/[pcre]/code/branches/pcre16/pcre_compile.c
ViewVC logotype

Diff of /code/branches/pcre16/pcre_compile.c

Parent Directory Parent Directory | Revision Log Revision Log | View Patch Patch

revision 79 by nigel, Sat Feb 24 21:40:52 2007 UTC revision 545 by ph10, Wed Jun 16 10:51:15 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 PCRE_DEBUG is defined, we need the pcre_printint() function, which is
57    also used by pcretest. PCRE_DEBUG is not defined when building a production
58    library. */
59    
60    #ifdef PCRE_DEBUG
61    #include "pcre_printint.src"
62    #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 BRASTACK_SIZE 200  #define COMPILE_WORK_SIZE (4096)
94    
95    /* 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 63  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 87  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 106  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    #define STRING(a)  # a
313    #define XSTRING(s) STRING(s)
314    
315  static const char *error_texts[] = {  /* The texts of compile-time error messages. These are "char *" because they
316    "no error",  are passed to the outside world. Do not ever re-use any error number, because
317    "\\ at end of pattern",  they are documented. Always add a new error instead. Messages marked DEAD below
318    "\\c at end of pattern",  are no longer used. This used to be a table of strings, but in order to reduce
319    "unrecognized character follows \\",  the number of relocations needed when a shared library is loaded dynamically,
320    "numbers out of order in {} quantifier",  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      ;
412    
413  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
414  patterns. Note that the tables in chartables are dependent on the locale, and  patterns. Note that the tables in chartables are dependent on the locale, and
# Line 220  For convenience, we use the same bit def Line 426  For convenience, we use the same bit def
426    
427  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
428    
429  #if !EBCDIC    /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC
430    
431    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
432    UTF-8 mode. */
433    
434  static const unsigned char digitab[] =  static const unsigned char digitab[] =
435    {    {
436    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
# Line 256  static const unsigned char digitab[] = Line 466  static const unsigned char digitab[] =
466    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
467    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
468    
469  #else          /* This is the "abnormal" case, for EBCDIC systems */  #else
470    
471    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
472    
473  static const unsigned char digitab[] =  static const unsigned char digitab[] =
474    {    {
475    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
# Line 270  static const unsigned char digitab[] = Line 483  static const unsigned char digitab[] =
483    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */
484    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */
485    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */
486    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88-     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88- 95    */
487    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */
488    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */
489    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */
# Line 304  static const unsigned char ebcdic_charta Line 517  static const unsigned char ebcdic_charta
517    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */
518    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */
519    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */
520    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88-  */    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88- 95 */
521    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */
522    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */
523    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */
# Line 331  static const unsigned char ebcdic_charta Line 544  static const unsigned char ebcdic_charta
544  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
545    
546  static BOOL  static BOOL
547    compile_regex(int, int, int *, uschar **, const uschar **, int *, BOOL, int,    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,
548      int *, int *, branch_chain *, compile_data *);      int *, int *, branch_chain *, compile_data *, int *);
549    
550    
551    
552    /*************************************************
553    *            Find an error text                  *
554    *************************************************/
555    
556    /* The error texts are now all in one long string, to save on relocations. As
557    some of the text is of unknown length, we can't use a table of offsets.
558    Instead, just count through the strings. This is not a performance issue
559    because it happens only when there has been a compilation error.
560    
561    Argument:   the error number
562    Returns:    pointer to the error string
563    */
564    
565    static const char *
566    find_error_text(int n)
567    {
568    const char *s = error_texts;
569    for (; n > 0; n--)
570      {
571      while (*s++ != 0) {};
572      if (*s == 0) return "Error text not found (please report)";
573      }
574    return s;
575    }
576    
577    
578  /*************************************************  /*************************************************
# Line 342  static BOOL Line 581  static BOOL
581    
582  /* 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
583  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
584  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
585  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
586  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,
587    ptr is pointing at the \. On exit, it is on the final character of the escape
588    sequence.
589    
590  Arguments:  Arguments:
591    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
# Line 355  Arguments: Line 596  Arguments:
596    
597  Returns:         zero or positive => a data character  Returns:         zero or positive => a data character
598                   negative => a special escape sequence                   negative => a special escape sequence
599                   on error, errorptr is set                   on error, errorcodeptr is set
600  */  */
601    
602  static int  static int
603  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,
604    int options, BOOL isclass)    int options, BOOL isclass)
605  {  {
606  const uschar *ptr = *ptrptr;  BOOL utf8 = (options & PCRE_UTF8) != 0;
607    const uschar *ptr = *ptrptr + 1;
608  int c, i;  int c, i;
609    
610    GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
611    ptr--;                            /* Set pointer back to the last byte */
612    
613  /* 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. */
614    
 c = *(++ptr);  
615  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
616    
617  /* 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
618  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.
619  Otherwise further processing may be required. */  Otherwise further processing may be required. */
620    
621  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
622  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */
623  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0) c = i;
624    
625  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
626  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */
627  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
628  #endif  #endif
629    
# Line 388  else if ((i = escapes[c - 0x48]) != 0) Line 632  else if ((i = escapes[c - 0x48]) != 0)
632  else  else
633    {    {
634    const uschar *oldptr;    const uschar *oldptr;
635      BOOL braced, negated;
636    
637    switch (c)    switch (c)
638      {      {
639      /* 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
640      error. */      error. */
641    
642      case 'l':      case CHAR_l:
643      case 'L':      case CHAR_L:
644      case 'N':      case CHAR_u:
645      case 'u':      case CHAR_U:
     case 'U':  
646      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
647      break;      break;
648    
649        /* \g must be followed by one of a number of specific things:
650    
651        (1) A number, either plain or braced. If positive, it is an absolute
652        backreference. If negative, it is a relative backreference. This is a Perl
653        5.10 feature.
654    
655        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
656        is part of Perl's movement towards a unified syntax for back references. As
657        this is synonymous with \k{name}, we fudge it up by pretending it really
658        was \k.
659    
660        (3) For Oniguruma compatibility we also support \g followed by a name or a
661        number either in angle brackets or in single quotes. However, these are
662        (possibly recursive) subroutine calls, _not_ backreferences. Just return
663        the -ESC_g code (cf \k). */
664    
665        case CHAR_g:
666        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
667          {
668          c = -ESC_g;
669          break;
670          }
671    
672        /* Handle the Perl-compatible cases */
673    
674        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
675          {
676          const uschar *p;
677          for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
678            if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
679          if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
680            {
681            c = -ESC_k;
682            break;
683            }
684          braced = TRUE;
685          ptr++;
686          }
687        else braced = FALSE;
688    
689        if (ptr[1] == CHAR_MINUS)
690          {
691          negated = TRUE;
692          ptr++;
693          }
694        else negated = FALSE;
695    
696        c = 0;
697        while ((digitab[ptr[1]] & ctype_digit) != 0)
698          c = c * 10 + *(++ptr) - CHAR_0;
699    
700        if (c < 0)   /* Integer overflow */
701          {
702          *errorcodeptr = ERR61;
703          break;
704          }
705    
706        if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
707          {
708          *errorcodeptr = ERR57;
709          break;
710          }
711    
712        if (c == 0)
713          {
714          *errorcodeptr = ERR58;
715          break;
716          }
717    
718        if (negated)
719          {
720          if (c > bracount)
721            {
722            *errorcodeptr = ERR15;
723            break;
724            }
725          c = bracount - (c - 1);
726          }
727    
728        c = -(ESC_REF + c);
729        break;
730    
731      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
732      starting with one that is not zero is not straightforward. By experiment,      starting with one that is not zero is not straightforward. By experiment,
733      the way Perl works seems to be as follows:      the way Perl works seems to be as follows:
# Line 413  else Line 740  else
740      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
741      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
742    
743      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:
744      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
745    
746      if (!isclass)      if (!isclass)
747        {        {
748        oldptr = ptr;        oldptr = ptr;
749        c -= '0';        c -= CHAR_0;
750        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
751          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - CHAR_0;
752          if (c < 0)    /* Integer overflow */
753            {
754            *errorcodeptr = ERR61;
755            break;
756            }
757        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
758          {          {
759          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 434  else Line 766  else
766      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.
767      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
768    
769      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
770        {        {
771        ptr--;        ptr--;
772        c = 0;        c = 0;
# Line 442  else Line 774  else
774        }        }
775    
776      /* \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
777      larger first octal digit. */      larger first octal digit. The original code used just to take the least
778        significant 8 bits of octal numbers (I think this is what early Perls used
779      case '0':      to do). Nowadays we allow for larger numbers in UTF-8 mode, but no more
780      c -= '0';      than 3 octal digits. */
781      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')  
782          c = c * 8 + *(++ptr) - '0';      case CHAR_0:
783      c &= 255;     /* Take least significant 8 bits */      c -= CHAR_0;
784        while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
785            c = c * 8 + *(++ptr) - CHAR_0;
786        if (!utf8 && c > 255) *errorcodeptr = ERR51;
787      break;      break;
788    
789      /* \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
790      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
791        treated as a data character. */
792    
793      case 'x':      case CHAR_x:
794  #ifdef SUPPORT_UTF8      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
     if (ptr[1] == '{' && (options & PCRE_UTF8) != 0)  
795        {        {
796        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
797        register int count = 0;        int count = 0;
798    
799        c = 0;        c = 0;
800        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
801          {          {
802          int cc = *pt++;          register int cc = *pt++;
803            if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
804          count++;          count++;
805  #if !EBCDIC    /* ASCII coding */  
806          if (cc >= 'a') cc -= 32;               /* Convert to upper case */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
807          c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
808  #else          /* EBCDIC coding */          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
809          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */  #else           /* EBCDIC coding */
810          c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
811            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
812  #endif  #endif
813          }          }
814        if (*pt == '}')  
815          if (*pt == CHAR_RIGHT_CURLY_BRACKET)
816          {          {
817          if (c < 0 || count > 8) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
818          ptr = pt;          ptr = pt;
819          break;          break;
820          }          }
821    
822        /* 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
823        recognize this construct; fall through to the normal \x handling. */        recognize this construct; fall through to the normal \x handling. */
824        }        }
 #endif  
825    
826      /* Read just a single hex char */      /* Read just a single-byte hex-defined char */
827    
828      c = 0;      c = 0;
829      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
830        {        {
831        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
832        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
833  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
834        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
835        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
836  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
837        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
838        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
839  #endif  #endif
840        }        }
841      break;      break;
842    
843      /* Other special escapes not starting with a digit are straightforward */      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
844        This coding is ASCII-specific, but then the whole concept of \cx is
845        ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
846    
847      case 'c':      case CHAR_c:
848      c = *(++ptr);      c = *(++ptr);
849      if (c == 0)      if (c == 0)
850        {        {
851        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
852        return 0;        break;
853        }        }
854    
855      /* A letter is upper-cased; then the 0x40 bit is flipped. This coding  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
856      is ASCII-specific, but then the whole concept of \cx is ASCII-specific.      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
     (However, an EBCDIC equivalent has now been added.) */  
   
 #if !EBCDIC    /* ASCII coding */  
     if (c >= 'a' && c <= 'z') c -= 32;  
857      c ^= 0x40;      c ^= 0x40;
858  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
859      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
860      c ^= 0xC0;      c ^= 0xC0;
861  #endif  #endif
862      break;      break;
863    
864      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
865      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
866      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
867      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
868      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
869    
870      default:      default:
871      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 541  else Line 878  else
878      }      }
879    }    }
880    
881    /* Perl supports \N{name} for character names, as well as plain \N for "not
882    newline". PCRE does not support \N{name}. */
883    
884    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET)
885      *errorcodeptr = ERR37;
886    
887    /* If PCRE_UCP is set, we change the values for \d etc. */
888    
889    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
890      c -= (ESC_DU - ESC_D);
891    
892    /* Set the pointer to the final character before returning. */
893    
894  *ptrptr = ptr;  *ptrptr = ptr;
895  return c;  return c;
