Paranoid check for integer overflow */
3071    
3072        if (OFLOW_MAX - *lengthptr < code - last_code)
3073          {
3074          *errorcodeptr = ERR20;
3075          goto FAILED;
3076          }
3077    
3078        *lengthptr += (int)(code - last_code);
3079        DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));
3080    
3081        /* If "previous" is set and it is not at the start of the work space, move
3082        it back to there, in order to avoid filling up the work space. Otherwise,
3083        if "previous" is NULL, reset the current code pointer to the start. */
3084    
3085        if (previous != NULL)
3086          {
3087          if (previous > orig_code)
3088            {
3089            memmove(orig_code, previous, code - previous);
3090            code -= previous - orig_code;
3091            previous = orig_code;
3092            }
3093          }
3094        else code = orig_code;
3095    
3096        /* Remember where this code item starts so we can pick up the length
3097        next time round. */
3098    
3099        last_code = code;
3100        }
3101    
3102      /* In the real compile phase, just check the workspace used by the forward
3103      reference list. */
3104    
3105      else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
3106        {
3107        *errorcodeptr = ERR52;
3108        goto FAILED;
3109        }
3110    
3111      /* If in \Q...\E, check for the end; if not, we have a literal */
3112    
3113      if (inescq && c != 0)
3114        {
3115        if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3116          {
3117          inescq = FALSE;
3118          ptr++;
3119          continue;
3120          }
3121        else
3122          {
3123          if (previous_callout != NULL)
3124            {
3125            if (lengthptr == NULL)  /* Don't attempt in pre-compile phase */
3126              complete_callout(previous_callout, ptr, cd);
3127            previous_callout = NULL;
3128            }
3129          if ((options & PCRE_AUTO_CALLOUT) != 0)
3130            {
3131            previous_callout = code;
3132            code = auto_callout(code, ptr, cd);
3133            }
3134          goto NORMAL_CHAR;
3135          }
3136        }
3137    
3138      /* Fill in length of a previous callout, except when the next thing is
3139      a quantifier. */
3140    
3141      is_quantifier =
3142        c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
3143        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
3144    
3145      if (!is_quantifier && previous_callout != NULL &&
3146           after_manual_callout-- <= 0)
3147        {
3148        if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */
3149          complete_callout(previous_callout, ptr, cd);
3150        previous_callout = NULL;
3151        }
3152    
3153      /* In extended mode, skip white space and comments */
3154    
3155      if ((options & PCRE_EXTENDED) != 0)
3156        {
3157        if ((cd->ctypes[c] & ctype_space) != 0) continue;
3158        if (c == CHAR_NUMBER_SIGN)
3159          {
3160          ptr++;
3161          while (*ptr != 0)
3162            {
3163            if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3164            ptr++;
3165    #ifdef SUPPORT_UTF8
3166            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
3167    #endif
3168            }
3169          if (*ptr != 0) continue;
3170    
3171          /* Else fall through to handle end of string */
3172          c = 0;
3173          }
3174        }
3175    
3176      /* No auto callout for quantifiers. */
3177    
3178      if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier)
3179        {
3180        previous_callout = code;
3181        code = auto_callout(code, ptr, cd);
3182        }
3183    
3184      switch(c)
3185        {
3186        /* ===================================================================*/
3187        case 0:                        /* The branch terminates at string end */
3188        case CHAR_VERTICAL_LINE:       /* or | or ) */
3189        case CHAR_RIGHT_PARENTHESIS:
3190        *firstbyteptr = firstbyte;
3191        *reqbyteptr = reqbyte;
3192        *codeptr = code;
3193        *ptrptr = ptr;
3194        if (lengthptr != NULL)
3195          {
3196          if (OFLOW_MAX - *lengthptr < code - last_code)
3197            {
3198            *errorcodeptr = ERR20;
3199            goto FAILED;
3200            }
3201          *lengthptr += (int)(code - last_code);   /* To include callout length */
3202          DPRINTF((">> end branch\n"));
3203          }
3204        return TRUE;
3205    
3206    
3207        /* ===================================================================*/
3208        /* Handle single-character metacharacters. In multiline mode, ^ disables
3209        the setting of any following char as a first character. */
3210    
3211        case CHAR_CIRCUMFLEX_ACCENT:
3212        if ((options & PCRE_MULTILINE) != 0)
3213          {
3214          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3215          }
3216        previous = NULL;
3217        *code++ = OP_CIRC;
3218        break;
3219    
3220        case CHAR_DOLLAR_SIGN:
3221        previous = NULL;
3222        *code++ = OP_DOLL;
3223        break;
3224    
3225        /* There can never be a first char if '.' is first, whatever happens about
3226        repeats. The value of reqbyte doesn't change either. */
3227    
3228        case CHAR_DOT:
3229        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3230        zerofirstbyte = firstbyte;
3231        zeroreqbyte = reqbyte;
3232        previous = code;
3233        *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
3234        break;
3235    
3236    
3237        /* ===================================================================*/
3238        /* Character classes. If the included characters are all < 256, we build a
3239        32-byte bitmap of the permitted characters, except in the special case
3240        where there is only one such character. For negated classes, we build the
3241        map as usual, then invert it at the end. However, we use a different opcode
3242        so that data characters > 255 can be handled correctly.
