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

Diff of /code/trunk/pcre_compile.c

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

revision 85 by nigel, Sat Feb 24 21:41:13 2007 UTC revision 745 by ph10, Mon Nov 14 11:41:03 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 DEBUG is defined, we need the pcre_printint() function, which is also  /* When PCRE_DEBUG is defined, we need the pcre_printint() function, which is
57  used by pcretest. DEBUG is not defined when building a production library. */  also used by pcretest. PCRE_DEBUG is not defined when building a production
58    library. */
59    
60  #ifdef DEBUG  #ifdef PCRE_DEBUG
61  #include "pcre_printint.src"  #include "pcre_printint.src"
62  #endif  #endif
63    
64    
65    /* Macro for setting individual bits in class bitmaps. */
66    
67    #define SETBIT(a,b) a[b/8] |= (1 << (b%8))
68    
69    /* Maximum length value to check against when making sure that the integer that
70    holds the compiled pattern length does not overflow. We make it a bit less than
71    INT_MAX to allow for adding in group terminating bytes, so that we don't have
72    to check them every time. */
73    
74    #define OFLOW_MAX (INT_MAX - 20)
75    
76    
77  /*************************************************  /*************************************************
78  *      Code parameters and static tables         *  *      Code parameters and static tables         *
79  *************************************************/  *************************************************/
80    
81  /* Maximum number of items on the nested bracket stacks at compile time. This  /* This value specifies the size of stack workspace that is used during the
82  applies to the nesting of all kinds of parentheses. It does not limit  first pre-compile phase that determines how much memory is required. The regex
83  un-nested, non-capturing parentheses. This number can be made bigger if  is partly compiled into this space, but the compiled parts are discarded as
84  necessary - it is used to dimension one int and one unsigned char vector at  soon as they can be, so that hopefully there will never be an overrun. The code
85  compile time. */  does, however, check for an overrun. The largest amount I've seen used is 218,
86    so this number is very generous.
87    
88    The same workspace is used during the second, actual compile phase for
89    remembering forward references to groups so that they can be filled in at the
90    end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE
91    is 4 there is plenty of room. */
92    
93    #define COMPILE_WORK_SIZE (4096)
94    
95  #define BRASTACK_SIZE 200  /* The overrun tests check for a slightly smaller size so that they detect the
96    overrun before it actually does run off the end of the data block. */
97    
98    #define WORK_SIZE_CHECK (COMPILE_WORK_SIZE - 100)
99    
100    
101  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
# Line 72  are simple data values; negative values Line 103  are simple data values; negative values
103  on. Zero means further processing is needed (for things like \x), or the escape  on. Zero means further processing is needed (for things like \x), or the escape
104  is invalid. */  is invalid. */
105    
106  #if !EBCDIC   /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC
107    
108    /* This is the "normal" table for ASCII systems or for EBCDIC systems running
109    in UTF-8 mode. */
110    
111  static const short int escapes[] = {  static const short int escapes[] = {
112       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,                       0,
113       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,                       0,
114     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */       0,                       0,
115       0,      0,      0,      0,      0,      0,      0,      0,   /* H - O */       0,                       0,
116  -ESC_P, -ESC_Q,      0, -ESC_S,      0,      0,      0, -ESC_W,   /* P - W */       0,                       0,
117  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */       CHAR_COLON,              CHAR_SEMICOLON,
118     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
119       0,      0,      0,      0,      0,      0,  ESC_n,      0,   /* h - o */       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
120  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0,      0, -ESC_w,   /* p - w */       CHAR_COMMERCIAL_AT,      -ESC_A,
121       0,      0, -ESC_z                                            /* x - z */       -ESC_B,                  -ESC_C,
122         -ESC_D,                  -ESC_E,
123         0,                       -ESC_G,
124         -ESC_H,                  0,
125         0,                       -ESC_K,
126         0,                       0,
127         -ESC_N,                  0,
128         -ESC_P,                  -ESC_Q,
129         -ESC_R,                  -ESC_S,
130         0,                       0,
131         -ESC_V,                  -ESC_W,
132         -ESC_X,                  0,
133         -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,
134         CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,
135         CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,
136         CHAR_GRAVE_ACCENT,       7,
137         -ESC_b,                  0,
138         -ESC_d,                  ESC_e,
139         ESC_f,                   0,
140         -ESC_h,                  0,
141         0,                       -ESC_k,
142         0,                       0,
143         ESC_n,                   0,
144         -ESC_p,                  0,
145         ESC_r,                   -ESC_s,
146         ESC_tee,                 0,
147         -ESC_v,                  -ESC_w,
148         0,                       0,
149         -ESC_z
150  };  };
151    
152  #else         /* This is the "abnormal" table for EBCDIC systems */  #else
153    
154    /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
155    
156  static const short int escapes[] = {  static const short int escapes[] = {
157  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
158  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 96  static const short int escapes[] = { Line 162  static const short int escapes[] = {
162  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,
163  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',
164  /*  80 */     0,     7, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,  /*  80 */     0,     7, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,
165  /*  88 */     0,     0,      0,     '{',      0,     0,      0,      0,  /*  88 */-ESC_h,     0,      0,     '{',      0,     0,      0,      0,
166  /*  90 */     0,     0,      0,     'l',      0, ESC_n,      0, -ESC_p,  /*  90 */     0,     0, -ESC_k,     'l',      0, ESC_n,      0, -ESC_p,
167  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,
168  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,     0, -ESC_w,      0,  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,-ESC_v, -ESC_w,      0,
169  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,
170  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,
171  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
172  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,
173  /*  C8 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
174  /*  D0 */   '}',     0,      0,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,-ESC_N,      0, -ESC_P,
175  /*  D8 */-ESC_Q,     0,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
176  /*  E0 */  '\\',     0, -ESC_S,       0,      0,     0, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
177  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
178  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,
179  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0
# Line 115  static const short int escapes[] = { Line 181  static const short int escapes[] = {
181  #endif  #endif
182    
183    
184  /* Tables of names of POSIX character classes and their lengths. The list is  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
185  terminated by a zero length entry. The first three must be alpha, upper, lower,  searched linearly. Put all the names into a single string, in order to reduce
186  as this is assumed for handling case independence. */  the number of relocations when a shared library is dynamically linked. The
187    string is built from string macros so that it works in UTF-8 mode on EBCDIC
188  static const char *const posix_names[] = {  platforms. */
189    "alpha", "lower", "upper",  
190    "alnum", "ascii", "blank", "cntrl", "digit", "graph",  typedef struct verbitem {
191    "print", "punct", "space", "word",  "xdigit" };    int   len;                 /* Length of verb name */
192      int   op;                  /* Op when no arg, or -1 if arg mandatory */
193      int   op_arg;              /* Op when arg present, or -1 if not allowed */
194    } verbitem;
195    
196    static const char verbnames[] =
197      "\0"                       /* Empty name is a shorthand for MARK */
198      STRING_MARK0
199      STRING_ACCEPT0
200      STRING_COMMIT0
201      STRING_F0
202      STRING_FAIL0
203      STRING_PRUNE0
204      STRING_SKIP0
205      STRING_THEN;
206    
207    static const verbitem verbs[] = {
208      { 0, -1,        OP_MARK },
209      { 4, -1,        OP_MARK },
210      { 6, OP_ACCEPT, -1 },
211      { 6, OP_COMMIT, -1 },
212      { 1, OP_FAIL,   -1 },
213      { 4, OP_FAIL,   -1 },
214      { 5, OP_PRUNE,  OP_PRUNE_ARG },
215      { 4, OP_SKIP,   OP_SKIP_ARG  },
216      { 4, OP_THEN,   OP_THEN_ARG  }
217    };
218    
219    static const int verbcount = sizeof(verbs)/sizeof(verbitem);
220    
221    
222    /* Tables of names of POSIX character classes and their lengths. The names are
223    now all in a single string, to reduce the number of relocations when a shared
224    library is dynamically loaded. The list of lengths is terminated by a zero
225    length entry. The first three must be alpha, lower, upper, as this is assumed
226    for handling case independence. */
227    
228    static const char posix_names[] =
229      STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
230      STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
231      STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
232      STRING_word0  STRING_xdigit;
233    
234  static const uschar posix_name_lengths[] = {  static const uschar posix_name_lengths[] = {
235    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
236    
237  /* Table of class bit maps for each POSIX class; up to three may be combined  /* Table of class bit maps for each POSIX class. Each class is formed from a
238  to form the class. The table for [:blank:] is dynamically modified to remove  base map, with an optional addition or removal of another map. Then, for some
239  the vertical space characters. */  classes, there is some additional tweaking: for [:blank:] the vertical space
240    characters are removed, and for [:alpha:] and [:alnum:] the underscore
241    character is removed. The triples in the table consist of the base map offset,
242    second map offset or -1 if no second map, and a non-negative value for map
243    addition or a negative value for map subtraction (if there are two maps). The
244    absolute value of the third field has these meanings: 0 => no tweaking, 1 =>
245    remove vertical space characters, 2 => remove underscore. */
246    
247  static const int posix_class_maps[] = {  static const int posix_class_maps[] = {
248    cbit_lower, cbit_upper, -1,             /* alpha */    cbit_word,  cbit_digit, -2,             /* alpha */
249    cbit_lower, -1,         -1,             /* lower */    cbit_lower, -1,          0,             /* lower */
250    cbit_upper, -1,         -1,             /* upper */    cbit_upper, -1,          0,             /* upper */
251    cbit_digit, cbit_lower, cbit_upper,     /* alnum */    cbit_word,  -1,          2,             /* alnum - word without underscore */
252    cbit_print, cbit_cntrl, -1,             /* ascii */    cbit_print, cbit_cntrl,  0,             /* ascii */
253    cbit_space, -1,         -1,             /* blank - a GNU extension */    cbit_space, -1,          1,             /* blank - a GNU extension */
254    cbit_cntrl, -1,         -1,             /* cntrl */    cbit_cntrl, -1,          0,             /* cntrl */
255    cbit_digit, -1,         -1,             /* digit */    cbit_digit, -1,          0,             /* digit */
256    cbit_graph, -1,         -1,             /* graph */    cbit_graph, -1,          0,             /* graph */
257    cbit_print, -1,         -1,             /* print */    cbit_print, -1,          0,             /* print */
258    cbit_punct, -1,         -1,             /* punct */    cbit_punct, -1,          0,             /* punct */
259    cbit_space, -1,         -1,             /* space */    cbit_space, -1,          0,             /* space */
260    cbit_word,  -1,         -1,             /* word - a Perl extension */    cbit_word,  -1,          0,             /* word - a Perl extension */
261    cbit_xdigit,-1,         -1              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
262  };  };
263    
264    /* Table of substitutes for \d etc when PCRE_UCP is set. The POSIX class
265    substitutes must be in the order of the names, defined above, and there are
266    both positive and negative cases. NULL means no substitute. */
267    
268  /* The texts of compile-time error messages. These are "char *" because they  #ifdef SUPPORT_UCP
269  are passed to the outside world. */  static const uschar *substitutes[] = {
270      (uschar *)"\\P{Nd}",    /* \D */
271      (uschar *)"\\p{Nd}",    /* \d */
272      (uschar *)"\\P{Xsp}",   /* \S */       /* NOTE: Xsp is Perl space */
273      (uschar *)"\\p{Xsp}",   /* \s */
274      (uschar *)"\\P{Xwd}",   /* \W */
275      (uschar *)"\\p{Xwd}"    /* \w */
276    };
277    
278    static const uschar *posix_substitutes[] = {
279      (uschar *)"\\p{L}",     /* alpha */
280      (uschar *)"\\p{Ll}",    /* lower */
281      (uschar *)"\\p{Lu}",    /* upper */
282      (uschar *)"\\p{Xan}",   /* alnum */
283      NULL,                   /* ascii */
284      (uschar *)"\\h",        /* blank */
285      NULL,                   /* cntrl */
286      (uschar *)"\\p{Nd}",    /* digit */
287      NULL,                   /* graph */
288      NULL,                   /* print */
289      NULL,                   /* punct */
290      (uschar *)"\\p{Xps}",   /* space */    /* NOTE: Xps is POSIX space */
291      (uschar *)"\\p{Xwd}",   /* word */
292      NULL,                   /* xdigit */
293      /* Negated cases */
294      (uschar *)"\\P{L}",     /* ^alpha */
295      (uschar *)"\\P{Ll}",    /* ^lower */
296      (uschar *)"\\P{Lu}",    /* ^upper */
297      (uschar *)"\\P{Xan}",   /* ^alnum */
298      NULL,                   /* ^ascii */
299      (uschar *)"\\H",        /* ^blank */
300      NULL,                   /* ^cntrl */
301      (uschar *)"\\P{Nd}",    /* ^digit */
302      NULL,                   /* ^graph */
303      NULL,                   /* ^print */
304      NULL,                   /* ^punct */
305      (uschar *)"\\P{Xps}",   /* ^space */   /* NOTE: Xps is POSIX space */
306      (uschar *)"\\P{Xwd}",   /* ^word */
307      NULL                    /* ^xdigit */
308    };
309    #define POSIX_SUBSIZE (sizeof(posix_substitutes)/sizeof(uschar *))
310    #endif
311    
312  static const char *error_texts[] = {  #define STRING(a)  # a
313    "no error",  #define XSTRING(s) STRING(s)
314    "\\ at end of pattern",  
315    "\\c at end of pattern",  /* The texts of compile-time error messages. These are "char *" because they
316    "unrecognized character follows \\",  are passed to the outside world. Do not ever re-use any error number, because
317    "numbers out of order in {} quantifier",  they are documented. Always add a new error instead. Messages marked DEAD below
318    are no longer used. This used to be a table of strings, but in order to reduce
319    the number of relocations needed when a shared library is loaded dynamically,
320    it is now one long string. We cannot use a table of offsets, because the
321    lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
322    simply count through to the one we want - this isn't a performance issue
323    because these strings are used only when there is a compilation error.
