/[pcre]/code/trunk/pcre_compile.c
ViewVC logotype

Diff of /code/trunk/pcre_compile.c

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

revision 81 by nigel, Sat Feb 24 21:40:59 2007 UTC revision 602 by ph10, Wed May 25 08:29:03 2011 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) 1997-2011 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      "\\c must be followed by an ASCII character\0"
412      ;
413    
414  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
415  patterns. Note that the tables in chartables are dependent on the locale, and  patterns. Note that the tables in chartables are dependent on the locale, and
# Line 220  For convenience, we use the same bit def Line 427  For convenience, we use the same bit def
427    
428  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
429    
430  #if !EBCDIC    /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC
431    
432    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
433    UTF-8 mode. */
434    
435  static const unsigned char digitab[] =  static const unsigned char digitab[] =
436    {    {
437    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
# Line 256  static const unsigned char digitab[] = Line 467  static const unsigned char digitab[] =
467    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
468    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
469    
470  #else          /* This is the "abnormal" case, for EBCDIC systems */  #else
471    
472    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
473    
474  static const unsigned char digitab[] =  static const unsigned char digitab[] =
475    {    {
476    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
# Line 270  static const unsigned char digitab[] = Line 484  static const unsigned char digitab[] =
484    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */
485    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */
486    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */
487    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88-     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88- 95    */
488    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */
489    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */
490    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */
# Line 304  static const unsigned char ebcdic_charta Line 518  static const unsigned char ebcdic_charta
518    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */
519    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */
520    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */
521    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88-  */    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88- 95 */
522    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */
523    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */
524    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */
# Line 331  static const unsigned char ebcdic_charta Line 545  static const unsigned char ebcdic_charta
545  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
546    
547  static BOOL  static BOOL
548    compile_regex(int, int, int *, uschar **, const uschar **, int *, BOOL, int,    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,
549      int *, int *, branch_chain *, compile_data *);      int *, int *, branch_chain *, compile_data *, int *);
550    
551    
552    
553    /*************************************************
554    *            Find an error text                  *
555    *************************************************/
556    
557    /* The error texts are now all in one long string, to save on relocations. As
558    some of the text is of unknown length, we can't use a table of offsets.
559    Instead, just count through the strings. This is not a performance issue
560    because it happens only when there has been a compilation error.
561    
562    Argument:   the error number
563    Returns:    pointer to the error string
564    */
565    
566    static const char *
567    find_error_text(int n)
568    {
569    const char *s = error_texts;
570    for (; n > 0; n--)
571      {
572      while (*s++ != 0) {};
573      if (*s == 0) return "Error text not found (please report)";
574      }
575    return s;
576    }
577    
578    
579  /*************************************************  /*************************************************
# Line 342  static BOOL Line 582  static BOOL
582    
583  /* This function is called when a \ has been encountered. It either returns a  /* This function is called when a \ has been encountered. It either returns a
584  positive value for a simple escape such as \n, or a negative value which  positive value for a simple escape such as \n, or a negative value which
585  encodes one of the more complicated things such as \d. When UTF-8 is enabled,  encodes one of the more complicated things such as \d. A backreference to group
586  a positive value greater than 255 may be returned. On entry, ptr is pointing at  n is returned as -(ESC_REF + n); ESC_REF is the highest ESC_xxx macro. When
587  the \. On exit, it is on the final character of the escape sequence.  UTF-8 is enabled, a positive value greater than 255 may be returned. On entry,
588    ptr is pointing at the \. On exit, it is on the final character of the escape
589    sequence.
590    
591  Arguments:  Arguments:
592    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
# Line 355  Arguments: Line 597  Arguments:
597    
598  Returns:         zero or positive => a data character  Returns:         zero or positive => a data character
599                   negative => a special escape sequence                   negative => a special escape sequence
600                   on error, errorptr is set                   on error, errorcodeptr is set
601  */  */
602    
603  static int  static int
604  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,
605    int options, BOOL isclass)    int options, BOOL isclass)
606  {  {
607  const uschar *ptr = *ptrptr;  BOOL utf8 = (options & PCRE_UTF8) != 0;
608    const uschar *ptr = *ptrptr + 1;
609  int c, i;  int c, i;
610    
611    GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
612    ptr--;                            /* Set pointer back to the last byte */
613    
614  /* If backslash is at the end of the pattern, it's an error. */  /* If backslash is at the end of the pattern, it's an error. */
615    
 c = *(++ptr);  
616  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
617    
618  /* Non-alphamerics are literals. For digits or letters, do an initial lookup in  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
619  a table. A non-zero result is something that can be returned immediately.  in a table. A non-zero result is something that can be returned immediately.
620  Otherwise further processing may be required. */  Otherwise further processing may be required. */
621    
622  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
623  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */
624  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0) c = i;
625    
626  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
627  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */
628  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
629  #endif  #endif
630    
# Line 388  else if ((i = escapes[c - 0x48]) != 0) Line 633  else if ((i = escapes[c - 0x48]) != 0)
633  else  else
634    {    {
635    const uschar *oldptr;    const uschar *oldptr;
636      BOOL braced, negated;
637    
638    switch (c)    switch (c)
639      {      {
640      /* A number of Perl escapes are not handled by PCRE. We give an explicit      /* A number of Perl escapes are not handled by PCRE. We give an explicit
641      error. */      error. */
642    
643      case 'l':      case CHAR_l:
644      case 'L':      case CHAR_L:
645      case 'N':      case CHAR_u:
646      case 'u':      case CHAR_U:
     case 'U':  
647      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
648      break;      break;
649    
650        /* \g must be followed by one of a number of specific things:
651    
652        (1) A number, either plain or braced. If positive, it is an absolute
653        backreference. If negative, it is a relative backreference. This is a Perl
654        5.10 feature.
655    
656        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
657        is part of Perl's movement towards a unified syntax for back references. As
658        this is synonymous with \k{name}, we fudge it up by pretending it really
659        was \k.
660    
661        (3) For Oniguruma compatibility we also support \g followed by a name or a
662        number either in angle brackets or in single quotes. However, these are
663        (possibly recursive) subroutine calls, _not_ backreferences. Just return
664        the -ESC_g code (cf \k). */
665    
666        case CHAR_g:
667        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
668          {
669          c = -ESC_g;
670          break;
671          }
672    
673        /* Handle the Perl-compatible cases */
674    
675        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
676          {
677          const uschar *p;
678          for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
679            if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
680          if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
681            {
682            c = -ESC_k;
683            break;
684            }
685          braced = TRUE;
686          ptr++;
687          }
688        else braced = FALSE;
689    
690        if (ptr[1] == CHAR_MINUS)
691          {
692          negated = TRUE;
693          ptr++;
694          }
695        else negated = FALSE;
696    
697        c = 0;
698        while ((digitab[ptr[1]] & ctype_digit) != 0)
699          c = c * 10 + *(++ptr) - CHAR_0;
700    
701        if (c < 0)   /* Integer overflow */
702          {
703          *errorcodeptr = ERR61;
704          break;
705          }
706    
707        if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
708          {
709          *errorcodeptr = ERR57;
710          break;
711          }
712    
713        if (c == 0)
714          {
715          *errorcodeptr = ERR58;
716          break;
717          }
718    
719        if (negated)
720          {
721          if (c > bracount)
722            {
723            *errorcodeptr = ERR15;
724            break;
725            }
726          c = bracount - (c - 1);
727          }
728    
729        c = -(ESC_REF + c);
730        break;
731    
732      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
733      starting with one that is not zero is not straightforward. By experiment,      starting with one that is not zero is not straightforward. By experiment,
734      the way Perl works seems to be as follows:      the way Perl works seems to be as follows:
# Line 413  else Line 741  else
741      value is greater than 377, the least significant 8 bits are taken. Inside a      value is greater than 377, the least significant 8 bits are taken. Inside a
742      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
743    
744      case '1': case '2': case '3': case '4': case '5':      case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:
745      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
746    
747      if (!isclass)      if (!isclass)
748        {        {
749        oldptr = ptr;        oldptr = ptr;
750        c -= '0';        c -= CHAR_0;
751        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
752          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - CHAR_0;
753          if (c < 0)    /* Integer overflow */
754            {
755            *errorcodeptr = ERR61;
756            break;
757            }
758        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
759          {          {
760          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 434  else Line 767  else
767      generates a binary zero byte and treats the digit as a following literal.      generates a binary zero byte and treats the digit as a following literal.
768      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
769    
770      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
771        {        {
772        ptr--;        ptr--;
773        c = 0;        c = 0;
# Line 442  else Line 775  else
775        }        }
776    
777      /* \0 always starts an octal number, but we may drop through to here with a      /* \0 always starts an octal number, but we may drop through to here with a
778      larger first octal digit. */      larger first octal digit. The original code used just to take the least
779        significant 8 bits of octal numbers (I think this is what early Perls used
780      case '0':      to do). Nowadays we allow for larger numbers in UTF-8 mode, but no more
781      c -= '0';      than 3 octal digits. */
782      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')  
783          c = c * 8 + *(++ptr) - '0';      case CHAR_0:
784      c &= 255;     /* Take least significant 8 bits */      c -= CHAR_0;
785        while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
786            c = c * 8 + *(++ptr) - CHAR_0;
787        if (!utf8 && c > 255) *errorcodeptr = ERR51;
788      break;      break;
789    
790      /* \x is complicated when UTF-8 is enabled. \x{ddd} is a character number      /* \x is complicated. \x{ddd} is a character number which can be greater
791      which can be greater than 0xff, but only if the ddd are hex digits. */      than 0xff in utf8 mode, but only if the ddd are hex digits. If not, { is
792        treated as a data character. */
793    
794      case 'x':      case CHAR_x:
795  #ifdef SUPPORT_UTF8      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
     if (ptr[1] == '{' && (options & PCRE_UTF8) != 0)  
796        {        {
797        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
798        register int count = 0;        int count = 0;
799    
800        c = 0;        c = 0;
801        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
802          {          {
803          int cc = *pt++;          register int cc = *pt++;
804            if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
805          count++;          count++;
806  #if !EBCDIC    /* ASCII coding */  
807          if (cc >= 'a') cc -= 32;               /* Convert to upper case */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
808          c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
809  #else          /* EBCDIC coding */          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
810          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */  #else           /* EBCDIC coding */
811          c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
812            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
813  #endif  #endif
814          }          }
815        if (*pt == '}')  
816          if (*pt == CHAR_RIGHT_CURLY_BRACKET)
817          {          {
818          if (c < 0 || count > 8) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
819          ptr = pt;          ptr = pt;
820          break;          break;
821          }          }
822    
823        /* If the sequence of hex digits does not end with '}', then we don't        /* If the sequence of hex digits does not end with '}', then we don't
824        recognize this construct; fall through to the normal \x handling. */        recognize this construct; fall through to the normal \x handling. */
825        }        }
 #endif  
826    
827      /* Read just a single hex char */      /* Read just a single-byte hex-defined char */
828    
829      c = 0;      c = 0;
830      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
831        {        {
832        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
833        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
834  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
835        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
836        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
837  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
838        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
839        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
840  #endif  #endif
841        }        }
842      break;      break;
843    
844      /* Other special escapes not starting with a digit are straightforward */      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
845        An error is given if the byte following \c is not an ASCII character. This
846        coding is ASCII-specific, but then the whole concept of \cx is
847        ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
848    
849      case 'c':      case CHAR_c:
850      c = *(++ptr);      c = *(++ptr);
851      if (c == 0)      if (c == 0)
852        {        {
853        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
854        return 0;        break;
855        }        }
856    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
857      /* A letter is upper-cased; then the 0x40 bit is flipped. This coding      if (c > 127)  /* Excludes all non-ASCII in either mode */
858      is ASCII-specific, but then the whole concept of \cx is ASCII-specific.        {
859      (However, an EBCDIC equivalent has now been added.) */        *errorcodeptr = ERR68;
860          break;
861  #if !EBCDIC    /* ASCII coding */        }
862      if (c >= 'a' && c <= 'z') c -= 32;      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
863      c ^= 0x40;      c ^= 0x40;
864  #else          /* EBCDIC coding */  #else             /* EBCDIC coding */
865      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
866      c ^= 0xC0;      c ^= 0xC0;
867  #endif  #endif
868      break;      break;
869    
870      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
871      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
872      for Perl compatibility, it is a literal. This code looks a bit odd, but      otherwise, for Perl compatibility, it is a literal. This code looks a bit
873      there used to be some cases other than the default, and there may be again      odd, but there used to be some cases other than the default, and there may
874      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
875    
876      default:      default:
877      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 541  else Line 884  else
884      }      }
885    }    }
886    
887    /* Perl supports \N{name} for character names, as well as plain \N for "not
888    newline". PCRE does not support \N{name}. */
889    
890    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET)
891      *errorcodeptr = ERR37;
892    
893    /* If PCRE_UCP is set, we change the values for \d etc. */
894    
895    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
896      c -= (ESC_DU - ESC_D);
897    
898    /* Set the pointer to the final character before returning. */
899    
900  *ptrptr = ptr;  *ptrptr = ptr;
901  return c;  return c;
