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

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

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

revision 81 by nigel, Sat Feb 24 21:40:59 2007 UTC revision 749 by ph10, Fri Nov 18 10:36:45 2011 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2005 University of Cambridge             Copyright (c) 1997-2011 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 42  POSSIBILITY OF SUCH DAMAGE. Line 42  POSSIBILITY OF SUCH DAMAGE.
42  supporting internal functions that are not used by other modules. */  supporting internal functions that are not used by other modules. */
43    
44    
45    #ifdef HAVE_CONFIG_H
46    #include "config.h"
47    #endif
48    
49    #define NLBLOCK cd             /* Block containing newline information */
50    #define PSSTART start_pattern  /* Field containing processed string start */
51    #define PSEND   end_pattern    /* Field containing processed string end */
52    
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56    /* When PCRE_DEBUG is defined, we need the pcre_printint() function, which is
57    also used by pcretest. PCRE_DEBUG is not defined when building a production
58    library. */
59    
60    #ifdef PCRE_DEBUG
61    #include "pcre_printint.src"
62    #endif
63    
64    
65    /* Macro for setting individual bits in class bitmaps. */
66    
67    #define SETBIT(a,b) a[b/8] |= (1 << (b%8))
68    
69    /* Maximum length value to check against when making sure that the integer that
70    holds the compiled pattern length does not overflow. We make it a bit less than
71    INT_MAX to allow for adding in group terminating bytes, so that we don't have
72    to check them every time. */
73    
74    #define OFLOW_MAX (INT_MAX - 20)
75    
76    
77  /*************************************************  /*************************************************
78  *      Code parameters and static tables         *  *      Code parameters and static tables         *
79  *************************************************/  *************************************************/
80    
81  /* Maximum number of items on the nested bracket stacks at compile time. This  /* This value specifies the size of stack workspace that is used during the
82  applies to the nesting of all kinds of parentheses. It does not limit  first pre-compile phase that determines how much memory is required. The regex
83  un-nested, non-capturing parentheses. This number can be made bigger if  is partly compiled into this space, but the compiled parts are discarded as
84  necessary - it is used to dimension one int and one unsigned char vector at  soon as they can be, so that hopefully there will never be an overrun. The code
85  compile time. */  does, however, check for an overrun. The largest amount I've seen used is 218,
86    so this number is very generous.
87    
88    The same workspace is used during the second, actual compile phase for
89    remembering forward references to groups so that they can be filled in at the
90    end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE
91    is 4 there is plenty of room. */
92    
93  #define BRASTACK_SIZE 200  #define COMPILE_WORK_SIZE (4096)
94    
95    /* The overrun tests check for a slightly smaller size so that they detect the
96    overrun before it actually does run off the end of the data block. */
97    
98    #define WORK_SIZE_CHECK (COMPILE_WORK_SIZE - 100)
99    
100    
101  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
# Line 63  are simple data values; negative values Line 103  are simple data values; negative values
103  on. Zero means further processing is needed (for things like \x), or the escape  on. Zero means further processing is needed (for things like \x), or the escape
104  is invalid. */  is invalid. */
105    
106  #if !EBCDIC   /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC
107    
108    /* This is the "normal" table for ASCII systems or for EBCDIC systems running
109    in UTF-8 mode. */
110    
111  static const short int escapes[] = {  static const short int escapes[] = {
112       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,                       0,
113       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,                       0,
114     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */       0,                       0,
115       0,      0,      0,      0,      0,      0,      0,      0,   /* H - O */       0,                       0,
116  -ESC_P, -ESC_Q,      0, -ESC_S,      0,      0,      0, -ESC_W,   /* P - W */       0,                       0,
117  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */       CHAR_COLON,              CHAR_SEMICOLON,
118     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
119       0,      0,      0,      0,      0,      0,  ESC_n,      0,   /* h - o */       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
120  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0,      0, -ESC_w,   /* p - w */       CHAR_COMMERCIAL_AT,      -ESC_A,
121       0,      0, -ESC_z                                            /* x - z */       -ESC_B,                  -ESC_C,
122         -ESC_D,                  -ESC_E,
123         0,                       -ESC_G,
124         -ESC_H,                  0,
125         0,                       -ESC_K,
126         0,                       0,
127         -ESC_N,                  0,
128         -ESC_P,                  -ESC_Q,
129         -ESC_R,                  -ESC_S,
130         0,                       0,
131         -ESC_V,                  -ESC_W,
132         -ESC_X,                  0,
133         -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,
134         CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,
135         CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,
136         CHAR_GRAVE_ACCENT,       7,
137         -ESC_b,                  0,
138         -ESC_d,                  ESC_e,
139         ESC_f,                   0,
140         -ESC_h,                  0,
141         0,                       -ESC_k,
142         0,                       0,
143         ESC_n,                   0,
144         -ESC_p,                  0,
145         ESC_r,                   -ESC_s,
146         ESC_tee,                 0,
147         -ESC_v,                  -ESC_w,
148         0,                       0,
149         -ESC_z
150  };  };
151    
152  #else         /* This is the "abnormal" table for EBCDIC systems */  #else
153    
154    /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
155    
156  static const short int escapes[] = {  static const short int escapes[] = {
157  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
158  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 87  static const short int escapes[] = { Line 162  static const short int escapes[] = {
162  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,
163  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',
164  /*  80 */     0,     7, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,  /*  80 */     0,     7, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,
165  /*  88 */     0,     0,      0,     '{',      0,     0,      0,      0,  /*  88 */-ESC_h,     0,      0,     '{',      0,     0,      0,      0,
166  /*  90 */     0,     0,      0,     'l',      0, ESC_n,      0, -ESC_p,  /*  90 */     0,     0, -ESC_k,     'l',      0, ESC_n,      0, -ESC_p,
167  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,
168  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,     0, -ESC_w,      0,  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,-ESC_v, -ESC_w,      0,
169  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,
170  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,
171  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
172  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,
173  /*  C8 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
174  /*  D0 */   '}',     0,      0,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,-ESC_N,      0, -ESC_P,
175  /*  D8 */-ESC_Q,     0,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
176  /*  E0 */  '\\',     0, -ESC_S,       0,      0,     0, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
177  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
178  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,
179  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0
# Line 106  static const short int escapes[] = { Line 181  static const short int escapes[] = {
181  #endif  #endif