324    
325    Each substring ends with \0 to insert a null character. This includes the final
326    substring, so that the whole string ends with \0\0, which can be detected when
327    counting through. */
328    
329    static const char error_texts[] =
330      "no error\0"
331      "\\ at end of pattern\0"
332      "\\c at end of pattern\0"
333      "unrecognized character follows \\\0"
334      "numbers out of order in {} quantifier\0"
335    /* 5 */    /* 5 */
336    "number too big in {} quantifier",    "number too big in {} quantifier\0"
337    "missing terminating ] for character class",    "missing terminating ] for character class\0"
338    "invalid escape sequence in character class",    "invalid escape sequence in character class\0"
339    "range out of order in character class",    "range out of order in character class\0"
340    "nothing to repeat",    "nothing to repeat\0"
341    /* 10 */    /* 10 */
342    "operand of unlimited repeat could match the empty string",    "operand of unlimited repeat could match the empty string\0"  /** DEAD **/
343    "internal error: unexpected repeat",    "internal error: unexpected repeat\0"
344    "unrecognized character after (?",    "unrecognized character after (? or (?-\0"
345    "POSIX named classes are supported only within a class",    "POSIX named classes are supported only within a class\0"
346    "missing )",    "missing )\0"
347    /* 15 */    /* 15 */
348    "reference to non-existent subpattern",    "reference to non-existent subpattern\0"
349    "erroffset passed as NULL",    "erroffset passed as NULL\0"
350    "unknown option bit(s) set",    "unknown option bit(s) set\0"
351    "missing ) after comment",    "missing ) after comment\0"
352    "parentheses nested too deeply",    "parentheses nested too deeply\0"  /** DEAD **/
353    /* 20 */    /* 20 */
354    "regular expression too large",    "regular expression is too large\0"
355    "failed to get memory",    "failed to get memory\0"
356    "unmatched parentheses",    "unmatched parentheses\0"
357    "internal error: code overflow",    "internal error: code overflow\0"
358    "unrecognized character after (?<",    "unrecognized character after (?<\0"
359    /* 25 */    /* 25 */
360    "lookbehind assertion is not fixed length",    "lookbehind assertion is not fixed length\0"
361    "malformed number after (?(",    "malformed number or name after (?(\0"
362    "conditional group contains more than two branches",    "conditional group contains more than two branches\0"
363    "assertion expected after (?(",    "assertion expected after (?(\0"
364    "(?R or (?digits must be followed by )",    "(?R or (?[+-]digits must be followed by )\0"
365    /* 30 */    /* 30 */
366    "unknown POSIX class name",    "unknown POSIX class name\0"
367    "POSIX collating elements are not supported",    "POSIX collating elements are not supported\0"
368    "this version of PCRE is not compiled with PCRE_UTF8 support",    "this version of PCRE is not compiled with PCRE_UTF8 support\0"
369    "spare error",    "spare error\0"  /** DEAD **/
370    "character value in \\x{...} sequence is too large",    "character value in \\x{...} sequence is too large\0"
371    /* 35 */    /* 35 */
372    "invalid condition (?(0)",    "invalid condition (?(0)\0"
373    "\\C not allowed in lookbehind assertion",    "\\C not allowed in lookbehind assertion\0"
374    "PCRE does not support \\L, \\l, \\N, \\U, or \\u",    "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
375    "number after (?C is > 255",    "number after (?C is > 255\0"
376    "closing ) for (?C expected",    "closing ) for (?C expected\0"
377    /* 40 */    /* 40 */
378    "recursive call could loop indefinitely",    "recursive call could loop indefinitely\0"
379    "unrecognized character after (?P",    "unrecognized character after (?P\0"
380    "syntax error after (?P",    "syntax error in subpattern name (missing terminator)\0"
381    "two named groups have the same name",    "two named subpatterns have the same name\0"
382    "invalid UTF-8 string",    "invalid UTF-8 string\0"
383    /* 45 */    /* 45 */
384    "support for \\P, \\p, and \\X has not been compiled",    "support for \\P, \\p, and \\X has not been compiled\0"
385    "malformed \\P or \\p sequence",    "malformed \\P or \\p sequence\0"
386    "unknown property name after \\P or \\p"    "unknown property name after \\P or \\p\0"
387  };    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
388      "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
389      /* 50 */
390      "repeated subpattern is too long\0"    /** DEAD **/
391      "octal value is greater than \\377 (not in UTF-8 mode)\0"
392      "internal error: overran compiling workspace\0"
393      "internal error: previously-checked referenced subpattern not found\0"
394      "DEFINE group contains more than one branch\0"
395      /* 55 */
396      "repeating a DEFINE group is not allowed\0"  /** DEAD **/
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      "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
413      ;
414    
415  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
416  patterns. Note that the tables in chartables are dependent on the locale, and  patterns. Note that the tables in chartables are dependent on the locale, and
# Line 229  For convenience, we use the same bit def Line 428  For convenience, we use the same bit def
428    
429  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
430    
431  #if !EBCDIC    /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC
432    
433    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
434    UTF-8 mode. */
435    
436  static const unsigned char digitab[] =  static const unsigned char digitab[] =
437    {    {
438    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
# Line 265  static const unsigned char digitab[] = Line 468  static const unsigned char digitab[] =
468    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
469    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
470    
471  #else          /* This is the "abnormal" case, for EBCDIC systems */  #else
472    
473    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
474    
475  static const unsigned char digitab[] =  static const unsigned char digitab[] =
476    {    {
477    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
# Line 279  static const unsigned char digitab[] = Line 485  static const unsigned char digitab[] =
485    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */
486    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */
487    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */
488    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88-     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88- 95    */
489    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */
490    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */
491    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */
# Line 313  static const unsigned char ebcdic_charta Line 519  static const unsigned char ebcdic_charta
519    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */
520    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */
521    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */
522    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88-  */    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88- 95 */
523    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */
524    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */
525    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */
# Line 340  static const unsigned char ebcdic_charta Line 546  static const unsigned char ebcdic_charta
546  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
547    
548  static BOOL  static BOOL
549    compile_regex(int, int, int *, uschar **, const uschar **, int *, BOOL, int,    compile_regex(int, uschar **, const uschar **, int *, BOOL, BOOL, int, int,
550      int *, int *, branch_chain *, compile_data *);      int *, int *, branch_chain *, compile_data *, int *);
551    
552    
553    
554    /*************************************************
555    *            Find an error text                  *
556    *************************************************/
557    
558    /* The error texts are now all in one long string, to save on relocations. As
559    some of the text is of unknown length, we can't use a table of offsets.
560    Instead, just count through the strings. This is not a performance issue
561    because it happens only when there has been a compilation error.
562    
563    Argument:   the error number
564    Returns:    pointer to the error string
565    */
566    
567    static const char *
568    find_error_text(int n)
569    {
570    const char *s = error_texts;
571    for (; n > 0; n--)
572      {
573      while (*s++ != 0) {};
574      if (*s == 0) return "Error text not found (please report)";
575      }
576    return s;
577    }
578    
579    
580    /*************************************************
581    *            Check for counted repeat            *
582    *************************************************/
583    
584    /* This function is called when a '{' is encountered in a place where it might
585    start a quantifier. It looks ahead to see if it really is a quantifier or not.
586    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
587    where the ddds are digits.
588    
589    Arguments:
590      p         pointer to the first char after '{'
591    
592    Returns:    TRUE or FALSE
593    */
594    
595    static BOOL
596    is_counted_repeat(const uschar *p)
597    {
598    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
599    while ((digitab[*p] & ctype_digit) != 0) p++;
600    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
601    
602    if (*p++ != CHAR_COMMA) return FALSE;
603    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
604    
605    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
606    while ((digitab[*p] & ctype_digit) != 0) p++;
607    
608    return (*p == CHAR_RIGHT_CURLY_BRACKET);
609    }
610    
611    
612    
# Line 351  static BOOL Line 616  static BOOL
616    
617  /* 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
618  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
619  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
620  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
621  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,
622    ptr is pointing at the \. On exit, it is on the final character of the escape
623    sequence.
624    
625  Arguments:  Arguments:
626    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
# Line 364  Arguments: Line 631  Arguments:
631    
632  Returns:         zero or positive => a data character  Returns:         zero or positive => a data character
633                   negative => a special escape sequence                   negative => a special escape sequence
634                   on error, errorptr is set                   on error, errorcodeptr is set
635  */  */
636    
637  static int  static int
638  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,
639    int options, BOOL isclass)    int options, BOOL isclass)
640  {  {
641  const uschar *ptr = *ptrptr;  BOOL utf8 = (options & PCRE_UTF8) != 0;
642    const uschar *ptr = *ptrptr + 1;
643  int c, i;  int c, i;
644    
645    GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
646    ptr--;                            /* Set pointer back to the last byte */
647    
648  /* 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. */
649    
 c = *(++ptr);  
650  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
651    
652  /* 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
653  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.
654  Otherwise further processing may be required. */  Otherwise further processing may be required. */
655    
656  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
657  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */
658  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0) c = i;
659    
660  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
661  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */
662  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
663  #endif  #endif
664    
# Line 397  else if ((i = escapes[c - 0x48]) != 0) Line 667  else if ((i = escapes[c - 0x48]) != 0)
667  else  else
668    {    {
669    const uschar *oldptr;    const uschar *oldptr;
670      BOOL braced, negated;
671    
672    switch (c)    switch (c)
673      {      {
674      /* 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
675      error. */      error. */
676    
677      case 'l':      case CHAR_l:
678      case 'L':      case CHAR_L:
     case 'N':  
     case 'u':  
     case 'U':  
679      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
680      break;      break;
681    
682        case CHAR_u:
683        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
684          {
685          /* In JavaScript, \u must be followed by four hexadecimal numbers.
686          Otherwise it is a lowercase u letter. */
687          if ((digitab[ptr[1]] & ctype_xdigit) != 0 && (digitab[ptr[2]] & ctype_xdigit) != 0
688               && (digitab[ptr[3]] & ctype_xdigit) != 0 && (digitab[ptr[4]] & ctype_xdigit) != 0)
689            {
690            c = 0;
691            for (i = 0; i < 4; ++i)
692              {
693              register int cc = *(++ptr);
694    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
695              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
696              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
697    #else           /* EBCDIC coding */
698              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
699              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
700    #endif
701              }
702            }
703          }
704        else
705          *errorcodeptr = ERR37;
706        break;
707    
708        case CHAR_U:
709        /* In JavaScript, \U is an uppercase U letter. */
710        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
711        break;
712    
713        /* In a character class, \g is just a literal "g". Outside a character
714        class, \g must be followed by one of a number of specific things:
715    
716        (1) A number, either plain or braced. If positive, it is an absolute
717        backreference. If negative, it is a relative backreference. This is a Perl
718        5.10 feature.
719    
720        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
721        is part of Perl's movement towards a unified syntax for back references. As
722        this is synonymous with \k{name}, we fudge it up by pretending it really
723        was \k.
724    
725        (3) For Oniguruma compatibility we also support \g followed by a name or a
726        number either in angle brackets or in single quotes. However, these are
727        (possibly recursive) subroutine calls, _not_ backreferences. Just return
728        the -ESC_g code (cf \k). */
729    
730        case CHAR_g:
731        if (isclass) break;
732        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
733          {
734          c = -ESC_g;
735          break;
736          }
737    
738        /* Handle the Perl-compatible cases */
739    
740        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
741          {
742          const uschar *p;
743          for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
744            if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
745          if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
746            {
747            c = -ESC_k;
748            break;
749            }
750          braced = TRUE;
751          ptr++;
752          }
753        else braced = FALSE;
754    
755        if (ptr[1] == CHAR_MINUS)
756          {
757          negated = TRUE;
758          ptr++;
759          }
760        else negated = FALSE;
761    
762        c = 0;
763        while ((digitab[ptr[1]] & ctype_digit) != 0)
764          c = c * 10 + *(++ptr) - CHAR_0;
765    
766        if (c < 0)   /* Integer overflow */
767          {
768          *errorcodeptr = ERR61;
769          break;
770          }
771    
772        if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
773          {
774          *errorcodeptr = ERR57;
775          break;
776          }
777    
778        if (c == 0)
779          {
780          *errorcodeptr = ERR58;
781          break;
782          }
783    
784        if (negated)
785          {
786          if (c > bracount)
787            {
788            *errorcodeptr = ERR15;
789            break;
790            }
791          c = bracount - (c - 1);
792          }
793    
794        c = -(ESC_REF + c);
795        break;
796    
797      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
798      starting with one that is not zero is not straightforward. By experiment,      starting with one that is not zero is not straightforward. By experiment,
799      the way Perl works seems to be as follows:      the way Perl works seems to be as follows:
# Line 422  else Line 806  else
806      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
807      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
808    
809      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:
810      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
811    
812      if (!isclass)      if (!isclass)
813        {        {
814        oldptr = ptr;        oldptr = ptr;
815        c -= '0';        c -= CHAR_0;
816        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
817          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - CHAR_0;
818          if (c < 0)    /* Integer overflow */
819            {
820            *errorcodeptr = ERR61;
821            break;
822            }
823        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
824          {          {
825          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 443  else Line 832  else
832      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.
833      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
834    
835      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
836        {        {
837        ptr--;        ptr--;
838        c = 0;        c = 0;
# Line 451  else Line 840  else
840        }        }
841    
842      /* \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
843      larger first octal digit. */      larger first octal digit. The original code used just to take the least
844        significant 8 bits of octal numbers (I think this is what early Perls used
845      case '0':      to do). Nowadays we allow for larger numbers in UTF-8 mode, but no more
846      c -= '0';      than 3 octal digits. */
847      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')  
848          c = c * 8 + *(++ptr) - '0';      case CHAR_0:
849      c &= 255;     /* Take least significant 8 bits */      c -= CHAR_0;
850        while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
851            c = c * 8 + *(++ptr) - CHAR_0;
852        if (!utf8 && c > 255) *errorcodeptr = ERR51;
853      break;      break;
854    
855      /* \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
856      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
857        treated as a data character. */
858    
859        case CHAR_x:
860        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
861          {
862          /* In JavaScript, \x must be followed by two hexadecimal numbers.