3243    
3244        If the class contains characters outside the 0-255 range, a different
3245        opcode is compiled. It may optionally have a bit map for characters < 256,
3246        but those above are are explicitly listed afterwards. A flag byte tells
3247        whether the bitmap is present, and whether this is a negated class or not.
3248    
3249        In JavaScript compatibility mode, an isolated ']' causes an error. In
3250        default (Perl) mode, it is treated as a data character. */
3251    
3252        case CHAR_RIGHT_SQUARE_BRACKET:
3253        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3254          {
3255          *errorcodeptr = ERR64;
3256          goto FAILED;
3257          }
3258        goto NORMAL_CHAR;
3259    
3260        case CHAR_LEFT_SQUARE_BRACKET:
3261        previous = code;
3262    
3263        /* PCRE supports POSIX class stuff inside a class. Perl gives an error if
3264        they are encountered at the top level, so we'll do that too. */
3265    
3266        if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3267             ptr[1] == CHAR_EQUALS_SIGN) &&
3268            check_posix_syntax(ptr, &tempptr))
3269          {
3270          *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
3271          goto FAILED;
3272          }
3273    
3274        /* If the first character is '^', set the negation flag and skip it. Also,
3275        if the first few characters (either before or after ^) are \Q\E or \E we
3276        skip them too. This makes for compatibility with Perl. */
3277    
3278        negate_class = FALSE;
3279        for (;;)
3280          {
3281          c = *(++ptr);
3282          if (c == CHAR_BACKSLASH)
3283            {
3284            if (ptr[1] == CHAR_E)
3285              ptr++;
3286            else if (strncmp((const char *)ptr+1,
3287                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
3288              ptr += 3;
3289            else
3290              break;
3291            }
3292          else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
3293            negate_class = TRUE;
3294          else break;
3295          }
3296    
3297        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
3298        an initial ']' is taken as a data character -- the code below handles
3299        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3300        [^] must match any character, so generate OP_ALLANY. */
3301    
3302        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3303            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3304          {
3305          *code++ = negate_class? OP_ALLANY : OP_FAIL;
3306          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3307          zerofirstbyte = firstbyte;
3308          break;
3309          }
3310    
3311        /* If a class contains a negative special such as \S, we need to flip the
3312        negation flag at the end, so that support for characters > 255 works
3313        correctly (they are all included in the class). */
3314    
3315        should_flip_negation = FALSE;
3316    
3317        /* Keep a count of chars with values < 256 so that we can optimize the case
3318        of just a single character (as long as it's < 256). However, For higher
3319        valued UTF-8 characters, we don't yet do any optimization. */
3320    
3321        class_charcount = 0;
3322        class_lastchar = -1;
3323    
3324        /* Initialize the 32-char bit map to all zeros. We build the map in a
3325        temporary bit of memory, in case the class contains only 1 character (less
3326        than 256), because in that case the compiled code doesn't use the bit map.