182    
183    
184  /* Tables of names of POSIX character classes and their lengths. The list is  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
185  terminated by a zero length entry. The first three must be alpha, upper, lower,  searched linearly. Put all the names into a single string, in order to reduce
186  as this is assumed for handling case independence. */  the number of relocations when a shared library is dynamically linked. The
187    string is built from string macros so that it works in UTF-8 mode on EBCDIC
188  static const char *const posix_names[] = {  platforms. */
189    "alpha", "lower", "upper",  
190    "alnum", "ascii", "blank", "cntrl", "digit", "graph",  typedef struct verbitem {
191    "print", "punct", "space", "word",  "xdigit" };    int   len;                 /* Length of verb name */
192      int   op;                  /* Op when no arg, or -1 if arg mandatory */
193      int   op_arg;              /* Op when arg present, or -1 if not allowed */
194    } verbitem;
195    
196    static const char verbnames[] =
197      "\0"                       /* Empty name is a shorthand for MARK */
198      STRING_MARK0
199      STRING_ACCEPT0
200      STRING_COMMIT0
201      STRING_F0
202      STRING_FAIL0
203      STRING_PRUNE0
204      STRING_SKIP0
205      STRING_THEN;
206    
207    static const verbitem verbs[] = {
208      { 0, -1,        OP_MARK },
209      { 4, -1,        OP_MARK },
210      { 6, OP_ACCEPT, -1 },
211      { 6, OP_COMMIT, -1 },
212      { 1, OP_FAIL,   -1 },
213      { 4, OP_FAIL,   -1 },
214      { 5, OP_PRUNE,  OP_PRUNE_ARG },
215      { 4, OP_SKIP,   OP_SKIP_ARG  },
216      { 4, OP_THEN,   OP_THEN_ARG  }
217    };
218    
219    static const int verbcount = sizeof(verbs)/sizeof(verbitem);
220    
221    
222    /* Tables of names of POSIX character classes and their lengths. The names are
223    now all in a single string, to reduce the number of relocations when a shared
224    library is dynamically loaded. The list of lengths is terminated by a zero
225    length entry. The first three must be alpha, lower, upper, as this is assumed
226    for handling case independence. */
227    
228    static const char posix_names[] =
229      STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
230      STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
231      STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
232      STRING_word0  STRING_xdigit;
233    
234  static const uschar posix_name_lengths[] = {  static const uschar posix_name_lengths[] = {
235    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
236    
237  /* Table of class bit maps for each POSIX class; up to three may be combined  /* Table of class bit maps for each POSIX class. Each class is formed from a
238  to form the class. The table for [:blank:] is dynamically modified to remove  base map, with an optional addition or removal of another map. Then, for some
239  the vertical space characters. */  classes, there is some additional tweaking: for [:blank:] the vertical space
240    characters are removed, and for [:alpha:] and [:alnum:] the underscore
241    character is removed. The triples in the table consist of the base map offset,
242    second map offset or -1 if no second map, and a non-negative value for map
243    addition or a negative value for map subtraction (if there are two maps). The
244    absolute value of the third field has these meanings: 0 => no tweaking, 1 =>
245    remove vertical space characters, 2 => remove underscore. */
246    
247  static const int posix_class_maps[] = {  static const int posix_class_maps[] = {
248    cbit_lower, cbit_upper, -1,             /* alpha */    cbit_word,  cbit_digit, -2,             /* alpha */
249    cbit_lower, -1,         -1,             /* lower */    cbit_lower, -1,          0,             /* lower */
250    cbit_upper, -1,         -1,             /* upper */    cbit_upper, -1,          0,             /* upper */
251    cbit_digit, cbit_lower, cbit_upper,     /* alnum */    cbit_word,  -1,          2,             /* alnum - word without underscore */
252    cbit_print, cbit_cntrl, -1,             /* ascii */    cbit_print, cbit_cntrl,  0,             /* ascii */
253    cbit_space, -1,         -1,             /* blank - a GNU extension */    cbit_space, -1,          1,             /* blank - a GNU extension */
254    cbit_cntrl, -1,         -1,             /* cntrl */    cbit_cntrl, -1,          0,             /* cntrl */
255    cbit_digit, -1,         -1,             /* digit */    cbit_digit, -1,          0,             /* digit */
256    cbit_graph, -1,         -1,             /* graph */    cbit_graph, -1,          0,             /* graph */
257    cbit_print, -1,         -1,             /* print */    cbit_print, -1,          0,             /* print */
258    cbit_punct, -1,         -1,             /* punct */    cbit_punct, -1,          0,             /* punct */
259    cbit_space, -1,         -1,             /* space */    cbit_space, -1,          0,             /* space */
260    cbit_word,  -1,         -1,             /* word - a Perl extension */    cbit_word,  -1,          0,             /* word - a Perl extension */
261    cbit_xdigit,-1,         -1              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
262  };  };
263    
264    /* Table of substitutes for \d etc when PCRE_UCP is set. The POSIX class
265    substitutes must be in the order of the names, defined above, and there are
266    both positive and negative cases. NULL means no substitute. */
267    
268  /* The texts of compile-time error messages. These are "char *" because they  #ifdef SUPPORT_UCP
269  are passed to the outside world. */  static const uschar *substitutes[] = {
270      (uschar *)"\\P{Nd}",    /* \D */
271      (uschar *)"\\p{Nd}",    /* \d */
272      (uschar *)"\\P{Xsp}",   /* \S */       /* NOTE: Xsp is Perl space */
273      (uschar *)"\\p{Xsp}",   /* \s */
274      (uschar *)"\\P{Xwd}",   /* \W */
275      (uschar *)"\\p{Xwd}"    /* \w */
276    };
277    
278    static const uschar *posix_substitutes[] = {
279      (uschar *)"\\p{L}",     /* alpha */
280      (uschar *)"\\p{Ll}",    /* lower */
281      (uschar *)"\\p{Lu}",    /* upper */
282      (uschar *)"\\p{Xan}",   /* alnum */
283      NULL,                   /* ascii */
284      (uschar *)"\\h",        /* blank */
285      NULL,                   /* cntrl */
286      (uschar *)"\\p{Nd}",    /* digit */
287      NULL,                   /* graph */
288      NULL,                   /* print */
289      NULL,                   /* punct */
290      (uschar *)"\\p{Xps}",   /* space */    /* NOTE: Xps is POSIX space */
291      (uschar *)"\\p{Xwd}",   /* word */
292      NULL,                   /* xdigit */
293      /* Negated cases */
294      (uschar *)"\\P{L}",     /* ^alpha */
295      (uschar *)"\\P{Ll}",    /* ^lower */
296      (uschar *)"\\P{Lu}",    /* ^upper */
297      (uschar *)"\\P{Xan}",   /* ^alnum */
298      NULL,                   /* ^ascii */
299      (uschar *)"\\H",        /* ^blank */
300      NULL,                   /* ^cntrl */
301      (uschar *)"\\P{Nd}",    /* ^digit */
302      NULL,                   /* ^graph */
303      NULL,                   /* ^print */
304      NULL,                   /* ^punct */
305      (uschar *)"\\P{Xps}",   /* ^space */   /* NOTE: Xps is POSIX space */
306      (uschar *)"\\P{Xwd}",   /* ^word */
307      NULL                    /* ^xdigit */
308    };
309    #define POSIX_SUBSIZE (sizeof(posix_substitutes)/sizeof(uschar *))
310    #endif
311    
312    #define STRING(a)  # a
313    #define XSTRING(s) STRING(s)
314    
315  static const char *error_texts[] = {  /* The texts of compile-time error messages. These are "char *" because they
316    "no error",  are passed to the outside world. Do not ever re-use any error number, because
317    "\\ at end of pattern",  they are documented. Always add a new error instead. Messages marked DEAD below
318    "\\c at end of pattern",  are no longer used. This used to be a table of strings, but in order to reduce
319    "unrecognized character follows \\",  the number of relocations needed when a shared library is loaded dynamically,
320    "numbers out of order in {} quantifier",  it is now one long string. We cannot use a table of offsets, because the
321    lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
322    simply count through to the one we want - this isn't a performance issue
323    because these strings are used only when there is a compilation error.