863          Otherwise it is a lowercase x letter. */
864          if ((digitab[ptr[1]] & ctype_xdigit) != 0 && (digitab[ptr[2]] & ctype_xdigit) != 0)
865            {
866            c = 0;
867            for (i = 0; i < 2; ++i)
868              {
869              register int cc = *(++ptr);
870    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
871              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
872              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
873    #else           /* EBCDIC coding */
874              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
875              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
876    #endif
877              }
878            }
879          break;
880          }
881    
882      case 'x':      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
 #ifdef SUPPORT_UTF8  
     if (ptr[1] == '{' && (options & PCRE_UTF8) != 0)  
883        {        {
884        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
885        register int count = 0;        int count = 0;
886    
887        c = 0;        c = 0;
888        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
889          {          {
890          int cc = *pt++;          register int cc = *pt++;
891            if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
892          count++;          count++;
893  #if !EBCDIC    /* ASCII coding */  
894          if (cc >= 'a') cc -= 32;               /* Convert to upper case */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
895          c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
896  #else          /* EBCDIC coding */          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
897          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */  #else           /* EBCDIC coding */
898          c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
899            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
900  #endif  #endif
901          }          }
902        if (*pt == '}')  
903          if (*pt == CHAR_RIGHT_CURLY_BRACKET)
904          {          {
905          if (c < 0 || count > 8) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
906          ptr = pt;          ptr = pt;
907          break;          break;
908          }          }
909    
910        /* 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
911        recognize this construct; fall through to the normal \x handling. */        recognize this construct; fall through to the normal \x handling. */
912        }        }
 #endif  
913    
914      /* Read just a single hex char */      /* Read just a single-byte hex-defined char */
915    
916      c = 0;      c = 0;
917      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
918        {        {
919        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
920        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
921  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
922        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
923        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
924  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
925        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
926        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
927  #endif  #endif
928        }        }
929      break;      break;
930    
931      /* Other special escapes not starting with a digit are straightforward */      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
932        An error is given if the byte following \c is not an ASCII character. This
933        coding is ASCII-specific, but then the whole concept of \cx is
934        ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
935    
936      case 'c':      case CHAR_c:
937      c = *(++ptr);      c = *(++ptr);
938      if (c == 0)      if (c == 0)
939        {        {
940        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
941        return 0;        break;
942        }        }
943    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
944      /* A letter is upper-cased; then the 0x40 bit is flipped. This coding      if (c > 127)  /* Excludes all non-ASCII in either mode */
945      is ASCII-specific, but then the whole concept of \cx is ASCII-specific.        {
946      (However, an EBCDIC equivalent has now been added.) */        *errorcodeptr = ERR68;
947          break;
948  #if !EBCDIC    /* ASCII coding */        }
949      if (c >= 'a' && c <= 'z') c -= 32;      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
950      c ^= 0x40;      c ^= 0x40;
951  #else          /* EBCDIC coding */  #else             /* EBCDIC coding */
952      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
953      c ^= 0xC0;      c ^= 0xC0;
954  #endif  #endif
955      break;      break;
956    
957      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
958      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
959      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
960      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
961      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
962    
963      default:      default:
964      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 550  else Line 971  else
971      }      }
972    }    }
973    
974    /* Perl supports \N{name} for character names, as well as plain \N for "not
975    newline". PCRE does not support \N{name}. However, it does support
976    quantification such as \N{2,3}. */
977    
978    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
979         !is_counted_repeat(ptr+2))
980      *errorcodeptr = ERR37;
981    
982    /* If PCRE_UCP is set, we change the values for \d etc. */
983    
984    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
985      c -= (ESC_DU - ESC_D);
986    
987    /* Set the pointer to the final character before returning. */
988    
989  *ptrptr = ptr;  *ptrptr = ptr;
990  return c;  return c;
991  }  }
# Line 569  escape sequence. Line 1005  escape sequence.
1005  Argument:  Argument:
1006    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
1007    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
1008      dptr           points to an int that is set to the detailed property value
1009    errorcodeptr   points to the error code variable    errorcodeptr   points to the error code variable
1010    
1011  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
1012  */  */
1013    
1014  static int  static int
1015  get_ucp(const uschar **ptrptr, BOOL *negptr, int *errorcodeptr)  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)
1016  {  {
1017  int c, i, bot, top;  int c, i, bot, top;
1018  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
1019  char name[4];  char name[32];
1020    
1021  c = *(++ptr);  c = *(++ptr);
1022  if (c == 0) goto ERROR_RETURN;  if (c == 0) goto ERROR_RETURN;
1023    
1024  *negptr = FALSE;  *negptr = FALSE;
1025    
1026  /* \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
1027  preceded by ^ for negation. */  negation. */
1028    
1029  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
1030    {    {
1031    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1032      {      {
1033      *negptr = TRUE;      *negptr = TRUE;
1034      ptr++;      ptr++;
1035      }      }
1036    for (i = 0; i <= 2; i++)    for (i = 0; i < (int)sizeof(name) - 1; i++)
1037      {      {
1038      c = *(++ptr);      c = *(++ptr);
1039      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
1040      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1041      name[i] = c;      name[i] = c;
1042      }      }
1043    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;  
     }  
1044    name[i] = 0;    name[i] = 0;
1045    }    }
1046    
# Line 628  top = _pcre_utt_size; Line 1061  top = _pcre_utt_size;
1061    
1062  while (bot < top)  while (bot < top)
1063    {    {
1064    i = (bot + top)/2;    i = (bot + top) >> 1;
1065    c = strcmp(name, _pcre_utt[i].name);    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);
1066    if (c == 0) return _pcre_utt[i].value;    if (c == 0)
1067        {
1068        *dptr = _pcre_utt[i].value;
1069        return _pcre_utt[i].type;
1070        }
1071    if (c > 0) bot = i + 1; else top = i;    if (c > 0) bot = i + 1; else top = i;
1072    }    }
1073    
 UNKNOWN_RETURN:  
1074  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
1075  *ptrptr = ptr;  *ptrptr = ptr;
1076  return -1;  return -1;
# Line 650  return -1; Line 1086  return -1;
1086    
1087    
1088  /*************************************************  /*************************************************
 *            Check for counted repeat            *  
 *************************************************/  
   
 /* This function is called when a '{' is encountered in a place where it might  
 start a quantifier. It looks ahead to see if it really is a quantifier or not.  
 It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}  
 where the ddds are digits.  
   
 Arguments:  
   p         pointer to the first char after '{'  
   
 Returns:    TRUE or FALSE  
 */  
   
 static BOOL  
 is_counted_repeat(const uschar *p)  
 {  
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
 if (*p == '}') return TRUE;  
   
 if (*p++ != ',') return FALSE;  
 if (*p == '}') return TRUE;  
   
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
   
 return (*p == '}');  
 }  
   
   
   
 /*************************************************  
1089  *         Read repeat counts                     *  *         Read repeat counts                     *
1090  *************************************************/  *************************************************/
1091    
# Line 710  int max = -1; Line 1113  int max = -1;
1113  /* 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
1114  an integer overflow. */  an integer overflow. */
1115    
1116  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;
1117  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1118    {    {
1119    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 720  if (min < 0 || min > 65535) Line 1123  if (min < 0 || min > 65535)
1123  /* 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.
1124  Also, max must not be less than min. */  Also, max must not be less than min. */
1125    
1126  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1127    {    {
1128    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1129      {      {
1130      max = 0;      max = 0;
1131      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;
1132      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1133        {        {
1134        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 750  return p; Line 1153  return p;
1153    
1154    
1155  /*************************************************  /*************************************************
1156  *      Find first significant op code            *  *  Subroutine for finding forward reference      *
1157  *************************************************/  *************************************************/
1158    
1159  /* This is called by several functions that scan a compiled expression looking  /* This recursive function is called only from find_parens() below. The
1160  for a fixed first character, or an anchoring op code etc. It skips over things  top-level call starts at the beginning of the pattern. All other calls must
1161  that do not influence this. For some calls, a change of option is important.  start at a parenthesis. It scans along a pattern's text looking for capturing
1162  For some calls, it makes sense to skip negative forward and all backward  subpatterns, and counting them. If it finds a named pattern that matches the
1163  assertions, and also the \b assertion; for others it does not.  name it is given, it returns its number. Alternatively, if the name is NULL, it
1164    returns when it reaches a given numbered subpattern. Recursion is used to keep
1165    track of subpatterns that reset the capturing group numbers - the (?| feature.
1166    
1167    This function was originally called only from the second pass, in which we know
1168    that if (?< or (?' or (?P< is encountered, the name will be correctly
1169    terminated because that is checked in the first pass. There is now one call to
1170    this function in the first pass, to check for a recursive back reference by
1171    name (so that we can make the whole group atomic). In this case, we need check
1172    only up to the current position in the pattern, and that is still OK because
1173    and previous occurrences will have been checked. To make this work, the test
1174    for "end of pattern" is a check against cd->end_pattern in the main loop,
1175    instead of looking for a binary zero. This means that the special first-pass
1176    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1177    processing items within the loop are OK, because afterwards the main loop will
1178    terminate.)
1179    
1180  Arguments:  Arguments:
1181    code         pointer to the start of the group    ptrptr       address of the current character pointer (updated)
1182    options      pointer to external options    cd           compile background data
1183    optbit       the option bit whose changing is significant, or    name         name to seek, or NULL if seeking a numbered subpattern
1184                   zero if none are    lorn         name length, or subpattern number if name is NULL
1185    skipassert   TRUE if certain assertions are to be skipped    xmode        TRUE if we are in /x mode
1186      utf8         TRUE if we are in UTF-8 mode
1187      count        pointer to the current capturing subpattern number (updated)
1188    
1189  Returns:       pointer to the first significant opcode  Returns:       the number of the named subpattern, or -1 if not found
1190  */  */
1191    
1192  static const uschar*  static int
1193  first_significant_code(const uschar *code, int *options, int optbit,  find_parens_sub(uschar **ptrptr, compile_data *cd, const uschar *name, int lorn,
1194    BOOL skipassert)    BOOL xmode, BOOL utf8, int *count)
1195  {  {
1196  for (;;)  uschar *ptr = *ptrptr;
1197    {  int start_count = *count;
1198    switch ((int)*code)  int hwm_count = start_count;
1199      {  BOOL dup_parens = FALSE;
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
1200    
1201      case OP_ASSERT_NOT:  /* If the first character is a parenthesis, check on the type of group we are
1202      case OP_ASSERTBACK:  dealing with. The very first call may not start with a parenthesis. */
     case OP_ASSERTBACK_NOT:  
     if (!skipassert) return code;  
     do code += GET(code, 1); while (*code == OP_ALT);  
     code += _pcre_OP_lengths[*code];  
     break;  
1203    
1204      case OP_WORD_BOUNDARY:  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1205      case OP_NOT_WORD_BOUNDARY:    {
1206      if (!skipassert) return code;    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
     /* Fall through */  
1207    
1208      case OP_CALLOUT:    if (ptr[1] == CHAR_ASTERISK) ptr += 2;
     case OP_CREF:  
     case OP_BRANUMBER:  
     code += _pcre_OP_lengths[*code];  
     break;  
1209    
1210      default:    /* Handle a normal, unnamed capturing parenthesis. */
1211      return code;  
1212      else if (ptr[1] != CHAR_QUESTION_MARK)
1213        {
1214        *count += 1;
1215        if (name == NULL && *count == lorn) return *count;
1216        ptr++;
1217      }      }
   }  
 /* Control never reaches here */  
 }  
1218    
1219      /* All cases now have (? at the start. Remember when we are in a group
1220      where the parenthesis numbers are duplicated. */
1221    
1222      else if (ptr[2] == CHAR_VERTICAL_LINE)
1223        {
1224        ptr += 3;
1225        dup_parens = TRUE;
1226        }
1227    
1228      /* Handle comments; all characters are allowed until a ket is reached. */
1229    
1230  /*************************************************    else if (ptr[2] == CHAR_NUMBER_SIGN)
1231  *        Find the fixed length of a pattern      *      {
1232  *************************************************/      for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1233        goto FAIL_EXIT;
1234        }
1235    
1236  /* Scan a pattern and compute the fixed length of subject that will match it,    /* Handle a condition. If it is an assertion, just carry on so that it
1237  if the length is fixed. This is needed for dealing with backward assertions.    is processed as normal. If not, skip to the closing parenthesis of the
1238  In UTF8 mode, the result is in characters rather than bytes.    condition (there can't be any nested parens). */
1239    
1240  Arguments:    else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1241    code     points to the start of the pattern (the bracket)      {
1242    options  the compiling options      ptr += 2;
1243        if (ptr[1] != CHAR_QUESTION_MARK)
1244          {
1245          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1246          if (*ptr != 0) ptr++;
1247          }
1248        }
1249    
1250  Returns:   the fixed length, or -1 if there is no fixed length,    /* Start with (? but not a condition. */
              or -2 if \C was encountered  
 */  
1251    
1252  static int    else
1253  find_fixedlength(uschar *code, int options)      {
1254  {      ptr += 2;
1255  int length = -1;      if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1256    
1257  register int branchlength = 0;      /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
 register uschar *cc = code + 1 + LINK_SIZE;  
1258    
1259  /* Scan along the opcodes for this branch. If we get to the end of the      if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1260  branch, check the length against that of the other branches. */          ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1261          {
1262          int term;
1263          const uschar *thisname;
1264          *count += 1;
1265          if (name == NULL && *count == lorn) return *count;
1266          term = *ptr++;
1267          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1268          thisname = ptr;
1269          while (*ptr != term) ptr++;
1270          if (name != NULL && lorn == ptr - thisname &&
1271              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1272            return *count;
1273          term++;
1274          }
1275        }
1276      }
1277    
1278  for (;;)  /* Past any initial parenthesis handling, scan for parentheses or vertical
1279    bars. Stop if we get to cd->end_pattern. Note that this is important for the
1280    first-pass call when this value is temporarily adjusted to stop at the current
1281    position. So DO NOT change this to a test for binary zero. */
1282    
1283    for (; ptr < cd->end_pattern; ptr++)
1284    {    {
1285    int d;    /* Skip over backslashed characters and also entire \Q...\E */
1286    register int op = *cc;  
1287    if (op >= OP_BRA) op = OP_BRA;    if (*ptr == CHAR_BACKSLASH)
1288        {
1289        if (*(++ptr) == 0) goto FAIL_EXIT;
1290        if (*ptr == CHAR_Q) for (;;)
1291          {
1292          while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1293          if (*ptr == 0) goto FAIL_EXIT;
1294          if (*(++ptr) == CHAR_E) break;
1295          }
1296        continue;
1297        }
1298    
1299      /* Skip over character classes; this logic must be similar to the way they
1300      are handled for real. If the first character is '^', skip it. Also, if the
1301      first few characters (either before or after ^) are \Q\E or \E we skip them
1302      too. This makes for compatibility with Perl. Note the use of STR macros to
1303      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1304    
1305      if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1306        {
1307        BOOL negate_class = FALSE;
1308        for (;;)
1309          {
1310          if (ptr[1] == CHAR_BACKSLASH)
1311            {
1312            if (ptr[2] == CHAR_E)
1313              ptr+= 2;
1314            else if (strncmp((const char *)ptr+2,
1315                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1316              ptr += 4;
1317            else
1318              break;
1319            }
1320          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1321            {
1322            negate_class = TRUE;
1323            ptr++;
1324            }
1325          else break;
1326          }
1327    
1328        /* If the next character is ']', it is a data character that must be
1329        skipped, except in JavaScript compatibility mode. */
1330    
1331        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1332            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1333          ptr++;
1334    
1335        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1336          {
1337          if (*ptr == 0) return -1;
1338          if (*ptr == CHAR_BACKSLASH)
1339            {
1340            if (*(++ptr) == 0) goto FAIL_EXIT;
1341            if (*ptr == CHAR_Q) for (;;)
1342              {
1343              while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1344              if (*ptr == 0) goto FAIL_EXIT;
1345              if (*(++ptr) == CHAR_E) break;
1346              }
1347            continue;
1348            }
1349          }
1350        continue;
1351        }
1352    
1353      /* Skip comments in /x mode */
1354    
1355      if (xmode && *ptr == CHAR_NUMBER_SIGN)
1356        {
1357        ptr++;
1358        while (*ptr != 0)
1359          {
1360          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1361          ptr++;
1362    #ifdef SUPPORT_UTF8
1363          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
1364    #endif
1365          }
1366        if (*ptr == 0) goto FAIL_EXIT;
1367        continue;
1368        }
1369    
1370      /* Check for the special metacharacters */
1371    
1372      if (*ptr == CHAR_LEFT_PARENTHESIS)
1373        {
1374        int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, count);
1375        if (rc > 0) return rc;
1376        if (*ptr == 0) goto FAIL_EXIT;
1377        }
1378    
1379      else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1380        {
1381        if (dup_parens && *count < hwm_count) *count = hwm_count;
1382        goto FAIL_EXIT;
1383        }
1384    
1385      else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1386        {
1387        if (*count > hwm_count) hwm_count = *count;
1388        *count = start_count;
1389        }
1390      }
1391    
1392    FAIL_EXIT:
1393    *ptrptr = ptr;
1394    return -1;
1395    }
1396    
1397    
1398    
1399    
1400    /*************************************************
1401    *       Find forward referenced subpattern       *
1402    *************************************************/
1403    
1404    /* This function scans along a pattern's text looking for capturing
1405    subpatterns, and counting them. If it finds a named pattern that matches the
1406    name it is given, it returns its number. Alternatively, if the name is NULL, it
1407    returns when it reaches a given numbered subpattern. This is used for forward
1408    references to subpatterns. We used to be able to start this scan from the
1409    current compiling point, using the current count value from cd->bracount, and
1410    do it all in a single loop, but the addition of the possibility of duplicate
1411    subpattern numbers means that we have to scan from the very start, in order to
1412    take account of such duplicates, and to use a recursive function to keep track
1413    of the different types of group.