902  }  }
# Line 560  escape sequence. Line 916  escape sequence.
916  Argument:  Argument:
917    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
918    negptr         points to a boolean that is set TRUE for negation else FALSE    negptr         points to a boolean that is set TRUE for negation else FALSE
919      dptr           points to an int that is set to the detailed property value
920    errorcodeptr   points to the error code variable    errorcodeptr   points to the error code variable
921    
922  Returns:     value from ucp_type_table, or -1 for an invalid type  Returns:         type value from ucp_type_table, or -1 for an invalid type
923  */  */
924    
925  static int  static int
926  get_ucp(const uschar **ptrptr, BOOL *negptr, int *errorcodeptr)  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)
927  {  {
928  int c, i, bot, top;  int c, i, bot, top;
929  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
930  char name[4];  char name[32];
931    
932  c = *(++ptr);  c = *(++ptr);
933  if (c == 0) goto ERROR_RETURN;  if (c == 0) goto ERROR_RETURN;
934    
935  *negptr = FALSE;  *negptr = FALSE;
936    
937  /* \P or \p can be followed by a one- or two-character name in {}, optionally  /* \P or \p can be followed by a name in {}, optionally preceded by ^ for
938  preceded by ^ for negation. */  negation. */
939    
940  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
941    {    {
942    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
943      {      {
944      *negptr = TRUE;      *negptr = TRUE;
945      ptr++;      ptr++;
946      }      }
947    for (i = 0; i <= 2; i++)    for (i = 0; i < (int)sizeof(name) - 1; i++)
948      {      {
949      c = *(++ptr);      c = *(++ptr);
950      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
951      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
952      name[i] = c;      name[i] = c;
953      }      }
954    if (c !='}')   /* Try to distinguish error cases */    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
     {  
     while (*(++ptr) != 0 && *ptr != '}');  
     if (*ptr == '}') goto UNKNOWN_RETURN; else goto ERROR_RETURN;  
     }  
955    name[i] = 0;    name[i] = 0;
956    }    }
957    
# Line 619  top = _pcre_utt_size; Line 972  top = _pcre_utt_size;
972    
973  while (bot < top)  while (bot < top)
974    {    {
975    i = (bot + top)/2;    i = (bot + top) >> 1;
976    c = strcmp(name, _pcre_utt[i].name);    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);
977    if (c == 0) return _pcre_utt[i].value;    if (c == 0)
978        {
979        *dptr = _pcre_utt[i].value;
980        return _pcre_utt[i].type;
981        }
982    if (c > 0) bot = i + 1; else top = i;    if (c > 0) bot = i + 1; else top = i;
983    }    }
984    
 UNKNOWN_RETURN:  
985  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
986  *ptrptr = ptr;  *ptrptr = ptr;
987  return -1;  return -1;
# Line 660  is_counted_repeat(const uschar *p) Line 1016  is_counted_repeat(const uschar *p)
1016  {  {
1017  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
1018  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
1019  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
1020    
1021  if (*p++ != ',') return FALSE;  if (*p++ != CHAR_COMMA) return FALSE;
1022  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
1023    
1024  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
1025  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
1026    
1027  return (*p == '}');  return (*p == CHAR_RIGHT_CURLY_BRACKET);
1028  }  }
1029    
1030    
# Line 701  int max = -1; Line 1057  int max = -1;
1057  /* Read the minimum value and do a paranoid check: a negative value indicates  /* Read the minimum value and do a paranoid check: a negative value indicates
1058  an integer overflow. */  an integer overflow. */
1059    
1060  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;
1061  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1062    {    {
1063    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 711  if (min < 0 || min > 65535) Line 1067  if (min < 0 || min > 65535)
1067  /* Read the maximum value if there is one, and again do a paranoid on its size.  /* Read the maximum value if there is one, and again do a paranoid on its size.
1068  Also, max must not be less than min. */  Also, max must not be less than min. */
1069    
1070  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1071    {    {
1072    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1073      {      {
1074      max = 0;      max = 0;
1075      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;
1076      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1077        {        {
1078        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 741  return p; Line 1097  return p;
1097    
1098    
1099  /*************************************************  /*************************************************
1100    *  Subroutine for finding forward reference      *
1101    *************************************************/
1102    
1103    /* This recursive function is called only from find_parens() below. The
1104    top-level call starts at the beginning of the pattern. All other calls must
1105    start at a parenthesis. It scans along a pattern's text looking for capturing
1106    subpatterns, and counting them. If it finds a named pattern that matches the
1107    name it is given, it returns its number. Alternatively, if the name is NULL, it
1108    returns when it reaches a given numbered subpattern. Recursion is used to keep
1109    track of subpatterns that reset the capturing group numbers - the (?| feature.
1110    
1111    This function was originally called only from the second pass, in which we know
1112    that if (?< or (?' or (?P< is encountered, the name will be correctly
1113    terminated because that is checked in the first pass. There is now one call to
1114    this function in the first pass, to check for a recursive back reference by
1115    name (so that we can make the whole group atomic). In this case, we need check
1116    only up to the current position in the pattern, and that is still OK because
1117    and previous occurrences will have been checked. To make this work, the test
1118    for "end of pattern" is a check against cd->end_pattern in the main loop,
1119    instead of looking for a binary zero. This means that the special first-pass
1120    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1121    processing items within the loop are OK, because afterwards the main loop will
1122    terminate.)
1123    
1124    Arguments:
1125      ptrptr       address of the current character pointer (updated)
1126      cd           compile background data
1127      name         name to seek, or NULL if seeking a numbered subpattern
1128      lorn         name length, or subpattern number if name is NULL
1129      xmode        TRUE if we are in /x mode
1130      utf8         TRUE if we are in UTF-8 mode
1131      count        pointer to the current capturing subpattern number (updated)
1132    
1133    Returns:       the number of the named subpattern, or -1 if not found
1134    */
1135    
1136    static int
1137    find_parens_sub(uschar **ptrptr, compile_data *cd, const uschar *name, int lorn,
1138      BOOL xmode, BOOL utf8, int *count)
1139    {
1140    uschar *ptr = *ptrptr;
1141    int start_count = *count;
1142    int hwm_count = start_count;
1143    BOOL dup_parens = FALSE;
1144    
1145    /* If the first character is a parenthesis, check on the type of group we are
1146    dealing with. The very first call may not start with a parenthesis. */
1147    
1148    if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1149      {
1150      /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1151    
1152      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1153    
1154      /* Handle a normal, unnamed capturing parenthesis. */
1155    
1156      else if (ptr[1] != CHAR_QUESTION_MARK)
1157        {
1158        *count += 1;
1159        if (name == NULL && *count == lorn) return *count;
1160        ptr++;
1161        }
1162    
1163      /* All cases now have (? at the start. Remember when we are in a group
1164      where the parenthesis numbers are duplicated. */
1165    
1166      else if (ptr[2] == CHAR_VERTICAL_LINE)
1167        {
1168        ptr += 3;
1169        dup_parens = TRUE;
1170        }
1171    
1172      /* Handle comments; all characters are allowed until a ket is reached. */
1173    
1174      else if (ptr[2] == CHAR_NUMBER_SIGN)
1175        {
1176        for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1177        goto FAIL_EXIT;
1178        }
1179    
1180      /* Handle a condition. If it is an assertion, just carry on so that it
1181      is processed as normal. If not, skip to the closing parenthesis of the
1182      condition (there can't be any nested parens). */
1183    
1184      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1185        {
1186        ptr += 2;
1187        if (ptr[1] != CHAR_QUESTION_MARK)
1188          {
1189          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1190          if (*ptr != 0) ptr++;
1191          }
1192        }
1193    
1194      /* Start with (? but not a condition. */
1195    
1196      else
1197        {
1198        ptr += 2;
1199        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1200    
1201        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1202    
1203        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1204            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1205          {
1206          int term;
1207          const uschar *thisname;
1208          *count += 1;
1209          if (name == NULL && *count == lorn) return *count;
1210          term = *ptr++;
1211          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1212          thisname = ptr;
1213          while (*ptr != term) ptr++;
1214          if (name != NULL && lorn == ptr - thisname &&
1215              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1216            return *count;
1217          term++;
1218          }
1219        }
1220      }
1221    
1222    /* Past any initial parenthesis handling, scan for parentheses or vertical
1223    bars. Stop if we get to cd->end_pattern. Note that this is important for the
1224    first-pass call when this value is temporarily adjusted to stop at the current
1225    position. So DO NOT change this to a test for binary zero. */
1226    
1227    for (; ptr < cd->end_pattern; ptr++)
1228      {
1229      /* Skip over backslashed characters and also entire \Q...\E */
1230    
1231      if (*ptr == CHAR_BACKSLASH)
1232        {
1233        if (*(++ptr) == 0) goto FAIL_EXIT;
1234        if (*ptr == CHAR_Q) for (;;)
1235          {
1236          while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1237          if (*ptr == 0) goto FAIL_EXIT;
1238          if (*(++ptr) == CHAR_E) break;
1239          }
1240        continue;
1241        }
1242    
1243      /* Skip over character classes; this logic must be similar to the way they
1244      are handled for real. If the first character is '^', skip it. Also, if the
1245      first few characters (either before or after ^) are \Q\E or \E we skip them
1246      too. This makes for compatibility with Perl. Note the use of STR macros to
1247      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1248    
1249      if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1250        {
1251        BOOL negate_class = FALSE;
1252        for (;;)
1253          {
1254          if (ptr[1] == CHAR_BACKSLASH)
1255            {
1256            if (ptr[2] == CHAR_E)
1257              ptr+= 2;
1258            else if (strncmp((const char *)ptr+2,
1259                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1260              ptr += 4;
1261            else
1262              break;
1263            }
1264          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1265            {
1266            negate_class = TRUE;
1267            ptr++;
1268            }
1269          else break;
1270          }
1271    
1272        /* If the next character is ']', it is a data character that must be
1273        skipped, except in JavaScript compatibility mode. */
1274    
1275        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1276            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1277          ptr++;
1278    
1279        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1280          {
1281          if (*ptr == 0) return -1;
1282          if (*ptr == CHAR_BACKSLASH)
1283            {
1284            if (*(++ptr) == 0) goto FAIL_EXIT;
1285            if (*ptr == CHAR_Q) for (;;)
1286              {
1287              while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1288              if (*ptr == 0) goto FAIL_EXIT;
1289              if (*(++ptr) == CHAR_E) break;
1290              }
1291            continue;
1292            }
1293          }
1294        continue;
1295        }
1296    
1297      /* Skip comments in /x mode */
1298    
1299      if (xmode && *ptr == CHAR_NUMBER_SIGN)
1300        {
1301        ptr++;
1302        while (*ptr != 0)
1303          {
1304          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1305          ptr++;
1306    #ifdef SUPPORT_UTF8
1307          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
1308    #endif
1309          }
1310        if (*ptr == 0) goto FAIL_EXIT;
1311        continue;
1312        }
1313    
1314      /* Check for the special metacharacters */
1315    
1316      if (*ptr == CHAR_LEFT_PARENTHESIS)
1317        {
1318        int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, count);
1319        if (rc > 0) return rc;
1320        if (*ptr == 0) goto FAIL_EXIT;
1321        }
1322    
1323      else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1324        {
1325        if (dup_parens && *count < hwm_count) *count = hwm_count;
1326        goto FAIL_EXIT;
1327        }
1328    
1329      else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1330        {
1331        if (*count > hwm_count) hwm_count = *count;
1332        *count = start_count;
1333        }
1334      }
1335    
1336    FAIL_EXIT:
1337    *ptrptr = ptr;
1338    return -1;
1339    }
1340    
1341    
1342    
1343    
1344    /*************************************************
1345    *       Find forward referenced subpattern       *
1346    *************************************************/
1347    
1348    /* This function scans along a pattern's text looking for capturing
1349    subpatterns, and counting them. If it finds a named pattern that matches the
1350    name it is given, it returns its number. Alternatively, if the name is NULL, it
1351    returns when it reaches a given numbered subpattern. This is used for forward
1352    references to subpatterns. We used to be able to start this scan from the
1353    current compiling point, using the current count value from cd->bracount, and
1354    do it all in a single loop, but the addition of the possibility of duplicate
1355    subpattern numbers means that we have to scan from the very start, in order to
1356    take account of such duplicates, and to use a recursive function to keep track
1357    of the different types of group.