896  }  }
# Line 560  escape sequence. Line 910  escape sequence.
910  Argument:  Argument:
911    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
912    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
913      dptr           points to an int that is set to the detailed property value
914    errorcodeptr   points to the error code variable    errorcodeptr   points to the error code variable
915    
916  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
917  */  */
918    
919  static int  static int
920  get_ucp(const uschar **ptrptr, BOOL *negptr, int *errorcodeptr)  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)
921  {  {
922  int c, i, bot, top;  int c, i, bot, top;
923  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
924  char name[4];  char name[32];
925    
926  c = *(++ptr);  c = *(++ptr);
927  if (c == 0) goto ERROR_RETURN;  if (c == 0) goto ERROR_RETURN;
928    
929  *negptr = FALSE;  *negptr = FALSE;
930    
931  /* \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
932  preceded by ^ for negation. */  negation. */
933    
934  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
935    {    {
936    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
937      {      {
938      *negptr = TRUE;      *negptr = TRUE;
939      ptr++;      ptr++;
940      }      }
941    for (i = 0; i <= 2; i++)    for (i = 0; i < (int)sizeof(name) - 1; i++)
942      {      {
943      c = *(++ptr);      c = *(++ptr);
944      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
945      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
946      name[i] = c;      name[i] = c;
947      }      }
948    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;  
     }  
949    name[i] = 0;    name[i] = 0;
950    }    }
951    
# Line 619  top = _pcre_utt_size; Line 966  top = _pcre_utt_size;
966    
967  while (bot < top)  while (bot < top)
968    {    {
969    i = (bot + top)/2;    i = (bot + top) >> 1;
970    c = strcmp(name, _pcre_utt[i].name);    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);
971    if (c == 0) return _pcre_utt[i].value;    if (c == 0)
972        {
973        *dptr = _pcre_utt[i].value;
974        return _pcre_utt[i].type;
975        }
976    if (c > 0) bot = i + 1; else top = i;    if (c > 0) bot = i + 1; else top = i;
977    }    }
978    
 UNKNOWN_RETURN:  
979  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
980  *ptrptr = ptr;  *ptrptr = ptr;
981  return -1;  return -1;
# Line 660  is_counted_repeat(const uschar *p) Line 1010  is_counted_repeat(const uschar *p)
1010  {  {
1011  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
1012  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
1013  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
1014    
1015  if (*p++ != ',') return FALSE;  if (*p++ != CHAR_COMMA) return FALSE;
1016  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
1017    
1018  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
1019  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
1020    
1021  return (*p == '}');  return (*p == CHAR_RIGHT_CURLY_BRACKET);
1022  }  }
1023    
1024    
# Line 698  read_repeat_counts(const uschar *p, int Line 1048  read_repeat_counts(const uschar *p, int
1048  int min = 0;  int min = 0;
1049  int max = -1;  int max = -1;
1050    
1051  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  /* Read the minimum value and do a paranoid check: a negative value indicates
1052    an integer overflow. */
1053    
1054    while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;
1055    if (min < 0 || min > 65535)
1056      {
1057      *errorcodeptr = ERR5;
1058      return p;
1059      }
1060    
1061    /* Read the maximum value if there is one, and again do a paranoid on its size.
1062    Also, max must not be less than min. */
1063    
1064  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1065    {    {
1066    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1067      {      {
1068      max = 0;      max = 0;
1069      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;
1070        if (max < 0 || max > 65535)
1071          {
1072          *errorcodeptr = ERR5;
1073          return p;
1074          }
1075      if (max < min)      if (max < min)
1076        {        {
1077        *errorcodeptr = ERR4;        *errorcodeptr = ERR4;
# Line 714  if (*p == '}') max = min; else Line 1080  if (*p == '}') max = min; else
1080      }      }
1081    }    }
1082    
1083  /* Do paranoid checks, then fill in the required variables, and pass back the  /* Fill in the required variables, and pass back the pointer to the terminating
1084  pointer to the terminating '}'. */  '}'. */
1085    
1086  if (min > 65535 || max > 65535)  *minp = min;
1087    *errorcodeptr = ERR5;  *maxp = max;
1088  else  return p;
1089    }
1090    
1091    
1092    
1093    /*************************************************
1094    *  Subroutine for finding forward reference      *
1095    *************************************************/
1096    
1097    /* This recursive function is called only from find_parens() below. The
1098    top-level call starts at the beginning of the pattern. All other calls must
1099    start at a parenthesis. It scans along a pattern's text looking for capturing
1100    subpatterns, and counting them. If it finds a named pattern that matches the
1101    name it is given, it returns its number. Alternatively, if the name is NULL, it
1102    returns when it reaches a given numbered subpattern. We know that if (?P< is
1103    encountered, the name will be terminated by '>' because that is checked in the
1104    first pass. Recursion is used to keep track of subpatterns that reset the
1105    capturing group numbers - the (?| feature.
1106    
1107    Arguments:
1108      ptrptr       address of the current character pointer (updated)
1109      cd           compile background data
1110      name         name to seek, or NULL if seeking a numbered subpattern
1111      lorn         name length, or subpattern number if name is NULL
1112      xmode        TRUE if we are in /x mode
1113      count        pointer to the current capturing subpattern number (updated)
1114    
1115    Returns:       the number of the named subpattern, or -1 if not found
1116    */
1117    
1118    static int
1119    find_parens_sub(uschar **ptrptr, compile_data *cd, const uschar *name, int lorn,
1120      BOOL xmode, int *count)
1121    {
1122    uschar *ptr = *ptrptr;
1123    int start_count = *count;
1124    int hwm_count = start_count;
1125    BOOL dup_parens = FALSE;
1126    
1127    /* If the first character is a parenthesis, check on the type of group we are
1128    dealing with. The very first call may not start with a parenthesis. */
1129    
1130    if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1131    {    {
1132    *minp = min;    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1133    *maxp = max;  
1134      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1135    
1136      /* Handle a normal, unnamed capturing parenthesis. */
1137    
1138      else if (ptr[1] != CHAR_QUESTION_MARK)
1139        {
1140        *count += 1;
1141        if (name == NULL && *count == lorn) return *count;
1142        ptr++;
1143        }
1144    
1145      /* All cases now have (? at the start. Remember when we are in a group
1146      where the parenthesis numbers are duplicated. */
1147    
1148      else if (ptr[2] == CHAR_VERTICAL_LINE)
1149        {
1150        ptr += 3;
1151        dup_parens = TRUE;
1152        }
1153    
1154      /* Handle comments; all characters are allowed until a ket is reached. */
1155    
1156      else if (ptr[2] == CHAR_NUMBER_SIGN)
1157        {
1158        for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1159        goto FAIL_EXIT;
1160        }
1161    
1162      /* Handle a condition. If it is an assertion, just carry on so that it
1163      is processed as normal. If not, skip to the closing parenthesis of the
1164      condition (there can't be any nested parens). */
1165    
1166      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1167        {
1168        ptr += 2;
1169        if (ptr[1] != CHAR_QUESTION_MARK)
1170          {
1171          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1172          if (*ptr != 0) ptr++;
1173          }
1174        }
1175    
1176      /* Start with (? but not a condition. */
1177    
1178      else
1179        {
1180        ptr += 2;
1181        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1182    
1183        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1184    
1185        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1186            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1187          {
1188          int term;
1189          const uschar *thisname;
1190          *count += 1;
1191          if (name == NULL && *count == lorn) return *count;
1192          term = *ptr++;
1193          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1194          thisname = ptr;
1195          while (*ptr != term) ptr++;
1196          if (name != NULL && lorn == ptr - thisname &&
1197              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1198            return *count;
1199          term++;
1200          }
1201        }
1202    }    }
1203  return p;  
1204    /* Past any initial parenthesis handling, scan for parentheses or vertical
1205    bars. */
1206    
1207    for (; *ptr != 0; ptr++)
1208      {
1209      /* Skip over backslashed characters and also entire \Q...\E */
1210    
1211      if (*ptr == CHAR_BACKSLASH)
1212        {
1213        if (*(++ptr) == 0) goto FAIL_EXIT;
1214        if (*ptr == CHAR_Q) for (;;)
1215          {
1216          while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1217          if (*ptr == 0) goto FAIL_EXIT;
1218          if (*(++ptr) == CHAR_E) break;
1219          }
1220        continue;
1221        }
1222    
1223      /* Skip over character classes; this logic must be similar to the way they
1224      are handled for real. If the first character is '^', skip it. Also, if the
1225      first few characters (either before or after ^) are \Q\E or \E we skip them
1226      too. This makes for compatibility with Perl. Note the use of STR macros to
1227      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1228    
1229      if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1230        {
1231        BOOL negate_class = FALSE;
1232        for (;;)
1233          {
1234          if (ptr[1] == CHAR_BACKSLASH)
1235            {
1236            if (ptr[2] == CHAR_E)
1237              ptr+= 2;
1238            else if (strncmp((const char *)ptr+2,
1239                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1240              ptr += 4;
1241            else
1242              break;
1243            }
1244          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1245            {
1246            negate_class = TRUE;
1247            ptr++;
1248            }
1249          else break;
1250          }
1251    
1252        /* If the next character is ']', it is a data character that must be
1253        skipped, except in JavaScript compatibility mode. */
1254    
1255        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1256            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1257          ptr++;
1258    
1259        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1260          {
1261          if (*ptr == 0) return -1;
1262          if (*ptr == CHAR_BACKSLASH)
1263            {
1264            if (*(++ptr) == 0) goto FAIL_EXIT;
1265            if (*ptr == CHAR_Q) for (;;)
1266              {
1267              while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1268              if (*ptr == 0) goto FAIL_EXIT;
1269              if (*(++ptr) == CHAR_E) break;
1270              }
1271            continue;
1272            }
1273          }
1274        continue;
1275        }
1276    
1277      /* Skip comments in /x mode */
1278    
1279      if (xmode && *ptr == CHAR_NUMBER_SIGN)
1280        {
1281        while (*(++ptr) != 0 && *ptr != CHAR_NL) {};
1282        if (*ptr == 0) goto FAIL_EXIT;
1283        continue;
1284        }
1285    
1286      /* Check for the special metacharacters */
1287    
1288      if (*ptr == CHAR_LEFT_PARENTHESIS)
1289        {
1290        int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, count);
1291        if (rc > 0) return rc;
1292        if (*ptr == 0) goto FAIL_EXIT;
1293        }
1294    
1295      else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1296        {
1297        if (dup_parens && *count < hwm_count) *count = hwm_count;
1298        goto FAIL_EXIT;
1299        }
1300    
1301      else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1302        {
1303        if (*count > hwm_count) hwm_count = *count;
1304        *count = start_count;
1305        }
1306      }
1307    
1308    FAIL_EXIT:
1309    *ptrptr = ptr;
1310    return -1;
1311    }
1312    
1313    
1314    
1315    
1316    /*************************************************
1317    *       Find forward referenced subpattern       *
1318    *************************************************/
1319    
1320    /* This function scans along a pattern's text looking for capturing
1321    subpatterns, and counting them. If it finds a named pattern that matches the
1322    name it is given, it returns its number. Alternatively, if the name is NULL, it
1323    returns when it reaches a given numbered subpattern. This is used for forward
1324    references to subpatterns. We used to be able to start this scan from the
1325    current compiling point, using the current count value from cd->bracount, and
1326    do it all in a single loop, but the addition of the possibility of duplicate
1327    subpattern numbers means that we have to scan from the very start, in order to
1328    take account of such duplicates, and to use a recursive function to keep track
1329    of the different types of group.
1330    
1331    Arguments:
1332      cd           compile background data
1333      name         name to seek, or NULL if seeking a numbered subpattern
1334      lorn         name length, or subpattern number if name is NULL
1335      xmode        TRUE if we are in /x mode
1336    
1337    Returns:       the number of the found subpattern, or -1 if not found
1338    */
1339    
1340    static int
1341    find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode)
1342    {
1343    uschar *ptr = (uschar *)cd->start_pattern;
1344    int count = 0;
1345    int rc;
1346    
1347    /* If the pattern does not start with an opening parenthesis, the first call
1348    to find_parens_sub() will scan right to the end (if necessary). However, if it
1349    does start with a parenthesis, find_parens_sub() will return when it hits the
1350    matching closing parens. That is why we have to have a loop. */
1351    
1352    for (;;)
1353      {
1354      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, &count);
1355      if (rc > 0 || *ptr++ == 0) break;
1356      }
1357    
1358    return rc;
1359  }  }
1360    
1361    
1362    
1363    
1364  /*************************************************  /*************************************************
1365  *      Find first significant op code            *  *      Find first significant op code            *
1366  *************************************************/  *************************************************/
# Line 778  for (;;) Line 1410  for (;;)
1410    
1411      case OP_CALLOUT:      case OP_CALLOUT:
1412      case OP_CREF:      case OP_CREF:
1413      case OP_BRANUMBER:      case OP_NCREF:
1414        case OP_RREF:
1415        case OP_NRREF:
1416        case OP_DEF:
1417      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1418      break;      break;
1419    
# Line 793  for (;;) Line 1428  for (;;)
1428    
1429    
1430  /*************************************************  /*************************************************
1431  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1432  *************************************************/  *************************************************/
1433    
1434  /* 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,
1435  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.
1436  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
1437    temporarily terminated with OP_END when this function is called.
1438    
1439    This function is called when a backward assertion is encountered, so that if it
1440    fails, the error message can point to the correct place in the pattern.