324    
325    Each substring ends with \0 to insert a null character. This includes the final
326    substring, so that the whole string ends with \0\0, which can be detected when
327    counting through. */
328    
329    static const char error_texts[] =
330      "no error\0"
331      "\\ at end of pattern\0"
332      "\\c at end of pattern\0"
333      "unrecognized character follows \\\0"
334      "numbers out of order in {} quantifier\0"
335    /* 5 */    /* 5 */
336    "number too big in {} quantifier",    "number too big in {} quantifier\0"
337    "missing terminating ] for character class",    "missing terminating ] for character class\0"
338    "invalid escape sequence in character class",    "invalid escape sequence in character class\0"
339    "range out of order in character class",    "range out of order in character class\0"
340    "nothing to repeat",    "nothing to repeat\0"
341    /* 10 */    /* 10 */
342    "operand of unlimited repeat could match the empty string",    "operand of unlimited repeat could match the empty string\0"  /** DEAD **/
343    "internal error: unexpected repeat",    "internal error: unexpected repeat\0"
344    "unrecognized character after (?",    "unrecognized character after (? or (?-\0"
345    "POSIX named classes are supported only within a class",    "POSIX named classes are supported only within a class\0"
346    "missing )",    "missing )\0"
347    /* 15 */    /* 15 */
348    "reference to non-existent subpattern",    "reference to non-existent subpattern\0"
349    "erroffset passed as NULL",    "erroffset passed as NULL\0"
350    "unknown option bit(s) set",    "unknown option bit(s) set\0"
351    "missing ) after comment",    "missing ) after comment\0"
352    "parentheses nested too deeply",    "parentheses nested too deeply\0"  /** DEAD **/
353    /* 20 */    /* 20 */
354    "regular expression too large",    "regular expression is too large\0"
355    "failed to get memory",    "failed to get memory\0"
356    "unmatched parentheses",    "unmatched parentheses\0"
357    "internal error: code overflow",    "internal error: code overflow\0"
358    "unrecognized character after (?<",    "unrecognized character after (?<\0"
359    /* 25 */    /* 25 */
360    "lookbehind assertion is not fixed length",    "lookbehind assertion is not fixed length\0"
361    "malformed number after (?(",    "malformed number or name after (?(\0"
362    "conditional group contains more than two branches",    "conditional group contains more than two branches\0"
363    "assertion expected after (?(",    "assertion expected after (?(\0"
364    "(?R or (?digits must be followed by )",    "(?R or (?[+-]digits must be followed by )\0"
365    /* 30 */    /* 30 */
366    "unknown POSIX class name",    "unknown POSIX class name\0"
367    "POSIX collating elements are not supported",    "POSIX collating elements are not supported\0"
368    "this version of PCRE is not compiled with PCRE_UTF8 support",    "this version of PCRE is not compiled with PCRE_UTF8 support\0"
369    "spare error",    "spare error\0"  /** DEAD **/
370    "character value in \\x{...} sequence is too large",    "character value in \\x{...} sequence is too large\0"
371    /* 35 */    /* 35 */
372    "invalid condition (?(0)",    "invalid condition (?(0)\0"
373    "\\C not allowed in lookbehind assertion",    "\\C not allowed in lookbehind assertion\0"
374    "PCRE does not support \\L, \\l, \\N, \\U, or \\u",    "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
375    "number after (?C is > 255",    "number after (?C is > 255\0"
376    "closing ) for (?C expected",    "closing ) for (?C expected\0"
377    /* 40 */    /* 40 */
378    "recursive call could loop indefinitely",    "recursive call could loop indefinitely\0"
379    "unrecognized character after (?P",    "unrecognized character after (?P\0"
380    "syntax error after (?P",    "syntax error in subpattern name (missing terminator)\0"
381    "two named groups have the same name",    "two named subpatterns have the same name\0"
382    "invalid UTF-8 string",    "invalid UTF-8 string\0"
383    /* 45 */    /* 45 */
384    "support for \\P, \\p, and \\X has not been compiled",    "support for \\P, \\p, and \\X has not been compiled\0"
385    "malformed \\P or \\p sequence",    "malformed \\P or \\p sequence\0"
386    "unknown property name after \\P or \\p"    "unknown property name after \\P or \\p\0"
387  };    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
388      "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
389      /* 50 */
390      "repeated subpattern is too long\0"    /** DEAD **/
391      "octal value is greater than \\377 (not in UTF-8 mode)\0"
392      "internal error: overran compiling workspace\0"
393      "internal error: previously-checked referenced subpattern not found\0"
394      "DEFINE group contains more than one branch\0"
395      /* 55 */
396      "repeating a DEFINE group is not allowed\0"  /** 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      /* 70 */
414      "internal error: unknown opcode in find_fixedlength()\0"
415      ;
416    
417  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
418  patterns. Note that the tables in chartables are dependent on the locale, and  patterns. Note that the tables in chartables are dependent on the locale, and
# Line 220  For convenience, we use the same bit def Line 430  For convenience, we use the same bit def
430    
431  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
432    
433  #if !EBCDIC    /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC
434    
435    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
436    UTF-8 mode. */
437    
438  static const unsigned char digitab[] =  static const unsigned char digitab[] =
439    {    {
440    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
# Line 256  static const unsigned char digitab[] = Line 470  static const unsigned char digitab[] =
470    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
471    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
472    
473  #else          /* This is the "abnormal" case, for EBCDIC systems */  #else
474    
475    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
476    
477  static const unsigned char digitab[] =  static const unsigned char digitab[] =
478    {    {
479    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
# Line 270  static const unsigned char digitab[] = Line 487  static const unsigned char digitab[] =
487    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */
488    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */
489    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */
490    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88-     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88- 95    */
491    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */
492    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */
493    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */
# Line 304  static const unsigned char ebcdic_charta Line 521  static const unsigned char ebcdic_charta
521    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */
522    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */
523    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */
524    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88-  */    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88- 95 */
525    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */
526    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */
527    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */
# Line 331  static const unsigned char ebcdic_charta Line 548  static const unsigned char ebcdic_charta
548  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
549    
550  static BOOL  static BOOL
551    compile_regex(int, int, int *, uschar **, const uschar **, int *, BOOL, int,    compile_regex(int, uschar **, const uschar **, int *, BOOL, BOOL, int, int,
552      int *, int *, branch_chain *, compile_data *);      int *, int *, branch_chain *, compile_data *, int *);
553    
554    
555    
556    /*************************************************
557    *            Find an error text                  *
558    *************************************************/
559    
560    /* The error texts are now all in one long string, to save on relocations. As
561    some of the text is of unknown length, we can't use a table of offsets.
562    Instead, just count through the strings. This is not a performance issue
563    because it happens only when there has been a compilation error.
564    
565    Argument:   the error number
566    Returns:    pointer to the error string
567    */
568    
569    static const char *
570    find_error_text(int n)
571    {
572    const char *s = error_texts;
573    for (; n > 0; n--)
574      {
575      while (*s++ != 0) {};
576      if (*s == 0) return "Error text not found (please report)";
577      }
578    return s;
579    }
580    
581    
582    /*************************************************
583    *            Check for counted repeat            *
584    *************************************************/
585    
586    /* This function is called when a '{' is encountered in a place where it might
587    start a quantifier. It looks ahead to see if it really is a quantifier or not.
588    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
589    where the ddds are digits.
590    
591    Arguments:
592      p         pointer to the first char after '{'
593    
594    Returns:    TRUE or FALSE
595    */
596    
597    static BOOL
598    is_counted_repeat(const uschar *p)
599    {
600    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
601    while ((digitab[*p] & ctype_digit) != 0) p++;
602    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
603    
604    if (*p++ != CHAR_COMMA) return FALSE;
605    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
606    
607    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
608    while ((digitab[*p] & ctype_digit) != 0) p++;
609    
610    return (*p == CHAR_RIGHT_CURLY_BRACKET);
611    }
612    
613    
614    
# Line 342  static BOOL Line 618  static BOOL
618    
619  /* 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
620  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
621  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
622  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
623  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,
624    ptr is pointing at the \. On exit, it is on the final character of the escape
625    sequence.