1358    
1359    Arguments:
1360      cd           compile background data
1361      name         name to seek, or NULL if seeking a numbered subpattern
1362      lorn         name length, or subpattern number if name is NULL
1363      xmode        TRUE if we are in /x mode
1364      utf8         TRUE if we are in UTF-8 mode
1365    
1366    Returns:       the number of the found subpattern, or -1 if not found
1367    */
1368    
1369    static int
1370    find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode,
1371      BOOL utf8)
1372    {
1373    uschar *ptr = (uschar *)cd->start_pattern;
1374    int count = 0;
1375    int rc;
1376    
1377    /* If the pattern does not start with an opening parenthesis, the first call
1378    to find_parens_sub() will scan right to the end (if necessary). However, if it
1379    does start with a parenthesis, find_parens_sub() will return when it hits the
1380    matching closing parens. That is why we have to have a loop. */
1381    
1382    for (;;)
1383      {
1384      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, &count);
1385      if (rc > 0 || *ptr++ == 0) break;
1386      }
1387    
1388    return rc;
1389    }
1390    
1391    
1392    
1393    
1394    /*************************************************
1395  *      Find first significant op code            *  *      Find first significant op code            *
1396  *************************************************/  *************************************************/
1397    
1398  /* This is called by several functions that scan a compiled expression looking  /* This is called by several functions that scan a compiled expression looking
1399  for a fixed first character, or an anchoring op code etc. It skips over things  for a fixed first character, or an anchoring op code etc. It skips over things
1400  that do not influence this. For some calls, a change of option is important.  that do not influence this. For some calls, it makes sense to skip negative
1401  For some calls, it makes sense to skip negative forward and all backward  forward and all backward assertions, and also the \b assertion; for others it
1402  assertions, and also the \b assertion; for others it does not.  does not.
1403    
1404  Arguments:  Arguments:
1405    code         pointer to the start of the group    code         pointer to the start of the group
# Line 768  for (;;) Line 1419  for (;;)
1419    {    {
1420    switch ((int)*code)    switch ((int)*code)
1421      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1422      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1423      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1424      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
# Line 789  for (;;) Line 1434  for (;;)
1434    
1435      case OP_CALLOUT:      case OP_CALLOUT:
1436      case OP_CREF:      case OP_CREF:
1437      case OP_BRANUMBER:      case OP_NCREF:
1438        case OP_RREF:
1439        case OP_NRREF:
1440        case OP_DEF:
1441      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1442      break;      break;
1443    
# Line 804  for (;;) Line 1452  for (;;)
1452    
1453    
1454  /*************************************************  /*************************************************
1455  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1456  *************************************************/  *************************************************/
1457    
1458  /* 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,
1459  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.
1460  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
1461    temporarily terminated with OP_END when this function is called.
1462    
1463    This function is called when a backward assertion is encountered, so that if it
1464    fails, the error message can point to the correct place in the pattern.
1465    However, we cannot do this when the assertion contains subroutine calls,
1466    because they can be forward references. We solve this by remembering this case
1467    and doing the check at the end; a flag specifies which mode we are running in.
1468    
1469  Arguments:  Arguments:
1470    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1471    options  the compiling options    options  the compiling options
1472      atend    TRUE if called when the pattern is complete
1473      cd       the "compile data" structure
1474    
1475  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1476                 or -1 if there is no fixed length,
1477               or -2 if \C was encountered               or -2 if \C was encountered
1478                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1479  */  */
1480    
1481  static int  static int
1482  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)
1483  {  {
1484  int length = -1;  int length = -1;
1485    
# Line 833  branch, check the length against that of Line 1492  branch, check the length against that of
1492  for (;;)  for (;;)
1493    {    {
1494    int d;    int d;
1495      uschar *ce, *cs;
1496    register int op = *cc;    register int op = *cc;
   if (op >= OP_BRA) op = OP_BRA;  
   
1497    switch (op)    switch (op)
1498      {      {
1499        case OP_CBRA:
1500      case OP_BRA:      case OP_BRA:
1501      case OP_ONCE:      case OP_ONCE:
1502      case OP_COND:      case OP_COND:
1503      d = find_fixedlength(cc, options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);
1504      if (d < 0) return d;      if (d < 0) return d;
1505      branchlength += d;      branchlength += d;
1506      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 864  for (;;) Line 1523  for (;;)
1523      branchlength = 0;      branchlength = 0;
1524      break;      break;
1525    
1526      /* Skip over assertive subpatterns */      /* A true recursion implies not fixed length, but a subroutine call may
1527        be OK. If the subroutine is a forward reference, we can't deal with
1528        it until the end of the pattern, so return -3. */
1529    
1530        case OP_RECURSE:
1531        if (!atend) return -3;
1532        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1533        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1534        if (cc > cs && cc < ce) return -1;                /* Recursion */
1535        d = find_fixedlength(cs + 2, options, atend, cd);
1536        if (d < 0) return d;
1537        branchlength += d;
1538        cc += 1 + LINK_SIZE;
1539        break;
1540    
1541        /* Skip over assertive subpatterns */
1542    
1543      case OP_ASSERT:      case OP_ASSERT:
1544      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1545      case OP_ASSERTBACK:      case OP_ASSERTBACK:
# Line 876  for (;;) Line 1550  for (;;)
1550      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1551    
1552      case OP_REVERSE:      case OP_REVERSE:
     case OP_BRANUMBER:  
1553      case OP_CREF:      case OP_CREF:
1554      case OP_OPT:      case OP_NCREF:
1555        case OP_RREF:
1556        case OP_NRREF:
1557        case OP_DEF:
1558      case OP_CALLOUT:      case OP_CALLOUT:
1559      case OP_SOD:      case OP_SOD:
1560      case OP_SOM:      case OP_SOM:
1561        case OP_SET_SOM:
1562      case OP_EOD:      case OP_EOD:
1563      case OP_EODN:      case OP_EODN:
1564      case OP_CIRC:      case OP_CIRC:
1565        case OP_CIRCM:
1566      case OP_DOLL:      case OP_DOLL:
1567        case OP_DOLLM:
1568      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1569      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1570      cc += _pcre_OP_lengths[*cc];      cc += _pcre_OP_lengths[*cc];
# Line 894  for (;;) Line 1573  for (;;)
1573      /* Handle literal characters */      /* Handle literal characters */
1574    
1575      case OP_CHAR:      case OP_CHAR:
1576      case OP_CHARNC:      case OP_CHARI:
1577        case OP_NOT:
1578        case OP_NOTI:
1579      branchlength++;      branchlength++;
1580      cc += 2;      cc += 2;
1581  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1582      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1583        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1584  #endif  #endif
1585      break;      break;
1586    
# Line 912  for (;;) Line 1591  for (;;)
1591      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1592      cc += 4;      cc += 4;
1593  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1594      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1595        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1596  #endif  #endif
1597      break;      break;
1598    
1599      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1600      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1601        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1602      cc += 4;      cc += 4;
1603      break;      break;
1604    
# Line 928  for (;;) Line 1606  for (;;)
1606    
1607      case OP_PROP:      case OP_PROP:
1608      case OP_NOTPROP:      case OP_NOTPROP:
1609      cc++;      cc += 2;
1610      /* Fall through */      /* Fall through */
1611    
1612      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
# Line 938  for (;;) Line 1616  for (;;)
1616      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1617      case OP_WORDCHAR:      case OP_WORDCHAR:
1618      case OP_ANY:      case OP_ANY:
1619        case OP_ALLANY:
1620      branchlength++;      branchlength++;
1621      cc++;      cc++;
1622      break;      break;
# Line 992  for (;;) Line 1671  for (;;)
1671    
1672    
1673  /*************************************************  /*************************************************
1674  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1675  *************************************************/  *************************************************/
1676    
1677  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1678  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1679    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1680    so that it can be called from pcre_study() when finding the minimum matching
1681    length.
1682    
1683  Arguments:  Arguments:
1684    code        points to start of expression    code        points to start of expression
1685    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1686    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1687    
1688  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
1689  */  */
1690    
1691  static const uschar *  const uschar *
1692  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1693  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
1694  for (;;)  for (;;)
1695    {    {
1696    register int c = *code;    register int c = *code;
1697    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1698    else if (c > OP_BRA)  
1699      /* XCLASS is used for classes that cannot be represented just by a bit
1700      map. This includes negated single high-valued characters. The length in
1701      the table is zero; the actual length is stored in the compiled code. */
1702    
1703      if (c == OP_XCLASS) code += GET(code, 1);
1704    
1705      /* Handle recursion */
1706    
1707      else if (c == OP_REVERSE)
1708        {
1709        if (number < 0) return (uschar *)code;
1710        code += _pcre_OP_lengths[c];
1711        }
1712    
1713      /* Handle capturing bracket */
1714    
1715      else if (c == OP_CBRA)
1716      {      {
1717      int n = c - OP_BRA;      int n = GET2(code, 1+LINK_SIZE);
     if (n > EXTRACT_BASIC_MAX) n = GET2(code, 2+LINK_SIZE);  
1718      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
1719      code += _pcre_OP_lengths[OP_BRA];      code += _pcre_OP_lengths[c];
1720      }      }
1721    
1722      /* Otherwise, we can get the item's length from the table, except that for
1723      repeated character types, we have to test for \p and \P, which have an extra
1724      two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1725      must add in its length. */
1726    
1727    else    else
1728      {      {
1729      code += _pcre_OP_lengths[c];      switch(c)
1730          {
1731          case OP_TYPESTAR:
1732          case OP_TYPEMINSTAR:
1733          case OP_TYPEPLUS:
1734          case OP_TYPEMINPLUS:
1735          case OP_TYPEQUERY:
1736          case OP_TYPEMINQUERY:
1737          case OP_TYPEPOSSTAR:
1738          case OP_TYPEPOSPLUS:
1739          case OP_TYPEPOSQUERY:
1740          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1741          break;
1742    
1743  #ifdef SUPPORT_UTF8        case OP_TYPEUPTO:
1744          case OP_TYPEMINUPTO:
1745          case OP_TYPEEXACT:
1746          case OP_TYPEPOSUPTO:
1747          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1748          break;
1749    
1750      /* In UTF-8 mode, opcodes that are followed by a character may be followed        case OP_MARK:
1751      by a multi-byte character. The length in the table is a minimum, so we have        case OP_PRUNE_ARG:
1752      to scan along to skip the extra bytes. All opcodes are less than 128, so we        case OP_SKIP_ARG:
1753      can use relatively efficient code. */        code += code[1];
1754          break;
1755    
1756          case OP_THEN_ARG:
1757          code += code[1+LINK_SIZE];
1758          break;
1759          }
1760    
1761        /* Add in the fixed length from the table */
1762    
1763        code += _pcre_OP_lengths[c];
1764    
1765      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1766      a multi-byte character. The length in the table is a minimum, so we have to
1767      arrange to skip the extra bytes. */
1768    
1769    #ifdef SUPPORT_UTF8
1770      if (utf8) switch(c)      if (utf8) switch(c)
1771        {        {
1772        case OP_CHAR:        case OP_CHAR:
1773        case OP_CHARNC:        case OP_CHARI:
1774        case OP_EXACT:        case OP_EXACT:
1775          case OP_EXACTI:
1776        case OP_UPTO:        case OP_UPTO:
1777          case OP_UPTOI:
1778        case OP_MINUPTO:        case OP_MINUPTO:
1779          case OP_MINUPTOI:
1780          case OP_POSUPTO:
1781          case OP_POSUPTOI:
1782        case OP_STAR:        case OP_STAR:
1783          case OP_STARI:
1784        case OP_MINSTAR:        case OP_MINSTAR:
1785          case OP_MINSTARI:
1786          case OP_POSSTAR:
1787          case OP_POSSTARI:
1788        case OP_PLUS:        case OP_PLUS:
1789          case OP_PLUSI:
1790        case OP_MINPLUS:        case OP_MINPLUS:
1791          case OP_MINPLUSI:
1792          case OP_POSPLUS:
1793          case OP_POSPLUSI:
1794        case OP_QUERY:        case OP_QUERY:
1795          case OP_QUERYI:
1796        case OP_MINQUERY:        case OP_MINQUERY:
1797        while ((*code & 0xc0) == 0x80) code++;        case OP_MINQUERYI:
1798        break;        case OP_POSQUERY:
1799          case OP_POSQUERYI:
1800        /* XCLASS is used for classes that cannot be represented just by a bit        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