1441    However, we cannot do this when the assertion contains subroutine calls,
1442    because they can be forward references. We solve this by remembering this case
1443    and doing the check at the end; a flag specifies which mode we are running in.
1444    
1445  Arguments:  Arguments:
1446    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1447    options  the compiling options    options  the compiling options
1448      atend    TRUE if called when the pattern is complete
1449      cd       the "compile data" structure
1450    
1451  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1452                 or -1 if there is no fixed length,
1453               or -2 if \C was encountered               or -2 if \C was encountered
1454                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1455  */  */
1456    
1457  static int  static int
1458  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)
1459  {  {
1460  int length = -1;  int length = -1;
1461    
# Line 822  branch, check the length against that of Line 1468  branch, check the length against that of
1468  for (;;)  for (;;)
1469    {    {
1470    int d;    int d;
1471      uschar *ce, *cs;
1472    register int op = *cc;    register int op = *cc;
   if (op >= OP_BRA) op = OP_BRA;  
   
1473    switch (op)    switch (op)
1474      {      {
1475        case OP_CBRA:
1476      case OP_BRA:      case OP_BRA:
1477      case OP_ONCE:      case OP_ONCE:
1478      case OP_COND:      case OP_COND:
1479      d = find_fixedlength(cc, options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);
1480      if (d < 0) return d;      if (d < 0) return d;
1481      branchlength += d;      branchlength += d;
1482      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 853  for (;;) Line 1499  for (;;)
1499      branchlength = 0;      branchlength = 0;
1500      break;      break;
1501    
1502        /* A true recursion implies not fixed length, but a subroutine call may
1503        be OK. If the subroutine is a forward reference, we can't deal with
1504        it until the end of the pattern, so return -3. */
1505    
1506        case OP_RECURSE:
1507        if (!atend) return -3;
1508        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1509        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1510        if (cc > cs && cc < ce) return -1;                /* Recursion */
1511        d = find_fixedlength(cs + 2, options, atend, cd);
1512        if (d < 0) return d;
1513        branchlength += d;
1514        cc += 1 + LINK_SIZE;
1515        break;
1516    
1517      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1518    
1519      case OP_ASSERT:      case OP_ASSERT:
# Line 865  for (;;) Line 1526  for (;;)
1526      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1527    
1528      case OP_REVERSE:      case OP_REVERSE:
     case OP_BRANUMBER:  
1529      case OP_CREF:      case OP_CREF:
1530        case OP_NCREF:
1531        case OP_RREF:
1532        case OP_NRREF:
1533        case OP_DEF:
1534      case OP_OPT:      case OP_OPT:
1535      case OP_CALLOUT:      case OP_CALLOUT:
1536      case OP_SOD:      case OP_SOD:
1537      case OP_SOM:      case OP_SOM:
1538        case OP_SET_SOM:
1539      case OP_EOD:      case OP_EOD:
1540      case OP_EODN:      case OP_EODN:
1541      case OP_CIRC:      case OP_CIRC:
# Line 884  for (;;) Line 1549  for (;;)
1549    
1550      case OP_CHAR:      case OP_CHAR:
1551      case OP_CHARNC:      case OP_CHARNC:
1552        case OP_NOT:
1553      branchlength++;      branchlength++;
1554      cc += 2;      cc += 2;
1555  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1556      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1557        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1558  #endif  #endif
1559      break;      break;
1560    
# Line 901  for (;;) Line 1565  for (;;)
1565      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1566      cc += 4;      cc += 4;
1567  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1568      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1569        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1570  #endif  #endif
1571      break;      break;
1572    
1573      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1574      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1575        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1576      cc += 4;      cc += 4;
1577      break;      break;
1578    
# Line 917  for (;;) Line 1580  for (;;)
1580    
1581      case OP_PROP:      case OP_PROP:
1582      case OP_NOTPROP:      case OP_NOTPROP:
1583      cc++;      cc += 2;
1584      /* Fall through */      /* Fall through */
1585    
1586      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
# Line 927  for (;;) Line 1590  for (;;)
1590      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1591      case OP_WORDCHAR:      case OP_WORDCHAR:
1592      case OP_ANY:      case OP_ANY:
1593        case OP_ALLANY:
1594      branchlength++;      branchlength++;
1595      cc++;      cc++;
1596      break;      break;
# Line 981  for (;;) Line 1645  for (;;)
1645    
1646    
1647  /*************************************************  /*************************************************
1648  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1649  *************************************************/  *************************************************/
1650    
1651  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1652  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1653    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1654    so that it can be called from pcre_study() when finding the minimum matching
1655    length.
1656    
1657  Arguments:  Arguments:
1658    code        points to start of expression    code        points to start of expression
1659    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1660    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1661    
1662  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
1663  */  */
1664    
1665  static const uschar *  const uschar *
1666  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1667  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
1668  for (;;)  for (;;)
1669    {    {
1670    register int c = *code;    register int c = *code;
1671    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1672    else if (c > OP_BRA)  
1673      /* XCLASS is used for classes that cannot be represented just by a bit
1674      map. This includes negated single high-valued characters. The length in
1675      the table is zero; the actual length is stored in the compiled code. */
1676    
1677      if (c == OP_XCLASS) code += GET(code, 1);
1678    
1679      /* Handle recursion */
1680    
1681      else if (c == OP_REVERSE)
1682        {
1683        if (number < 0) return (uschar *)code;
1684        code += _pcre_OP_lengths[c];
1685        }
1686    
1687      /* Handle capturing bracket */
1688    
1689      else if (c == OP_CBRA)
1690      {      {
1691      int n = c - OP_BRA;      int n = GET2(code, 1+LINK_SIZE);
     if (n > EXTRACT_BASIC_MAX) n = GET2(code, 2+LINK_SIZE);  
1692      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
1693      code += _pcre_OP_lengths[OP_BRA];      code += _pcre_OP_lengths[c];
1694      }      }
1695    
1696      /* Otherwise, we can get the item's length from the table, except that for
1697      repeated character types, we have to test for \p and \P, which have an extra
1698      two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1699      must add in its length. */
1700    
1701    else    else
1702      {      {
1703      code += _pcre_OP_lengths[c];      switch(c)
1704          {
1705          case OP_TYPESTAR:
1706          case OP_TYPEMINSTAR:
1707          case OP_TYPEPLUS:
1708          case OP_TYPEMINPLUS:
1709          case OP_TYPEQUERY:
1710          case OP_TYPEMINQUERY:
1711          case OP_TYPEPOSSTAR:
1712          case OP_TYPEPOSPLUS:
1713          case OP_TYPEPOSQUERY:
1714          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1715          break;
1716    
1717  #ifdef SUPPORT_UTF8        case OP_TYPEUPTO:
1718          case OP_TYPEMINUPTO:
1719          case OP_TYPEEXACT:
1720          case OP_TYPEPOSUPTO:
1721          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1722          break;
1723    
1724      /* In UTF-8 mode, opcodes that are followed by a character may be followed        case OP_MARK:
1725      by a multi-byte character. The length in the table is a minimum, so we have        case OP_PRUNE_ARG:
1726      to scan along to skip the extra bytes. All opcodes are less than 128, so we        case OP_SKIP_ARG:
1727      can use relatively efficient code. */        case OP_THEN_ARG:
1728          code += code[1];
1729          break;
1730          }
1731    
1732        /* Add in the fixed length from the table */
1733    
1734        code += _pcre_OP_lengths[c];
1735    
1736      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1737      a multi-byte character. The length in the table is a minimum, so we have to
1738      arrange to skip the extra bytes. */
1739    
1740    #ifdef SUPPORT_UTF8
1741      if (utf8) switch(c)      if (utf8) switch(c)
1742        {        {
1743        case OP_CHAR:        case OP_CHAR:
# Line 1031  for (;;) Line 1745  for (;;)
1745        case OP_EXACT:        case OP_EXACT:
1746        case OP_UPTO:        case OP_UPTO:
1747        case OP_MINUPTO:        case OP_MINUPTO:
1748          case OP_POSUPTO:
1749        case OP_STAR:        case OP_STAR:
1750        case OP_MINSTAR:        case OP_MINSTAR:
1751          case OP_POSSTAR:
1752        case OP_PLUS:        case OP_PLUS:
1753        case OP_MINPLUS:        case OP_MINPLUS:
1754          case OP_POSPLUS:
1755        case OP_QUERY:        case OP_QUERY:
1756        case OP_MINQUERY:        case OP_MINQUERY:
1757        while ((*code & 0xc0) == 0x80) code++;        case OP_POSQUERY:
1758        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;  
1759        break;        break;
1760        }        }
1761    #else
1762        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1763  #endif  #endif
1764      }      }
1765    }    }
# Line 1072  Returns:      pointer to the opcode for Line 1784  Returns:      pointer to the opcode for
1784  static const uschar *  static const uschar *
1785  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const uschar *code, BOOL utf8)
1786  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
1787  for (;;)  for (;;)
1788    {    {
1789    register int c = *code;    register int c = *code;
1790    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1791    else if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
1792    else if (c > OP_BRA)  
1793      {    /* XCLASS is used for classes that cannot be represented just by a bit
1794      code += _pcre_OP_lengths[OP_BRA];    map. This includes negated single high-valued characters. The length in
1795      }    the table is zero; the actual length is stored in the compiled code. */
1796    
1797      if (c == OP_XCLASS) code += GET(code, 1);
1798    
1799      /* Otherwise, we can get the item's length from the table, except that for
1800      repeated character types, we have to test for \p and \P, which have an extra
1801      two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1802      must add in its length. */
1803    
1804    else    else
1805      {      {
1806      code += _pcre_OP_lengths[c];      switch(c)
1807          {
1808          case OP_TYPESTAR:
1809          case OP_TYPEMINSTAR:
1810          case OP_TYPEPLUS:
1811          case OP_TYPEMINPLUS:
1812          case OP_TYPEQUERY:
1813          case OP_TYPEMINQUERY:
1814          case OP_TYPEPOSSTAR:
1815          case OP_TYPEPOSPLUS:
1816          case OP_TYPEPOSQUERY:
1817          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1818          break;
1819    
1820  #ifdef SUPPORT_UTF8        case OP_TYPEPOSUPTO:
1821          case OP_TYPEUPTO:
1822          case OP_TYPEMINUPTO:
1823          case OP_TYPEEXACT:
1824          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1825          break;
1826    
1827          case OP_MARK:
1828          case OP_PRUNE_ARG:
1829          case OP_SKIP_ARG:
1830          case OP_THEN_ARG:
1831          code += code[1];
1832          break;
1833          }
1834    
1835        /* Add in the fixed length from the table */
1836    
1837        code += _pcre_OP_lengths[c];
1838    
1839      /* 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
1840      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
1841      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. */  
1842    
1843    #ifdef SUPPORT_UTF8
1844      if (utf8) switch(c)      if (utf8) switch(c)
1845        {        {
1846        case OP_CHAR:        case OP_CHAR:
# Line 1103  for (;;) Line 1848  for (;;)
1848        case OP_EXACT:        case OP_EXACT:
1849        case OP_UPTO:        case OP_UPTO:
1850        case OP_MINUPTO:        case OP_MINUPTO:
1851          case OP_POSUPTO:
1852        case OP_STAR:        case OP_STAR:
1853        case OP_MINSTAR:        case OP_MINSTAR:
1854          case OP_POSSTAR:
1855        case OP_PLUS:        case OP_PLUS:
1856        case OP_MINPLUS:        case OP_MINPLUS:
1857          case OP_POSPLUS:
1858        case OP_QUERY:        case OP_QUERY:
1859        case OP_MINQUERY:        case OP_MINQUERY:
1860        while ((*code & 0xc0) == 0x80) code++;        case OP_POSQUERY:
1861        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;  
1862        break;        break;
1863        }        }
1864    #else
1865        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1866  #endif  #endif
1867      }      }
1868    }    }
# Line 1132  for (;;) Line 1875  for (;;)
1875  *************************************************/  *************************************************/
1876    
1877  /* 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
1878  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()
1879  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
1880  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
1881  whose current branch will already have been scanned.  backward and negative forward assertions when its final argument is TRUE. If we
1882    hit an unclosed bracket, we return "empty" - this means we've struck an inner
1883    bracket whose current branch will already have been scanned.
1884    
1885  Arguments:  Arguments:
1886    code        points to start of search    code        points to start of search
1887    endcode     points to where to stop    endcode     points to where to stop
1888    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
1889      cd          contains pointers to tables etc.