1414    
1415    Arguments:
1416      cd           compile background data
1417      name         name to seek, or NULL if seeking a numbered subpattern
1418      lorn         name length, or subpattern number if name is NULL
1419      xmode        TRUE if we are in /x mode
1420      utf8         TRUE if we are in UTF-8 mode
1421    
1422    Returns:       the number of the found subpattern, or -1 if not found
1423    */
1424    
1425    static int
1426    find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode,
1427      BOOL utf8)
1428    {
1429    uschar *ptr = (uschar *)cd->start_pattern;
1430    int count = 0;
1431    int rc;
1432    
1433    /* If the pattern does not start with an opening parenthesis, the first call
1434    to find_parens_sub() will scan right to the end (if necessary). However, if it
1435    does start with a parenthesis, find_parens_sub() will return when it hits the
1436    matching closing parens. That is why we have to have a loop. */
1437    
1438    for (;;)
1439      {
1440      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, &count);
1441      if (rc > 0 || *ptr++ == 0) break;
1442      }
1443    
1444    return rc;
1445    }
1446    
1447    
1448    
1449    
1450    /*************************************************
1451    *      Find first significant op code            *
1452    *************************************************/
1453    
1454    /* This is called by several functions that scan a compiled expression looking
1455    for a fixed first character, or an anchoring op code etc. It skips over things
1456    that do not influence this. For some calls, it makes sense to skip negative
1457    forward and all backward assertions, and also the \b assertion; for others it
1458    does not.
1459    
1460    Arguments:
1461      code         pointer to the start of the group
1462      skipassert   TRUE if certain assertions are to be skipped
1463    
1464    Returns:       pointer to the first significant opcode
1465    */
1466    
1467    static const uschar*
1468    first_significant_code(const uschar *code, BOOL skipassert)
1469    {
1470    for (;;)
1471      {
1472      switch ((int)*code)
1473        {
1474        case OP_ASSERT_NOT:
1475        case OP_ASSERTBACK:
1476        case OP_ASSERTBACK_NOT:
1477        if (!skipassert) return code;
1478        do code += GET(code, 1); while (*code == OP_ALT);
1479        code += _pcre_OP_lengths[*code];
1480        break;
1481    
1482        case OP_WORD_BOUNDARY:
1483        case OP_NOT_WORD_BOUNDARY:
1484        if (!skipassert) return code;
1485        /* Fall through */
1486    
1487        case OP_CALLOUT:
1488        case OP_CREF:
1489        case OP_NCREF:
1490        case OP_RREF:
1491        case OP_NRREF:
1492        case OP_DEF:
1493        code += _pcre_OP_lengths[*code];
1494        break;
1495    
1496        default:
1497        return code;
1498        }
1499      }
1500    /* Control never reaches here */
1501    }
1502    
1503    
1504    
1505    
1506    /*************************************************
1507    *        Find the fixed length of a branch       *
1508    *************************************************/
1509    
1510    /* Scan a branch and compute the fixed length of subject that will match it,
1511    if the length is fixed. This is needed for dealing with backward assertions.
1512    In UTF8 mode, the result is in characters rather than bytes. The branch is
1513    temporarily terminated with OP_END when this function is called.
1514    
1515    This function is called when a backward assertion is encountered, so that if it
1516    fails, the error message can point to the correct place in the pattern.
1517    However, we cannot do this when the assertion contains subroutine calls,
1518    because they can be forward references. We solve this by remembering this case
1519    and doing the check at the end; a flag specifies which mode we are running in.
1520    
1521    Arguments:
1522      code     points to the start of the pattern (the bracket)
1523      utf8     TRUE in UTF-8 mode
1524      atend    TRUE if called when the pattern is complete
1525      cd       the "compile data" structure
1526    
1527    Returns:   the fixed length,
1528                 or -1 if there is no fixed length,
1529                 or -2 if \C was encountered
1530                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1531    */
1532    
1533    static int
1534    find_fixedlength(uschar *code, BOOL utf8, BOOL atend, compile_data *cd)
1535    {
1536    int length = -1;
1537    
1538    register int branchlength = 0;
1539    register uschar *cc = code + 1 + LINK_SIZE;
1540    
1541    /* Scan along the opcodes for this branch. If we get to the end of the
1542    branch, check the length against that of the other branches. */
1543    
1544    for (;;)
1545      {
1546      int d;
1547      uschar *ce, *cs;
1548      register int op = *cc;
1549    switch (op)    switch (op)
1550      {      {
1551        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1552        OP_BRA (normal non-capturing bracket) because the other variants of these
1553        opcodes are all concerned with unlimited repeated groups, which of course
1554        are not of fixed length. They will cause a -1 response from the default
1555        case of this switch. */
1556    
1557        case OP_CBRA:
1558      case OP_BRA:      case OP_BRA:
1559      case OP_ONCE:      case OP_ONCE:
1560        case OP_ONCE_NC:
1561      case OP_COND:      case OP_COND:
1562      d = find_fixedlength(cc, options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), utf8, atend, cd);
1563      if (d < 0) return d;      if (d < 0) return d;
1564      branchlength += d;      branchlength += d;
1565      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 859  for (;;) Line 1568  for (;;)
1568    
1569      /* Reached end of a branch; if it's a ket it is the end of a nested      /* Reached end of a branch; if it's a ket it is the end of a nested
1570      call. If it's ALT it is an alternation in a nested call. If it is      call. If it's ALT it is an alternation in a nested call. If it is
1571      END it's the end of the outer call. All can be handled by the same code. */      END it's the end of the outer call. All can be handled by the same code.
1572        Note that we must not include the OP_KETRxxx opcodes here, because they
1573        all imply an unlimited repeat. */
1574    
1575      case OP_ALT:      case OP_ALT:
1576      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1577      case OP_END:      case OP_END:
1578      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1579        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
# Line 873  for (;;) Line 1582  for (;;)
1582      branchlength = 0;      branchlength = 0;
1583      break;      break;
1584    
1585        /* A true recursion implies not fixed length, but a subroutine call may
1586        be OK. If the subroutine is a forward reference, we can't deal with
1587        it until the end of the pattern, so return -3. */
1588    
1589        case OP_RECURSE:
1590        if (!atend) return -3;
1591        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1592        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1593        if (cc > cs && cc < ce) return -1;                /* Recursion */
1594        d = find_fixedlength(cs + 2, utf8, atend, cd);
1595        if (d < 0) return d;
1596        branchlength += d;
1597        cc += 1 + LINK_SIZE;
1598        break;
1599    
1600      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1601    
1602      case OP_ASSERT:      case OP_ASSERT:
# Line 885  for (;;) Line 1609  for (;;)
1609      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1610    
1611      case OP_REVERSE:      case OP_REVERSE:
     case OP_BRANUMBER:  
1612      case OP_CREF:      case OP_CREF:
1613      case OP_OPT:      case OP_NCREF:
1614        case OP_RREF:
1615        case OP_NRREF:
1616        case OP_DEF:
1617      case OP_CALLOUT:      case OP_CALLOUT:
1618      case OP_SOD:      case OP_SOD:
1619      case OP_SOM:      case OP_SOM:
1620        case OP_SET_SOM:
1621      case OP_EOD:      case OP_EOD:
1622      case OP_EODN:      case OP_EODN:
1623      case OP_CIRC:      case OP_CIRC:
1624        case OP_CIRCM:
1625      case OP_DOLL:      case OP_DOLL:
1626        case OP_DOLLM:
1627      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1628      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1629      cc += _pcre_OP_lengths[*cc];      cc += _pcre_OP_lengths[*cc];
# Line 903  for (;;) Line 1632  for (;;)
1632      /* Handle literal characters */      /* Handle literal characters */
1633    
1634      case OP_CHAR:      case OP_CHAR:
1635      case OP_CHARNC:      case OP_CHARI:
1636        case OP_NOT:
1637        case OP_NOTI:
1638      branchlength++;      branchlength++;
1639      cc += 2;      cc += 2;
1640  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1641      if ((options & PCRE_UTF8) != 0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       {  
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1642  #endif  #endif
1643      break;      break;
1644    
# Line 921  for (;;) Line 1649  for (;;)
1649      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1650      cc += 4;      cc += 4;
1651  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1652      if ((options & PCRE_UTF8) != 0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       {  
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1653  #endif  #endif
1654      break;      break;
1655    
1656      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1657      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1658        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1659      cc += 4;      cc += 4;
1660      break;      break;
1661    
# Line 937  for (;;) Line 1663  for (;;)
1663    
1664      case OP_PROP:      case OP_PROP:
1665      case OP_NOTPROP:      case OP_NOTPROP:
1666      cc++;      cc += 2;
1667      /* Fall through */      /* Fall through */
1668    
1669      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
# Line 947  for (;;) Line 1673  for (;;)
1673      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1674      case OP_WORDCHAR:      case OP_WORDCHAR:
1675      case OP_ANY:      case OP_ANY:
1676        case OP_ALLANY:
1677      branchlength++;      branchlength++;
1678      cc++;      cc++;
1679      break;      break;
# Line 1001  for (;;) Line 1728  for (;;)
1728    
1729    
1730  /*************************************************  /*************************************************
1731  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1732  *************************************************/  *************************************************/
1733    
1734  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1735  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1736    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1737    so that it can be called from pcre_study() when finding the minimum matching
1738    length.