       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;  
1801        break;        break;
1802        }        }
1803    #else
1804        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1805  #endif  #endif
1806      }      }
1807    }    }
# Line 1083  Returns:      pointer to the opcode for Line 1826  Returns:      pointer to the opcode for
1826  static const uschar *  static const uschar *
1827  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const uschar *code, BOOL utf8)
1828  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
1829  for (;;)  for (;;)
1830    {    {
1831    register int c = *code;    register int c = *code;
1832    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1833    else if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
1834    else if (c > OP_BRA)  
1835      {    /* XCLASS is used for classes that cannot be represented just by a bit
1836      code += _pcre_OP_lengths[OP_BRA];    map. This includes negated single high-valued characters. The length in
1837      }    the table is zero; the actual length is stored in the compiled code. */
1838    
1839      if (c == OP_XCLASS) code += GET(code, 1);
1840    
1841      /* Otherwise, we can get the item's length from the table, except that for
1842      repeated character types, we have to test for \p and \P, which have an extra
1843      two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1844      must add in its length. */
1845    
1846    else    else
1847      {      {
1848      code += _pcre_OP_lengths[c];      switch(c)
1849          {
1850          case OP_TYPESTAR:
1851          case OP_TYPEMINSTAR:
1852          case OP_TYPEPLUS:
1853          case OP_TYPEMINPLUS:
1854          case OP_TYPEQUERY:
1855          case OP_TYPEMINQUERY:
1856          case OP_TYPEPOSSTAR:
1857          case OP_TYPEPOSPLUS:
1858          case OP_TYPEPOSQUERY:
1859          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1860          break;
1861    
1862  #ifdef SUPPORT_UTF8        case OP_TYPEPOSUPTO:
1863          case OP_TYPEUPTO:
1864          case OP_TYPEMINUPTO:
1865          case OP_TYPEEXACT:
1866          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1867          break;
1868    
1869          case OP_MARK:
1870          case OP_PRUNE_ARG:
1871          case OP_SKIP_ARG:
1872          code += code[1];
1873          break;
1874    
1875          case OP_THEN_ARG:
1876          code += code[1+LINK_SIZE];
1877          break;
1878          }
1879    
1880        /* Add in the fixed length from the table */
1881    
1882        code += _pcre_OP_lengths[c];
1883    
1884      /* 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
1885      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
1886      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. */  
1887    
1888    #ifdef SUPPORT_UTF8
1889      if (utf8) switch(c)      if (utf8) switch(c)
1890        {        {
1891        case OP_CHAR:        case OP_CHAR:
1892        case OP_CHARNC:        case OP_CHARI:
1893        case OP_EXACT:        case OP_EXACT:
1894          case OP_EXACTI:
1895        case OP_UPTO:        case OP_UPTO:
1896          case OP_UPTOI:
1897        case OP_MINUPTO:        case OP_MINUPTO:
1898          case OP_MINUPTOI:
1899          case OP_POSUPTO:
1900          case OP_POSUPTOI:
1901        case OP_STAR:        case OP_STAR:
1902          case OP_STARI:
1903        case OP_MINSTAR:        case OP_MINSTAR:
1904          case OP_MINSTARI:
1905          case OP_POSSTAR:
1906          case OP_POSSTARI:
1907        case OP_PLUS:        case OP_PLUS:
1908          case OP_PLUSI:
1909        case OP_MINPLUS:        case OP_MINPLUS:
1910          case OP_MINPLUSI:
1911          case OP_POSPLUS:
1912          case OP_POSPLUSI:
1913        case OP_QUERY:        case OP_QUERY:
1914          case OP_QUERYI:
1915        case OP_MINQUERY:        case OP_MINQUERY:
1916        while ((*code & 0xc0) == 0x80) code++;        case OP_MINQUERYI:
1917        break;        case OP_POSQUERY:
1918          case OP_POSQUERYI:
1919        /* XCLASS is used for classes that cannot be represented just by a bit        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
       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;  
1920        break;        break;
1921        }        }
1922    #else
1923        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1924  #endif  #endif
1925      }      }
1926    }    }
# Line 1143  for (;;) Line 1933  for (;;)
1933  *************************************************/  *************************************************/
1934    
1935  /* 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
1936  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()
1937  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
1938  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
1939  whose current branch will already have been scanned.  backward and negative forward assertions when its final argument is TRUE. If we
1940    hit an unclosed bracket, we return "empty" - this means we've struck an inner
1941    bracket whose current branch will already have been scanned.
1942    
1943  Arguments:  Arguments:
1944    code        points to start of search    code        points to start of search
1945    endcode     points to where to stop    endcode     points to where to stop
1946    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
1947      cd          contains pointers to tables etc.
1948    
1949  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
1950  */  */
1951    
1952  static BOOL  static BOOL
1953  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
1954      compile_data *cd)
1955  {  {
1956  register int c;  register int c;
1957  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);
1958       code < endcode;       code < endcode;
1959       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))
1960    {    {
# Line 1168  for (code = first_significant_code(code Line 1962  for (code = first_significant_code(code
1962    
1963    c = *code;    c = *code;
1964    
1965    if (c >= OP_BRA)    /* Skip over forward assertions; the other assertions are skipped by
1966      first_significant_code() with a TRUE final argument. */
1967    
1968      if (c == OP_ASSERT)
1969      {      {
1970      BOOL empty_branch;      do code += GET(code, 1); while (*code == OP_ALT);
1971      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      c = *code;
1972        continue;
1973        }
1974    
1975      /* Scan a closed bracket */    /* Groups with zero repeats can of course be empty; skip them. */
1976    
1977      empty_branch = FALSE;    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)
1978        {
1979        code += _pcre_OP_lengths[c];
1980        do code += GET(code, 1); while (*code == OP_ALT);
1981        c = *code;
1982        continue;
1983        }
1984    
1985      /* For a recursion/subroutine call, if its end has been reached, which
1986      implies a subroutine call, we can scan it. */
1987    
1988      if (c == OP_RECURSE)
1989        {
1990        BOOL empty_branch = FALSE;
1991        const uschar *scode = cd->start_code + GET(code, 1);
1992        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
1993      do      do
1994        {        {
1995        if (!empty_branch && could_be_empty_branch(code, endcode, utf8))        if (could_be_empty_branch(scode, endcode, utf8, cd))
1996            {
1997          empty_branch = TRUE;          empty_branch = TRUE;
1998            break;
1999            }
2000          scode += GET(scode, 1);
2001          }
2002        while (*scode == OP_ALT);
2003        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2004        continue;
2005        }
2006    
2007      /* For other groups, scan the branches. */
2008    
2009      if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)
2010        {
2011        BOOL empty_branch;
2012        if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
2013    
2014        /* If a conditional group has only one branch, there is a second, implied,
2015        empty branch, so just skip over the conditional, because it could be empty.
2016        Otherwise, scan the individual branches of the group. */
2017    
2018        if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
2019        code += GET(code, 1);        code += GET(code, 1);
2020        else
2021          {
2022          empty_branch = FALSE;
2023          do
2024            {
2025            if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
2026              empty_branch = TRUE;
2027            code += GET(code, 1);
2028            }
2029          while (*code == OP_ALT);
2030          if (!empty_branch) return FALSE;   /* All branches are non-empty */
2031        }        }
2032      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
     code += 1 + LINK_SIZE;  
2033      c = *code;      c = *code;
2034        continue;
2035      }      }
2036    
2037    else switch (c)    /* Handle the other opcodes */
2038    
2039      switch (c)
2040      {      {
2041      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
2042        cannot be represented just by a bit map. This includes negated single
2043        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
2044        actual length is stored in the compiled code, so we must update "code"
2045        here. */
2046    
2047  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2048      case OP_XCLASS:      case OP_XCLASS:
2049      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
2050      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
2051  #endif  #endif
2052    
# Line 1238  for (code = first_significant_code(code Line 2090  for (code = first_significant_code(code
2090      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2091      case OP_WORDCHAR:      case OP_WORDCHAR:
2092      case OP_ANY:      case OP_ANY:
2093        case OP_ALLANY:
2094      case OP_ANYBYTE:      case OP_ANYBYTE:
2095      case OP_CHAR:      case OP_CHAR:
2096      case OP_CHARNC:      case OP_CHARI:
2097      case OP_NOT:      case OP_NOT:
2098        case OP_NOTI:
2099      case OP_PLUS:      case OP_PLUS:
2100      case OP_MINPLUS:      case OP_MINPLUS:
2101        case OP_POSPLUS:
2102      case OP_EXACT:      case OP_EXACT:
2103      case OP_NOTPLUS:      case OP_NOTPLUS:
2104      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
2105        case OP_NOTPOSPLUS:
2106      case OP_NOTEXACT:      case OP_NOTEXACT:
2107      case OP_TYPEPLUS:      case OP_TYPEPLUS:
2108      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
2109        case OP_TYPEPOSPLUS:
2110      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2111      return FALSE;      return FALSE;
2112    
2113        /* These are going to continue, as they may be empty, but we have to
2114        fudge the length for the \p and \P cases. */
2115    
2116        case OP_TYPESTAR:
2117        case OP_TYPEMINSTAR:
2118        case OP_TYPEPOSSTAR:
2119        case OP_TYPEQUERY:
2120        case OP_TYPEMINQUERY:
2121        case OP_TYPEPOSQUERY:
2122        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2123        break;
2124    
2125        /* Same for these */
2126    
2127        case OP_TYPEUPTO:
2128        case OP_TYPEMINUPTO:
2129        case OP_TYPEPOSUPTO:
2130        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
2131        break;
2132    
2133      /* End of branch */      /* End of branch */
2134    
2135      case OP_KET:      case OP_KET:
# Line 1261  for (code = first_significant_code(code Line 2138  for (code = first_significant_code(code
2138      case OP_ALT:      case OP_ALT:
2139      return TRUE;      return TRUE;
2140    
2141      /* In UTF-8 mode, STAR, MINSTAR, QUERY, MINQUERY, UPTO, and MINUPTO  may be      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2142      followed by a multibyte character */      MINUPTO, and POSUPTO may be followed by a multibyte character */
2143    
2144  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2145      case OP_STAR:      case OP_STAR:
2146        case OP_STARI:
2147      case OP_MINSTAR:      case OP_MINSTAR:
2148        case OP_MINSTARI:
2149        case OP_POSSTAR:
2150        case OP_POSSTARI:
2151      case OP_QUERY:      case OP_QUERY:
2152        case OP_QUERYI:
2153      case OP_MINQUERY:      case OP_MINQUERY:
2154        case OP_MINQUERYI:
2155        case OP_POSQUERY:
2156        case OP_POSQUERYI:
2157        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2158        break;
2159    
2160      case OP_UPTO:      case OP_UPTO:
2161        case OP_UPTOI:
2162      case OP_MINUPTO:      case OP_MINUPTO:
2163      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_MINUPTOI:
2164        case OP_POSUPTO:
2165        case OP_POSUPTOI:
2166        if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2167      break;      break;
2168  #endif  #endif
2169    
2170        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2171        string. */
2172    
2173        case OP_MARK:
2174        case OP_PRUNE_ARG:
2175        case OP_SKIP_ARG:
2176        code += code[1];
2177        break;
2178    
2179        case OP_THEN_ARG:
2180        code += code[1+LINK_SIZE];
2181        break;
2182    
2183        /* None of the remaining opcodes are required to match a character. */
2184    
2185        default:
2186        break;
2187      }      }
2188    }    }
2189    
# Line 1296  Arguments: Line 2206  Arguments:
2206    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2207    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2208    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2209      cd          pointers to tables etc
2210    
2211  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2212  */  */
2213    
2214  static BOOL  static BOOL
2215  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2216    BOOL utf8)    BOOL utf8, compile_data *cd)
2217  {  {
2218  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2219    {    {
2220    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2221        return FALSE;
2222    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2223    }    }
2224  return TRUE;  return TRUE;
# Line 1319  return TRUE; Line 2231  return TRUE;
2231  *************************************************/  *************************************************/
2232    
2233  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2234  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
2235  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2236  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2237    
2238    Originally, this function only recognized a sequence of letters between the
2239    terminators, but it seems that Perl recognizes any sequence of characters,
2240    though of course unknown POSIX names are subsequently rejected. Perl gives an
2241    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2242    didn't consider this to be a POSIX class. Likewise for [:1234:].