626    
627  Arguments:  Arguments:
628    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
# Line 355  Arguments: Line 633  Arguments:
633    
634  Returns:         zero or positive => a data character  Returns:         zero or positive => a data character
635                   negative => a special escape sequence                   negative => a special escape sequence
636                   on error, errorptr is set                   on error, errorcodeptr is set
637  */  */
638    
639  static int  static int
640  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,
641    int options, BOOL isclass)    int options, BOOL isclass)
642  {  {
643  const uschar *ptr = *ptrptr;  BOOL utf8 = (options & PCRE_UTF8) != 0;
644    const uschar *ptr = *ptrptr + 1;
645  int c, i;  int c, i;
646    
647    GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
648    ptr--;                            /* Set pointer back to the last byte */
649    
650  /* 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. */
651    
 c = *(++ptr);  
652  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
653    
654  /* 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
655  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.
656  Otherwise further processing may be required. */  Otherwise further processing may be required. */
657    
658  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
659  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */
660  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0) c = i;
661    
662  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
663  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */
664  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
665  #endif  #endif
666    
# Line 388  else if ((i = escapes[c - 0x48]) != 0) Line 669  else if ((i = escapes[c - 0x48]) != 0)
669  else  else
670    {    {
671    const uschar *oldptr;    const uschar *oldptr;
672      BOOL braced, negated;
673    
674    switch (c)    switch (c)
675      {      {
676      /* 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
677      error. */      error. */
678    
679      case 'l':      case CHAR_l:
680      case 'L':      case CHAR_L:
     case 'N':  
     case 'u':  
     case 'U':  
681      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
682      break;      break;
683    
684        case CHAR_u:
685        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
686          {
687          /* In JavaScript, \u must be followed by four hexadecimal numbers.
688          Otherwise it is a lowercase u letter. */
689          if ((digitab[ptr[1]] & ctype_xdigit) != 0 && (digitab[ptr[2]] & ctype_xdigit) != 0
690               && (digitab[ptr[3]] & ctype_xdigit) != 0 && (digitab[ptr[4]] & ctype_xdigit) != 0)
691            {
692            c = 0;
693            for (i = 0; i < 4; ++i)
694              {
695              register int cc = *(++ptr);
696    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
697              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
698              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
699    #else           /* EBCDIC coding */
700              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
701              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
702    #endif
703              }
704            }
705          }
706        else
707          *errorcodeptr = ERR37;
708        break;
709    
710        case CHAR_U:
711        /* In JavaScript, \U is an uppercase U letter. */
712        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
713        break;
714    
715        /* In a character class, \g is just a literal "g". Outside a character
716        class, \g must be followed by one of a number of specific things:
717    
718        (1) A number, either plain or braced. If positive, it is an absolute
719        backreference. If negative, it is a relative backreference. This is a Perl
720        5.10 feature.
721    
722        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
723        is part of Perl's movement towards a unified syntax for back references. As
724        this is synonymous with \k{name}, we fudge it up by pretending it really
725        was \k.
726    
727        (3) For Oniguruma compatibility we also support \g followed by a name or a
728        number either in angle brackets or in single quotes. However, these are
729        (possibly recursive) subroutine calls, _not_ backreferences. Just return
730        the -ESC_g code (cf \k). */
731    
732        case CHAR_g:
733        if (isclass) break;
734        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
735          {
736          c = -ESC_g;
737          break;
738          }
739    
740        /* Handle the Perl-compatible cases */
741    
742        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
743          {
744          const uschar *p;
745          for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
746            if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
747          if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
748            {
749            c = -ESC_k;
750            break;
751            }
752          braced = TRUE;
753          ptr++;
754          }
755        else braced = FALSE;
756    
757        if (ptr[1] == CHAR_MINUS)
758          {
759          negated = TRUE;
760          ptr++;
761          }
762        else negated = FALSE;
763    
764        c = 0;
765        while ((digitab[ptr[1]] & ctype_digit) != 0)
766          c = c * 10 + *(++ptr) - CHAR_0;
767    
768        if (c < 0)   /* Integer overflow */
769          {
770          *errorcodeptr = ERR61;
771          break;
772          }
773    
774        if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
775          {
776          *errorcodeptr = ERR57;
777          break;
778          }
779    
780        if (c == 0)
781          {
782          *errorcodeptr = ERR58;
783          break;
784          }
785    
786        if (negated)
787          {
788          if (c > bracount)
789            {
790            *errorcodeptr = ERR15;
791            break;
792            }
793          c = bracount - (c - 1);
794          }
795    
796        c = -(ESC_REF + c);
797        break;
798    
799      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
800      starting with one that is not zero is not straightforward. By experiment,      starting with one that is not zero is not straightforward. By experiment,
801      the way Perl works seems to be as follows:      the way Perl works seems to be as follows:
# Line 413  else Line 808  else
808      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
809      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
810    
811      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:
812      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
813    
814      if (!isclass)      if (!isclass)
815        {        {
816        oldptr = ptr;        oldptr = ptr;
817        c -= '0';        c -= CHAR_0;
818        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
819          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - CHAR_0;
820          if (c < 0)    /* Integer overflow */
821            {
822            *errorcodeptr = ERR61;
823            break;
824            }
825        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
826          {          {
827          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 434  else Line 834  else
834      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.
835      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
836    
837      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
838        {        {
839        ptr--;        ptr--;
840        c = 0;        c = 0;
# Line 442  else Line 842  else
842        }        }
843    
844      /* \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
845      larger first octal digit. */      larger first octal digit. The original code used just to take the least
846        significant 8 bits of octal numbers (I think this is what early Perls used
847      case '0':      to do). Nowadays we allow for larger numbers in UTF-8 mode, but no more
848      c -= '0';      than 3 octal digits. */
849      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')  
850          c = c * 8 + *(++ptr) - '0';      case CHAR_0:
851      c &= 255;     /* Take least significant 8 bits */      c -= CHAR_0;
852        while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
853            c = c * 8 + *(++ptr) - CHAR_0;
854        if (!utf8 && c > 255) *errorcodeptr = ERR51;
855      break;      break;
856    
857      /* \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
858      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
859        treated as a data character. */
860    
861        case CHAR_x:
862        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
863          {
864          /* In JavaScript, \x must be followed by two hexadecimal numbers.