3327        */
3328    
3329        memset(classbits, 0, 32 * sizeof(uschar));
3330    
3331    #ifdef SUPPORT_UTF8
3332        class_utf8 = FALSE;                       /* No chars >= 256 */
3333        class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
3334        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
3335    #endif
3336    
3337        /* Process characters until ] is reached. By writing this as a "do" it
3338        means that an initial ] is taken as a data character. At the start of the
3339        loop, c contains the first byte of the character. */
3340    
3341        if (c != 0) do
3342          {
3343          const uschar *oldptr;
3344    
3345    #ifdef SUPPORT_UTF8
3346          if (utf8 && c > 127)
3347            {                           /* Braces are required because the */
3348            GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3349            }
3350    
3351          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
3352          data and reset the pointer. This is so that very large classes that
3353          contain a zillion UTF-8 characters no longer overwrite the work space
3354          (which is on the stack). */
3355    
3356          if (lengthptr != NULL)
3357            {
3358            *lengthptr += class_utf8data - class_utf8data_base;
3359            class_utf8data = class_utf8data_base;
3360            }
3361    
3362    #endif
3363    
3364          /* Inside \Q...\E everything is literal except \E */
3365    
3366          if (inescq)
3367            {
3368            if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3369              {
3370              inescq = FALSE;                   /* Reset literal state */
3371              ptr++;                            /* Skip the 'E' */
3372              continue;                         /* Carry on with next */
3373              }
3374            goto CHECK_RANGE;                   /* Could be range if \E follows */
3375            }
3376    
3377          /* Handle POSIX class names. Perl allows a negation extension of the
3378          form [:^name:]. A square bracket that doesn't match the syntax is
3379          treated as a literal. We also recognize the POSIX constructions
3380          [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3381          5.6 and 5.8 do. */
3382    
3383          if (c == CHAR_LEFT_SQUARE_BRACKET &&
3384              (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3385               ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3386            {
3387            BOOL local_negate = FALSE;
3388            int posix_class, taboffset, tabopt;
3389            register const uschar *cbits = cd->cbits;
3390            uschar pbits[32];
3391    
3392            if (ptr[1] != CHAR_COLON)
3393              {
3394              *errorcodeptr = ERR31;
3395              goto FAILED;
3396              }
3397    
3398            ptr += 2;
3399            if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3400              {
3401              local_negate = TRUE;
3402              should_flip_negation = TRUE;  /* Note negative special */
3403              ptr++;
3404              }
3405    
3406            posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3407            if (posix_class < 0)
3408              {
3409              *errorcodeptr = ERR30;
3410              goto FAILED;
3411              }
3412    
3413            /* If matching is caseless, upper and lower are converted to
3414            alpha. This relies on the fact that the class table starts with
3415            alpha, lower, upper as the first 3 entries. */
3416    
3417            if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3418              posix_class = 0;
3419    
3420            /* When PCRE_UCP is set, some of the POSIX classes are converted to
3421            different escape sequences that use Unicode properties. */
3422    
3423    #ifdef SUPPORT_UCP
3424            if ((options & PCRE_UCP) != 0)
3425              {
3426              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3427              if (posix_substitutes[pc] != NULL)
3428                {
3429                nestptr = tempptr + 1;
3430                ptr = posix_substitutes[pc] - 1;
3431                continue;
3432                }
3433              }
3434    #endif
3435            /* In the non-UCP case, we build the bit map for the POSIX class in a
3436            chunk of local store because we may be adding and subtracting from it,
3437            and we don't want to subtract bits that may be in the main map already.