2243    
2244    The problem in trying to be exactly like Perl is in the handling of escapes. We
2245    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2246    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2247    below handles the special case of \], but does not try to do any other escape
2248    processing. This makes it different from Perl for cases such as [:l\ower:]
2249    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2250    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2251    I think.
2252    
2253  Argument:  Arguments:
2254    ptr      pointer to the initial [    ptr      pointer to the initial [
2255    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2256    
2257  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2258  */  */
2259    
2260  static BOOL  static BOOL
2261  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
2262  {  {
2263  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2264  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2265  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2266    {    {
2267    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else
2268    return TRUE;      {
2269        if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2270        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2271          {
2272          *endptr = ptr;
2273          return TRUE;
2274          }
2275        }
2276    }    }
2277  return FALSE;  return FALSE;
2278  }  }
# Line 1366  Returns:     a value representing the na Line 2297  Returns:     a value representing the na
2297  static int  static int
2298  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
2299  {  {
2300    const char *pn = posix_names;
2301  register int yield = 0;  register int yield = 0;
2302  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2303    {    {
2304    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2305      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
2306      pn += posix_name_lengths[yield] + 1;
2307    yield++;    yield++;
2308    }    }
2309  return -1;  return -1;
# Line 1385  return -1; Line 2318  return -1;
2318  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2319  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2320  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
2321  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
2322  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
2323  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
2324  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
2325  partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2326    OP_END.
2327    
2328    This function has been extended with the possibility of forward references for
2329    recursions and subroutine calls. It must also check the list of such references
2330    for the group we are dealing with. If it finds that one of the recursions in
2331    the current group is on this list, it adjusts the offset in the list, not the
2332    value in the reference (which is a group number).
2333    
2334  Arguments:  Arguments:
2335    group      points to the start of the group    group      points to the start of the group
2336    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
2337    utf8       TRUE in UTF-8 mode    utf8       TRUE in UTF-8 mode
2338    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
2339      save_hwm   the hwm forward reference pointer at the start of the group
2340    
2341  Returns:     nothing  Returns:     nothing
2342  */  */
2343    
2344  static void  static void
2345  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd)  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,
2346      uschar *save_hwm)
2347  {  {
2348  uschar *ptr = group;  uschar *ptr = group;
2349    
2350  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
2351    {    {
2352    int offset = GET(ptr, 1);    int offset;
2353    if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);    uschar *hc;
2354    
2355      /* See if this recursion is on the forward reference list. If so, adjust the
2356      reference. */
2357    
2358      for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
2359        {
2360        offset = GET(hc, 0);
2361        if (cd->start_code + offset == ptr + 1)
2362          {
2363          PUT(hc, 0, offset + adjust);
2364          break;
2365          }
2366        }
2367    
2368      /* Otherwise, adjust the recursion offset if it's after the start of this
2369      group. */
2370    
2371      if (hc >= cd->hwm)
2372        {
2373        offset = GET(ptr, 1);
2374        if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
2375        }
2376    
2377    ptr += 1 + LINK_SIZE;    ptr += 1 + LINK_SIZE;
2378    }    }
2379  }  }
# Line 1434  auto_callout(uschar *code, const uschar Line 2400  auto_callout(uschar *code, const uschar
2400  {  {
2401  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2402  *code++ = 255;  *code++ = 255;
2403  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2404  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2405  return code + 2*LINK_SIZE;  return code + 2*LINK_SIZE;
2406  }  }
2407    
# Line 1460  Returns:             nothing Line 2426  Returns:             nothing
2426  static void  static void
2427  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)
2428  {  {
2429  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2430  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2431  }  }
2432    
# Line 1486  Yield:        TRUE when range returned; Line 2452  Yield:        TRUE when range returned;
2452  */  */
2453    
2454  static BOOL  static BOOL
2455  get_othercase_range(int *cptr, int d, int *ocptr, int *odptr)  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,
2456      unsigned int *odptr)
2457  {  {
2458  int c, chartype, othercase, next;  unsigned int c, othercase, next;
2459    
2460  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2461    {    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
   if (_pcre_ucp_findchar(c, &chartype, &othercase) == ucp_L && othercase != 0)  
     break;  
   }  
2462    
2463  if (c > d) return FALSE;  if (c > d) return FALSE;
2464    
# Line 1503  next = othercase + 1; Line 2467  next = othercase + 1;
2467    
2468  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2469    {    {
2470    if (_pcre_ucp_findchar(c, &chartype, &othercase) != ucp_L ||    if (UCD_OTHERCASE(c) != next) break;
         othercase != next)  
     break;  
2471    next++;    next++;
2472    }    }
2473    
# Line 1514  for (++c; c <= d; c++) Line 2476  for (++c; c <= d; c++)
2476    
2477  return TRUE;  return TRUE;
2478  }  }
2479  #endif  /* SUPPORT_UCP */  
2480    
2481    
2482  /*************************************************  /*************************************************
2483  *           Compile one branch                   *  *        Check a character and a property        *
2484  *************************************************/  *************************************************/
2485    
2486  /* 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
2487  changed during the branch, the pointer is used to change the external options  is adjacent to a fixed character.
 bits.  
2488    
2489  Arguments:  Arguments:
2490    optionsptr     pointer to the option bits    c            the character
2491    brackets       points to number of extracting brackets used    ptype        the property type
2492    codeptr        points to the pointer to the current code point    pdata        the data for the type
2493    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.  
2494    
2495  Returns:         TRUE on success  Returns:       TRUE if auto-possessifying is OK
                  FALSE, with *errorcodeptr set non-zero on error  
2496  */  */
2497    
2498  static BOOL  static BOOL
2499  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)  
2500  {  {
2501  int repeat_type, op_type;  const ucd_record *prop = GET_UCD(c);
2502  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  switch(ptype)
2503  int bravalue = 0;    {
2504  int greedy_default, greedy_non_default;    case PT_LAMP:
2505  int firstbyte, reqbyte;    return (prop->chartype == ucp_Lu ||
2506  int zeroreqbyte, zerofirstbyte;            prop->chartype == ucp_Ll ||
2507  int req_caseopt, reqvary, tempreqvary;            prop->chartype == ucp_Lt) == negated;
2508  int condcount = 0;  
2509  int options = *optionsptr;    case PT_GC:
2510  int after_manual_callout = 0;    return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;
2511  register int c;  
2512  register uschar *code = *codeptr;    case PT_PC:
2513  uschar *tempcode;    return (pdata == prop->chartype) == negated;
2514  BOOL inescq = FALSE;  
2515  BOOL groupsetfirstbyte = FALSE;    case PT_SC:
2516  const uschar *ptr = *ptrptr;    return (pdata == prop->script) == negated;
2517  const uschar *tempptr;  
2518  uschar *previous = NULL;    /* These are specials */
2519  uschar *previous_callout = NULL;  
2520  uschar classbits[32];    case PT_ALNUM:
2521      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2522  #ifdef SUPPORT_UTF8            _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;
2523  BOOL class_utf8;  
2524  BOOL utf8 = (options & PCRE_UTF8) != 0;    case PT_SPACE:    /* Perl space */
2525  uschar *class_utf8data;    return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2526  uschar utf8_char[6];            c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2527  #else            == negated;
2528  BOOL utf8 = FALSE;  
2529  #endif    case PT_PXSPACE:  /* POSIX space */
2530      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2531              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2532              c == CHAR_FF || c == CHAR_CR)
2533              == negated;
2534    
2535      case PT_WORD:
2536      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2537              _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2538              c == CHAR_UNDERSCORE) == negated;
2539      }
2540    return FALSE;
2541    }
2542    #endif  /* SUPPORT_UCP */
2543    
 /* Set up the default and non-default settings for greediness */  
2544    
 greedy_default = ((options & PCRE_UNGREEDY) != 0);  
 greedy_non_default = greedy_default ^ 1;  
2545    
2546  /* Initialize no first byte, no required byte. REQ_UNSET means "no char  /*************************************************
2547  matching encountered yet". It gets changed to REQ_NONE if we hit something that  *     Check if auto-possessifying is possible    *
2548  matches a non-fixed char first char; reqbyte just remains unset if we never  *************************************************/
 find one.  
2549    
2550  When we hit a repeat whose minimum is zero, we may have to adjust these values  /* This function is called for unlimited repeats of certain items, to see
2551  to take the zero repeat into account. This is implemented by setting them to  whether the next thing could possibly match the repeated item. If not, it makes
2552  zerofirstbyte and zeroreqbyte when such a repeat is encountered. The individual  sense to automatically possessify the repeated item.
 item types that can be repeated set these backoff variables appropriately. */  
2553    
2554  firstbyte = reqbyte = zerofirstbyte = zeroreqbyte = REQ_UNSET;  Arguments:
2555      previous      pointer to the repeated opcode
2556      utf8          TRUE in UTF-8 mode
2557      ptr           next character in pattern
2558      options       options bits
2559      cd            contains pointers to tables etc.