865          Otherwise it is a lowercase x letter. */
866          if ((digitab[ptr[1]] & ctype_xdigit) != 0 && (digitab[ptr[2]] & ctype_xdigit) != 0)
867            {
868            c = 0;
869            for (i = 0; i < 2; ++i)
870              {
871              register int cc = *(++ptr);
872    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
873              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
874              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
875    #else           /* EBCDIC coding */
876              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
877              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
878    #endif
879              }
880            }
881          break;
882          }
883    
884      case 'x':      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
 #ifdef SUPPORT_UTF8  
     if (ptr[1] == '{' && (options & PCRE_UTF8) != 0)  
885        {        {
886        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
887        register int count = 0;        int count = 0;
888    
889        c = 0;        c = 0;
890        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
891          {          {
892          int cc = *pt++;          register int cc = *pt++;
893            if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
894          count++;          count++;
895  #if !EBCDIC    /* ASCII coding */  
896          if (cc >= 'a') cc -= 32;               /* Convert to upper case */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
897          c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
898  #else          /* EBCDIC coding */          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
899          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */  #else           /* EBCDIC coding */
900          c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
901            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
902  #endif  #endif
903          }          }
904        if (*pt == '}')  
905          if (*pt == CHAR_RIGHT_CURLY_BRACKET)
906          {          {
907          if (c < 0 || count > 8) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
908          ptr = pt;          ptr = pt;
909          break;          break;
910          }          }
911    
912        /* 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
913        recognize this construct; fall through to the normal \x handling. */        recognize this construct; fall through to the normal \x handling. */
914        }        }
 #endif  
915    
916      /* Read just a single hex char */      /* Read just a single-byte hex-defined char */
917    
918      c = 0;      c = 0;
919      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
920        {        {
921        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
922        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
923  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
924        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
925        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
926  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
927        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
928        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
929  #endif  #endif
930        }        }
931      break;      break;
932    
933      /* Other special escapes not starting with a digit are straightforward */      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
934        An error is given if the byte following \c is not an ASCII character. This
935        coding is ASCII-specific, but then the whole concept of \cx is
936        ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
937    
938      case 'c':      case CHAR_c:
939      c = *(++ptr);      c = *(++ptr);
940      if (c == 0)      if (c == 0)
941        {        {
942        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
943        return 0;        break;
944        }        }
945    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
946      /* A letter is upper-cased; then the 0x40 bit is flipped. This coding      if (c > 127)  /* Excludes all non-ASCII in either mode */
947      is ASCII-specific, but then the whole concept of \cx is ASCII-specific.        {
948      (However, an EBCDIC equivalent has now been added.) */        *errorcodeptr = ERR68;
949          break;
950  #if !EBCDIC    /* ASCII coding */        }
951      if (c >= 'a' && c <= 'z') c -= 32;      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
952      c ^= 0x40;      c ^= 0x40;
953  #else          /* EBCDIC coding */  #else             /* EBCDIC coding */
954      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
955      c ^= 0xC0;      c ^= 0xC0;
956  #endif  #endif
957      break;      break;
958    
959      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
960      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
961      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
962      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
963      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
964    
965      default:      default:
966      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 541  else Line 973  else
973      }      }
974    }    }
975    
976    /* Perl supports \N{name} for character names, as well as plain \N for "not
977    newline". PCRE does not support \N{name}. However, it does support
978    quantification such as \N{2,3}. */
979    
980    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
981         !is_counted_repeat(ptr+2))
982      *errorcodeptr = ERR37;
983    
984    /* If PCRE_UCP is set, we change the values for \d etc. */
985    
986    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
987      c -= (ESC_DU - ESC_D);
988    
989    /* Set the pointer to the final character before returning. */
990    
991  *ptrptr = ptr;  *ptrptr = ptr;
992  return c;  return c;
993  }  }
# Line 560  escape sequence. Line 1007  escape sequence.
1007  Argument:  Argument:
1008    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
1009    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
1010      dptr           points to an int that is set to the detailed property value
1011    errorcodeptr   points to the error code variable    errorcodeptr   points to the error code variable
1012    
1013  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
1014  */  */
1015    
1016  static int  static int
1017  get_ucp(const uschar **ptrptr, BOOL *negptr, int *errorcodeptr)  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)
1018  {  {
1019  int c, i, bot, top;  int c, i, bot, top;
1020  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
1021  char name[4];  char name[32];
1022    
1023  c = *(++ptr);  c = *(++ptr);
1024  if (c == 0) goto ERROR_RETURN;  if (c == 0) goto ERROR_RETURN;
1025    
1026  *negptr = FALSE;  *negptr = FALSE;
1027    
1028  /* \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
1029  preceded by ^ for negation. */  negation. */
1030    
1031  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
1032    {    {
1033    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1034      {      {
1035      *negptr = TRUE;      *negptr = TRUE;
1036      ptr++;      ptr++;
1037      }      }
1038    for (i = 0; i <= 2; i++)    for (i = 0; i < (int)sizeof(name) - 1; i++)
1039      {      {
1040      c = *(++ptr);      c = *(++ptr);
1041      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
1042      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1043      name[i] = c;      name[i] = c;
1044      }      }
1045    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;  
     }  
1046    name[i] = 0;    name[i] = 0;
1047    }    }
1048    
# Line 619  top = _pcre_utt_size; Line 1063  top = _pcre_utt_size;
1063    
1064  while (bot < top)  while (bot < top)
1065    {    {
1066    i = (bot + top)/2;    i = (bot + top) >> 1;
1067    c = strcmp(name, _pcre_utt[i].name);    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);
1068    if (c == 0) return _pcre_utt[i].value;    if (c == 0)
1069        {
1070        *dptr = _pcre_utt[i].value;
1071        return _pcre_utt[i].type;
1072        }
1073    if (c > 0) bot = i + 1; else top = i;    if (c > 0) bot = i + 1; else top = i;
1074    }    }
1075    
 UNKNOWN_RETURN:  
1076  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
1077  *ptrptr = ptr;  *ptrptr = ptr;
1078  return -1;  return -1;
# Line 641  return -1; Line 1088  return -1;
1088    
1089    
1090  /*************************************************  /*************************************************
 *            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 == '}');  
 }  
   
   
   
 /*************************************************  
1091  *         Read repeat counts                     *  *         Read repeat counts                     *
1092  *************************************************/  *************************************************/
1093    
# Line 701  int max = -1; Line 1115  int max = -1;
1115  /* 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
1116  an integer overflow. */  an integer overflow. */
1117    
1118  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;
1119  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1120    {    {
1121    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 711  if (min < 0 || min > 65535) Line 1125  if (min < 0 || min > 65535)
1125  /* 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.
1126  Also, max must not be less than min. */  Also, max must not be less than min. */
1127    
1128  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1129    {    {
1130    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1131      {      {
1132      max = 0;      max = 0;
1133      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;
1134      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1135        {        {
1136        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 741  return p; Line 1155  return p;
1155    
1156    
1157  /*************************************************  /*************************************************
1158  *      Find first significant op code            *  *  Subroutine for finding forward reference      *
1159  *************************************************/  *************************************************/
1160    
1161  /* This is called by several functions that scan a compiled expression looking  /* This recursive function is called only from find_parens() below. The
1162  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
1163  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
1164  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
1165  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
1166    returns when it reaches a given numbered subpattern. Recursion is used to keep
1167    track of subpatterns that reset the capturing group numbers - the (?| feature.
1168    
1169    This function was originally called only from the second pass, in which we know
1170    that if (?< or (?' or (?P< is encountered, the name will be correctly
1171    terminated because that is checked in the first pass. There is now one call to
1172    this function in the first pass, to check for a recursive back reference by
1173    name (so that we can make the whole group atomic). In this case, we need check
1174    only up to the current position in the pattern, and that is still OK because
1175    and previous occurrences will have been checked. To make this work, the test
1176    for "end of pattern" is a check against cd->end_pattern in the main loop,
1177    instead of looking for a binary zero. This means that the special first-pass
1178    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1179    processing items within the loop are OK, because afterwards the main loop will
1180    terminate.)