1890    
1891  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
1892  */  */
1893    
1894  static BOOL  static BOOL
1895  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
1896      compile_data *cd)
1897  {  {
1898  register int c;  register int c;
1899  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);
1900       code < endcode;       code < endcode;
1901       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))
1902    {    {
# Line 1157  for (code = first_significant_code(code Line 1904  for (code = first_significant_code(code
1904    
1905    c = *code;    c = *code;
1906    
1907    if (c >= OP_BRA)    /* Skip over forward assertions; the other assertions are skipped by
1908      first_significant_code() with a TRUE final argument. */
1909    
1910      if (c == OP_ASSERT)
1911      {      {
1912      BOOL empty_branch;      do code += GET(code, 1); while (*code == OP_ALT);
1913      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      c = *code;
1914        continue;
1915        }
1916    
1917      /* Groups with zero repeats can of course be empty; skip them. */
1918    
1919      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)
1920        {
1921        code += _pcre_OP_lengths[c];
1922        do code += GET(code, 1); while (*code == OP_ALT);
1923        c = *code;
1924        continue;
1925        }
1926    
1927      /* Scan a closed bracket */    /* For a recursion/subroutine call, if its end has been reached, which
1928      implies a subroutine call, we can scan it. */
1929    
1930      empty_branch = FALSE;    if (c == OP_RECURSE)
1931        {
1932        BOOL empty_branch = FALSE;
1933        const uschar *scode = cd->start_code + GET(code, 1);
1934        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
1935      do      do
1936        {        {
1937        if (!empty_branch && could_be_empty_branch(code, endcode, utf8))        if (could_be_empty_branch(scode, endcode, utf8, cd))
1938            {
1939          empty_branch = TRUE;          empty_branch = TRUE;
1940            break;
1941            }
1942          scode += GET(scode, 1);
1943          }
1944        while (*scode == OP_ALT);
1945        if (!empty_branch) return FALSE;  /* All branches are non-empty */
1946        continue;
1947        }
1948    
1949      /* For other groups, scan the branches. */
1950    
1951      if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)
1952        {
1953        BOOL empty_branch;
1954        if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
1955    
1956        /* If a conditional group has only one branch, there is a second, implied,
1957        empty branch, so just skip over the conditional, because it could be empty.
1958        Otherwise, scan the individual branches of the group. */
1959    
1960        if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
1961        code += GET(code, 1);        code += GET(code, 1);
1962        else
1963          {
1964          empty_branch = FALSE;
1965          do
1966            {
1967            if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
1968              empty_branch = TRUE;
1969            code += GET(code, 1);
1970            }
1971          while (*code == OP_ALT);
1972          if (!empty_branch) return FALSE;   /* All branches are non-empty */
1973        }        }
1974      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
     code += 1 + LINK_SIZE;  
1975      c = *code;      c = *code;
1976        continue;
1977      }      }
1978    
1979    else switch (c)    /* Handle the other opcodes */
1980    
1981      switch (c)
1982      {      {
1983      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
1984        cannot be represented just by a bit map. This includes negated single
1985        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
1986        actual length is stored in the compiled code, so we must update "code"
1987        here. */
1988    
1989  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1990      case OP_XCLASS:      case OP_XCLASS:
1991      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
1992      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
1993  #endif  #endif
1994    
# Line 1227  for (code = first_significant_code(code Line 2032  for (code = first_significant_code(code
2032      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2033      case OP_WORDCHAR:      case OP_WORDCHAR:
2034      case OP_ANY:      case OP_ANY:
2035        case OP_ALLANY:
2036      case OP_ANYBYTE:      case OP_ANYBYTE:
2037      case OP_CHAR:      case OP_CHAR:
2038      case OP_CHARNC:      case OP_CHARNC:
2039      case OP_NOT:      case OP_NOT:
2040      case OP_PLUS:      case OP_PLUS:
2041      case OP_MINPLUS:      case OP_MINPLUS:
2042        case OP_POSPLUS:
2043      case OP_EXACT:      case OP_EXACT:
2044      case OP_NOTPLUS:      case OP_NOTPLUS:
2045      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
2046        case OP_NOTPOSPLUS:
2047      case OP_NOTEXACT:      case OP_NOTEXACT:
2048      case OP_TYPEPLUS:      case OP_TYPEPLUS:
2049      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
2050        case OP_TYPEPOSPLUS:
2051      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2052      return FALSE;      return FALSE;
2053    
2054        /* These are going to continue, as they may be empty, but we have to
2055        fudge the length for the \p and \P cases. */
2056    
2057        case OP_TYPESTAR:
2058        case OP_TYPEMINSTAR:
2059        case OP_TYPEPOSSTAR:
2060        case OP_TYPEQUERY:
2061        case OP_TYPEMINQUERY:
2062        case OP_TYPEPOSQUERY:
2063        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2064        break;
2065    
2066        /* Same for these */
2067    
2068        case OP_TYPEUPTO:
2069        case OP_TYPEMINUPTO:
2070        case OP_TYPEPOSUPTO:
2071        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
2072        break;
2073    
2074      /* End of branch */      /* End of branch */
2075    
2076      case OP_KET:      case OP_KET:
# Line 1250  for (code = first_significant_code(code Line 2079  for (code = first_significant_code(code
2079      case OP_ALT:      case OP_ALT:
2080      return TRUE;      return TRUE;
2081    
2082      /* In UTF-8 mode, STAR, MINSTAR, QUERY, MINQUERY, UPTO, and MINUPTO  may be      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2083      followed by a multibyte character */      MINUPTO, and POSUPTO may be followed by a multibyte character */
2084    
2085  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2086      case OP_STAR:      case OP_STAR:
2087      case OP_MINSTAR:      case OP_MINSTAR:
2088        case OP_POSSTAR:
2089      case OP_QUERY:      case OP_QUERY:
2090      case OP_MINQUERY:      case OP_MINQUERY:
2091        case OP_POSQUERY:
2092        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2093        break;
2094    
2095      case OP_UPTO:      case OP_UPTO:
2096      case OP_MINUPTO:      case OP_MINUPTO:
2097      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_POSUPTO:
2098        if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2099      break;      break;
2100  #endif  #endif
2101    
2102        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2103        string. */
2104    
2105        case OP_MARK:
2106        case OP_PRUNE_ARG:
2107        case OP_SKIP_ARG:
2108        case OP_THEN_ARG:
2109        code += code[1];
2110        break;
2111    
2112        /* None of the remaining opcodes are required to match a character. */
2113    
2114        default:
2115        break;
2116      }      }
2117    }    }
2118    
# Line 1285  Arguments: Line 2135  Arguments:
2135    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2136    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2137    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2138      cd          pointers to tables etc
2139    
2140  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2141  */  */
2142    
2143  static BOOL  static BOOL
2144  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2145    BOOL utf8)    BOOL utf8, compile_data *cd)
2146  {  {
2147  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2148    {    {
2149    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2150        return FALSE;
2151    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2152    }    }
2153  return TRUE;  return TRUE;
# Line 1308  return TRUE; Line 2160  return TRUE;
2160  *************************************************/  *************************************************/
2161    
2162  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2163  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
2164  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2165  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2166    
2167    Originally, this function only recognized a sequence of letters between the
2168    terminators, but it seems that Perl recognizes any sequence of characters,
2169    though of course unknown POSIX names are subsequently rejected. Perl gives an
2170    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2171    didn't consider this to be a POSIX class. Likewise for [:1234:].
2172    
2173    The problem in trying to be exactly like Perl is in the handling of escapes. We
2174    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2175    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2176    below handles the special case of \], but does not try to do any other escape
2177    processing. This makes it different from Perl for cases such as [:l\ower:]
2178    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2179    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2180    I think.
2181    
2182  Argument:  Arguments:
2183    ptr      pointer to the initial [    ptr      pointer to the initial [
2184    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2185    
2186  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2187  */  */
2188    
2189  static BOOL  static BOOL
2190  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
2191  {  {
2192  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2193  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2194  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2195    {    {
2196    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else
2197    return TRUE;      {
2198        if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2199        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2200          {
2201          *endptr = ptr;
2202          return TRUE;
2203          }
2204        }
2205    }    }
2206  return FALSE;  return FALSE;
2207  }  }
# Line 1355  Returns:     a value representing the na Line 2226  Returns:     a value representing the na
2226  static int  static int
2227  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
2228  {  {
2229    const char *pn = posix_names;
2230  register int yield = 0;  register int yield = 0;
2231  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2232    {    {
2233    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2234      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
2235      pn += posix_name_lengths[yield] + 1;
2236    yield++;    yield++;
2237    }    }
2238  return -1;  return -1;
# Line 1374  return -1; Line 2247  return -1;
2247  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2248  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2249  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
2250  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
2251  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
2252  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
2253  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
2254  partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2255    OP_END.
2256    
2257    This function has been extended with the possibility of forward references for
2258    recursions and subroutine calls. It must also check the list of such references
2259    for the group we are dealing with. If it finds that one of the recursions in
2260    the current group is on this list, it adjusts the offset in the list, not the
2261    value in the reference (which is a group number).
2262    
2263  Arguments:  Arguments:
2264    group      points to the start of the group    group      points to the start of the group
2265    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
2266    utf8       TRUE in UTF-8 mode    utf8       TRUE in UTF-8 mode
2267    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
2268      save_hwm   the hwm forward reference pointer at the start of the group
2269    
2270  Returns:     nothing  Returns:     nothing
2271  */  */
2272    
2273  static void  static void
2274  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd)  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,
2275      uschar *save_hwm)
2276  {  {
2277  uschar *ptr = group;  uschar *ptr = group;
2278    
2279  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
2280    {    {
2281    int offset = GET(ptr, 1);    int offset;
2282    if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);    uschar *hc;
2283    
2284      /* See if this recursion is on the forward reference list. If so, adjust the
2285      reference. */
2286    
2287      for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
2288        {
2289        offset = GET(hc, 0);
2290        if (cd->start_code + offset == ptr + 1)
2291          {
2292          PUT(hc, 0, offset + adjust);
2293          break;
2294          }
2295        }
2296    
2297      /* Otherwise, adjust the recursion offset if it's after the start of this
2298      group. */
2299    
2300      if (hc >= cd->hwm)
2301        {
2302        offset = GET(ptr, 1);
2303        if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
2304        }
2305    
2306    ptr += 1 + LINK_SIZE;    ptr += 1 + LINK_SIZE;
2307    }    }
2308  }  }
# Line 1423  auto_callout(uschar *code, const uschar Line 2329  auto_callout(uschar *code, const uschar
2329  {  {
2330  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2331  *code++ = 255;  *code++ = 255;
2332  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2333  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2334  return code + 2*LINK_SIZE;  return code + 2*LINK_SIZE;
2335  }  }
2336    
# Line 1449  Returns:             nothing Line 2355  Returns:             nothing
2355  static void  static void
2356  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)
2357  {  {
2358  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2359  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2360  }  }
2361    
# Line 1475  Yield:        TRUE when range returned; Line 2381  Yield:        TRUE when range returned;
2381  */  */
2382    
2383  static BOOL  static BOOL
2384  get_othercase_range(int *cptr, int d, int *ocptr, int *odptr)  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,
2385      unsigned int *odptr)
2386  {  {
2387  int c, chartype, othercase, next;  unsigned int c, othercase, next;
2388    
2389  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2390    {    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
   if (_pcre_ucp_findchar(c, &chartype, &othercase) == ucp_L && othercase != 0)  
     break;  
   }  
2391    
2392  if (c > d) return FALSE;  if (c > d) return FALSE;
2393    
# Line 1492  next = othercase + 1; Line 2396  next = othercase + 1;
2396    
2397  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2398    {    {
2399    if (_pcre_ucp_findchar(c, &chartype, &othercase) != ucp_L ||    if (UCD_OTHERCASE(c) != next) break;
         othercase != next)  
     break;  
2400    next++;    next++;
2401    }    }
2402    
# Line 1503  for (++c; c <= d; c++) Line 2405  for (++c; c <= d; c++)
2405    
2406  return TRUE;  return TRUE;
2407  }  }
2408  #endif  /* SUPPORT_UCP */  
2409    
2410    
2411  /*************************************************  /*************************************************
2412  *           Compile one branch                   *  *        Check a character and a property        *
2413  *************************************************/  *************************************************/
2414    
2415  /* 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
2416  changed during the branch, the pointer is used to change the external options  is adjacent to a fixed character.
 bits.  
2417    
2418  Arguments:  Arguments:
2419    optionsptr     pointer to the option bits    c            the character
2420    brackets       points to number of extracting brackets used    ptype        the property type
2421    codeptr        points to the pointer to the current code point    pdata        the data for the type
2422    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.  