3438            At the end we or the result into the bit map that is being built. */
3439    
3440            posix_class *= 3;
3441    
3442            /* Copy in the first table (always present) */
3443    
3444            memcpy(pbits, cbits + posix_class_maps[posix_class],
3445              32 * sizeof(uschar));
3446    
3447            /* If there is a second table, add or remove it as required. */
3448    
3449            taboffset = posix_class_maps[posix_class + 1];
3450            tabopt = posix_class_maps[posix_class + 2];
3451    
3452            if (taboffset >= 0)
3453              {
3454              if (tabopt >= 0)
3455                for (c = 0; c < 32; c++) pbits[c] |= cbits[c + taboffset];
3456              else
3457                for (c = 0; c < 32; c++) pbits[c] &= ~cbits[c + taboffset];
3458              }
3459    
3460            /* Not see if we need to remove any special characters. An option
3461            value of 1 removes vertical space and 2 removes underscore. */
3462    
3463            if (tabopt < 0) tabopt = -tabopt;
3464            if (tabopt == 1) pbits[1] &= ~0x3c;
3465              else if (tabopt == 2) pbits[11] &= 0x7f;
3466    
3467            /* Add the POSIX table or its complement into the main table that is
3468            being built and we are done. */
3469    
3470            if (local_negate)
3471              for (c = 0; c < 32; c++) classbits[c] |= ~pbits[c];
3472            else
3473              for (c = 0; c < 32; c++) classbits[c] |= pbits[c];
3474    
3475            ptr = tempptr + 1;
3476            class_charcount = 10;  /* Set > 1; assumes more than 1 per class */
3477            continue;    /* End of POSIX syntax handling */
3478            }
3479    
3480          /* Backslash may introduce a single character, or it may introduce one
3481          of the specials, which just set a flag. The sequence \b is a special
3482          case. Inside a class (and only there) it is treated as backspace. We
3483          assume that other escapes have more than one character in them, so set
3484          class_charcount bigger than one. Unrecognized escapes fall through and
3485          are either treated as literal characters (by default), or are faulted if
3486          PCRE_EXTRA is set. */
3487    
3488          if (c == CHAR_BACKSLASH)
3489            {
3490            c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3491            if (*errorcodeptr != 0) goto FAILED;
3492    
3493            if (-c == ESC_b) c = CHAR_BS;    /* \b is backspace in a class */
3494            else if (-c == ESC_Q)            /* Handle start of quoted string */
3495              {
3496              if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3497                {
3498                ptr += 2; /* avoid empty string */
3499                }
3500              else inescq = TRUE;
3501              continue;
3502              }
3503            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
3504    
3505            if (c < 0)
3506              {
3507              register const uschar *cbits = cd->cbits;
3508              class_charcount += 2;     /* Greater than 1 is what matters */
3509    
3510              switch (-c)
3511                {
3512    #ifdef SUPPORT_UCP
3513                case ESC_du:     /* These are the values given for \d etc */
3514                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3515                case ESC_wu:     /* escape sequence with an appropriate \p */
3516                case ESC_WU:     /* or \P to test Unicode properties instead */
3517                case ESC_su:     /* of the default ASCII testing. */
3518                case ESC_SU:
3519                nestptr = ptr;
3520                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3521                class_charcount -= 2;                /* Undo! */
3522                continue;
3523    #endif
3524                case ESC_d:
3525                for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3526                continue;
3527    
3528                case ESC_D:
3529                should_flip_negation = TRUE;
3530                for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
3531                continue;
3532    
3533                case ESC_w:
3534                for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_word];
3535                continue;
3536    
3537                case ESC_W:
3538                should_flip_negation = TRUE;
3539                for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3540                continue;
3541    
3542                /* Perl 5.004 onwards omits VT from \s, but we must preserve it
3543                if it was previously set by something earlier in the character
3544                class. */
3545    
3546                case ESC_s:
3547                classbits[0] |= cbits[cbit_space];
3548                classbits[1] |= cbits[cbit_space+1] & ~0x08;
3549                for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3550                continue;
3551    
3552                case ESC_S:
3553                should_flip_negation = TRUE;
3554                for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
3555                classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3556                continue;
3557    
3558                case ESC_h:
3559                SETBIT(classbits, 0x09); /* VT */
3560                SETBIT(classbits, 0x20); /* SPACE */
3561                SETBIT(classbits, 0xa0); /* NSBP */
3562    #ifdef SUPPORT_UTF8
3563                if (utf8)
3564                  {
3565                  class_utf8 = TRUE;
3566                  *class_utf8data++ = XCL_SINGLE;
3567                  class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);
3568                  *class_utf8data++ = XCL_SINGLE;
3569                  class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);
3570                  *class_utf8data++ = XCL_RANGE;
3571                  class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);
3572                  class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);
3573                  *class_utf8data++ = XCL_SINGLE;
3574                  class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);
3575                  *class_utf8data++ = XCL_SINGLE;
3576                  class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);
3577                  *class_utf8data++ = XCL_SINGLE;
3578                  class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);