1181    
1182  Arguments:  Arguments:
1183    code         pointer to the start of the group    ptrptr       address of the current character pointer (updated)
1184    options      pointer to external options    cd           compile background data
1185    optbit       the option bit whose changing is significant, or    name         name to seek, or NULL if seeking a numbered subpattern
1186                   zero if none are    lorn         name length, or subpattern number if name is NULL
1187    skipassert   TRUE if certain assertions are to be skipped    xmode        TRUE if we are in /x mode
1188      utf8         TRUE if we are in UTF-8 mode
1189      count        pointer to the current capturing subpattern number (updated)
1190    
1191  Returns:       pointer to the first significant opcode  Returns:       the number of the named subpattern, or -1 if not found
1192  */  */
1193    
1194  static const uschar*  static int
1195  first_significant_code(const uschar *code, int *options, int optbit,  find_parens_sub(uschar **ptrptr, compile_data *cd, const uschar *name, int lorn,
1196    BOOL skipassert)    BOOL xmode, BOOL utf8, int *count)
1197  {  {
1198  for (;;)  uschar *ptr = *ptrptr;
1199    int start_count = *count;
1200    int hwm_count = start_count;
1201    BOOL dup_parens = FALSE;
1202    
1203    /* If the first character is a parenthesis, check on the type of group we are
1204    dealing with. The very first call may not start with a parenthesis. */
1205    
1206    if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1207    {    {
1208    switch ((int)*code)    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
     {  
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
1209    
1210      case OP_ASSERT_NOT:    if (ptr[1] == CHAR_ASTERISK) ptr += 2;
     case OP_ASSERTBACK:  
     case OP_ASSERTBACK_NOT:  
     if (!skipassert) return code;  
     do code += GET(code, 1); while (*code == OP_ALT);  
     code += _pcre_OP_lengths[*code];  
     break;  
1211    
1212      case OP_WORD_BOUNDARY:    /* Handle a normal, unnamed capturing parenthesis. */
     case OP_NOT_WORD_BOUNDARY:  
     if (!skipassert) return code;  
     /* Fall through */  
1213    
1214      case OP_CALLOUT:    else if (ptr[1] != CHAR_QUESTION_MARK)
1215      case OP_CREF:      {
1216      case OP_BRANUMBER:      *count += 1;
1217      code += _pcre_OP_lengths[*code];      if (name == NULL && *count == lorn) return *count;
1218      break;      ptr++;
1219        }
1220    
1221      default:    /* All cases now have (? at the start. Remember when we are in a group
1222      return code;    where the parenthesis numbers are duplicated. */
1223    
1224      else if (ptr[2] == CHAR_VERTICAL_LINE)
1225        {
1226        ptr += 3;
1227        dup_parens = TRUE;
1228      }      }
   }  
 /* Control never reaches here */  
 }  
1229    
1230      /* Handle comments; all characters are allowed until a ket is reached. */
1231    
1232      else if (ptr[2] == CHAR_NUMBER_SIGN)
1233        {
1234        for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1235        goto FAIL_EXIT;
1236        }
1237    
1238      /* Handle a condition. If it is an assertion, just carry on so that it
1239      is processed as normal. If not, skip to the closing parenthesis of the
1240      condition (there can't be any nested parens). */
1241    
1242  /*************************************************    else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1243  *        Find the fixed length of a pattern      *      {
1244  *************************************************/      ptr += 2;
1245        if (ptr[1] != CHAR_QUESTION_MARK)
1246          {
1247          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1248          if (*ptr != 0) ptr++;
1249          }
1250        }
1251    
1252  /* Scan a pattern and compute the fixed length of subject that will match it,    /* Start with (? but not a condition. */
 if the length is fixed. This is needed for dealing with backward assertions.  
 In UTF8 mode, the result is in characters rather than bytes.  
1253    
1254  Arguments:    else
1255    code     points to the start of the pattern (the bracket)      {
1256    options  the compiling options      ptr += 2;
1257        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1258    
1259  Returns:   the fixed length, or -1 if there is no fixed length,      /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
              or -2 if \C was encountered  
 */  
1260    
1261  static int      if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1262  find_fixedlength(uschar *code, int options)          ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1263  {        {
1264  int length = -1;        int term;
1265          const uschar *thisname;
1266          *count += 1;
1267          if (name == NULL && *count == lorn) return *count;
1268          term = *ptr++;
1269          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1270          thisname = ptr;
1271          while (*ptr != term) ptr++;
1272          if (name != NULL && lorn == ptr - thisname &&
1273              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1274            return *count;
1275          term++;
1276          }
1277        }
1278      }
1279    
1280  register int branchlength = 0;  /* Past any initial parenthesis handling, scan for parentheses or vertical
1281  register uschar *cc = code + 1 + LINK_SIZE;  bars. Stop if we get to cd->end_pattern. Note that this is important for the
1282    first-pass call when this value is temporarily adjusted to stop at the current
1283    position. So DO NOT change this to a test for binary zero. */
1284    
1285    for (; ptr < cd->end_pattern; ptr++)
1286      {
1287      /* Skip over backslashed characters and also entire \Q...\E */
1288    
1289      if (*ptr == CHAR_BACKSLASH)
1290        {
1291        if (*(++ptr) == 0) goto FAIL_EXIT;
1292        if (*ptr == CHAR_Q) for (;;)
1293          {
1294          while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1295          if (*ptr == 0) goto FAIL_EXIT;
1296          if (*(++ptr) == CHAR_E) break;
1297          }
1298        continue;
1299        }
1300    
1301      /* Skip over character classes; this logic must be similar to the way they
1302      are handled for real. If the first character is '^', skip it. Also, if the
1303      first few characters (either before or after ^) are \Q\E or \E we skip them
1304      too. This makes for compatibility with Perl. Note the use of STR macros to
1305      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1306    
1307      if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1308        {
1309        BOOL negate_class = FALSE;
1310        for (;;)
1311          {
1312          if (ptr[1] == CHAR_BACKSLASH)
1313            {
1314            if (ptr[2] == CHAR_E)
1315              ptr+= 2;
1316            else if (strncmp((const char *)ptr+2,
1317                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1318              ptr += 4;
1319            else
1320              break;
1321            }
1322          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1323            {
1324            negate_class = TRUE;
1325            ptr++;
1326            }
1327          else break;
1328          }
1329    
1330        /* If the next character is ']', it is a data character that must be
1331        skipped, except in JavaScript compatibility mode. */
1332    
1333        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1334            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1335          ptr++;
1336    
1337        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1338          {
1339          if (*ptr == 0) return -1;
1340          if (*ptr == CHAR_BACKSLASH)
1341            {
1342            if (*(++ptr) == 0) goto FAIL_EXIT;
1343            if (*ptr == CHAR_Q) for (;;)
1344              {
1345              while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1346              if (*ptr == 0) goto FAIL_EXIT;
1347              if (*(++ptr) == CHAR_E) break;
1348              }
1349            continue;
1350            }
1351          }
1352        continue;
1353        }
1354    
1355      /* Skip comments in /x mode */
1356    
1357      if (xmode && *ptr == CHAR_NUMBER_SIGN)
1358        {
1359        ptr++;
1360        while (*ptr != 0)
1361          {
1362          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1363          ptr++;
1364    #ifdef SUPPORT_UTF8
1365          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
1366    #endif
1367          }
1368        if (*ptr == 0) goto FAIL_EXIT;
1369        continue;
1370        }
1371    
1372      /* Check for the special metacharacters */
1373    
1374      if (*ptr == CHAR_LEFT_PARENTHESIS)
1375        {
1376        int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, count);
1377        if (rc > 0) return rc;
1378        if (*ptr == 0) goto FAIL_EXIT;
1379        }
1380    
1381      else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1382        {
1383        if (dup_parens && *count < hwm_count) *count = hwm_count;
1384        goto FAIL_EXIT;
1385        }
1386    
1387      else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1388        {
1389        if (*count > hwm_count) hwm_count = *count;
1390        *count = start_count;
1391        }
1392      }
1393    
1394    FAIL_EXIT:
1395    *ptrptr = ptr;
1396    return -1;
1397    }
1398    
1399    
1400    
1401    
1402    /*************************************************
1403    *       Find forward referenced subpattern       *
1404    *************************************************/
1405    
1406    /* This function scans along a pattern's text looking for capturing
1407    subpatterns, and counting them. If it finds a named pattern that matches the
1408    name it is given, it returns its number. Alternatively, if the name is NULL, it
1409    returns when it reaches a given numbered subpattern. This is used for forward
1410    references to subpatterns. We used to be able to start this scan from the
1411    current compiling point, using the current count value from cd->bracount, and
1412    do it all in a single loop, but the addition of the possibility of duplicate
1413    subpattern numbers means that we have to scan from the very start, in order to
1414    take account of such duplicates, and to use a recursive function to keep track
1415    of the different types of group.