2423    
2424  Returns:         TRUE on success  Returns:       TRUE if auto-possessifying is OK
                  FALSE, with *errorcodeptr set non-zero on error  
2425  */  */
2426    
2427  static BOOL  static BOOL
2428  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)  
2429  {  {
2430  int repeat_type, op_type;  const ucd_record *prop = GET_UCD(c);
2431  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  switch(ptype)
2432  int bravalue = 0;    {
2433  int greedy_default, greedy_non_default;    case PT_LAMP:
2434  int firstbyte, reqbyte;    return (prop->chartype == ucp_Lu ||
2435  int zeroreqbyte, zerofirstbyte;            prop->chartype == ucp_Ll ||
2436  int req_caseopt, reqvary, tempreqvary;            prop->chartype == ucp_Lt) == negated;
2437  int condcount = 0;  
2438  int options = *optionsptr;    case PT_GC:
2439  int after_manual_callout = 0;    return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;
2440  register int c;  
2441  register uschar *code = *codeptr;    case PT_PC:
2442  uschar *tempcode;    return (pdata == prop->chartype) == negated;
2443  BOOL inescq = FALSE;  
2444  BOOL groupsetfirstbyte = FALSE;    case PT_SC:
2445  const uschar *ptr = *ptrptr;    return (pdata == prop->script) == negated;
2446  const uschar *tempptr;  
2447  uschar *previous = NULL;    /* These are specials */
2448  uschar *previous_callout = NULL;  
2449  uschar classbits[32];    case PT_ALNUM:
2450      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2451              _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;
2452    
2453      case PT_SPACE:    /* Perl space */
2454      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2455              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2456              == negated;
2457    
2458      case PT_PXSPACE:  /* POSIX space */
2459      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2460              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2461              c == CHAR_FF || c == CHAR_CR)
2462              == negated;
2463    
2464      case PT_WORD:
2465      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2466              _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2467              c == CHAR_UNDERSCORE) == negated;
2468      }
2469    return FALSE;
2470    }
2471    #endif  /* SUPPORT_UCP */
2472    
 #ifdef SUPPORT_UTF8  
 BOOL class_utf8;  
 BOOL utf8 = (options & PCRE_UTF8) != 0;  
 uschar *class_utf8data;  
 uschar utf8_char[6];  
 #else  
 BOOL utf8 = FALSE;  
 #endif  
2473    
 /* Set up the default and non-default settings for greediness */  
2474    
2475  greedy_default = ((options & PCRE_UNGREEDY) != 0);  /*************************************************
2476  greedy_non_default = greedy_default ^ 1;  *     Check if auto-possessifying is possible    *
2477    *************************************************/
2478    
2479  /* Initialize no first byte, no required byte. REQ_UNSET means "no char  /* This function is called for unlimited repeats of certain items, to see
2480  matching encountered yet". It gets changed to REQ_NONE if we hit something that  whether the next thing could possibly match the repeated item. If not, it makes
2481  matches a non-fixed char first char; reqbyte just remains unset if we never  sense to automatically possessify the repeated item.
 find one.  
2482    
2483  When we hit a repeat whose minimum is zero, we may have to adjust these values  Arguments:
2484  to take the zero repeat into account. This is implemented by setting them to    previous      pointer to the repeated opcode
2485  zerofirstbyte and zeroreqbyte when such a repeat is encountered. The individual    utf8          TRUE in UTF-8 mode
2486  item types that can be repeated set these backoff variables appropriately. */    ptr           next character in pattern
2487      options       options bits
2488      cd            contains pointers to tables etc.
2489    
2490  firstbyte = reqbyte = zerofirstbyte = zeroreqbyte = REQ_UNSET;  Returns:        TRUE if possessifying is wanted
2491    */
2492    
2493  /* The variable req_caseopt contains either the REQ_CASELESS value or zero,  static BOOL
2494  according to the current setting of the caseless flag. REQ_CASELESS is a bit  check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
2495  value > 255. It is added into the firstbyte or reqbyte variables to record the    int options, compile_data *cd)
2496  case status of the value. This is used only for ASCII characters. */  {
2497    int c, next;
2498    int op_code = *previous++;
2499    
2500  req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;  /* Skip whitespace and comments in extended mode */
2501    
2502  /* Switch on next character until the end of the branch */  if ((options & PCRE_EXTENDED) != 0)
2503      {
2504      for (;;)
2505        {
2506        while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2507        if (*ptr == CHAR_NUMBER_SIGN)
2508          {
2509          while (*(++ptr) != 0)
2510            if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2511          }
2512        else break;
2513        }
2514      }
2515    
2516  for (;; ptr++)  /* If the next item is one that we can handle, get its value. A non-negative
2517    value is a character, a negative value is an escape value. */
2518    
2519    if (*ptr == CHAR_BACKSLASH)
2520    {    {
2521    BOOL negate_class;    int temperrorcode = 0;
2522    BOOL possessive_quantifier;    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
2523    BOOL is_quantifier;    if (temperrorcode != 0) return FALSE;
2524    int class_charcount;    ptr++;    /* Point after the escape sequence */
2525    int class_lastchar;    }
   int newoptions;  
   int recno;  
   int skipbytes;  
   int subreqbyte;  
   int subfirstbyte;  
   int mclength;  
   uschar mcbuffer[8];  
2526    
2527    /* Next byte in the pattern */  else if ((cd->ctypes[*ptr] & ctype_meta) == 0)
2528      {
2529    #ifdef SUPPORT_UTF8
2530      if (utf8) { GETCHARINC(next, ptr); } else
2531    #endif
2532      next = *ptr++;
2533      }
2534    
2535    c = *ptr;  else return FALSE;
2536    
2537    /* If in \Q...\E, check for the end; if not, we have a literal */  /* Skip whitespace and comments in extended mode */
2538    
2539    if (inescq && c != 0)  if ((options & PCRE_EXTENDED) != 0)
2540      {
2541      for (;;)
2542      {      {
2543      if (c == '\\' && ptr[1] == 'E')      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2544        {      if (*ptr == CHAR_NUMBER_SIGN)
       inescq = FALSE;  
       ptr++;  
       continue;  
       }  
     else  
2545        {        {
2546        if (previous_callout != NULL)        while (*(++ptr) != 0)
2547          {          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
         complete_callout(previous_callout, ptr, cd);  
         previous_callout = NULL;  
         }  
       if ((options & PCRE_AUTO_CALLOUT) != 0)  
         {  
         previous_callout = code;  
         code = auto_callout(code, ptr, cd);  
         }  
       goto NORMAL_CHAR;  
2548        }        }
2549        else break;
2550      }      }
2551      }
2552    
2553    /* Fill in length of a previous callout, except when the next thing is  /* If the next thing is itself optional, we have to give up. */
   a quantifier. */  
2554    
2555    is_quantifier = c == '*' || c == '+' || c == '?' ||  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2556      (c == '{' && is_counted_repeat(ptr+1));    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2557        return FALSE;
2558    
2559    if (!is_quantifier && previous_callout != NULL &&  /* Now compare the next item with the previous opcode. First, handle cases when
2560         after_manual_callout-- <= 0)  the next item is a character. */
2561    
2562    if (next >= 0) switch(op_code)
2563      {
2564      case OP_CHAR:
2565    #ifdef SUPPORT_UTF8
2566      GETCHARTEST(c, previous);
2567    #else
2568      c = *previous;
2569    #endif
2570      return c != next;
2571    
2572      /* For CHARNC (caseless character) we must check the other case. If we have
2573      Unicode property support, we can use it to test the other case of
2574      high-valued characters. */
2575    
2576      case OP_CHARNC:
2577    #ifdef SUPPORT_UTF8
2578      GETCHARTEST(c, previous);
2579    #else
2580      c = *previous;
2581    #endif
2582      if (c == next) return FALSE;
2583    #ifdef SUPPORT_UTF8
2584      if (utf8)
2585      {      {
2586      complete_callout(previous_callout, ptr, cd);      unsigned int othercase;
2587      previous_callout = NULL;      if (next < 128) othercase = cd->fcc[next]; else
2588    #ifdef SUPPORT_UCP
2589        othercase = UCD_OTHERCASE((unsigned int)next);
2590    #else
2591        othercase = NOTACHAR;
2592    #endif
2593        return (unsigned int)c != othercase;
2594      }      }
2595      else
2596    #endif  /* SUPPORT_UTF8 */
2597      return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2598    
2599    /* In extended mode, skip white space and comments */    /* For OP_NOT, its data is always a single-byte character. */
2600    
2601    if ((options & PCRE_EXTENDED) != 0)    case OP_NOT:
2602      if ((c = *previous) == next) return TRUE;
2603      if ((options & PCRE_CASELESS) == 0) return FALSE;
2604    #ifdef SUPPORT_UTF8
2605      if (utf8)
2606      {      {
2607      if ((cd->ctypes[c] & ctype_space) != 0) continue;      unsigned int othercase;
2608      if (c == '#')      if (next < 128) othercase = cd->fcc[next]; else
2609        {  #ifdef SUPPORT_UCP
2610        /* The space before the ; is to avoid a warning on a silly compiler      othercase = UCD_OTHERCASE(next);
2611        on the Macintosh. */  #else
2612        while ((c = *(++ptr)) != 0 && c != NEWLINE) ;      othercase = NOTACHAR;
2613        if (c != 0) continue;   /* Else fall through to handle end of string */  #endif
2614        }      return (unsigned int)c == othercase;
2615      }      }
2616      else
2617    #endif  /* SUPPORT_UTF8 */
2618      return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2619    
2620    /* No auto callout for quantifiers. */    /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2621      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
2622    
2623    if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier)    case OP_DIGIT:
2624      return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
2625    
2626      case OP_NOT_DIGIT:
2627      return next <= 127 && (cd->ctypes[next] & ctype_digit) != 0;
2628    
2629      case OP_WHITESPACE:
2630      return next > 127 || (cd->ctypes[next] & ctype_space) == 0;
2631    
2632      case OP_NOT_WHITESPACE:
2633      return next <= 127 && (cd->ctypes[next] & ctype_space) != 0;
2634    
2635      case OP_WORDCHAR:
2636      return next > 127 || (cd->ctypes[next] & ctype_word) == 0;
2637    
2638      case OP_NOT_WORDCHAR:
2639      return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;
2640    
2641      case OP_HSPACE:
2642      case OP_NOT_HSPACE:
2643      switch(next)
2644      {      {
2645      previous_callout = code;      case 0x09:
2646      code = auto_callout(code, ptr, cd);      case 0x20:
2647        case 0xa0:
2648        case 0x1680:
2649        case 0x180e:
2650        case 0x2000:
2651        case 0x2001:
2652        case 0x2002:
2653        case 0x2003:
2654        case 0x2004:
2655        case 0x2005:
2656        case 0x2006:
2657        case 0x2007:
2658        case 0x2008:
2659        case 0x2009:
2660        case 0x200A:
2661        case 0x202f:
2662        case 0x205f:
2663        case 0x3000:
2664        return op_code == OP_NOT_HSPACE;
2665        default:
2666        return op_code != OP_NOT_HSPACE;
2667      }      }
2668    
2669    switch(c)    case OP_ANYNL:
2670      case OP_VSPACE:
2671      case OP_NOT_VSPACE:
2672      switch(next)
2673      {      {
2674      /* The branch terminates at end of string, |, or ). */      case 0x0a:
2675        case 0x0b:
2676      case 0:      case 0x0c:
2677      case '|':      case 0x0d:
2678      case ')':      case 0x85:
2679      *firstbyteptr = firstbyte;      case 0x2028:
2680      *reqbyteptr = reqbyte;      case 0x2029:
2681      *codeptr = code;      return op_code == OP_NOT_VSPACE;
2682      *ptrptr = ptr;      default:
2683      return TRUE;      return op_code != OP_NOT_VSPACE;
2684        }
2685    
2686      /* Handle single-character metacharacters. In multiline mode, ^ disables  #ifdef SUPPORT_UCP
2687      the setting of any following char as a first character. */    case OP_PROP:
2688      return check_char_prop(next, previous[0], previous[1], FALSE);
2689    
2690      case '^':    case OP_NOTPROP:
2691      if ((options & PCRE_MULTILINE) != 0)    return check_char_prop(next, previous[0], previous[1], TRUE);
2692        {  #endif
       if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;  
       }  
     previous = NULL;  
     *code++ = OP_CIRC;  
     break;  
2693    
2694      case '$':    default:
2695      previous = NULL;    return FALSE;
2696      *code++ = OP_DOLL;    }
     break;  
2697    
     /* There can never be a first char if '.' is first, whatever happens about  
     repeats. The value of reqbyte doesn't change either. */  
2698    
2699      case '.':  /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP
2700      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;  is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
2701      zerofirstbyte = firstbyte;  generated only when PCRE_UCP is *not* set, that is, when only ASCII
2702      zeroreqbyte = reqbyte;  characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
2703      previous = code;  replaced by OP_PROP codes when PCRE_UCP is set. */
     *code++ = OP_ANY;  
     break;  
2704    
2705      /* Character classes. If the included characters are all < 255 in value, we  switch(op_code)
2706      build a 32-byte bitmap of the permitted characters, except in the special    {
2707      case where there is only one such character. For negated classes, we build    case OP_CHAR:
2708      the map as usual, then invert it at the end. However, we use a different    case OP_CHARNC:
2709      opcode so that data characters > 255 can be handled correctly.  #ifdef SUPPORT_UTF8
2710      GETCHARTEST(c, previous);
2711    #else
2712      c = *previous;
2713    #endif
2714      switch(-next)
2715        {
2716        case ESC_d:
2717        return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
2718    
2719      If the class contains characters outside the 0-255 range, a different      case ESC_D:
2720      opcode is compiled. It may optionally have a bit map for characters < 256,      return c <= 127 && (cd->ctypes[c] & 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.  