3579                  }
3580    #endif
3581                continue;
3582    
3583                case ESC_H:
3584                for (c = 0; c < 32; c++)
3585                  {
3586                  int x = 0xff;
3587                  switch (c)
3588                    {
3589                    case 0x09/8: x ^= 1 << (0x09%8); break;
3590                    case 0x20/8: x ^= 1 << (0x20%8); break;
3591                    case 0xa0/8: x ^= 1 << (0xa0%8); break;
3592                    default: break;
3593                    }
3594                  classbits[c] |= x;
3595                  }
3596    
3597    #ifdef SUPPORT_UTF8
3598                if (utf8)
3599                  {
3600                  class_utf8 = TRUE;
3601                  *class_utf8data++ = XCL_RANGE;
3602                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3603                  class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);
3604                  *class_utf8data++ = XCL_RANGE;
3605                  class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);
3606                  class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);
3607                  *class_utf8data++ = XCL_RANGE;
3608                  class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);
3609                  class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);
3610                  *class_utf8data++ = XCL_RANGE;
3611                  class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);
3612                  class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);
3613                  *class_utf8data++ = XCL_RANGE;
3614                  class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);
3615                  class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);
3616                  *class_utf8data++ = XCL_RANGE;
3617                  class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);
3618                  class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);
3619                  *class_utf8data++ = XCL_RANGE;
3620                  class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);
3621                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3622                  }
3623    #endif
3624                continue;
3625    
3626                case ESC_v:
3627                SETBIT(classbits, 0x0a); /* LF */
3628                SETBIT(classbits, 0x0b); /* VT */
3629                SETBIT(classbits, 0x0c); /* FF */
3630                SETBIT(classbits, 0x0d); /* CR */
3631                SETBIT(classbits, 0x85); /* NEL */
3632    #ifdef SUPPORT_UTF8
3633                if (utf8)
3634                  {
3635                  class_utf8 = TRUE;
3636                  *class_utf8data++ = XCL_RANGE;
3637                  class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);
3638                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3639                  }
3640    #endif
3641                continue;
3642    
3643                case ESC_V:
3644                for (c = 0; c < 32; c++)
3645                  {
3646                  int x = 0xff;
3647                  switch (c)
3648                    {
3649                    case 0x0a/8: x ^= 1 << (0x0a%8);
3650                                 x ^= 1 << (0x0b%8);
3651                                 x ^= 1 << (0x0c%8);
3652                                 x ^= 1 << (0x0d%8);
3653                                 break;
3654                    case 0x85/8: x ^= 1 << (0x85%8); break;
3655                    default: break;
3656                    }
3657                  classbits[c] |= x;
3658                  }
3659    
3660    #ifdef SUPPORT_UTF8
3661                if (utf8)
3662                  {
3663                  class_utf8 = TRUE;
3664                  *class_utf8data++ = XCL_RANGE;
3665                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3666                  class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);
3667                  *class_utf8data++ = XCL_RANGE;
3668                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3669                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3670                  }
3671    #endif
3672                continue;
3673    
3674    #ifdef SUPPORT_UCP
3675                case ESC_p:
3676                case ESC_P:
3677                  {
3678                  BOOL negated;
3679                  int pdata;
3680                  int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3681                  if (ptype < 0) goto FAILED;
3682                  class_utf8 = TRUE;
3683                  *class_utf8data++ = ((-c == ESC_p) != negated)?
3684                    XCL_PROP : XCL_NOTPROP;
3685                  *class_utf8data++ = ptype;
3686                  *class_utf8data++ = pdata;
3687                  class_charcount -= 2;   /* Not a < 256 character */
3688                  continue;
3689                  }
3690    #endif
3691                /* Unrecognized escapes are faulted if PCRE is running in its
3692                strict mode. By default, for compatibility with Perl, they are
3693                treated as literals. */
3694    
3695                default:
3696                if ((options & PCRE_EXTRA) != 0)
3697                  {
3698                  *errorcodeptr = ERR7;
3699                  goto FAILED;
3700                  }
3701                class_charcount -= 2;  /* Undo the default count from above */
3702                c = *ptr;              /* Get the final character and fall through */
3703                break;
3704                }
3705              }
3706    
3707            /* Fall through if we have a single character (c >= 0). This may be
3708            greater than 256 in UTF-8 mode. */
3709    
3710            }   /* End of backslash handling */
3711    
3712          /* A single character may be followed by '-' to form a range. However,
3713          Perl does not permit ']' to be the end of the range. A '-' character
3714          at the end is treated as a literal. Perl ignores orphaned \E sequences
3715          entirely. The code for handling \Q and \E is messy. */
3716    
3717          CHECK_RANGE:
3718          while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3719            {
3720            inescq = FALSE;
3721            ptr += 2;
3722            }
3723    
3724          oldptr = ptr;
3725    
3726          /* Remember \r or \n */
3727    
3728          if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3729    
3730