1416    
1417    Arguments:
1418      cd           compile background data
1419      name         name to seek, or NULL if seeking a numbered subpattern
1420      lorn         name length, or subpattern number if name is NULL
1421      xmode        TRUE if we are in /x mode
1422      utf8         TRUE if we are in UTF-8 mode
1423    
1424    Returns:       the number of the found subpattern, or -1 if not found
1425    */
1426    
1427    static int
1428    find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode,
1429      BOOL utf8)
1430    {
1431    uschar *ptr = (uschar *)cd->start_pattern;
1432    int count = 0;
1433    int rc;
1434    
1435    /* If the pattern does not start with an opening parenthesis, the first call
1436    to find_parens_sub() will scan right to the end (if necessary). However, if it
1437    does start with a parenthesis, find_parens_sub() will return when it hits the
1438    matching closing parens. That is why we have to have a loop. */
1439    
1440    for (;;)
1441      {
1442      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, &count);
1443      if (rc > 0 || *ptr++ == 0) break;
1444      }
1445    
1446    return rc;
1447    }
1448    
1449    
1450    
1451    
1452    /*************************************************
1453    *      Find first significant op code            *
1454    *************************************************/
1455    
1456    /* This is called by several functions that scan a compiled expression looking
1457    for a fixed first character, or an anchoring op code etc. It skips over things
1458    that do not influence this. For some calls, it makes sense to skip negative
1459    forward and all backward assertions, and also the \b assertion; for others it
1460    does not.
1461    
1462    Arguments:
1463      code         pointer to the start of the group
1464      skipassert   TRUE if certain assertions are to be skipped
1465    
1466    Returns:       pointer to the first significant opcode
1467    */
1468    
1469    static const uschar*
1470    first_significant_code(const uschar *code, BOOL skipassert)
1471    {
1472    for (;;)
1473      {
1474      switch ((int)*code)
1475        {
1476        case OP_ASSERT_NOT:
1477        case OP_ASSERTBACK:
1478        case OP_ASSERTBACK_NOT:
1479        if (!skipassert) return code;
1480        do code += GET(code, 1); while (*code == OP_ALT);
1481        code += _pcre_OP_lengths[*code];
1482        break;
1483    
1484        case OP_WORD_BOUNDARY:
1485        case OP_NOT_WORD_BOUNDARY:
1486        if (!skipassert) return code;
1487        /* Fall through */
1488    
1489        case OP_CALLOUT:
1490        case OP_CREF:
1491        case OP_NCREF:
1492        case OP_RREF:
1493        case OP_NRREF:
1494        case OP_DEF:
1495        code += _pcre_OP_lengths[*code];
1496        break;
1497    
1498        default:
1499        return code;
1500        }
1501      }
1502    /* Control never reaches here */
1503    }
1504    
1505    
1506    
1507    
1508    /*************************************************
1509    *        Find the fixed length of a branch       *
1510    *************************************************/
1511    
1512    /* Scan a branch and compute the fixed length of subject that will match it,
1513    if the length is fixed. This is needed for dealing with backward assertions.
1514    In UTF8 mode, the result is in characters rather than bytes. The branch is
1515    temporarily terminated with OP_END when this function is called.
1516    
1517    This function is called when a backward assertion is encountered, so that if it
1518    fails, the error message can point to the correct place in the pattern.
1519    However, we cannot do this when the assertion contains subroutine calls,
1520    because they can be forward references. We solve this by remembering this case
1521    and doing the check at the end; a flag specifies which mode we are running in.
1522    
1523    Arguments:
1524      code     points to the start of the pattern (the bracket)
1525      utf8     TRUE in UTF-8 mode
1526      atend    TRUE if called when the pattern is complete
1527      cd       the "compile data" structure
1528    
1529    Returns:   the fixed length,
1530                 or -1 if there is no fixed length,
1531                 or -2 if \C was encountered
1532                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1533                 or -4 if an unknown opcode was encountered (internal error)
1534    */
1535    
1536    static int
1537    find_fixedlength(uschar *code, BOOL utf8, BOOL atend, compile_data *cd)
1538    {
1539    int length = -1;
1540    
1541    register int branchlength = 0;
1542    register uschar *cc = code + 1 + LINK_SIZE;
1543    
1544  /* Scan along the opcodes for this branch. If we get to the end of the  /* Scan along the opcodes for this branch. If we get to the end of the
1545  branch, check the length against that of the other branches. */  branch, check the length against that of the other branches. */
# Line 833  branch, check the length against that of Line 1547  branch, check the length against that of
1547  for (;;)  for (;;)
1548    {    {
1549    int d;    int d;
1550      uschar *ce, *cs;
1551    register int op = *cc;    register int op = *cc;
   if (op >= OP_BRA) op = OP_BRA;  
   
1552    switch (op)    switch (op)
1553      {      {
1554        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1555        OP_BRA (normal non-capturing bracket) because the other variants of these
1556        opcodes are all concerned with unlimited repeated groups, which of course
1557        are not of fixed length. */
1558    
1559        case OP_CBRA:
1560      case OP_BRA:      case OP_BRA:
1561      case OP_ONCE:      case OP_ONCE:
1562        case OP_ONCE_NC:
1563      case OP_COND:      case OP_COND:
1564      d = find_fixedlength(cc, options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), utf8, atend, cd);
1565      if (d < 0) return d;      if (d < 0) return d;
1566      branchlength += d;      branchlength += d;
1567      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1568      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1569      break;      break;
1570    
1571      /* Reached end of a branch; if it's a ket it is the end of a nested      /* Reached end of a branch; if it's a ket it is the end of a nested call.