1739    
1740  Arguments:  Arguments:
1741    code        points to start of expression    code        points to start of expression
1742    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1743    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1744    
1745  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
1746  */  */
1747    
1748  static const uschar *  const uschar *
1749  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1750  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
1751  for (;;)  for (;;)
1752    {    {
1753    register int c = *code;    register int c = *code;
1754    
1755    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1756    else if (c > OP_BRA)  
1757      /* XCLASS is used for classes that cannot be represented just by a bit
1758      map. This includes negated single high-valued characters. The length in
1759      the table is zero; the actual length is stored in the compiled code. */
1760    
1761      if (c == OP_XCLASS) code += GET(code, 1);
1762    
1763      /* Handle recursion */
1764    
1765      else if (c == OP_REVERSE)
1766        {
1767        if (number < 0) return (uschar *)code;
1768        code += _pcre_OP_lengths[c];
1769        }
1770    
1771      /* Handle capturing bracket */
1772    
1773      else if (c == OP_CBRA || c == OP_SCBRA ||
1774               c == OP_CBRAPOS || c == OP_SCBRAPOS)
1775      {      {
1776      int n = c - OP_BRA;      int n = GET2(code, 1+LINK_SIZE);
     if (n > EXTRACT_BASIC_MAX) n = GET2(code, 2+LINK_SIZE);  
1777      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
1778      code += _pcre_OP_lengths[OP_BRA];      code += _pcre_OP_lengths[c];
1779      }      }
1780    
1781      /* Otherwise, we can get the item's length from the table, except that for
1782      repeated character types, we have to test for \p and \P, which have an extra
1783      two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1784      must add in its length. */
1785    
1786    else    else
1787      {      {
1788      code += _pcre_OP_lengths[c];      switch(c)
1789          {
1790          case OP_TYPESTAR:
1791          case OP_TYPEMINSTAR:
1792          case OP_TYPEPLUS:
1793          case OP_TYPEMINPLUS:
1794          case OP_TYPEQUERY:
1795          case OP_TYPEMINQUERY:
1796          case OP_TYPEPOSSTAR:
1797          case OP_TYPEPOSPLUS:
1798          case OP_TYPEPOSQUERY:
1799          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1800          break;
1801    
1802  #ifdef SUPPORT_UTF8        case OP_TYPEUPTO:
1803          case OP_TYPEMINUPTO:
1804          case OP_TYPEEXACT:
1805          case OP_TYPEPOSUPTO:
1806          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1807          break;
1808    
1809      /* In UTF-8 mode, opcodes that are followed by a character may be followed        case OP_MARK:
1810      by a multi-byte character. The length in the table is a minimum, so we have        case OP_PRUNE_ARG:
1811      to scan along to skip the extra bytes. All opcodes are less than 128, so we        case OP_SKIP_ARG:
1812      can use relatively efficient code. */        code += code[1];
1813          break;
1814    
1815          case OP_THEN_ARG:
1816          code += code[1];
1817          break;
1818          }
1819    
1820        /* Add in the fixed length from the table */
1821    
1822        code += _pcre_OP_lengths[c];
1823    
1824      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1825      a multi-byte character. The length in the table is a minimum, so we have to
1826      arrange to skip the extra bytes. */
1827    
1828    #ifdef SUPPORT_UTF8
1829      if (utf8) switch(c)      if (utf8) switch(c)
1830        {        {
1831        case OP_CHAR:        case OP_CHAR:
1832        case OP_CHARNC:        case OP_CHARI:
1833        case OP_EXACT:        case OP_EXACT:
1834          case OP_EXACTI:
1835        case OP_UPTO:        case OP_UPTO:
1836          case OP_UPTOI:
1837        case OP_MINUPTO:        case OP_MINUPTO:
1838          case OP_MINUPTOI:
1839          case OP_POSUPTO:
1840          case OP_POSUPTOI:
1841        case OP_STAR:        case OP_STAR:
1842          case OP_STARI:
1843        case OP_MINSTAR:        case OP_MINSTAR:
1844          case OP_MINSTARI:
1845          case OP_POSSTAR:
1846          case OP_POSSTARI:
1847        case OP_PLUS:        case OP_PLUS:
1848          case OP_PLUSI:
1849        case OP_MINPLUS:        case OP_MINPLUS:
1850          case OP_MINPLUSI:
1851          case OP_POSPLUS:
1852          case OP_POSPLUSI:
1853        case OP_QUERY:        case OP_QUERY:
1854          case OP_QUERYI:
1855        case OP_MINQUERY:        case OP_MINQUERY:
1856        while ((*code & 0xc0) == 0x80) code++;        case OP_MINQUERYI:
1857        break;        case OP_POSQUERY:
1858          case OP_POSQUERYI:
1859        /* 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;  
1860        break;        break;
1861        }        }
1862    #else
1863        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1864  #endif  #endif
1865      }      }
1866    }    }
# Line 1092  Returns:      pointer to the opcode for Line 1885  Returns:      pointer to the opcode for
1885  static const uschar *  static const uschar *
1886  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const uschar *code, BOOL utf8)
1887  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
1888  for (;;)  for (;;)
1889    {    {
1890    register int c = *code;    register int c = *code;
1891    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1892    else if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
1893    else if (c > OP_BRA)  
1894      {    /* XCLASS is used for classes that cannot be represented just by a bit
1895      code += _pcre_OP_lengths[OP_BRA];    map. This includes negated single high-valued characters. The length in
1896      }    the table is zero; the actual length is stored in the compiled code. */
1897    
1898      if (c == OP_XCLASS) code += GET(code, 1);
1899    
1900      /* Otherwise, we can get the item's length from the table, except that for
1901      repeated character types, we have to test for \p and \P, which have an extra
1902      two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1903      must add in its length. */
1904    
1905    else    else
1906      {      {
1907      code += _pcre_OP_lengths[c];      switch(c)
1908          {
1909          case OP_TYPESTAR:
1910          case OP_TYPEMINSTAR:
1911          case OP_TYPEPLUS:
1912          case OP_TYPEMINPLUS:
1913          case OP_TYPEQUERY:
1914          case OP_TYPEMINQUERY:
1915          case OP_TYPEPOSSTAR:
1916          case OP_TYPEPOSPLUS:
1917          case OP_TYPEPOSQUERY:
1918          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1919          break;
1920    
1921  #ifdef SUPPORT_UTF8        case OP_TYPEPOSUPTO:
1922          case OP_TYPEUPTO:
1923          case OP_TYPEMINUPTO:
1924          case OP_TYPEEXACT:
1925          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1926          break;
1927    
1928          case OP_MARK:
1929          case OP_PRUNE_ARG:
1930          case OP_SKIP_ARG:
1931          code += code[1];
1932          break;
1933    
1934          case OP_THEN_ARG:
1935          code += code[1];
1936          break;
1937          }
1938    
1939        /* Add in the fixed length from the table */
1940    
1941        code += _pcre_OP_lengths[c];
1942    
1943      /* 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
1944      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
1945      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. */  
1946    
1947    #ifdef SUPPORT_UTF8
1948      if (utf8) switch(c)      if (utf8) switch(c)
1949        {        {
1950        case OP_CHAR:        case OP_CHAR:
1951        case OP_CHARNC:        case OP_CHARI:
1952        case OP_EXACT:        case OP_EXACT:
1953          case OP_EXACTI:
1954        case OP_UPTO:        case OP_UPTO:
1955          case OP_UPTOI:
1956        case OP_MINUPTO:        case OP_MINUPTO:
1957          case OP_MINUPTOI:
1958          case OP_POSUPTO:
1959          case OP_POSUPTOI:
1960        case OP_STAR:        case OP_STAR:
1961          case OP_STARI:
1962        case OP_MINSTAR:        case OP_MINSTAR:
1963          case OP_MINSTARI:
1964          case OP_POSSTAR:
1965          case OP_POSSTARI:
1966        case OP_PLUS:        case OP_PLUS:
1967          case OP_PLUSI:
1968        case OP_MINPLUS:        case OP_MINPLUS:
1969          case OP_MINPLUSI:
1970          case OP_POSPLUS:
1971          case OP_POSPLUSI:
1972        case OP_QUERY:        case OP_QUERY:
1973          case OP_QUERYI:
1974        case OP_MINQUERY:        case OP_MINQUERY:
1975        while ((*code & 0xc0) == 0x80) code++;        case OP_MINQUERYI:
1976        break;        case OP_POSQUERY:
1977          case OP_POSQUERYI:
1978        /* 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;  
1979        break;        break;
1980        }        }
1981    #else
1982        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1983  #endif  #endif
1984      }      }
1985    }    }
# Line 1152  for (;;) Line 1992  for (;;)
1992  *************************************************/  *************************************************/
1993    
1994  /* 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
1995  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()
1996  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
1997  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
1998  whose current branch will already have been scanned.  backward and negative forward assertions when its final argument is TRUE. If we
1999    hit an unclosed bracket, we return "empty" - this means we've struck an inner
2000    bracket whose current branch will already have been scanned.
2001    
2002  Arguments:  Arguments:
2003    code        points to start of search    code        points to start of search
2004    endcode     points to where to stop    endcode     points to where to stop
2005    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
2006      cd          contains pointers to tables etc.
2007    
2008  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2009  */  */
2010    
2011  static BOOL  static BOOL
2012  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
2013      compile_data *cd)
2014  {  {
2015  register int c;  register int c;
2016  for (code = first_significant_code(code + 1 + LINK_SIZE, NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], TRUE);
2017       code < endcode;       code < endcode;
2018       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], TRUE))
2019    {    {
2020    const uschar *ccode;    const uschar *ccode;
2021    
2022    c = *code;    c = *code;
2023    
2024    if (c >= OP_BRA)    /* Skip over forward assertions; the other assertions are skipped by
2025      first_significant_code() with a TRUE final argument. */
2026    
2027      if (c == OP_ASSERT)
2028        {
2029        do code += GET(code, 1); while (*code == OP_ALT);
2030        c = *code;
2031        continue;
2032        }
2033    
2034      /* For a recursion/subroutine call, if its end has been reached, which
2035      implies a backward reference subroutine call, we can scan it. If it's a
2036      forward reference subroutine call, we can't. To detect forward reference
2037      we have to scan up the list that is kept in the workspace. This function is
2038      called only when doing the real compile, not during the pre-compile that
2039      measures the size of the compiled pattern. */
2040    
2041      if (c == OP_RECURSE)
2042      {      {
2043        const uschar *scode;
2044      BOOL empty_branch;      BOOL empty_branch;
     if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */  
2045    
2046      /* Scan a closed bracket */      /* Test for forward reference */
2047    
2048        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
2049          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
2050    
2051        /* Not a forward reference, test for completed backward reference */
2052    
2053      empty_branch = FALSE;      empty_branch = FALSE;
2054        scode = cd->start_code + GET(code, 1);
2055        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2056    
2057        /* Completed backwards reference */
2058    
2059      do      do
2060        {        {
2061        if (!empty_branch && could_be_empty_branch(code, endcode, utf8))        if (could_be_empty_branch(scode, endcode, utf8, cd))
2062            {
2063          empty_branch = TRUE;          empty_branch = TRUE;
2064        code += GET(code, 1);          break;
2065            }
2066          scode += GET(scode, 1);
2067        }        }
2068      while (*code == OP_ALT);      while (*scode == OP_ALT);
2069      if (!empty_branch) return FALSE;   /* All branches are non-empty */  
2070      code += 1 + LINK_SIZE;      if (!empty_branch) return FALSE;  /* All branches are non-empty */
2071      c = *code;      continue;
2072      }      }
2073    
2074    else switch (c)    /* Groups with zero repeats can of course be empty; skip them. */
     {  
     /* Check for quantifiers after a class */  
2075    
2076  #ifdef SUPPORT_UTF8    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2077      case OP_XCLASS:        c == OP_BRAPOSZERO)
2078      ccode = code + GET(code, 1);      {
2079        code += _pcre_OP_lengths[c];
2080        do code += GET(code, 1); while (*code == OP_ALT);
2081        c = *code;
2082        continue;
2083        }
2084    
2085      /* A nested group that is already marked as "could be empty" can just be
2086      skipped. */
2087    
2088      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2089          c == OP_SCBRA || c == OP_SCBRAPOS)
2090        {
2091        do code += GET(code, 1); while (*code == OP_ALT);
2092        c = *code;
2093        continue;
2094        }
2095    
2096      /* For other groups, scan the branches. */
2097    
2098      if (c == OP_BRA  || c == OP_BRAPOS ||
2099          c == OP_CBRA || c == OP_CBRAPOS ||
2100          c == OP_ONCE || c == OP_ONCE_NC ||
2101          c == OP_COND)
2102        {
2103        BOOL empty_branch;
2104        if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
2105    
2106        /* If a conditional group has only one branch, there is a second, implied,
2107        empty branch, so just skip over the conditional, because it could be empty.