     */  
2721    
2722      case '[':      case ESC_s:
2723      previous = code;      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
2724    
2725      /* PCRE supports POSIX class stuff inside a class. Perl gives an error if      case ESC_S:
2726      they are encountered at the top level, so we'll do that too. */      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
2727    
2728        case ESC_w:
2729        return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
2730    
2731        case ESC_W:
2732        return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
2733    
2734        case ESC_h:
2735        case ESC_H:
2736        switch(c)
2737          {
2738          case 0x09:
2739          case 0x20:
2740          case 0xa0:
2741          case 0x1680:
2742          case 0x180e:
2743          case 0x2000:
2744          case 0x2001:
2745          case 0x2002:
2746          case 0x2003:
2747          case 0x2004:
2748          case 0x2005:
2749          case 0x2006:
2750          case 0x2007:
2751          case 0x2008:
2752          case 0x2009:
2753          case 0x200A:
2754          case 0x202f:
2755          case 0x205f:
2756          case 0x3000:
2757          return -next != ESC_h;
2758          default:
2759          return -next == ESC_h;
2760          }
2761    
2762      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      case ESC_v:
2763          check_posix_syntax(ptr, &tempptr, cd))      case ESC_V:
2764        {      switch(c)
2765        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        {
2766        goto FAILED;        case 0x0a:
2767          case 0x0b:
2768          case 0x0c:
2769          case 0x0d:
2770          case 0x85:
2771          case 0x2028:
2772          case 0x2029:
2773          return -next != ESC_v;
2774          default:
2775          return -next == ESC_v;
2776        }        }
2777    
2778      /* If the first character is '^', set the negation flag and skip it. */      /* When PCRE_UCP is set, these values get generated for \d etc. Find
2779        their substitutions and process them. The result will always be either
2780        -ESC_p or -ESC_P. Then fall through to process those values. */
2781    
2782      if ((c = *(++ptr)) == '^')  #ifdef SUPPORT_UCP
2783        {      case ESC_du:
2784        negate_class = TRUE;      case ESC_DU:
2785        c = *(++ptr);      case ESC_wu:
2786        case ESC_WU:
2787        case ESC_su:
2788        case ESC_SU:
2789          {
2790          int temperrorcode = 0;
2791          ptr = substitutes[-next - ESC_DU];
2792          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
2793          if (temperrorcode != 0) return FALSE;
2794          ptr++;    /* For compatibility */
2795        }        }
2796      else      /* Fall through */
2797    
2798        case ESC_p:
2799        case ESC_P:
2800        {        {
2801        negate_class = FALSE;        int ptype, pdata, errorcodeptr;
2802        }        BOOL negated;
2803    
2804      /* Keep a count of chars with values < 256 so that we can optimize the case        ptr--;      /* Make ptr point at the p or P */
2805      of just a single character (as long as it's < 256). For higher valued UTF-8        ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
2806      characters, we don't yet do any optimization. */        if (ptype < 0) return FALSE;
2807          ptr++;      /* Point past the final curly ket */
2808    
2809          /* If the property item is optional, we have to give up. (When generated
2810          from \d etc by PCRE_UCP, this test will have been applied much earlier,
2811          to the original \d etc. At this point, ptr will point to a zero byte. */
2812    
2813          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2814            strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2815              return FALSE;
2816    
2817      class_charcount = 0;        /* Do the property check. */
     class_lastchar = -1;  
2818    
2819  #ifdef SUPPORT_UTF8        return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
2820      class_utf8 = FALSE;                       /* No chars >= 256 */        }
     class_utf8data = code + LINK_SIZE + 34;   /* For UTF-8 items */  
2821  #endif  #endif
2822    
2823      /* Initialize the 32-char bit map to all zeros. We have to build the      default:
2824      map in a temporary bit of store, in case the class contains only 1      return FALSE;
2825      character (< 256), because in that case the compiled code doesn't use the      }
     bit map. */  
2826    
2827      memset(classbits, 0, 32 * sizeof(uschar));    /* In principle, support for Unicode properties should be integrated here as
2828      well. It means re-organizing the above code so as to get hold of the property
2829      values before switching on the op-code. However, I wonder how many patterns
2830      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
2831      these op-codes are never generated.) */
2832    
2833      /* Process characters until ] is reached. By writing this as a "do" it    case OP_DIGIT:
2834      means that an initial ] is taken as a data character. The first pass    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2835      through the regex checked the overall syntax, so we don't need to be very           next == -ESC_h || next == -ESC_v || next == -ESC_R;
     strict here. At the start of the loop, c contains the first byte of the  
     character. */  
2836    
2837      do    case OP_NOT_DIGIT:
2838        {    return next == -ESC_d;
 #ifdef SUPPORT_UTF8  
       if (utf8 && c > 127)  
         {                           /* Braces are required because the */  
         GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */  
         }  
 #endif  
2839    
2840        /* Inside \Q...\E everything is literal except \E */    case OP_WHITESPACE:
2841      return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
2842    
2843        if (inescq)    case OP_NOT_WHITESPACE:
2844          {    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
         if (c == '\\' && ptr[1] == 'E')  
           {  
           inescq = FALSE;  
           ptr++;  
           continue;  
           }  
         else goto LONE_SINGLE_CHARACTER;  
         }  
2845    
2846        /* Handle POSIX class names. Perl allows a negation extension of the    case OP_HSPACE:
2847        form [:^name:]. A square bracket that doesn't match the syntax is    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
2848        treated as a literal. We also recognize the POSIX constructions           next == -ESC_w || next == -ESC_v || next == -ESC_R;
       [.ch.] and [=ch=] ("collating elements") and fault them, as Perl  
       5.6 and 5.8 do. */  
2849    
2850        if (c == '[' &&    case OP_NOT_HSPACE:
2851            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&    return next == -ESC_h;
           check_posix_syntax(ptr, &tempptr, cd))  
         {  
         BOOL local_negate = FALSE;  
         int posix_class, i;  
         register const uschar *cbits = cd->cbits;  
2852    
2853          if (ptr[1] != ':')    /* Can't have \S in here because VT matches \S (Perl anomaly) */
2854            {    case OP_ANYNL:
2855            *errorcodeptr = ERR31;    case OP_VSPACE:
2856            goto FAILED;    return next == -ESC_V || next == -ESC_d || next == -ESC_w;
           }  
2857    
2858          ptr += 2;    case OP_NOT_VSPACE:
2859          if (*ptr == '^')    return next == -ESC_v || next == -ESC_R;
           {  
           local_negate = TRUE;  
           ptr++;  
           }  
2860    
2861          posix_class = check_posix_name(ptr, tempptr - ptr);    case OP_WORDCHAR:
2862          if (posix_class < 0)    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
2863            {           next == -ESC_v || next == -ESC_R;
           *errorcodeptr = ERR30;  
           goto FAILED;  
           }  
2864    
2865          /* If matching is caseless, upper and lower are converted to    case OP_NOT_WORDCHAR:
2866          alpha. This relies on the fact that the class table starts with    return next == -ESC_w || next == -ESC_d;
         alpha, lower, upper as the first 3 entries. */  
2867    
2868          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)    default:
2869            posix_class = 0;    return FALSE;
2870      }
2871    
2872          /* Or into the map we are building up to 3 of the static class  /* Control does not reach here */
2873          tables, or their negations. The [:blank:] class sets up the same  }
         chars as the [:space:] class (all white space). We remove the vertical  
         white space chars afterwards. */  
2874    
         posix_class *= 3;  
         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;  
             }  
           }  
2875    
         ptr = tempptr + 1;  
         class_charcount = 10;  /* Set > 1; assumes more than 1 per class */  
         continue;    /* End of POSIX syntax handling */  
         }  
2876    
2877        /* Backslash may introduce a single character, or it may introduce one  /*************************************************
2878        of the specials, which just set a flag. Escaped items are checked for  *           Compile one branch                   *
2879        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. */  
2880    
2881        if (c == '\\')  /* Scan the pattern, compiling it into the a vector. If the options are
2882          {  changed during the branch, the pointer is used to change the external options
2883          c = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);  bits. This function is used during the pre-compile phase when we are trying
2884    to find out the amount of memory needed, as well as during the real compile
2885    phase. The value of lengthptr distinguishes the two phases.
2886    
2887          if (-c == ESC_b) c = '\b';       /* \b is backslash in a class */  Arguments:
2888          else if (-c == ESC_X) c = 'X';   /* \X is literal X in a class */    optionsptr     pointer to the option bits
2889          else if (-c == ESC_Q)            /* Handle start of quoted string */    codeptr        points to the pointer to the current code point
2890            {    ptrptr         points to the current pattern pointer
2891            if (ptr[1] == '\\' && ptr[2] == 'E')    errorcodeptr   points to error code variable
2892              {    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
2893              ptr += 2; /* avoid empty string */    reqbyteptr     set to the last literal character required, else < 0
2894              }    bcptr          points to current branch chain
2895            else inescq = TRUE;    cd             contains pointers to tables etc.
2896            continue;    lengthptr      NULL during the real compile phase
2897            }                   points to length accumulator during pre-compile phase
2898    
2899          if (c < 0)  Returns:         TRUE on success
2900            {                   FALSE, with *errorcodeptr set non-zero on error
2901            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;  
2902    
2903              case ESC_D:  static BOOL
2904              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
2905              continue;    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
2906      compile_data *cd, int *lengthptr)
2907    {
2908    int repeat_type, op_type;
2909    int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
2910    int bravalue = 0;
2911    int greedy_default, greedy_non_default;
2912    int firstbyte, reqbyte;
2913    int zeroreqbyte, zerofirstbyte;
2914    int req_caseopt, reqvary, tempreqvary;
2915    int options = *optionsptr;
2916    int after_manual_callout = 0;
2917    int length_prevgroup = 0;
2918    register int c;
2919    register uschar *code = *codeptr;
2920    uschar *last_code = code;
2921    uschar *orig_code = code;
2922    uschar *tempcode;
2923    BOOL inescq = FALSE;
2924    BOOL groupsetfirstbyte = FALSE;
2925    const uschar *ptr = *ptrptr;
2926    const uschar *tempptr;
2927    const uschar *nestptr = NULL;
2928    uschar *previous = NULL;
2929    uschar *previous_callout = NULL;
2930    uschar *save_hwm = NULL;
2931    uschar classbits[32];
2932    
2933              case ESC_w:  #ifdef SUPPORT_UTF8
2934              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_word];  BOOL class_utf8;
2935              continue;  BOOL utf8 = (options & PCRE_UTF8) != 0;
2936    uschar *class_utf8data;
2937    uschar *class_utf8data_base;
2938    uschar utf8_char[6];
2939    #else
2940    BOOL utf8 = FALSE;
2941    uschar *utf8_char = NULL;
2942    #endif
2943    
2944              case ESC_W:  #ifdef PCRE_DEBUG
2945              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2946              continue;  #endif
2947    
2948              case ESC_s:  /* Set up the default and non-default settings for greediness */
             for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];  
             classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */  
             continue;  
2949    
2950              case ESC_S:  greedy_default = ((options & PCRE_UNGREEDY) != 0);
2951              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];  greedy_non_default = greedy_default ^ 1;
             classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */  
             continue;  
2952    
2953  #ifdef SUPPORT_UCP  /* Initialize no first byte, no required byte. REQ_UNSET means "no char
2954              case ESC_p:  matching encountered yet". It gets changed to REQ_NONE if we hit something that
2955              case ESC_P:  matches a non-fixed char first char; reqbyte just remains unset if we never
2956                {  find one.