2560    
2561  /* The variable req_caseopt contains either the REQ_CASELESS value or zero,  Returns:        TRUE if possessifying is wanted
2562  according to the current setting of the caseless flag. REQ_CASELESS is a bit  */
 value > 255. It is added into the firstbyte or reqbyte variables to record the  
 case status of the value. This is used only for ASCII characters. */  
2563    
2564  req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;  static BOOL
2565    check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
2566      int options, compile_data *cd)
2567    {
2568    int c, next;
2569    int op_code = *previous++;
2570    
2571  /* Switch on next character until the end of the branch */  /* Skip whitespace and comments in extended mode */
2572    
2573  for (;; ptr++)  if ((options & PCRE_EXTENDED) != 0)
2574    {    {
2575    BOOL negate_class;    for (;;)
   BOOL possessive_quantifier;  
   BOOL is_quantifier;  
   int class_charcount;  
   int class_lastchar;  
   int newoptions;  
   int recno;  
   int skipbytes;  
   int subreqbyte;  
   int subfirstbyte;  
   int mclength;  
   uschar mcbuffer[8];  
   
   /* Next byte in the pattern */  
   
   c = *ptr;  
   
   /* If in \Q...\E, check for the end; if not, we have a literal */  
   
   if (inescq && c != 0)  
2576      {      {
2577      if (c == '\\' && ptr[1] == 'E')      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2578        if (*ptr == CHAR_NUMBER_SIGN)
2579        {        {
       inescq = FALSE;  
2580        ptr++;        ptr++;
2581        continue;        while (*ptr != 0)
       }  
     else  
       {  
       if (previous_callout != NULL)  
2582          {          {
2583          complete_callout(previous_callout, ptr, cd);          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2584          previous_callout = NULL;          ptr++;
2585    #ifdef SUPPORT_UTF8
2586            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2587    #endif
2588          }          }
       if ((options & PCRE_AUTO_CALLOUT) != 0)  
         {  
         previous_callout = code;  
         code = auto_callout(code, ptr, cd);  
         }  
       goto NORMAL_CHAR;  
2589        }        }
2590        else break;
2591      }      }
2592      }
2593    
2594    /* Fill in length of a previous callout, except when the next thing is  /* If the next item is one that we can handle, get its value. A non-negative
2595    a quantifier. */  value is a character, a negative value is an escape value. */
2596    
2597    is_quantifier = c == '*' || c == '+' || c == '?' ||  if (*ptr == CHAR_BACKSLASH)
2598      (c == '{' && is_counted_repeat(ptr+1));    {
2599      int temperrorcode = 0;
2600      next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
2601      if (temperrorcode != 0) return FALSE;
2602      ptr++;    /* Point after the escape sequence */
2603      }
2604    
2605    if (!is_quantifier && previous_callout != NULL &&  else if ((cd->ctypes[*ptr] & ctype_meta) == 0)
2606         after_manual_callout-- <= 0)    {
2607      {  #ifdef SUPPORT_UTF8
2608      complete_callout(previous_callout, ptr, cd);    if (utf8) { GETCHARINC(next, ptr); } else
2609      previous_callout = NULL;  #endif
2610      }    next = *ptr++;
2611      }
2612    
2613    /* In extended mode, skip white space and comments */  else return FALSE;
2614    
2615    if ((options & PCRE_EXTENDED) != 0)  /* Skip whitespace and comments in extended mode */
2616    
2617    if ((options & PCRE_EXTENDED) != 0)
2618      {
2619      for (;;)
2620      {      {
2621      if ((cd->ctypes[c] & ctype_space) != 0) continue;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2622      if (c == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2623        {        {
2624        /* The space before the ; is to avoid a warning on a silly compiler        ptr++;
2625        on the Macintosh. */        while (*ptr != 0)
2626        while ((c = *(++ptr)) != 0 && c != NEWLINE) ;          {
2627        if (c != 0) continue;   /* Else fall through to handle end of string */          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2628            ptr++;
2629    #ifdef SUPPORT_UTF8
2630            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2631    #endif
2632            }
2633        }        }
2634        else break;
2635      }      }
2636      }
2637    
2638    /* No auto callout for quantifiers. */  /* If the next thing is itself optional, we have to give up. */
   
   if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier)  
     {  
     previous_callout = code;  
     code = auto_callout(code, ptr, cd);  
     }  
2639    
2640    switch(c)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2641      {    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2642      /* The branch terminates at end of string, |, or ). */      return FALSE;
2643    
2644      case 0:  /* Now compare the next item with the previous opcode. First, handle cases when
2645      case '|':  the next item is a character. */
     case ')':  
     *firstbyteptr = firstbyte;  
     *reqbyteptr = reqbyte;  
     *codeptr = code;  
     *ptrptr = ptr;  
     return TRUE;  
2646    
2647      /* Handle single-character metacharacters. In multiline mode, ^ disables  if (next >= 0) switch(op_code)
2648      the setting of any following char as a first character. */    {
2649      case OP_CHAR:
2650    #ifdef SUPPORT_UTF8
2651      GETCHARTEST(c, previous);
2652    #else
2653      c = *previous;
2654    #endif
2655      return c != next;
2656    
2657      case '^':    /* For CHARI (caseless character) we must check the other case. If we have
2658      if ((options & PCRE_MULTILINE) != 0)    Unicode property support, we can use it to test the other case of
2659        {    high-valued characters. */
       if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;  
       }  
     previous = NULL;  
     *code++ = OP_CIRC;  
     break;  
2660    
2661      case '$':    case OP_CHARI:
2662      previous = NULL;  #ifdef SUPPORT_UTF8
2663      *code++ = OP_DOLL;    GETCHARTEST(c, previous);
2664      break;  #else
2665      c = *previous;
2666    #endif
2667      if (c == next) return FALSE;
2668    #ifdef SUPPORT_UTF8
2669      if (utf8)
2670        {
2671        unsigned int othercase;
2672        if (next < 128) othercase = cd->fcc[next]; else
2673    #ifdef SUPPORT_UCP
2674        othercase = UCD_OTHERCASE((unsigned int)next);
2675    #else
2676        othercase = NOTACHAR;
2677    #endif
2678        return (unsigned int)c != othercase;
2679        }
2680      else
2681    #endif  /* SUPPORT_UTF8 */
2682      return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2683    
2684      /* There can never be a first char if '.' is first, whatever happens about    /* For OP_NOT and OP_NOTI, the data is always a single-byte character. These
2685      repeats. The value of reqbyte doesn't change either. */    opcodes are not used for multi-byte characters, because they are coded using
2686      an XCLASS instead. */
2687    
2688      case OP_NOT:
2689      return (c = *previous) == next;
2690    
2691      case OP_NOTI:
2692      if ((c = *previous) == next) return TRUE;
2693    #ifdef SUPPORT_UTF8
2694      if (utf8)
2695        {
2696        unsigned int othercase;
2697        if (next < 128) othercase = cd->fcc[next]; else
2698    #ifdef SUPPORT_UCP
2699        othercase = UCD_OTHERCASE(next);
2700    #else
2701        othercase = NOTACHAR;
2702    #endif
2703        return (unsigned int)c == othercase;
2704        }
2705      else
2706    #endif  /* SUPPORT_UTF8 */
2707      return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2708    
2709      case '.':    /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2710      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;    When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
     zerofirstbyte = firstbyte;  
     zeroreqbyte = reqbyte;  
     previous = code;  
     *code++ = OP_ANY;  
     break;  
2711    
2712      /* Character classes. If the included characters are all < 255 in value, we    case OP_DIGIT:
2713      build a 32-byte bitmap of the permitted characters, except in the special    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
     case where there is only one such character. For negated classes, we build  
     the map as usual, then invert it at the end. However, we use a different  
     opcode so that data characters > 255 can be handled correctly.  
2714    
2715      If the class contains characters outside the 0-255 range, a different    case OP_NOT_DIGIT:
2716      opcode is compiled. It may optionally have a bit map for characters < 256,    return next <= 127 && (cd->ctypes[next] & ctype_digit) != 0;
     but those above are are explicitly listed afterwards. A flag byte tells  
     whether the bitmap is present, and whether this is a negated class or not.  
     */  
2717    
2718      case '[':    case OP_WHITESPACE:
2719      previous = code;    return next > 127 || (cd->ctypes[next] & ctype_space) == 0;
2720    
2721      /* PCRE supports POSIX class stuff inside a class. Perl gives an error if    case OP_NOT_WHITESPACE:
2722      they are encountered at the top level, so we'll do that too. */    return next <= 127 && (cd->ctypes[next] & ctype_space) != 0;
2723    
2724      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&    case OP_WORDCHAR:
2725          check_posix_syntax(ptr, &tempptr, cd))    return next > 127 || (cd->ctypes[next] & ctype_word) == 0;
       {  
       *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;  
       goto FAILED;  
       }  
2726    
2727      /* If the first character is '^', set the negation flag and skip it. */    case OP_NOT_WORDCHAR:
2728      return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;
2729    
2730      if ((c = *(++ptr)) == '^')    case OP_HSPACE:
2731        {    case OP_NOT_HSPACE:
2732        negate_class = TRUE;    switch(next)
2733        c = *(++ptr);      {
2734        }      case 0x09:
2735      else      case 0x20:
2736        {      case 0xa0:
2737        negate_class = FALSE;      case 0x1680:
2738        }      case 0x180e:
2739        case 0x2000:
2740        case 0x2001:
2741        case 0x2002:
2742        case 0x2003:
2743        case 0x2004:
2744        case 0x2005:
2745        case 0x2006:
2746        case 0x2007:
2747        case 0x2008:
2748        case 0x2009:
2749        case 0x200A:
2750        case 0x202f:
2751        case 0x205f:
2752        case 0x3000:
2753        return op_code == OP_NOT_HSPACE;
2754        default:
2755        return op_code != OP_NOT_HSPACE;
2756        }
2757    
2758      /* Keep a count of chars with values < 256 so that we can optimize the case    case OP_ANYNL:
2759      of just a single character (as long as it's < 256). For higher valued UTF-8    case OP_VSPACE:
2760      characters, we don't yet do any optimization. */    case OP_NOT_VSPACE:
2761      switch(next)
2762        {
2763        case 0x0a:
2764        case 0x0b:
2765        case 0x0c:
2766        case 0x0d:
2767        case 0x85:
2768        case 0x2028:
2769        case 0x2029:
2770        return op_code == OP_NOT_VSPACE;
2771        default:
2772        return op_code != OP_NOT_VSPACE;
2773        }
2774    
2775      class_charcount = 0;  #ifdef SUPPORT_UCP
2776      class_lastchar = -1;    case OP_PROP:
2777      return check_char_prop(next, previous[0], previous[1], FALSE);
2778    
2779  #ifdef SUPPORT_UTF8    case OP_NOTPROP:
2780      class_utf8 = FALSE;                       /* No chars >= 256 */    return check_char_prop(next, previous[0], previous[1], TRUE);
     class_utf8data = code + LINK_SIZE + 34;   /* For UTF-8 items */  
2781  #endif  #endif
2782    
2783      /* Initialize the 32-char bit map to all zeros. We have to build the    default:
2784      map in a temporary bit of store, in case the class contains only 1    return FALSE;
2785      character (< 256), because in that case the compiled code doesn't use the    }
     bit map. */  
2786    
     memset(classbits, 0, 32 * sizeof(uschar));  
2787    
2788      /* Process characters until ] is reached. By writing this as a "do" it  /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP
2789      means that an initial ] is taken as a data character. The first pass  is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
2790      through the regex checked the overall syntax, so we don't need to be very  generated only when PCRE_UCP is *not* set, that is, when only ASCII
2791      strict here. At the start of the loop, c contains the first byte of the  characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
2792      character. */  replaced by OP_PROP codes when PCRE_UCP is set. */
2793    
2794      do  switch(op_code)
2795        {    {
2796      case OP_CHAR:
2797      case OP_CHARI:
2798  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2799        if (utf8 && c > 127)    GETCHARTEST(c, previous);
2800          {                           /* Braces are required because the */  #else
2801          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */    c = *previous;
         }  
2802  #endif  #endif
2803      switch(-next)
2804        {
2805        case ESC_d:
2806        return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
2807    
2808        /* Inside \Q...\E everything is literal except \E */      case ESC_D:
2809        return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
2810    
2811        if (inescq)      case ESC_s:
2812          {      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
         if (c == '\\' && ptr[1] == 'E')  
           {  
           inescq = FALSE;  
           ptr++;  
           continue;  
           }  
         else goto LONE_SINGLE_CHARACTER;  
         }  
2813    
2814        /* Handle POSIX class names. Perl allows a negation extension of the      case ESC_S:
2815        form [:^name:]. A square bracket that doesn't match the syntax is      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
2816        treated as a literal. We also recognize the POSIX constructions  
2817        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl      case ESC_w:
2818        5.6 and 5.8 do. */      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
2819    
2820        case ESC_W:
2821        return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
2822    
2823        case ESC_h:
2824        case ESC_H:
2825        switch(c)
2826          {
2827          case 0x09:
2828          case 0x20:
2829          case 0xa0:
2830          case 0x1680:
2831          case 0x180e:
2832          case 0x2000:
2833          case 0x2001:
2834          case 0x2002:
2835          case 0x2003:
2836          case 0x2004:
2837          case 0x2005:
2838          case 0x2006:
2839          case 0x2007:
2840          case 0x2008:
2841          case 0x2009:
2842          case 0x200A:
2843          case 0x202f:
2844          case 0x205f:
2845          case 0x3000:
2846          return -next != ESC_h;
2847          default:
2848          return -next == ESC_h;
2849          }
2850    
2851        if (c == '[' &&      case ESC_v:
2852            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      case ESC_V:
2853            check_posix_syntax(ptr, &tempptr, cd))      switch(c)
2854          {        {
2855          BOOL local_negate = FALSE;        case 0x0a:
2856          int posix_class, i;        case 0x0b:
2857          register const uschar *cbits = cd->cbits;        case 0x0c:
2858          case 0x0d:
2859          case 0x85:
2860          case 0x2028:
2861          case 0x2029:
2862          return -next != ESC_v;
2863          default:
2864          return -next == ESC_v;
2865          }
2866    
2867          if (ptr[1] != ':')      /* When PCRE_UCP is set, these values get generated for \d etc. Find
2868            {      their substitutions and process them. The result will always be either
2869            *errorcodeptr = ERR31;      -ESC_p or -ESC_P. Then fall through to process those values. */
           goto FAILED;  
           }  
2870    
2871          ptr += 2;  #ifdef SUPPORT_UCP
2872          if (*ptr == '^')      case ESC_du:
2873            {      case ESC_DU:
2874            local_negate = TRUE;      case ESC_wu:
2875            ptr++;      case ESC_WU:
2876            }      case ESC_su:
2877        case ESC_SU:
2878          {
2879          int temperrorcode = 0;
2880          ptr = substitutes[-next - ESC_DU];
2881          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
2882          if (temperrorcode != 0) return FALSE;
2883          ptr++;    /* For compatibility */
2884          }
2885        /* Fall through */
2886    
2887          posix_class = check_posix_name(ptr, tempptr - ptr);      case ESC_p:
2888          if (posix_class < 0)      case ESC_P:
2889            {        {
2890            *errorcodeptr = ERR30;        int ptype, pdata, errorcodeptr;
2891            goto FAILED;        BOOL negated;
           }  
2892    
2893          /* If matching is caseless, upper and lower are converted to        ptr--;      /* Make ptr point at the p or P */
2894          alpha. This relies on the fact that the class table starts with        ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
2895          alpha, lower, upper as the first 3 entries. */        if (ptype < 0) return FALSE;
2896          ptr++;      /* Point past the final curly ket */
2897    
2898          /* If the property item is optional, we have to give up. (When generated
2899          from \d etc by PCRE_UCP, this test will have been applied much earlier,
2900          to the original \d etc. At this point, ptr will point to a zero byte. */
2901    
2902          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2903            strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2904              return FALSE;
2905    
2906          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)        /* Do the property check. */
           posix_class = 0;  
2907    
2908          /* Or into the map we are building up to 3 of the static class        return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
2909          tables, or their negations. The [:blank:] class sets up the same        }
2910          chars as the [:space:] class (all white space). We remove the vertical  #endif
         white space chars afterwards. */  
2911    
2912          posix_class *= 3;      default:
2913          for (i = 0; i < 3; i++)      return FALSE;
2914            {      }
           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;  
             }  
           }  
2915    
2916          ptr = tempptr + 1;    /* In principle, support for Unicode properties should be integrated here as
2917          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */    well. It means re-organizing the above code so as to get hold of the property
2918          continue;    /* End of POSIX syntax handling */    values before switching on the op-code. However, I wonder how many patterns
2919          }    combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
2920      these op-codes are never generated.) */
2921    
2922        /* Backslash may introduce a single character, or it may introduce one    case OP_DIGIT:
2923        of the specials, which just set a flag. Escaped items are checked for    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2924        validity in the pre-compiling pass. The sequence \b is a special case.           next == -ESC_h || next == -ESC_v || next == -ESC_R;
       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. */  
2925    
2926        if (c == '\\')    case OP_NOT_DIGIT:
2927          {    return next == -ESC_d;
         c = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);  
2928    
2929          if (-c == ESC_b) c = '\b';       /* \b is backslash in a class */    case OP_WHITESPACE:
2930          else if (-c == ESC_X) c = 'X';   /* \X is literal X in a class */    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
         else if (-c == ESC_Q)            /* Handle start of quoted string */  
           {  
           if (ptr[1] == '\\' && ptr[2] == 'E')  
             {  
             ptr += 2; /* avoid empty string */  
             }  
           else inescq = TRUE;  
           continue;  
           }  
2931    
2932          if (c < 0)    case OP_NOT_WHITESPACE:
2933            {    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
           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;  
2934    
2935              case ESC_D:    case OP_HSPACE:
2936              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
2937              continue;           next == -ESC_w || next == -ESC_v || next == -ESC_R;
2938    
2939              case ESC_w:    case OP_NOT_HSPACE:
2940              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_word];    return next == -ESC_h;
             continue;  
2941    
2942              case ESC_W:    /* Can't have \S in here because VT matches \S (Perl anomaly) */
2943              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];    case OP_ANYNL:
2944              continue;    case OP_VSPACE:
2945      return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2946    
2947              case ESC_s:    case OP_NOT_VSPACE:
2948              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];    return next == -ESC_v || next == -ESC_R;
             classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */  
             continue;  
2949    
2950              case ESC_S:    case OP_WORDCHAR:
2951              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
2952              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */           next == -ESC_v || next == -ESC_R;
             continue;  
2953    
2954  #ifdef SUPPORT_UCP    case OP_NOT_WORDCHAR:
2955              case ESC_p:    return next == -ESC_w || next == -ESC_d;
             case ESC_P:  
               {  
               BOOL negated;  
               int property = get_ucp(&ptr, &negated, errorcodeptr);  
               if (property < 0) goto FAILED;  
               class_utf8 = TRUE;  
               *class_utf8data++ = ((-c == ESC_p) != negated)?  
                 XCL_PROP : XCL_NOTPROP;  
               *class_utf8data++ = property;  
               class_charcount -= 2;   /* Not a < 256 character */  
               }  
             continue;  
 #endif  
2956    
2957              /* Unrecognized escapes are faulted if PCRE is running in its    default:
2958              strict mode. By default, for compatibility with Perl, they are    return FALSE;
2959              treated as literals. */    }
2960    
2961              default:  /* Control does not reach here */
2962              if ((options & PCRE_EXTRA) != 0)  }
               {  
               *errorcodeptr = ERR7;  
               goto FAILED;  
               }  
             c = *ptr;              /* The final character */  
             class_charcount -= 2;  /* Undo the default count from above */  
             }  
           }  
2963    
         /* Fall through if we have a single character (c >= 0). This may be  
         > 256 in UTF-8 mode. */  
2964    
         }   /* End of backslash handling */  
2965    
2966        /* A single character may be followed by '-' to form a range. However,  /*************************************************
2967        Perl does not permit ']' to be the end of the range. A '-' character  *           Compile one branch                   *
2968        here is treated as a literal. */  *************************************************/
2969    
2970        if (ptr[1] == '-' && ptr[2] != ']')  /* Scan the pattern, compiling it into the a vector. If the options are
2971          {  changed during the branch, the pointer is used to change the external options
2972          int d;  bits. This function is used during the pre-compile phase when we are trying
2973          ptr += 2;  to find out the amount of memory needed, as well as during the real compile
2974    phase. The value of lengthptr distinguishes the two phases.
2975    
2976    Arguments:
2977      optionsptr     pointer to the option bits
2978      codeptr        points to the pointer to the current code point
2979      ptrptr         points to the current pattern pointer
2980      errorcodeptr   points to error code variable
2981      firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
2982      reqbyteptr     set to the last literal character required, else < 0
2983      bcptr          points to current branch chain
2984      cd             contains pointers to tables etc.
2985      lengthptr      NULL during the real compile phase
2986                     points to length accumulator during pre-compile phase
2987    
2988    Returns:         TRUE on success
2989                     FALSE, with *errorcodeptr set non-zero on error
2990    */
2991    
2992    static BOOL
2993    compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
2994      int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
2995      compile_data *cd, int *lengthptr)
2996    {
2997    int repeat_type, op_type;
2998    int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
2999    int bravalue = 0;
3000    int greedy_default, greedy_non_default;
3001    int firstbyte, reqbyte;
3002    int zeroreqbyte, zerofirstbyte;
3003    int req_caseopt, reqvary, tempreqvary;
3004    int options = *optionsptr;
3005    int after_manual_callout = 0;
3006    int length_prevgroup = 0;
3007    register int c;
3008    register uschar *code = *codeptr;
3009    uschar *last_code = code;
3010    uschar *orig_code = code;
3011    uschar *tempcode;
3012    BOOL inescq = FALSE;
3013    BOOL groupsetfirstbyte = FALSE;
3014    const uschar *ptr = *ptrptr;
3015    const uschar *tempptr;
3016    const uschar *nestptr = NULL;
3017    uschar *previous = NULL;
3018    uschar *previous_callout = NULL;
3019    uschar *save_hwm = NULL;
3020    uschar classbits[32];
3021    
3022    #ifdef SUPPORT_UTF8
3023    BOOL class_utf8;
3024    BOOL utf8 = (options & PCRE_UTF8) != 0;
3025    uschar *class_utf8data;
3026    uschar *class_utf8data_base;
3027    uschar utf8_char[6];
3028    #else
3029    BOOL utf8 = FALSE;
3030    uschar *utf8_char = NULL;
3031    #endif
3032    
3033    #ifdef PCRE_DEBUG
3034    if (lengthptr != NULL) DPRINTF((">> start branch\n"));
3035    #endif
3036    
3037    /* Set up the default and non-default settings for greediness */
3038    
3039    greedy_default = ((options & PCRE_UNGREEDY) != 0);
3040    greedy_non_default = greedy_default ^ 1;
3041    
3042    /* Initialize no first byte, no required byte. REQ_UNSET means "no char
3043    matching encountered yet". It gets changed to REQ_NONE if we hit something that
3044    matches a non-fixed char first char; reqbyte just remains unset if we never
3045    find one.
3046    
3047    When we hit a repeat whose minimum is zero, we may have to adjust these values
3048    to take the zero repeat into account. This is implemented by setting them to
3049    zerofirstbyte and zeroreqbyte when such a repeat is encountered. The individual
3050    item types that can be repeated set these backoff variables appropriately. */
3051    
3052    firstbyte = reqbyte = zerofirstbyte = zeroreqbyte = REQ_UNSET;
3053    
3054    /* The variable req_caseopt contains either the REQ_CASELESS value or zero,
3055    according to the current setting of the caseless flag. REQ_CASELESS is a bit
3056    value > 255. It is added into the firstbyte or reqbyte variables to record the
3057    case status of the value. This is used only for ASCII characters. */
3058    
3059    req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
3060    
3061    /* Switch on next character until the end of the branch */
3062    
3063    for (;; ptr++)
3064      {
3065      BOOL negate_class;
3066      BOOL should_flip_negation;
3067      BOOL possessive_quantifier;
3068      BOOL is_quantifier;
3069      BOOL is_recurse;
3070      BOOL reset_bracount;
3071      int class_charcount;
3072      int class_lastchar;
3073      int newoptions;
3074      int recno;
3075      int refsign;
3076      int skipbytes;
3077      int subreqbyte;
3078      int subfirstbyte;
3079      int terminator;
3080      int mclength;
3081      uschar mcbuffer[8];
3082    
3083      /* Get next byte in the pattern */
3084    
3085      c = *ptr;
3086    
3087      /* If we are at the end of a nested substitution, revert to the outer level
3088      string. Nesting only happens one level deep. */
3089    
3090      if (c == 0 && nestptr != NULL)
3091        {
3092        ptr = nestptr;
3093        nestptr = NULL;
3094        c = *ptr;
3095        }
3096    
3097      /* If we are in the pre-compile phase, accumulate the length used for the
3098      previous cycle of this loop. */
3099    
3100      if (lengthptr != NULL)
3101        {
3102    #ifdef PCRE_DEBUG
3103        if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3104    #endif
3105        if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
3106          {
3107          *errorcodeptr = ERR52;
3108          goto FAILED;
3109          }
3110    
3111        /* There is at least one situation where code goes backwards: this is the
3112        case of a zero quantifier after a class (e.g. [ab]{0}). At compile time,
3113        the class is simply eliminated. However, it is created first, so we have to
3114        allow memory for it. Therefore, don't ever reduce the length at this point.