2108        Otherwise, scan the individual branches of the group. */
2109    
2110        if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
2111          code += GET(code, 1);
2112        else
2113          {
2114          empty_branch = FALSE;
2115          do
2116            {
2117            if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
2118              empty_branch = TRUE;
2119            code += GET(code, 1);
2120            }
2121          while (*code == OP_ALT);
2122          if (!empty_branch) return FALSE;   /* All branches are non-empty */
2123          }
2124    
2125        c = *code;
2126        continue;
2127        }
2128    
2129      /* Handle the other opcodes */
2130    
2131      switch (c)
2132        {
2133        /* Check for quantifiers after a class. XCLASS is used for classes that
2134        cannot be represented just by a bit map. This includes negated single
2135        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
2136        actual length is stored in the compiled code, so we must update "code"
2137        here. */
2138    
2139    #ifdef SUPPORT_UTF8
2140        case OP_XCLASS:
2141        ccode = code += GET(code, 1);
2142      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
2143  #endif  #endif
2144    
# Line 1247  for (code = first_significant_code(code Line 2182  for (code = first_significant_code(code
2182      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2183      case OP_WORDCHAR:      case OP_WORDCHAR:
2184      case OP_ANY:      case OP_ANY:
2185        case OP_ALLANY:
2186      case OP_ANYBYTE:      case OP_ANYBYTE:
2187      case OP_CHAR:      case OP_CHAR:
2188      case OP_CHARNC:      case OP_CHARI:
2189      case OP_NOT:      case OP_NOT:
2190        case OP_NOTI:
2191      case OP_PLUS:      case OP_PLUS:
2192      case OP_MINPLUS:      case OP_MINPLUS:
2193        case OP_POSPLUS:
2194      case OP_EXACT:      case OP_EXACT:
2195      case OP_NOTPLUS:      case OP_NOTPLUS:
2196      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
2197        case OP_NOTPOSPLUS:
2198      case OP_NOTEXACT:      case OP_NOTEXACT:
2199      case OP_TYPEPLUS:      case OP_TYPEPLUS:
2200      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
2201        case OP_TYPEPOSPLUS:
2202      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2203      return FALSE;      return FALSE;
2204    
2205        /* These are going to continue, as they may be empty, but we have to
2206        fudge the length for the \p and \P cases. */
2207    
2208        case OP_TYPESTAR:
2209        case OP_TYPEMINSTAR:
2210        case OP_TYPEPOSSTAR:
2211        case OP_TYPEQUERY:
2212        case OP_TYPEMINQUERY:
2213        case OP_TYPEPOSQUERY:
2214        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2215        break;
2216    
2217        /* Same for these */
2218    
2219        case OP_TYPEUPTO:
2220        case OP_TYPEMINUPTO:
2221        case OP_TYPEPOSUPTO:
2222        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
2223        break;
2224    
2225      /* End of branch */      /* End of branch */
2226    
2227      case OP_KET:      case OP_KET:
2228      case OP_KETRMAX:      case OP_KETRMAX:
2229      case OP_KETRMIN:      case OP_KETRMIN:
2230        case OP_KETRPOS:
2231      case OP_ALT:      case OP_ALT:
2232      return TRUE;      return TRUE;
2233    
2234      /* In UTF-8 mode, STAR, MINSTAR, QUERY, MINQUERY, UPTO, and MINUPTO  may be      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2235      followed by a multibyte character */      MINUPTO, and POSUPTO may be followed by a multibyte character */
2236    
2237  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2238      case OP_STAR:      case OP_STAR:
2239        case OP_STARI:
2240      case OP_MINSTAR:      case OP_MINSTAR:
2241        case OP_MINSTARI:
2242        case OP_POSSTAR:
2243        case OP_POSSTARI:
2244      case OP_QUERY:      case OP_QUERY:
2245        case OP_QUERYI:
2246      case OP_MINQUERY:      case OP_MINQUERY:
2247        case OP_MINQUERYI:
2248        case OP_POSQUERY:
2249        case OP_POSQUERYI:
2250        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2251        break;
2252    
2253      case OP_UPTO:      case OP_UPTO:
2254        case OP_UPTOI:
2255      case OP_MINUPTO:      case OP_MINUPTO:
2256      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_MINUPTOI:
2257        case OP_POSUPTO:
2258        case OP_POSUPTOI:
2259        if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2260      break;      break;
2261  #endif  #endif
2262    
2263        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2264        string. */
2265    
2266        case OP_MARK:
2267        case OP_PRUNE_ARG:
2268        case OP_SKIP_ARG:
2269        code += code[1];
2270        break;
2271    
2272        case OP_THEN_ARG:
2273        code += code[1];
2274        break;
2275    
2276        /* None of the remaining opcodes are required to match a character. */
2277    
2278        default:
2279        break;
2280      }      }
2281    }    }
2282    
# Line 1299  return TRUE; Line 2293  return TRUE;
2293  the current branch of the current pattern to see if it could match the empty  the current branch of the current pattern to see if it could match the empty
2294  string. If it could, we must look outwards for branches at other levels,  string. If it could, we must look outwards for branches at other levels,
2295  stopping when we pass beyond the bracket which is the subject of the recursion.  stopping when we pass beyond the bracket which is the subject of the recursion.
2296    This function is called only during the real compile, not during the
2297    pre-compile.
2298    
2299  Arguments:  Arguments:
2300    code        points to start of the recursion    code        points to start of the recursion
2301    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2302    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2303    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2304      cd          pointers to tables etc
2305    
2306  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2307  */  */
2308    
2309  static BOOL  static BOOL
2310  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2311    BOOL utf8)    BOOL utf8, compile_data *cd)
2312  {  {
2313  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2314    {    {
2315    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2316        return FALSE;
2317    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2318    }    }
2319  return TRUE;  return TRUE;
# Line 1328  return TRUE; Line 2326  return TRUE;
2326  *************************************************/  *************************************************/
2327    
2328  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2329  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
2330  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2331  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2332    
2333    Originally, this function only recognized a sequence of letters between the
2334    terminators, but it seems that Perl recognizes any sequence of characters,
2335    though of course unknown POSIX names are subsequently rejected. Perl gives an
2336    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2337    didn't consider this to be a POSIX class. Likewise for [:1234:].
2338    
2339    The problem in trying to be exactly like Perl is in the handling of escapes. We
2340    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2341    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2342    below handles the special case of \], but does not try to do any other escape
2343    processing. This makes it different from Perl for cases such as [:l\ower:]
2344    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2345    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2346    I think.
2347    
2348    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2349    It seems that the appearance of a nested POSIX class supersedes an apparent
2350    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2351    a digit.
2352    
2353    In Perl, unescaped square brackets may also appear as part of class names. For
2354    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2355    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2356    seem right at all. PCRE does not allow closing square brackets in POSIX class
2357    names.
2358    
2359  Argument:  Arguments:
2360    ptr      pointer to the initial [    ptr      pointer to the initial [
2361    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2362    
2363  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2364  */  */
2365    
2366  static BOOL  static BOOL
2367  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
2368  {  {
2369  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2370  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2371  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2372    {    {
2373    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2374    return TRUE;      ptr++;
2375      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2376      else
2377        {
2378        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2379          {
2380          *endptr = ptr;
2381          return TRUE;
2382          }
2383        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2384             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2385              ptr[1] == CHAR_EQUALS_SIGN) &&
2386            check_posix_syntax(ptr, endptr))
2387          return FALSE;
2388        }
2389    }    }
2390  return FALSE;  return FALSE;
2391  }  }
# Line 1375  Returns:     a value representing the na Line 2410  Returns:     a value representing the na
2410  static int  static int
2411  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
2412  {  {
2413    const char *pn = posix_names;
2414  register int yield = 0;  register int yield = 0;
2415  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2416    {    {
2417    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2418      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
2419      pn += posix_name_lengths[yield] + 1;
2420    yield++;    yield++;
2421    }    }
2422  return -1;  return -1;
# Line 1394  return -1; Line 2431  return -1;
2431  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2432  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2433  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
2434  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
2435  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
2436  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
2437  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
2438  partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2439    OP_END.
2440    
2441    This function has been extended with the possibility of forward references for
2442    recursions and subroutine calls. It must also check the list of such references
2443    for the group we are dealing with. If it finds that one of the recursions in
2444    the current group is on this list, it adjusts the offset in the list, not the
2445    value in the reference (which is a group number).
2446    
2447  Arguments:  Arguments:
2448    group      points to the start of the group    group      points to the start of the group
2449    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
2450    utf8       TRUE in UTF-8 mode    utf8       TRUE in UTF-8 mode
2451    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
2452      save_hwm   the hwm forward reference pointer at the start of the group
2453    
2454  Returns:     nothing  Returns:     nothing
2455  */  */
2456    
2457  static void  static void
2458  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd)  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,
2459      uschar *save_hwm)
2460  {  {
2461  uschar *ptr = group;  uschar *ptr = group;
2462    
2463  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
2464    {    {
2465    int offset = GET(ptr, 1);    int offset;
2466    if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);    uschar *hc;
2467    
2468      /* See if this recursion is on the forward reference list. If so, adjust the
2469      reference. */
2470    
2471      for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
2472        {
2473        offset = GET(hc, 0);
2474        if (cd->start_code + offset == ptr + 1)
2475          {
2476          PUT(hc, 0, offset + adjust);
2477          break;
2478          }
2479        }
2480    
2481      /* Otherwise, adjust the recursion offset if it's after the start of this
2482      group. */
2483    
2484      if (hc >= cd->hwm)
2485        {
2486        offset = GET(ptr, 1);
2487        if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
2488        }
2489    
2490    ptr += 1 + LINK_SIZE;    ptr += 1 + LINK_SIZE;
2491    }    }
2492  }  }
# Line 1443  auto_callout(uschar *code, const uschar Line 2513  auto_callout(uschar *code, const uschar
2513  {  {
2514  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2515  *code++ = 255;  *code++ = 255;
2516  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2517  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2518  return code + 2*LINK_SIZE;  return code + 2*LINK_SIZE;
2519  }  }
2520    
# Line 1469  Returns:             nothing Line 2539  Returns:             nothing
2539  static void  static void
2540  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)
2541  {  {
2542  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2543  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2544  }  }
2545    
# Line 1495  Yield:        TRUE when range returned; Line 2565  Yield:        TRUE when range returned;
2565  */  */
2566    
2567  static BOOL  static BOOL
2568  get_othercase_range(int *cptr, int d, int *ocptr, int *odptr)  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,
2569      unsigned int *odptr)
2570  {  {
2571  int c, chartype, othercase, next;  unsigned int c, othercase, next;
2572    
2573  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2574    {    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
   if (_pcre_ucp_findchar(c, &chartype, &othercase) == ucp_L && othercase != 0)  
     break;  
   }  
2575    
2576  if (c > d) return FALSE;  if (c > d) return FALSE;
2577    
# Line 1512  next = othercase + 1; Line 2580  next = othercase + 1;
2580    
2581  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2582    {    {
2583    if (_pcre_ucp_findchar(c, &chartype, &othercase) != ucp_L ||    if (UCD_OTHERCASE(c) != next) break;
         othercase != next)  
     break;  
2584    next++;    next++;
2585    }    }
2586    
# Line 1523  for (++c; c <= d; c++) Line 2589  for (++c; c <= d; c++)
2589    
2590  return TRUE;  return TRUE;
2591  }  }
2592    
2593    
2594    
2595    /*************************************************
2596    *        Check a character and a property        *
2597    *************************************************/
2598    
2599    /* This function is called by check_auto_possessive() when a property item
2600    is adjacent to a fixed character.
2601    
2602    Arguments:
2603      c            the character
2604      ptype        the property type
2605      pdata        the data for the type
2606      negated      TRUE if it's a negated property (\P or \p{^)
2607    
2608    Returns:       TRUE if auto-possessifying is OK
2609    */
2610    
2611    static BOOL
2612    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2613    {
2614    const ucd_record *prop = GET_UCD(c);
2615    switch(ptype)
2616      {
2617      case PT_LAMP:
2618      return (prop->chartype == ucp_Lu ||
2619              prop->chartype == ucp_Ll ||
2620              prop->chartype == ucp_Lt) == negated;
2621    
2622      case PT_GC:
2623      return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;
2624    
2625      case PT_PC:
2626      return (pdata == prop->chartype) == negated;
2627    
2628      case PT_SC:
2629      return (pdata == prop->script) == negated;
2630    
2631      /* These are specials */
2632    
2633      case PT_ALNUM:
2634      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2635              _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;
2636    
2637      case PT_SPACE:    /* Perl space */
2638      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2639              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2640              == negated;
2641    
2642      case PT_PXSPACE:  /* POSIX space */
2643      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2644              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2645              c == CHAR_FF || c == CHAR_CR)
2646              == negated;
2647    
2648      case PT_WORD:
2649      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2650              _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2651              c == CHAR_UNDERSCORE) == negated;
2652      }
2653    return FALSE;
2654    }
2655  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2656    
2657    
2658    
2659    /*************************************************
2660    *     Check if auto-possessifying is possible    *
2661    *************************************************/
2662    
2663    /* This function is called for unlimited repeats of certain items, to see
2664    whether the next thing could possibly match the repeated item. If not, it makes
2665    sense to automatically possessify the repeated item.
2666    
2667    Arguments:
2668      previous      pointer to the repeated opcode
2669      utf8          TRUE in UTF-8 mode
2670      ptr           next character in pattern
2671      options       options bits
2672      cd            contains pointers to tables etc.