2957    
2958    When we hit a repeat whose minimum is zero, we may have to adjust these values
2959    to take the zero repeat into account. This is implemented by setting them to
2960    zerofirstbyte and zeroreqbyte when such a repeat is encountered. The individual
2961    item types that can be repeated set these backoff variables appropriately. */
2962    
2963    firstbyte = reqbyte = zerofirstbyte = zeroreqbyte = REQ_UNSET;
2964    
2965    /* The variable req_caseopt contains either the REQ_CASELESS value or zero,
2966    according to the current setting of the caseless flag. REQ_CASELESS is a bit
2967    value > 255. It is added into the firstbyte or reqbyte variables to record the
2968    case status of the value. This is used only for ASCII characters. */
2969    
2970    req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
2971    
2972    /* Switch on next character until the end of the branch */
2973    
2974    for (;; ptr++)
2975      {
2976      BOOL negate_class;
2977      BOOL should_flip_negation;
2978      BOOL possessive_quantifier;
2979      BOOL is_quantifier;
2980      BOOL is_recurse;
2981      BOOL reset_bracount;
2982      int class_charcount;
2983      int class_lastchar;
2984      int newoptions;
2985      int recno;
2986      int refsign;
2987      int skipbytes;
2988      int subreqbyte;
2989      int subfirstbyte;
2990      int terminator;
2991      int mclength;
2992      uschar mcbuffer[8];
2993    
2994      /* Get next byte in the pattern */
2995    
2996      c = *ptr;
2997    
2998      /* If we are at the end of a nested substitution, revert to the outer level
2999      string. Nesting only happens one level deep. */
3000    
3001      if (c == 0 && nestptr != NULL)
3002        {
3003        ptr = nestptr;
3004        nestptr = NULL;
3005        c = *ptr;
3006        }
3007    
3008      /* If we are in the pre-compile phase, accumulate the length used for the
3009      previous cycle of this loop. */
3010    
3011      if (lengthptr != NULL)
3012        {
3013    #ifdef PCRE_DEBUG
3014        if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3015    #endif
3016        if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
3017          {
3018          *errorcodeptr = ERR52;
3019          goto FAILED;
3020          }
3021    
3022        /* There is at least one situation where code goes backwards: this is the
3023        case of a zero quantifier after a class (e.g. [ab]{0}). At compile time,
3024        the class is simply eliminated. However, it is created first, so we have to
3025        allow memory for it. Therefore, don't ever reduce the length at this point.
3026        */
3027    
3028        if (code < last_code) code = last_code;
3029    
3030        /* Paranoid check for integer overflow */
3031    
3032        if (OFLOW_MAX - *lengthptr < code - last_code)
3033          {
3034          *errorcodeptr = ERR20;
3035          goto FAILED;
3036          }
3037    
3038        *lengthptr += (int)(code - last_code);
3039        DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));
3040    
3041        /* If "previous" is set and it is not at the start of the work space, move
3042        it back to there, in order to avoid filling up the work space. Otherwise,
3043        if "previous" is NULL, reset the current code pointer to the start. */
3044    
3045        if (previous != NULL)
3046          {
3047          if (previous > orig_code)
3048            {
3049            memmove(orig_code, previous, code - previous);
3050            code -= previous - orig_code;
3051            previous = orig_code;
3052            }
3053          }
3054        else code = orig_code;
3055    
3056        /* Remember where this code item starts so we can pick up the length
3057        next time round. */
3058    
3059        last_code = code;
3060        }
3061    
3062      /* In the real compile phase, just check the workspace used by the forward
3063      reference list. */
3064    
3065      else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
3066        {
3067        *errorcodeptr = ERR52;
3068        goto FAILED;
3069        }
3070    
3071      /* If in \Q...\E, check for the end; if not, we have a literal */
3072    
3073      if (inescq && c != 0)
3074        {
3075        if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3076          {
3077          inescq = FALSE;
3078          ptr++;
3079          continue;
3080          }
3081        else
3082          {
3083          if (previous_callout != NULL)
3084            {
3085            if (lengthptr == NULL)  /* Don't attempt in pre-compile phase */
3086              complete_callout(previous_callout, ptr, cd);
3087            previous_callout = NULL;
3088            }
3089          if ((options & PCRE_AUTO_CALLOUT) != 0)
3090            {
3091            previous_callout = code;
3092            code = auto_callout(code, ptr, cd);
3093            }
3094          goto NORMAL_CHAR;
3095          }
3096        }
3097    
3098      /* Fill in length of a previous callout, except when the next thing is
3099      a quantifier. */
3100    
3101      is_quantifier =
3102        c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
3103        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
3104    
3105      if (!is_quantifier && previous_callout != NULL &&
3106           after_manual_callout-- <= 0)
3107        {
3108        if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */
3109          complete_callout(previous_callout, ptr, cd);
3110        previous_callout = NULL;
3111        }
3112    
3113      /* In extended mode, skip white space and comments */
3114    
3115      if ((options & PCRE_EXTENDED) != 0)
3116        {
3117        if ((cd->ctypes[c] & ctype_space) != 0) continue;
3118        if (c == CHAR_NUMBER_SIGN)
3119          {
3120          while (*(++ptr) != 0)
3121            {
3122            if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3123            }
3124          if (*ptr != 0) continue;
3125    
3126          /* Else fall through to handle end of string */
3127          c = 0;
3128          }
3129        }
3130    
3131      /* No auto callout for quantifiers. */
3132    
3133      if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier)
3134        {
3135        previous_callout = code;
3136        code = auto_callout(code, ptr, cd);
3137        }
3138    
3139      switch(c)
3140        {
3141        /* ===================================================================*/
3142        case 0:                        /* The branch terminates at string end */
3143        case CHAR_VERTICAL_LINE:       /* or | or ) */
3144        case CHAR_RIGHT_PARENTHESIS:
3145        *firstbyteptr = firstbyte;
3146        *reqbyteptr = reqbyte;
3147        *codeptr = code;
3148        *ptrptr = ptr;
3149        if (lengthptr != NULL)
3150          {
3151          if (OFLOW_MAX - *lengthptr < code - last_code)
3152            {
3153            *errorcodeptr = ERR20;
3154            goto FAILED;
3155            }
3156          *lengthptr += (int)(code - last_code);   /* To include callout length */
3157          DPRINTF((">> end branch\n"));
3158          }
3159        return TRUE;
3160    
3161    
3162        /* ===================================================================*/
3163        /* Handle single-character metacharacters. In multiline mode, ^ disables
3164        the setting of any following char as a first character. */
3165    
3166        case CHAR_CIRCUMFLEX_ACCENT:
3167        if ((options & PCRE_MULTILINE) != 0)
3168          {
3169          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3170          }
3171        previous = NULL;
3172        *code++ = OP_CIRC;
3173        break;
3174    
3175        case CHAR_DOLLAR_SIGN:
3176        previous = NULL;
3177        *code++ = OP_DOLL;
3178        break;
3179    
3180        /* There can never be a first char if '.' is first, whatever happens about
3181        repeats. The value of reqbyte doesn't change either. */
3182    
3183        case CHAR_DOT:
3184        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3185        zerofirstbyte = firstbyte;
3186        zeroreqbyte = reqbyte;
3187        previous = code;
3188        *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
3189        break;
3190    
3191    
3192        /* ===================================================================*/
3193        /* Character classes. If the included characters are all < 256, we build a
3194        32-byte bitmap of the permitted characters, except in the special case
3195        where there is only one such character. For negated classes, we build the
3196        map as usual, then invert it at the end. However, we use a different opcode
3197        so that data characters > 255 can be handled correctly.
3198    
3199        If the class contains characters outside the 0-255 range, a different
3200        opcode is compiled. It may optionally have a bit map for characters < 256,
3201        but those above are are explicitly listed afterwards. A flag byte tells
3202        whether the bitmap is present, and whether this is a negated class or not.
3203    
3204        In JavaScript compatibility mode, an isolated ']' causes an error. In
3205        default (Perl) mode, it is treated as a data character. */
3206    
3207        case CHAR_RIGHT_SQUARE_BRACKET:
3208        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3209          {
3210          *errorcodeptr = ERR64;
3211          goto FAILED;
3212          }
3213        goto NORMAL_CHAR;
3214    
3215        case CHAR_LEFT_SQUARE_BRACKET:
3216        previous = code;
3217    
3218        /* PCRE supports POSIX class stuff inside a class. Perl gives an error if
3219        they are encountered at the top level, so we'll do that too. */
3220    
3221        if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3222             ptr[1] == CHAR_EQUALS_SIGN) &&
3223            check_posix_syntax(ptr, &tempptr))
3224          {
3225          *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
3226          goto FAILED;
3227          }
3228    
3229        /* If the first character is '^', set the negation flag and skip it. Also,
3230        if the first few characters (either before or after ^) are \Q\E or \E we
3231        skip them too. This makes for compatibility with Perl. */
3232    
3233        negate_class = FALSE;
3234        for (;;)
3235          {
3236          c = *(++ptr);
3237          if (c == CHAR_BACKSLASH)
3238            {
3239            if (ptr[1] == CHAR_E)
3240              ptr++;
3241            else if (strncmp((const char *)ptr+1,
3242                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
3243              ptr += 3;
3244            else
3245              break;
3246            }
3247          else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
3248            negate_class = TRUE;
3249          else break;
3250          }
3251    
3252        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
3253        an initial ']' is taken as a data character -- the code below handles
3254        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3255        [^] must match any character, so generate OP_ALLANY. */
3256    
3257        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3258            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3259          {
3260          *code++ = negate_class? OP_ALLANY : OP_FAIL;
3261          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3262          zerofirstbyte = firstbyte;
3263          break;
3264          }
3265    
3266        /* If a class contains a negative special such as \S, we need to flip the
3267        negation flag at the end, so that support for characters > 255 works
3268        correctly (they are all included in the class). */
3269    
3270        should_flip_negation = FALSE;
3271    
3272        /* Keep a count of chars with values < 256 so that we can optimize the case
3273        of just a single character (as long as it's < 256). However, For higher
3274        valued UTF-8 characters, we don't yet do any optimization. */
3275    
3276        class_charcount = 0;
3277        class_lastchar = -1;
3278    
3279        /* Initialize the 32-char bit map to all zeros. We build the map in a
3280        temporary bit of memory, in case the class contains only 1 character (less
3281        than 256), because in that case the compiled code doesn't use the bit map.
3282        */
3283    
3284        memset(classbits, 0, 32 * sizeof(uschar));
3285    
3286    #ifdef SUPPORT_UTF8
3287        class_utf8 = FALSE;                       /* No chars >= 256 */
3288        class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
3289        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
3290    #endif
3291    
3292        /* Process characters until ] is reached. By writing this as a "do" it
3293        means that an initial ] is taken as a data character. At the start of the
3294        loop, c contains the first byte of the character. */
3295    
3296        if (c != 0) do
3297          {
3298          const uschar *oldptr;
3299    
3300    #ifdef SUPPORT_UTF8
3301          if (utf8 && c > 127)
3302            {                           /* Braces are required because the */
3303            GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3304            }
3305    
3306          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
3307          data and reset the pointer. This is so that very large classes that
3308          contain a zillion UTF-8 characters no longer overwrite the work space
3309          (which is on the stack). */
3310    
3311          if (lengthptr != NULL)
3312            {
3313            *lengthptr += class_utf8data - class_utf8data_base;
3314            class_utf8data = class_utf8data_base;
3315            }
3316    
3317    #endif
3318    
3319          /* Inside \Q...\E everything is literal except \E */
3320    
3321          if (inescq)
3322            {
3323            if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3324              {
3325              inescq = FALSE;                   /* Reset literal state */
3326              ptr++;                            /* Skip the 'E' */
3327              continue;                         /* Carry on with next */
3328              }
3329            goto CHECK_RANGE;                   /* Could be range if \E follows */
3330            }
3331    
3332          /* Handle POSIX class names. Perl allows a negation extension of the
3333          form [:^name:]. A square bracket that doesn't match the syntax is
3334          treated as a literal. We also recognize the POSIX constructions
3335          [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3336          5.6 and 5.8 do. */
3337    
3338          if (c == CHAR_LEFT_SQUARE_BRACKET &&
3339              (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3340               ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3341            {
3342            BOOL local_negate = FALSE;
3343            int posix_class, taboffset, tabopt;
3344            register const uschar *cbits = cd->cbits;
3345            uschar pbits[32];
3346    
3347            if (ptr[1] != CHAR_COLON)
3348              {
3349              *errorcodeptr = ERR31;
3350              goto FAILED;
3351              }
3352    
3353            ptr += 2;
3354            if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3355              {
3356              local_negate = TRUE;
3357              should_flip_negation = TRUE;  /* Note negative special */
3358              ptr++;
3359              }
3360    
3361            posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3362            if (posix_class < 0)
3363              {
3364              *errorcodeptr = ERR30;
3365              goto FAILED;
3366              }
3367    
3368            /* If matching is caseless, upper and lower are converted to
3369            alpha. This relies on the fact that the class table starts with
3370            alpha, lower, upper as the first 3 entries. */
3371    
3372            if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3373              posix_class = 0;
3374    
3375            /* When PCRE_UCP is set, some of the POSIX classes are converted to
3376            different escape sequences that use Unicode properties. */
3377    
3378    #ifdef SUPPORT_UCP
3379            if ((options & PCRE_UCP) != 0)
3380              {
3381              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3382              if (posix_substitutes[pc] != NULL)
3383                {
3384                nestptr = tempptr + 1;
3385                ptr = posix_substitutes[pc] - 1;
3386                continue;
3387                }
3388              }
3389    #endif
3390            /* In the non-UCP case, we build the bit map for the POSIX class in a
3391            chunk of local store because we may be adding and subtracting from it,
3392            and we don't want to subtract bits that may be in the main map already.