3115        */
3116    
3117        if (code < last_code) code = last_code;
3118    
3119        /* Paranoid check for integer overflow */
3120    
3121        if (OFLOW_MAX - *lengthptr < code - last_code)
3122          {
3123          *errorcodeptr = ERR20;
3124          goto FAILED;
3125          }
3126    
3127        *lengthptr += (int)(code - last_code);
3128        DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));
3129    
3130        /* If "previous" is set and it is not at the start of the work space, move
3131        it back to there, in order to avoid filling up the work space. Otherwise,
3132        if "previous" is NULL, reset the current code pointer to the start. */
3133    
3134        if (previous != NULL)
3135          {
3136          if (previous > orig_code)
3137            {
3138            memmove(orig_code, previous, code - previous);
3139            code -= previous - orig_code;
3140            previous = orig_code;
3141            }
3142          }
3143        else code = orig_code;
3144    
3145        /* Remember where this code item starts so we can pick up the length
3146        next time round. */
3147    
3148        last_code = code;
3149        }
3150    
3151      /* In the real compile phase, just check the workspace used by the forward
3152      reference list. */
3153    
3154      else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
3155        {
3156        *errorcodeptr = ERR52;
3157        goto FAILED;
3158        }
3159    
3160      /* If in \Q...\E, check for the end; if not, we have a literal */
3161    
3162      if (inescq && c != 0)
3163        {
3164        if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3165          {
3166          inescq = FALSE;
3167          ptr++;
3168          continue;
3169          }
3170        else
3171          {
3172          if (previous_callout != NULL)
3173            {
3174            if (lengthptr == NULL)  /* Don't attempt in pre-compile phase */
3175              complete_callout(previous_callout, ptr, cd);
3176            previous_callout = NULL;
3177            }
3178          if ((options & PCRE_AUTO_CALLOUT) != 0)
3179            {
3180            previous_callout = code;
3181            code = auto_callout(code, ptr, cd);
3182            }
3183          goto NORMAL_CHAR;
3184          }
3185        }
3186    
3187      /* Fill in length of a previous callout, except when the next thing is
3188      a quantifier. */
3189    
3190      is_quantifier =
3191        c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
3192        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
3193    
3194      if (!is_quantifier && previous_callout != NULL &&
3195           after_manual_callout-- <= 0)
3196        {
3197        if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */
3198          complete_callout(previous_callout, ptr, cd);
3199        previous_callout = NULL;
3200        }
3201    
3202      /* In extended mode, skip white space and comments */
3203    
3204      if ((options & PCRE_EXTENDED) != 0)
3205        {
3206        if ((cd->ctypes[c] & ctype_space) != 0) continue;
3207        if (c == CHAR_NUMBER_SIGN)
3208          {
3209          ptr++;
3210          while (*ptr != 0)
3211            {
3212            if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3213            ptr++;
3214    #ifdef SUPPORT_UTF8
3215            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
3216    #endif
3217            }
3218          if (*ptr != 0) continue;
3219    
3220          /* Else fall through to handle end of string */
3221          c = 0;
3222          }
3223        }
3224    
3225      /* No auto callout for quantifiers. */
3226    
3227      if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier)
3228        {
3229        previous_callout = code;
3230        code = auto_callout(code, ptr, cd);
3231        }
3232    
3233      switch(c)
3234        {
3235        /* ===================================================================*/
3236        case 0:                        /* The branch terminates at string end */
3237        case CHAR_VERTICAL_LINE:       /* or | or ) */
3238        case CHAR_RIGHT_PARENTHESIS:
3239        *firstbyteptr = firstbyte;
3240        *reqbyteptr = reqbyte;
3241        *codeptr = code;
3242        *ptrptr = ptr;
3243        if (lengthptr != NULL)
3244          {
3245          if (OFLOW_MAX - *lengthptr < code - last_code)
3246            {
3247            *errorcodeptr = ERR20;
3248            goto FAILED;
3249            }
3250          *lengthptr += (int)(code - last_code);   /* To include callout length */
3251          DPRINTF((">> end branch\n"));
3252          }
3253        return TRUE;
3254    
3255    
3256        /* ===================================================================*/
3257        /* Handle single-character metacharacters. In multiline mode, ^ disables
3258        the setting of any following char as a first character. */
3259    
3260        case CHAR_CIRCUMFLEX_ACCENT:
3261        previous = NULL;
3262        if ((options & PCRE_MULTILINE) != 0)
3263          {
3264          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3265          *code++ = OP_CIRCM;
3266          }
3267        else *code++ = OP_CIRC;
3268        break;
3269    
3270        case CHAR_DOLLAR_SIGN:
3271        previous = NULL;
3272        *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
3273        break;
3274    
3275        /* There can never be a first char if '.' is first, whatever happens about
3276        repeats. The value of reqbyte doesn't change either. */
3277    
3278        case CHAR_DOT:
3279        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3280        zerofirstbyte = firstbyte;
3281        zeroreqbyte = reqbyte;
3282        previous = code;
3283        *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
3284        break;
3285    
3286    
3287        /* ===================================================================*/
3288        /* Character classes. If the included characters are all < 256, we build a
3289        32-byte bitmap of the permitted characters, except in the special case
3290        where there is only one such character. For negated classes, we build the
3291        map as usual, then invert it at the end. However, we use a different opcode
3292        so that data characters > 255 can be handled correctly.
3293    
3294        If the class contains characters outside the 0-255 range, a different
3295        opcode is compiled. It may optionally have a bit map for characters < 256,
3296        but those above are are explicitly listed afterwards. A flag byte tells
3297        whether the bitmap is present, and whether this is a negated class or not.
3298    
3299        In JavaScript compatibility mode, an isolated ']' causes an error. In
3300        default (Perl) mode, it is treated as a data character. */
3301    
3302        case CHAR_RIGHT_SQUARE_BRACKET:
3303        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3304          {
3305          *errorcodeptr = ERR64;
3306          goto FAILED;
3307          }
3308        goto NORMAL_CHAR;
3309    
3310        case CHAR_LEFT_SQUARE_BRACKET:
3311        previous = code;
3312    
3313        /* PCRE supports POSIX class stuff inside a class. Perl gives an error if
3314        they are encountered at the top level, so we'll do that too. */
3315    
3316        if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3317             ptr[1] == CHAR_EQUALS_SIGN) &&
3318            check_posix_syntax(ptr, &tempptr))
3319          {
3320          *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
3321          goto FAILED;
3322          }
3323    
3324        /* If the first character is '^', set the negation flag and skip it. Also,
3325        if the first few characters (either before or after ^) are \Q\E or \E we
3326        skip them too. This makes for compatibility with Perl. */
3327    
3328        negate_class = FALSE;
3329        for (;;)
3330          {
3331          c = *(++ptr);
3332          if (c == CHAR_BACKSLASH)
3333            {
3334            if (ptr[1] == CHAR_E)
3335              ptr++;
3336            else if (strncmp((const char *)ptr+1,
3337                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
3338              ptr += 3;
3339            else
3340              break;
3341            }
3342          else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
3343            negate_class = TRUE;
3344          else break;
3345          }
3346    
3347        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
3348        an initial ']' is taken as a data character -- the code below handles
3349        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3350        [^] must match any character, so generate OP_ALLANY. */
3351    
3352        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3353            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3354          {
3355          *code++ = negate_class? OP_ALLANY : OP_FAIL;
3356          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3357          zerofirstbyte = firstbyte;
3358          break;
3359          }
3360    
3361        /* If a class contains a negative special such as \S, we need to flip the
3362        negation flag at the end, so that support for characters > 255 works
3363        correctly (they are all included in the class). */
3364    
3365        should_flip_negation = FALSE;
3366    
3367        /* Keep a count of chars with values < 256 so that we can optimize the case
3368        of just a single character (as long as it's < 256). However, For higher
3369        valued UTF-8 characters, we don't yet do any optimization. */
3370    
3371        class_charcount = 0;
3372        class_lastchar = -1;
3373    
3374        /* Initialize the 32-char bit map to all zeros. We build the map in a
3375        temporary bit of memory, in case the class contains only 1 character (less
3376        than 256), because in that case the compiled code doesn't use the bit map.
3377        */
3378    
3379        memset(classbits, 0, 32 * sizeof(uschar));
3380    
3381    #ifdef SUPPORT_UTF8
3382        class_utf8 = FALSE;                       /* No chars >= 256 */
3383        class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
3384        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
3385    #endif
3386    
3387        /* Process characters until ] is reached. By writing this as a "do" it
3388        means that an initial ] is taken as a data character. At the start of the
3389        loop, c contains the first byte of the character. */
3390    
3391        if (c != 0) do
3392          {
3393          const uschar *oldptr;
3394    
3395    #ifdef SUPPORT_UTF8
3396          if (utf8 && c > 127)
3397            {                           /* Braces are required because the */
3398            GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3399            }
3400    
3401          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
3402          data and reset the pointer. This is so that very large classes that
3403          contain a zillion UTF-8 characters no longer overwrite the work space
3404          (which is on the stack). */
3405    
3406          if (lengthptr != NULL)
3407            {
3408            *lengthptr += class_utf8data - class_utf8data_base;
3409            class_utf8data = class_utf8data_base;
3410            }
3411    
3412    #endif
3413    
3414          /* Inside \Q...\E everything is literal except \E */
3415    
3416          if (inescq)
3417            {
3418            if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3419              {
3420              inescq = FALSE;                   /* Reset literal state */
3421              ptr++;                            /* Skip the 'E' */
3422              continue;                         /* Carry on with next */
3423              }
3424            goto CHECK_RANGE;                   /* Could be range if \E follows */
3425            }
3426    
3427          /* Handle POSIX class names. Perl allows a negation extension of the
3428          form [:^name:]. A square bracket that doesn't match the syntax is
3429          treated as a literal. We also recognize the POSIX constructions
3430          [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3431          5.6 and 5.8 do. */
3432    
3433          if (c == CHAR_LEFT_SQUARE_BRACKET &&
3434              (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3435               ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3436            {
3437            BOOL local_negate = FALSE;
3438            int posix_class, taboffset, tabopt;
3439            register const uschar *cbits = cd->cbits;
3440            uschar pbits[32];
3441    
3442            if (ptr[1] != CHAR_COLON)
3443              {
3444              *errorcodeptr = ERR31;
3445              goto FAILED;
3446              }
3447    
3448            ptr += 2;
3449            if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3450              {
3451              local_negate = TRUE;
3452              should_flip_negation = TRUE;  /* Note negative special */
3453              ptr++;
3454              }
3455    
3456            posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3457            if (posix_class < 0)
3458              {
3459              *errorcodeptr = ERR30;
3460              goto FAILED;
3461              }
3462    
3463            /* If matching is caseless, upper and lower are converted to
3464            alpha. This relies on the fact that the class table starts with
3465            alpha, lower, upper as the first 3 entries. */
3466    
3467            if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3468              posix_class = 0;
3469    
3470            /* When PCRE_UCP is set, some of the POSIX classes are converted to
3471            different escape sequences that use Unicode properties. */
3472    
3473    #ifdef SUPPORT_UCP
3474            if ((options & PCRE_UCP) != 0)
3475              {
3476              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3477              if (posix_substitutes[pc] != NULL)
3478                {
3479                nestptr = tempptr + 1;
3480                ptr = posix_substitutes[pc] - 1;
3481                continue;
3482                }
3483              }
3484    #endif
3485            /* In the non-UCP case, we build the bit map for the POSIX class in a
3486            chunk of local store because we may be adding and subtracting from it,
3487            and we don't want to subtract bits that may be in the main map already.
3488            At the end we or the result into the bit map that is being built. */
3489    
3490            posix_class *= 3;
3491    
3492            /* Copy in the first table (always present) */
3493    
3494            memcpy(pbits, cbits + posix_class_maps[posix_class],
3495              32 * sizeof(uschar));
3496    
3497            /* If there is a second table, add or remove it as required. */
3498    
3499            taboffset = posix_class_maps[posix_class + 1];
3500            tabopt = posix_class_maps[posix_class + 2];
3501    
3502            if (taboffset >= 0)
3503              {
3504              if (tabopt >= 0)
3505                for (c = 0; c < 32; c++) pbits[c] |= cbits[c + taboffset];
3506              else
3507                for (c = 0; c < 32; c++) pbits[c] &= ~cbits[c + taboffset];
3508              }
3509    
3510            /* Not see if we need to remove any special characters. An option
3511            value of 1 removes vertical space and 2 removes underscore. */
3512    
3513            if (tabopt < 0) tabopt = -tabopt;
3514            if (tabopt == 1) pbits[1] &= ~0x3c;
3515              else if (tabopt == 2) pbits[11] &= 0x7f;
3516    
3517            /* Add the POSIX table or its complement into the main table that is
3518            being built and we are done. */
3519    
3520            if (local_negate)
3521              for (c = 0; c < 32; c++) classbits[c] |= ~pbits[c];
3522            else
3523              for (c = 0; c < 32; c++) classbits[c] |= pbits[c];
3524    
3525            ptr = tempptr + 1;
3526            class_charcount = 10;  /* Set > 1; assumes more than 1 per class */
3527            continue;    /* End of POSIX syntax handling */
3528            }
3529    
3530          /* Backslash may introduce a single character, or it may introduce one
3531          of the specials, which just set a flag. The sequence \b is a special
3532          case. Inside a class (and only there) it is treated as backspace. We
3533          assume that other escapes have more than one character in them, so set
3534          class_charcount bigger than one. Unrecognized escapes fall through and
3535          are either treated as literal characters (by default), or are faulted if
3536          PCRE_EXTRA is set. */
3537    
3538          if (c == CHAR_BACKSLASH)
3539            {
3540            c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3541            if (*errorcodeptr != 0) goto FAILED;
3542    
3543            if (-c == ESC_b) c = CHAR_BS;    /* \b is backspace in a class */
3544            else if (-c == ESC_Q)            /* Handle start of quoted string */
3545              {
3546              if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3547                {
3548                ptr += 2; /* avoid empty string */
3549                }
3550              else inescq = TRUE;
3551              continue;
3552              }
3553            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
3554    
3555            if (c < 0)
3556              {
3557              register const uschar *cbits = cd->cbits;
3558