2673    
2674    Returns:        TRUE if possessifying is wanted
2675    */
2676    
2677    static BOOL
2678    check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
2679      int options, compile_data *cd)
2680    {
2681    int c, next;
2682    int op_code = *previous++;
2683    
2684    /* Skip whitespace and comments in extended mode */
2685    
2686    if ((options & PCRE_EXTENDED) != 0)
2687      {
2688      for (;;)
2689        {
2690        while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2691        if (*ptr == CHAR_NUMBER_SIGN)
2692          {
2693          ptr++;
2694          while (*ptr != 0)
2695            {
2696            if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2697            ptr++;
2698    #ifdef SUPPORT_UTF8
2699            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2700    #endif
2701            }
2702          }
2703        else break;
2704        }
2705      }
2706    
2707    /* If the next item is one that we can handle, get its value. A non-negative
2708    value is a character, a negative value is an escape value. */
2709    
2710    if (*ptr == CHAR_BACKSLASH)
2711      {
2712      int temperrorcode = 0;
2713      next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
2714      if (temperrorcode != 0) return FALSE;
2715      ptr++;    /* Point after the escape sequence */
2716      }
2717    
2718    else if ((cd->ctypes[*ptr] & ctype_meta) == 0)
2719      {
2720    #ifdef SUPPORT_UTF8
2721      if (utf8) { GETCHARINC(next, ptr); } else
2722    #endif
2723      next = *ptr++;
2724      }
2725    
2726    else return FALSE;
2727    
2728    /* Skip whitespace and comments in extended mode */
2729    
2730    if ((options & PCRE_EXTENDED) != 0)
2731      {
2732      for (;;)
2733        {
2734        while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2735        if (*ptr == CHAR_NUMBER_SIGN)
2736          {
2737          ptr++;
2738          while (*ptr != 0)
2739            {
2740            if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2741            ptr++;
2742    #ifdef SUPPORT_UTF8
2743            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2744    #endif
2745            }
2746          }
2747        else break;
2748        }
2749      }
2750    
2751    /* If the next thing is itself optional, we have to give up. */
2752    
2753    if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2754      strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2755        return FALSE;
2756    
2757    /* Now compare the next item with the previous opcode. First, handle cases when
2758    the next item is a character. */
2759    
2760    if (next >= 0) switch(op_code)
2761      {
2762      case OP_CHAR:
2763    #ifdef SUPPORT_UTF8
2764      GETCHARTEST(c, previous);
2765    #else
2766      c = *previous;
2767    #endif
2768      return c != next;
2769    
2770      /* For CHARI (caseless character) we must check the other case. If we have
2771      Unicode property support, we can use it to test the other case of
2772      high-valued characters. */
2773    
2774      case OP_CHARI:
2775    #ifdef SUPPORT_UTF8
2776      GETCHARTEST(c, previous);
2777    #else
2778      c = *previous;
2779    #endif
2780      if (c == next) return FALSE;
2781    #ifdef SUPPORT_UTF8
2782      if (utf8)
2783        {
2784        unsigned int othercase;
2785        if (next < 128) othercase = cd->fcc[next]; else
2786    #ifdef SUPPORT_UCP
2787        othercase = UCD_OTHERCASE((unsigned int)next);
2788    #else
2789        othercase = NOTACHAR;
2790    #endif
2791        return (unsigned int)c != othercase;
2792        }
2793      else
2794    #endif  /* SUPPORT_UTF8 */
2795      return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2796    
2797      /* For OP_NOT and OP_NOTI, the data is always a single-byte character. These
2798      opcodes are not used for multi-byte characters, because they are coded using
2799      an XCLASS instead. */
2800    
2801      case OP_NOT:
2802      return (c = *previous) == next;
2803    
2804      case OP_NOTI:
2805      if ((c = *previous) == next) return TRUE;
2806    #ifdef SUPPORT_UTF8
2807      if (utf8)
2808        {
2809        unsigned int othercase;
2810        if (next < 128) othercase = cd->fcc[next]; else
2811    #ifdef SUPPORT_UCP
2812        othercase = UCD_OTHERCASE(next);
2813    #else
2814        othercase = NOTACHAR;
2815    #endif
2816        return (unsigned int)c == othercase;
2817        }
2818      else
2819    #endif  /* SUPPORT_UTF8 */
2820      return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2821    
2822      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2823      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
2824    
2825      case OP_DIGIT:
2826      return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
2827    
2828      case OP_NOT_DIGIT:
2829      return next <= 127 && (cd->ctypes[next] & ctype_digit) != 0;
2830    
2831      case OP_WHITESPACE:
2832      return next > 127 || (cd->ctypes[next] & ctype_space) == 0;
2833    
2834      case OP_NOT_WHITESPACE:
2835      return next <= 127 && (cd->ctypes[next] & ctype_space) != 0;
2836    
2837      case OP_WORDCHAR:
2838      return next > 127 || (cd->ctypes[next] & ctype_word) == 0;
2839    
2840      case OP_NOT_WORDCHAR:
2841      return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;
2842    
2843      case OP_HSPACE:
2844      case OP_NOT_HSPACE:
2845      switch(next)
2846        {
2847        case 0x09:
2848        case 0x20:
2849        case 0xa0:
2850        case 0x1680:
2851        case 0x180e:
2852        case 0x2000:
2853        case 0x2001:
2854        case 0x2002:
2855        case 0x2003:
2856        case 0x2004:
2857        case 0x2005:
2858        case 0x2006:
2859        case 0x2007:
2860        case 0x2008:
2861        case 0x2009:
2862        case 0x200A:
2863        case 0x202f:
2864        case 0x205f:
2865        case 0x3000:
2866        return op_code == OP_NOT_HSPACE;
2867        default:
2868        return op_code != OP_NOT_HSPACE;
2869        }
2870    
2871      case OP_ANYNL:
2872      case OP_VSPACE:
2873      case OP_NOT_VSPACE:
2874      switch(next)
2875        {
2876        case 0x0a:
2877        case 0x0b:
2878        case 0x0c:
2879        case 0x0d:
2880        case 0x85:
2881        case 0x2028:
2882        case 0x2029:
2883        return op_code == OP_NOT_VSPACE;
2884        default:
2885        return op_code != OP_NOT_VSPACE;
2886        }
2887    
2888    #ifdef SUPPORT_UCP
2889      case OP_PROP:
2890      return check_char_prop(next, previous[0], previous[1], FALSE);
2891    
2892      case OP_NOTPROP:
2893      return check_char_prop(next, previous[0], previous[1], TRUE);
2894    #endif
2895    
2896      default:
2897      return FALSE;
2898      }
2899    
2900    
2901    /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP
2902    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
2903    generated only when PCRE_UCP is *not* set, that is, when only ASCII
2904    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
2905    replaced by OP_PROP codes when PCRE_UCP is set. */
2906    
2907    switch(op_code)
2908      {
2909      case OP_CHAR:
2910      case OP_CHARI:
2911    #ifdef SUPPORT_UTF8
2912      GETCHARTEST(c, previous);
2913    #else
2914      c = *previous;
2915    #endif
2916      switch(-next)
2917        {
2918        case ESC_d:
2919        return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
2920    
2921        case ESC_D:
2922        return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
2923    
2924        case ESC_s:
2925        return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
2926    
2927        case ESC_S:
2928        return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
2929    
2930        case ESC_w:
2931        return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
2932    
2933        case ESC_W:
2934        return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
2935    
2936        case ESC_h:
2937        case ESC_H:
2938        switch(c)
2939          {
2940          case 0x09:
2941          case 0x20:
2942          case 0xa0:
2943          case 0x1680:
2944          case 0x180e:
2945          case 0x2000:
2946          case 0x2001:
2947          case 0x2002:
2948          case 0x2003:
2949          case 0x2004:
2950          case 0x2005:
2951          case 0x2006:
2952          case 0x2007:
2953          case 0x2008:
2954          case 0x2009:
2955          case 0x200A:
2956          case 0x202f:
2957          case 0x205f:
2958          case 0x3000:
2959          return -next != ESC_h;
2960          default:
2961          return -next == ESC_h;
2962          }
2963    
2964        case ESC_v:
2965        case ESC_V:
2966        switch(c)
2967          {
2968          case 0x0a:
2969          case 0x0b:
2970          case 0x0c:
2971          case 0x0d:
2972          case 0x85:
2973          case 0x2028:
2974          case 0x2029:
2975          return -next != ESC_v;
2976          default:
2977          return -next == ESC_v;
2978          }
2979    
2980        /* When PCRE_UCP is set, these values get generated for \d etc. Find
2981        their substitutions and process them. The result will always be either
2982        -ESC_p or -ESC_P. Then fall through to process those values. */
2983    
2984    #ifdef SUPPORT_UCP
2985        case ESC_du:
2986        case ESC_DU:
2987        case ESC_wu:
2988        case ESC_WU:
2989        case ESC_su:
2990        case ESC_SU:
2991          {
2992          int temperrorcode = 0;
2993          ptr = substitutes[-next - ESC_DU];
2994          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
2995          if (temperrorcode != 0) return FALSE;
2996          ptr++;    /* For compatibility */
2997          }
2998        /* Fall through */
2999    
3000        case ESC_p:
3001        case ESC_P:
3002          {
3003          int ptype, pdata, errorcodeptr;
3004          BOOL negated;
3005    
3006          ptr--;      /* Make ptr point at the p or P */
3007          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
3008          if (ptype < 0) return FALSE;
3009          ptr++;      /* Point past the final curly ket */
3010    
3011          /* If the property item is optional, we have to give up. (When generated
3012          from \d etc by PCRE_UCP, this test will have been applied much earlier,
3013          to the original \d etc. At this point, ptr will point to a zero byte. */
3014    
3015          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3016            strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3017              return FALSE;
3018    
3019          /* Do the property check. */
3020    
3021          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
3022          }
3023    #endif
3024    
3025        default:
3026        return FALSE;
3027        }
3028    
3029      /* In principle, support for Unicode properties should be integrated here as
3030      well. It means re-organizing the above code so as to get hold of the property
3031      values before switching on the op-code. However, I wonder how many patterns
3032      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
3033      these op-codes are never generated.) */
3034    
3035      case OP_DIGIT:
3036      return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
3037             next == -ESC_h || next == -ESC_v || next == -ESC_R;
3038    
3039      case OP_NOT_DIGIT:
3040      return next == -ESC_d;
3041    
3042      case OP_WHITESPACE:
3043      return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
3044    
3045      case OP_NOT_WHITESPACE:
3046      return next == -ESC_s || next == -ESC_h || next == -ESC_v;
3047    
3048      case OP_HSPACE:
3049      return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
3050             next == -ESC_w || next == -ESC_v || next == -ESC_R;
3051    
3052      case OP_NOT_HSPACE:
3053      return next == -ESC_h;
3054    
3055      /* Can't have \S in here because VT matches \S (Perl anomaly) */
3056      case OP_ANYNL:
3057      case OP_VSPACE:
3058      return next == -ESC_V || next == -ESC_d || next == -ESC_w;
3059    
3060      case OP_NOT_VSPACE:
3061      return next == -ESC_v || next == -ESC_R;
3062    
3063      case OP_WORDCHAR:
3064      return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
3065             next == -ESC_v || next == -ESC_R;
3066    
3067      case OP_NOT_WORDCHAR:
3068      return next == -ESC_w || next == -ESC_d;
3069    
3070      default:
3071      return FALSE;
3072      }
3073    
3074    /* Control does not reach here */
3075    }
3076    
3077    
3078    
3079  /*************************************************  /*************************************************
3080  *           Compile one branch                   *  *           Compile one branch                   *
3081  *************************************************/  *************************************************/
3082    
3083  /* Scan the pattern, compiling it into the code vector. If the options are  /* Scan the pattern, compiling it into the a vector. If the options are
3084  changed during the branch, the pointer is used to change the external options  changed during the branch, the pointer is used to change the external options
3085  bits.  bits. This function is used during the pre-compile phase when we are trying
3086    to find out the amount of memory needed, as well as during the real compile
3087    phase. The value of lengthptr distinguishes the two phases.
3088    
3089  Arguments:  Arguments:
3090    optionsptr     pointer to the option bits    optionsptr     pointer to the option bits
   brackets       points to number of extracting brackets used  
3091    codeptr        points to the pointer to the current code point    codeptr        points to the pointer to the current code point
3092    ptrptr         points to the current pattern pointer    ptrptr         points to the current pattern pointer
3093    errorcodeptr   points to error code variable    errorcodeptr   points to error code variable
3094    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
3095    reqbyteptr     set to the last literal character required, else < 0    reqbyteptr     set to the last literal character required, else < 0
3096    bcptr          points to current branch chain    bcptr          points to current branch chain
3097      cond_depth     conditional nesting depth
3098    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
3099      lengthptr      NULL during the real compile phase
3100                     points to length accumulator during pre-compile phase
3101    
3102  Returns:         TRUE on success  Returns:         TRUE on success
3103                   FALSE, with *errorcodeptr set non-zero on error                   FALSE, with *errorcodeptr set non-zero on error
3104  */  */
3105    
3106  static BOOL  static BOOL
3107  compile_branch(int *optionsptr, int *brackets, uschar **codeptr,  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
3108    const uschar **ptrptr, int *errorcodeptr, int *firstbyteptr,    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
3109    int *reqbyteptr, branch_chain *bcptr, compile_data *cd)    int cond_depth, compile_data *cd, int *lengthptr)
3110  {  {
3111  int repeat_type, op_type;  int repeat_type, op_type;
3112  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
# Line 1561  int greedy_default, greedy_non_default; Line 3115  int greedy_default, greedy_non_default;
3115  int firstbyte, reqbyte;  int firstbyte, reqbyte;
3116  int zeroreqbyte, zerofirstbyte;  int zeroreqbyte, zerofirstbyte;
3117  int req_caseopt, reqvary, tempreqvary;  int req_caseopt, reqvary, tempreqvary;
3118  int condcount = 0;  int options = *optionsptr;               /* May change dynamically */
 int options = *optionsptr;  
3119  int after_manual_callout = 0;  int after_manual_callout = 0;
3120    int length_prevgroup = 0;
3121  register int c;  register int c;
3122  register uschar *code = *codeptr;  register uschar *code = *codeptr;
3123    uschar *last_code = code;
3124    uschar *orig_code = code;
3125  uschar *tempcode;  uschar *tempcode;
3126  BOOL inescq = FALSE;  BOOL inescq = FALSE;
3127  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
3128  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
3129  const uschar *tempptr;  const uschar *tempptr;
3130    const uschar *nestptr = NULL;
3131  uschar *previous = NULL;  uschar *previous = NULL;
3132  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
3133    uschar *save_hwm = NULL;
3134  uschar classbits[32];  uschar classbits[32];
3135    
3136    /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3137    must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3138    dynamically as we process the pattern. */
3139    
3140  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3141  BOOL class_utf8;  BOOL class_utf8;
3142  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
3143  uschar *class_utf8data;  uschar *class_utf8data;
3144    uschar *class_utf8data_base;
3145  uschar utf8_char[6];  uschar utf8_char[6];
3146  #else  #else
3147  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
3148  #endif  #endif
3149    
3150    #ifdef PCRE_DEBUG
3151    if (lengthptr != NULL) DPRINTF((">> start branch\n"));
3152    #endif
3153    
3154  /* Set up the default and non-default settings for greediness */  /* Set up the default and non-default settings for greediness */
3155    
3156  greedy_default = ((options & PCRE_UNGREEDY) != 0);  greedy_default = ((options & PCRE_UNGREEDY) != 0);
# Line 1613  req_caseopt = ((options & PCRE_CASELESS) Line 3180  req_caseopt = ((options & PCRE_CASELESS)
3180  for (;; ptr++)  for (;; ptr++)
3181    {    {
3182    BOOL negate_class;    BOOL negate_class;
3183      BOOL should_flip_negation;
3184    BOOL possessive_quantifier;    BOOL possessive_quantifier;
3185    BOOL is_quantifier;    BOOL is_quantifier;
3186      BOOL is_recurse;
3187      BOOL reset_bracount;
3188    int class_charcount;    int class_charcount;
3189    int class_lastchar;    int class_lastchar;
3190    int newoptions;    int newoptions;
3191    int recno;    int recno;
3192      int refsign;
3193    int skipbytes;    int skipbytes;
3194    int subreqbyte;    int subreqbyte;
3195    int subfirstbyte;    int subfirstbyte;
3196      int terminator;
3197    int mclength;    int mclength;
3198      int tempbracount;
3199    uschar mcbuffer[8];    uschar mcbuffer[8];
3200    
3201    /* Next byte in the pattern */    /* Get next byte in the pattern */
3202    
3203    c = *ptr;    c = *ptr;
3204    
3205      /* If we are at the end of a nested substitution, revert to the outer level
3206      string. Nesting only happens one level deep. */
3207    
3208      if (c == 0 && nestptr != NULL)
3209        {
3210        ptr = nestptr;
3211        nestptr = NULL;
3212        c = *ptr;
3213        }
3214    
3215      /* If we are in the pre-compile phase, accumulate the length used for the
3216      previous cycle of this loop. */
3217    
3218      if (lengthptr != NULL)
3219        {
3220    #ifdef PCRE_DEBUG
3221        if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3222    #endif
3223        if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
3224          {
3225          *errorcodeptr = ERR52;
3226          goto FAILED;
3227          }
3228    
3229        /* There is at least one situation where code goes backwards: this is the
3230        case of a zero quantifier after a class (e.g. [ab]{0}). At compile time,
3231        the class is simply eliminated. However, it is created first, so we have to
3232        allow memory for it. Therefore, don't ever reduce the length at this point.