3393            At the end we or the result into the bit map that is being built. */
3394    
3395            posix_class *= 3;
3396    
3397            /* Copy in the first table (always present) */
3398    
3399            memcpy(pbits, cbits + posix_class_maps[posix_class],
3400              32 * sizeof(uschar));
3401    
3402            /* If there is a second table, add or remove it as required. */
3403    
3404            taboffset = posix_class_maps[posix_class + 1];
3405            tabopt = posix_class_maps[posix_class + 2];
3406    
3407            if (taboffset >= 0)
3408              {
3409              if (tabopt >= 0)
3410                for (c = 0; c < 32; c++) pbits[c] |= cbits[c + taboffset];
3411              else
3412                for (c = 0; c < 32; c++) pbits[c] &= ~cbits[c + taboffset];
3413              }
3414    
3415            /* Not see if we need to remove any special characters. An option
3416            value of 1 removes vertical space and 2 removes underscore. */
3417    
3418            if (tabopt < 0) tabopt = -tabopt;
3419            if (tabopt == 1) pbits[1] &= ~0x3c;
3420              else if (tabopt == 2) pbits[11] &= 0x7f;
3421    
3422            /* Add the POSIX table or its complement into the main table that is
3423            being built and we are done. */
3424    
3425            if (local_negate)
3426              for (c = 0; c < 32; c++) classbits[c] |= ~pbits[c];
3427            else
3428              for (c = 0; c < 32; c++) classbits[c] |= pbits[c];
3429    
3430            ptr = tempptr + 1;
3431            class_charcount = 10;  /* Set > 1; assumes more than 1 per class */
3432            continue;    /* End of POSIX syntax handling */
3433            }
3434    
3435          /* Backslash may introduce a single character, or it may introduce one
3436          of the specials, which just set a flag. The sequence \b is a special
3437          case. Inside a class (and only there) it is treated as backspace. We
3438          assume that other escapes have more than one character in them, so set
3439          class_charcount bigger than one. Unrecognized escapes fall through and
3440          are either treated as literal characters (by default), or are faulted if
3441          PCRE_EXTRA is set. */
3442    
3443          if (c == CHAR_BACKSLASH)
3444            {
3445            c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3446            if (*errorcodeptr != 0) goto FAILED;
3447    
3448            if (-c == ESC_b) c = CHAR_BS;    /* \b is backspace in a class */
3449            else if (-c == ESC_Q)            /* Handle start of quoted string */
3450              {
3451              if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3452                {
3453                ptr += 2; /* avoid empty string */
3454                }
3455              else inescq = TRUE;
3456              continue;
3457              }
3458            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
3459    
3460            if (c < 0)
3461              {
3462              register const uschar *cbits = cd->cbits;
3463              class_charcount += 2;     /* Greater than 1 is what matters */
3464    
3465              switch (-c)
3466                {
3467    #ifdef SUPPORT_UCP
3468                case ESC_du:     /* These are the values given for \d etc */
3469                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3470                case ESC_wu:     /* escape sequence with an appropriate \p */
3471                case ESC_WU:     /* or \P to test Unicode properties instead */
3472                case ESC_su:     /* of the default ASCII testing. */
3473                case ESC_SU:
3474                nestptr = ptr;
3475                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3476                class_charcount -= 2;                /* Undo! */
3477                continue;
3478    #endif
3479                case ESC_d:
3480                for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3481                continue;
3482    
3483                case ESC_D:
3484                should_flip_negation = TRUE;
3485                for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
3486                continue;
3487    
3488                case ESC_w:
3489                for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_word];
3490                continue;
3491    
3492                case ESC_W:
3493                should_flip_negation = TRUE;
3494                for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3495                continue;
3496    
3497                case ESC_s:
3498                for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3499                classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */
3500                continue;
3501    
3502                case ESC_S:
3503                should_flip_negation = TRUE;
3504                for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
3505                classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3506                continue;
3507    
3508                case ESC_h:
3509                SETBIT(classbits, 0x09); /* VT */
3510                SETBIT(classbits, 0x20); /* SPACE */
3511                SETBIT(classbits, 0xa0); /* NSBP */
3512    #ifdef SUPPORT_UTF8
3513                if (utf8)
3514                  {
3515                  class_utf8 = TRUE;
3516                  *class_utf8data++ = XCL_SINGLE;
3517                  class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);
3518                  *class_utf8data++ = XCL_SINGLE;
3519                  class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);
3520                  *class_utf8data++ = XCL_RANGE;
3521                  class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);
3522                  class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);
3523                  *class_utf8data++ = XCL_SINGLE;
3524                  class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);
3525                  *class_utf8data++ = XCL_SINGLE;
3526                  class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);
3527                  *class_utf8data++ = XCL_SINGLE;
3528                  class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);
3529                  }
3530    #endif
3531                continue;
3532    
3533                case ESC_H:
3534                for (c = 0; c < 32; c++)
3535                  {
3536                  int x = 0xff;
3537                  switch (c)
3538                    {
3539                    case 0x09/8: x ^= 1 << (0x09%8); break;
3540                    case 0x20/8: x ^= 1 << (0x20%8); break;
3541                    case 0xa0/8: x ^= 1 << (0xa0%8); break;
3542                    default: break;
3543                    }
3544                  classbits[c] |= x;
3545                  }
3546    
3547    #ifdef SUPPORT_UTF8
3548                if (utf8)
3549                  {
3550                  class_utf8 = TRUE;
3551                  *class_utf8data++ = XCL_RANGE;
3552                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3553                  class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);
3554                  *class_utf8data++ = XCL_RANGE;
3555                  class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);
3556                  class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);
3557                  *class_utf8data++ = XCL_RANGE;
3558                  class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);
3559                  class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);
3560                  *class_utf8data++ = XCL_RANGE;
3561                  class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);
3562                  class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);
3563                  *class_utf8data++ = XCL_RANGE;
3564                  class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);
3565                  class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);
3566                  *class_utf8data++ = XCL_RANGE;
3567                  class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);
3568                  class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);
3569                  *class_utf8data++ = XCL_RANGE;
3570                  class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);
3571                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3572                  }
3573    #endif
3574                continue;
3575    
3576                case ESC_v:
3577                SETBIT(classbits, 0x0a); /* LF */
3578                SETBIT(classbits, 0x0b); /* VT */
3579                SETBIT(classbits, 0x0c); /* FF */
3580                SETBIT(classbits, 0x0d); /* CR */
3581                SETBIT(classbits, 0x85); /* NEL */
3582    #ifdef SUPPORT_UTF8
3583                if (utf8)
3584                  {
3585                  class_utf8 = TRUE;
3586                  *class_utf8data++ = XCL_RANGE;
3587                  class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);
3588                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3589                  }
3590    #endif
3591                continue;
3592    
3593                case ESC_V:
3594                for (c = 0; c < 32; c++)
3595                  {
3596                  int x = 0xff;
3597                  switch (c)
3598                    {
3599                    case 0x0a/8: x ^= 1 << (0x0a%8);
3600                                 x ^= 1 << (0x0b%8);
3601                                 x ^= 1 << (0x0c%8);
3602                                 x ^= 1 << (0x0d%8);
3603                                 break;
3604                    case 0x85/8: x ^= 1 << (0x85%8); break;
3605                    default: break;
3606                    }
3607                  classbits[c] |= x;
3608                  }
3609    
3610    #ifdef SUPPORT_UTF8
3611                if (utf8)
3612                  {
3613                  class_utf8 = TRUE;
3614                  *class_utf8data++ = XCL_RANGE;
3615                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3616                  class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);
3617                  *class_utf8data++ = XCL_RANGE;
3618                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3619                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3620                  }
3621    #endif
3622                continue;
3623    
3624    #ifdef SUPPORT_UCP
3625                case ESC_p:
3626                case ESC_P:
3627                  {
3628                BOOL negated;                BOOL negated;
3629                int property = get_ucp(&ptr, &negated, errorcodeptr);                int pdata;
3630                if (property < 0) goto FAILED;                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3631                  if (ptype < 0) goto FAILED;
3632                class_utf8 = TRUE;                class_utf8 = TRUE;
3633                *class_utf8data++ = ((-c == ESC_p) != negated)?                *class_utf8data++ = ((-c == ESC_p) != negated)?
3634                  XCL_PROP : XCL_NOTPROP;                  XCL_PROP : XCL_NOTPROP;
3635                *class_utf8data++ = property;                *class_utf8data++ = ptype;
3636                  *class_utf8data++ = pdata;
3637                class_charcount -= 2;   /* Not a < 256 character */                class_charcount -= 2;   /* Not a < 256 character */
3638                  continue;
3639                }                }
             continue;  
3640  #endif  #endif
   
3641              /* Unrecognized escapes are faulted if PCRE is running in its              /* Unrecognized escapes are faulted if PCRE is running in its
3642              strict mode. By default, for compatibility with Perl, they are              strict mode. By default, for compatibility with Perl, they are
3643              treated as literals. */              treated as literals. */
# Line 1949  for (;; ptr++) Line 3648  for (;; ptr++)
3648                *errorcodeptr = ERR7;                *errorcodeptr = ERR7;
3649                goto FAILED;                goto FAILED;
3650                }                }
             c = *ptr;              /* The final character */  
3651              class_charcount -= 2;  /* Undo the default count from above */              class_charcount -= 2;  /* Undo the default count from above */
3652                c = *ptr;              /* Get the final character and fall through */
3653                break;
3654              }              }
3655            }            }
3656    
3657          /* Fall through if we have a single character (c >= 0). This may be          /* Fall through if we have a single character (c >= 0). This may be
3658          > 256 in UTF-8 mode. */          greater than 256 in UTF-8 mode. */
3659    
3660          }   /* End of backslash handling */          }   /* End of backslash handling */
3661    
3662        /* A single character may be followed by '-' to form a range. However,        /* A single character may be followed by '-' to form a range. However,
3663        Perl does not permit ']' to be the end of the range. A '-' character        Perl does not permit ']' to be the end of the range. A '-' character
3664        here is treated as a literal. */        at the end is treated as a literal. Perl ignores orphaned \E sequences
3665          entirely. The code for handling \Q and \E is messy. */
3666    
3667        if (ptr[1] == '-' && ptr[2] != ']')        CHECK_RANGE:
3668          while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3669            {
3670            inescq = FALSE;
3671            ptr += 2;
3672            }
3673    
3674          oldptr = ptr;
3675    
3676          /* Remember \r or \n */
3677    
3678          if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3679    
3680          /* Check for range */
3681    
3682          if (!inescq && ptr[1] == CHAR_MINUS)
3683          {          {
3684          int d;          int d;
3685          ptr += 2;          ptr += 2;
3686            while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3687    
3688            /* If we hit \Q (not followed by \E) at this point, go into escaped
3689            mode. */
3690    
3691            while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3692              {
3693              ptr += 2;
3694              if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3695                { ptr += 2; continue; }
3696              inescq = TRUE;
3697              break;
3698              }
3699    
3700            if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3701              {
3702              ptr = oldptr;
3703              goto LONE_SINGLE_CHARACTER;
3704              }
3705    
3706  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3707          if (utf8)          if (utf8)
# Line 1981  for (;; ptr++) Line 3716  for (;; ptr++)
3716          not any of the other escapes. Perl 5.6 treats a hyphen as a literal          not any of the other escapes. Perl 5.6 treats a hyphen as a literal
3717          in such circumstances. */          in such circumstances. */
3718    
3719          if (d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3720            {            {
3721            const uschar *oldptr = ptr;            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3722            d = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);            if (*errorcodeptr != 0) goto FAILED;
3723    
3724            /* \b is backslash; \X is literal X; any other special means the '-'            /* \b is backspace; any other special means the '-' was literal */
           was literal */  
3725    
3726            if (d < 0)            if (d < 0)
3727              {              {
3728              if (d == -ESC_b) d = '\b';              if (d == -ESC_b) d = CHAR_BS; else
             else if (d == -ESC_X) d = 'X'; else  
3729                {                {
3730                ptr = oldptr -