3233        */
3234    
3235        if (code < last_code) code = last_code;
3236    
3237        /* Paranoid check for integer overflow */
3238    
3239        if (OFLOW_MAX - *lengthptr < code - last_code)
3240          {
3241          *errorcodeptr = ERR20;
3242          goto FAILED;
3243          }
3244    
3245        *lengthptr += (int)(code - last_code);
3246        DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));
3247    
3248        /* If "previous" is set and it is not at the start of the work space, move
3249        it back to there, in order to avoid filling up the work space. Otherwise,
3250        if "previous" is NULL, reset the current code pointer to the start. */
3251    
3252        if (previous != NULL)
3253          {
3254          if (previous > orig_code)
3255            {
3256            memmove(orig_code, previous, code - previous);
3257            code -= previous - orig_code;
3258            previous = orig_code;
3259            }
3260          }
3261        else code = orig_code;
3262    
3263        /* Remember where this code item starts so we can pick up the length
3264        next time round. */
3265    
3266        last_code = code;
3267        }
3268    
3269      /* In the real compile phase, just check the workspace used by the forward
3270      reference list. */
3271    
3272      else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
3273        {
3274        *errorcodeptr = ERR52;
3275        goto FAILED;
3276        }
3277    
3278    /* If in \Q...\E, check for the end; if not, we have a literal */    /* If in \Q...\E, check for the end; if not, we have a literal */
3279    
3280    if (inescq && c != 0)    if (inescq && c != 0)
3281      {      {
3282      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3283        {        {
3284        inescq = FALSE;        inescq = FALSE;
3285        ptr++;        ptr++;
# Line 1643  for (;; ptr++) Line 3289  for (;; ptr++)
3289        {        {
3290        if (previous_callout != NULL)        if (previous_callout != NULL)
3291          {          {
3292          complete_callout(previous_callout, ptr, cd);          if (lengthptr == NULL)  /* Don't attempt in pre-compile phase */
3293              complete_callout(previous_callout, ptr, cd);
3294          previous_callout = NULL;          previous_callout = NULL;
3295          }          }
3296        if ((options & PCRE_AUTO_CALLOUT) != 0)        if ((options & PCRE_AUTO_CALLOUT) != 0)
# Line 1658  for (;; ptr++) Line 3305  for (;; ptr++)
3305    /* Fill in length of a previous callout, except when the next thing is    /* Fill in length of a previous callout, except when the next thing is
3306    a quantifier. */    a quantifier. */
3307    
3308    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
3309      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
3310        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
3311    
3312    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
3313         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
3314      {      {
3315      complete_callout(previous_callout, ptr, cd);      if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */
3316          complete_callout(previous_callout, ptr, cd);
3317      previous_callout = NULL;      previous_callout = NULL;
3318      }      }
3319    
3320    /* In extended mode, skip white space and comments */    /* In extended mode, skip white space and comments. */
3321    
3322    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3323      {      {
3324      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
3325      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
3326        {        {
3327        /* The space before the ; is to avoid a warning on a silly compiler        ptr++;
3328        on the Macintosh. */        while (*ptr != 0)
3329        while ((c = *(++ptr)) != 0 && c != NEWLINE) ;          {
3330        if (c != 0) continue;   /* Else fall through to handle end of string */          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3331            ptr++;
3332    #ifdef SUPPORT_UTF8
3333            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
3334    #endif
3335            }
3336          if (*ptr != 0) continue;
3337    
3338          /* Else fall through to handle end of string */
3339          c = 0;
3340        }        }
3341      }      }
3342    
# Line 1692  for (;; ptr++) Line 3350  for (;; ptr++)
3350    
3351    switch(c)    switch(c)
3352      {      {
3353      /* The branch terminates at end of string, |, or ). */      /* ===================================================================*/
3354        case 0:                        /* The branch terminates at string end */
3355      case 0:      case CHAR_VERTICAL_LINE:       /* or | or ) */
3356      case '|':      case CHAR_RIGHT_PARENTHESIS:
     case ')':  
3357      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
3358      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
3359      *codeptr = code;      *codeptr = code;
3360      *ptrptr = ptr;      *ptrptr = ptr;
3361        if (lengthptr != NULL)
3362          {
3363          if (OFLOW_MAX - *lengthptr < code - last_code)
3364            {
3365            *errorcodeptr = ERR20;
3366            goto FAILED;
3367            }
3368          *lengthptr += (int)(code - last_code);   /* To include callout length */
3369          DPRINTF((">> end branch\n"));
3370          }
3371      return TRUE;      return TRUE;
3372    
3373    
3374        /* ===================================================================*/
3375      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
3376      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3377    
3378      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
3379        previous = NULL;
3380      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3381        {        {
3382        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3383          *code++ = OP_CIRCM;
3384        }        }
3385      previous = NULL;      else *code++ = OP_CIRC;
     *code++ = OP_CIRC;  
3386      break;      break;
3387    
3388      case '$':      case CHAR_DOLLAR_SIGN:
3389      previous = NULL;      previous = NULL;
3390      *code++ = OP_DOLL;      *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
3391      break;      break;
3392    
3393      /* There can never be a first char if '.' is first, whatever happens about      /* There can never be a first char if '.' is first, whatever happens about
3394      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
3395    
3396      case '.':      case CHAR_DOT:
3397      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3398      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
3399      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3400      previous = code;      previous = code;
3401      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
3402      break;      break;
3403    
3404      /* Character classes. If the included characters are all < 255 in value, we  
3405      build a 32-byte bitmap of the permitted characters, except in the special      /* ===================================================================*/
3406      case where there is only one such character. For negated classes, we build      /* Character classes. If the included characters are all < 256, we build a
3407      the map as usual, then invert it at the end. However, we use a different      32-byte bitmap of the permitted characters, except in the special case
3408      opcode so that data characters > 255 can be handled correctly.      where there is only one such character. For negated classes, we build the
3409        map as usual, then invert it at the end. However, we use a different opcode
3410        so that data characters > 255 can be handled correctly.
3411    
3412      If the class contains characters outside the 0-255 range, a different      If the class contains characters outside the 0-255 range, a different
3413      opcode is compiled. It may optionally have a bit map for characters < 256,      opcode is compiled. It may optionally have a bit map for characters < 256,
3414      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
3415      whether the bitmap is present, and whether this is a negated class or not.      whether the bitmap is present, and whether this is a negated class or not.
     */  
3416    
3417      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
3418        default (Perl) mode, it is treated as a data character. */
3419    
3420        case CHAR_RIGHT_SQUARE_BRACKET:
3421        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3422          {
3423          *errorcodeptr = ERR64;
3424          goto FAILED;
3425          }
3426        goto NORMAL_CHAR;
3427    
3428        case CHAR_LEFT_SQUARE_BRACKET:
3429      previous = code;      previous = code;
3430    
3431      /* PCRE supports POSIX class stuff inside a class. Perl gives an error if      /* PCRE supports POSIX class stuff inside a class. Perl gives an error if
3432      they are encountered at the top level, so we'll do that too. */      they are encountered at the top level, so we'll do that too. */
3433    
3434      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3435          check_posix_syntax(ptr, &tempptr, cd))           ptr[1] == CHAR_EQUALS_SIGN) &&
3436            check_posix_syntax(ptr, &tempptr))
3437        {        {
3438        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
3439        goto FAILED;        goto FAILED;
3440        }        }
3441    
3442      /* If the first character is '^', set the negation flag and skip it. */      /* If the first character is '^', set the negation flag and skip it. Also,
3443        if the first few characters (either before or after ^) are \Q\E or \E we
3444        skip them too. This makes for compatibility with Perl. */
3445    
3446      if ((c = *(++ptr)) == '^')      negate_class = FALSE;
3447        for (;;)
3448        {        {
       negate_class = TRUE;  
3449        c = *(++ptr);        c = *(++ptr);
3450          if (c == CHAR_BACKSLASH)
3451            {
3452            if (ptr[1] == CHAR_E)
3453              ptr++;
3454            else if (strncmp((const char *)ptr+1,
3455                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
3456              ptr += 3;
3457            else
3458              break;
3459            }
3460          else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
3461            negate_class = TRUE;
3462          else break;
3463        }        }
3464      else  
3465        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
3466        an initial ']' is taken as a data character -- the code below handles
3467        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3468        [^] must match any character, so generate OP_ALLANY. */
3469    
3470        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3471            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3472        {        {
3473        negate_class = FALSE;        *code++ = negate_class? OP_ALLANY : OP_FAIL;
3474          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3475          zerofirstbyte = firstbyte;
3476          break;
3477        }        }
3478    
3479        /* If a class contains a negative special such as \S, we need to flip the
3480        negation flag at the end, so that support for characters > 255 works
3481        correctly (they are all included in the class). */
3482    
3483        should_flip_negation = FALSE;
3484    
3485      /* Keep a count of chars with values < 256 so that we can optimize the case      /* Keep a count of chars with values < 256 so that we can optimize the case
3486      of just a single character (as long as it's < 256). For higher valued UTF-8      of just a single character (as long as it's < 256). However, For higher
3487      characters, we don't yet do any optimization. */      valued UTF-8 characters, we don't yet do any optimization. */
3488    
3489      class_charcount = 0;      class_charcount = 0;
3490      class_lastchar = -1;      class_lastchar = -1;
3491    
3492        /* Initialize the 32-char bit map to all zeros. We build the map in a
3493        temporary bit of memory, in case the class contains only 1 character (less
3494        than 256), because in that case the compiled code doesn't use the bit map.
3495        */
3496    
3497        memset(classbits, 0, 32 * sizeof(uschar));
3498    
3499  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3500      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
3501      class_utf8data = code + LINK_SIZE + 34;   /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
3502        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
3503  #endif  #endif
3504    
     /* Initialize the 32-char bit map to all zeros. We have to build the  
     map in a temporary bit of store, in case the class contains only 1  
     character (< 256), because in that case the compiled code doesn't use the  
     bit map. */  
   
     memset(classbits, 0, 32 * sizeof(uschar));  
   
3505      /* Process characters until ] is reached. By writing this as a "do" it      /* Process characters until ] is reached. By writing this as a "do" it
3506      means that an initial ] is taken as a data character. The first pass      means that an initial ] is taken as a data character. At the start of the
3507      through the regex checked the overall syntax, so we don't need to be very      loop, c contains the first byte of the character. */
     strict here. At the start of the loop, c contains the first byte of the  
     character. */  
3508    
3509      do      if (c != 0) do
3510        {        {
3511          const uschar *oldptr;
3512    
3513  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3514        if (utf8 && c > 127)        if (utf8 && c > 127)
3515          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3516          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3517          }          }
3518    
3519          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
3520          data and reset the pointer. This is so that very large classes that
3521          contain a zillion UTF-8 characters no longer overwrite the work space
3522          (which is on the stack). */
3523    
3524          if (lengthptr != NULL)
3525            {
3526            *lengthptr += class_utf8data - class_utf8data_base;
3527            class_utf8data = class_utf8data_base;
3528            }
3529    
3530  #endif  #endif
3531    
3532        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
3533    
3534        if (inescq)        if (inescq)
3535          {          {
3536          if (c == '\\' && ptr[1] == 'E')          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3537            {            {
3538            inescq = FALSE;            inescq = FALSE;                   /* Reset literal state */
3539            ptr++;            ptr++;                            /* Skip the 'E' */
3540            continue;            continue;                         /* Carry on with next */
3541            }            }
3542          else goto LONE_SINGLE_CHARACTER;          goto CHECK_RANGE;                   /* Could be range if \E follows */
3543          }          }
3544    
3545        /* Handle POSIX class names. Perl allows a negation extension of the        /* Handle POSIX class names. Perl allows a negation extension of the
# Line 1821  for (;; ptr++) Line 3548  for (;; ptr++)
3548        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3549        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3550    
3551        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3552            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3553            check_posix_syntax(ptr, &tempptr, cd))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3554          {          {
3555          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3556          int posix_class, i;          int posix_class, taboffset, tabopt;
3557          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3558            uschar pbits[32];
3559    
3560          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3561            {            {
3562            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3563            goto FAILED;            goto FAILED;
3564            }            }
3565    
3566          ptr += 2;          ptr += 2;
3567          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3568            {            {
3569            local_negate = TRUE;            local_negate = TRUE;
3570              should_flip_negation = TRUE;  /* Note negative special */
3571            ptr++;            ptr++;
3572            }            }
3573    
3574          posix_class = check_posix_name(ptr, tempptr - ptr);          posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3575          if (posix_class < 0)          if (posix_class < 0)
3576            {            {
3577            *errorcodeptr = ERR30;            *errorcodeptr = ERR30;
# Line 1856  for (;; ptr++) Line 3585  for (;; ptr++)
3585          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3586            posix_class = 0;            posix_class = 0;
3587    
3588          /* Or into the map we are building up to 3 of the static class          /* When PCRE_UCP is set, some of the POSIX classes are converted to
3589          tables, or their negations. The [:blank:] class sets up the same          different escape sequences that use Unicode properties. */
         chars as the [:space:] class (all white space). We remove the vertical  
         white space chars afterwards. */  
3590    
3591          posix_class *= 3;  #ifdef SUPPORT_UCP
3592          for (i = 0; i < 3; i++)          if ((options & PCRE_UCP) != 0)
3593            {            {
3594            BOOL blankclass = strncmp((char *)ptr, "blank", 5) == 0;            int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3595            int taboffset = posix_class_maps[posix_class + i];            if (posix_substitutes[pc] != NULL)
           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  
3596              {              {
3597              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+taboffset];              nestptr = tempptr + 1;
3598              if (blankclass) classbits[1] &= ~0x3c;              ptr = posix_substitutes[pc] - 1;
3599                continue;
3600              }              }
3601            }            }
3602