1572      call. If it's ALT it is an alternation in a nested call. If it is      If it's ALT it is an alternation in a nested call. An ACCEPT is effectively
1573      END it's the end of the outer call. All can be handled by the same code. */      an ALT. If it is END it's the end of the outer call. All can be handled by
1574        the same code. Note that we must not include the OP_KETRxxx opcodes here,
1575        because they all imply an unlimited repeat. */
1576    
1577      case OP_ALT:      case OP_ALT:
1578      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1579      case OP_END:      case OP_END:
1580        case OP_ACCEPT:
1581        case OP_ASSERT_ACCEPT:
1582      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1583        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
1584      if (*cc != OP_ALT) return length;      if (*cc != OP_ALT) return length;
# Line 864  for (;;) Line 1586  for (;;)
1586      branchlength = 0;      branchlength = 0;
1587      break;      break;
1588    
1589        /* A true recursion implies not fixed length, but a subroutine call may
1590        be OK. If the subroutine is a forward reference, we can't deal with
1591        it until the end of the pattern, so return -3. */
1592    
1593        case OP_RECURSE:
1594        if (!atend) return -3;
1595        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1596        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1597        if (cc > cs && cc < ce) return -1;                /* Recursion */
1598        d = find_fixedlength(cs + 2, utf8, atend, cd);
1599        if (d < 0) return d;
1600        branchlength += d;
1601        cc += 1 + LINK_SIZE;
1602        break;
1603    
1604      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1605    
1606      case OP_ASSERT:      case OP_ASSERT:
# Line 875  for (;;) Line 1612  for (;;)
1612    
1613      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1614    
1615      case OP_REVERSE:      case OP_MARK:
1616      case OP_BRANUMBER:      case OP_PRUNE_ARG:
1617      case OP_CREF:      case OP_SKIP_ARG:
1618      case OP_OPT:      case OP_THEN_ARG:
1619        cc += cc[1] + _pcre_OP_lengths[*cc];
1620        break;
1621    
1622      case OP_CALLOUT:      case OP_CALLOUT:
     case OP_SOD:  
     case OP_SOM:  
     case OP_EOD:  
     case OP_EODN:  
1623      case OP_CIRC:      case OP_CIRC:
1624        case OP_CIRCM:
1625        case OP_CLOSE:
1626        case OP_COMMIT:
1627        case OP_CREF:
1628        case OP_DEF:
1629      case OP_DOLL:      case OP_DOLL:
1630        case OP_DOLLM:
1631        case OP_EOD:
1632        case OP_EODN:
1633        case OP_FAIL:
1634        case OP_NCREF:
1635        case OP_NRREF:
1636      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1637        case OP_PRUNE:
1638        case OP_REVERSE:
1639        case OP_RREF:
1640        case OP_SET_SOM:
1641        case OP_SKIP:
1642        case OP_SOD:
1643        case OP_SOM:
1644        case OP_THEN:
1645      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1646      cc += _pcre_OP_lengths[*cc];      cc += _pcre_OP_lengths[*cc];
1647      break;      break;
# Line 894  for (;;) Line 1649  for (;;)
1649      /* Handle literal characters */      /* Handle literal characters */
1650    
1651      case OP_CHAR:      case OP_CHAR:
1652      case OP_CHARNC:      case OP_CHARI:
1653        case OP_NOT:
1654        case OP_NOTI:
1655      branchlength++;      branchlength++;
1656      cc += 2;      cc += 2;
1657  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1658      if ((options & PCRE_UTF8) != 0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       {  
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1659  #endif  #endif
1660      break;      break;
1661    
# Line 909  for (;;) Line 1663  for (;;)
1663      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1664    
1665      case OP_EXACT:      case OP_EXACT:
1666        case OP_EXACTI:
1667        case OP_NOTEXACT:
1668        case OP_NOTEXACTI:
1669      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1670      cc += 4;      cc += 4;
1671  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1672      if ((options & PCRE_UTF8) != 0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       {  
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1673  #endif  #endif
1674      break;      break;
1675    
1676      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1677      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1678        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1679      cc += 4;      cc += 4;
1680      break;      break;
1681    
# Line 928  for (;;) Line 1683  for (;;)
1683    
1684      case OP_PROP:      case OP_PROP:
1685      case OP_NOTPROP:      case OP_NOTPROP:
1686      cc++;      cc += 2;
1687      /* Fall through */      /* Fall through */
1688    
1689        case OP_HSPACE:
1690        case OP_VSPACE:
1691        case OP_NOT_HSPACE:
1692        case OP_NOT_VSPACE:
1693      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
1694      case OP_DIGIT:      case OP_DIGIT:
1695      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
# Line 938  for (;;) Line 1697  for (;;)
1697      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1698      case OP_WORDCHAR:      case OP_WORDCHAR:
1699      case OP_ANY:      case OP_ANY:
1700        case OP_ALLANY:
1701      branchlength++;      branchlength++;
1702      cc++;      cc++;
1703      break;      break;
# Line 961  for (;;) Line 1721  for (;;)
1721    
1722      switch (*cc)      switch (*cc)
1723        {        {
1724          case OP_CRPLUS:
1725          case OP_CRMINPLUS:
1726        case OP_CRSTAR:        case OP_CRSTAR:
1727        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1728        case OP_CRQUERY:        case OP_CRQUERY:
# Line 981  for (;;) Line 1743  for (;;)
1743    
1744      /* Anything else is variable length */      /* Anything else is variable length */
1745    
1746      default:      case OP_ANYNL:
1747        case OP_BRAMINZERO:
1748        case OP_BRAPOS:
1749        case OP_BRAPOSZERO:
1750        case OP_BRAZERO:
1751        case OP_CBRAPOS:
1752        case OP_EXTUNI:
1753        case OP_KETRMAX:
1754        case OP_KETRMIN:
1755        case OP_KETRPOS:
1756        case OP_MINPLUS:
1757        case OP_MINPLUSI:
1758        case OP_MINQUERY:
1759        case OP_MINQUERYI:
1760        case OP_MINSTAR:
1761        case OP_MINSTARI:
1762        case OP_MINUPTO:
1763        case OP_MINUPTOI:
1764        case OP_NOTMINPLUS:
1765        case OP_NOTMINPLUSI:
1766        case OP_NOTMINQUERY:
1767        case OP_NOTMINQUERYI:
1768        case OP_NOTMINSTAR:
1769        case OP_NOTMINSTARI:
1770        case OP_NOTMINUPTO:
1771        case OP_NOTMINUPTOI:
1772        case OP_NOTPLUS:
1773        case OP_NOTPLUSI:
1774        case OP_NOTPOSPLUS:
1775        case OP_NOTPOSPLUSI:
1776        case OP_NOTPOSQUERY:
1777        case OP_NOTPOSQUERYI:
1778        case OP_NOTPOSSTAR:
1779        case OP_NOTPOSSTARI:
1780        case OP_NOTPOSUPTO:
1781        case OP_NOTPOSUPTOI:
1782        case OP_NOTQUERY:
1783        case OP_NOTQUERYI:
1784        case OP_NOTSTAR:
1785        case OP_NOTSTARI:
1786        case OP_NOTUPTO:
1787        case OP_NOTUPTOI:
1788        case OP_PLUS:
1789        case OP_PLUSI:
1790        case OP_POSPLUS:
1791        case OP_POSPLUSI:
1792        case OP_POSQUERY:
1793        case OP_POSQUERYI:
1794        case OP_POSSTAR:
1795        case OP_POSSTARI:
1796        case OP_POSUPTO:
1797        case OP_POSUPTOI:
1798        case OP_QUERY:
1799        case OP_QUERYI:
1800        case OP_REF:
1801        case OP_REFI:
1802        case OP_SBRA:
1803        case OP_SBRAPOS:
1804        case OP_SCBRA:
1805        case OP_SCBRAPOS:
1806        case OP_SCOND:
1807        case OP_SKIPZERO:
1808        case OP_STAR:
1809        case OP_STARI:
1810        case OP_TYPEMINPLUS:
1811        case OP_TYPEMINQUERY:
1812        case OP_TYPEMINSTAR:
1813        case OP_TYPEMINUPTO:
1814        case OP_TYPEPLUS:
1815        case OP_TYPEPOSPLUS:
1816        case OP_TYPEPOSQUERY:
1817        case OP_TYPEPOSSTAR:
1818        case OP_TYPEPOSUPTO:
1819        case OP_TYPEQUERY:
1820        case OP_TYPESTAR:
1821        case OP_TYPEUPTO:
1822        case OP_UPTO:
1823        case OP_UPTOI:
1824      return -1;      return -1;
1825    
1826        /* Catch unrecognized opcodes so that when new ones are added they
1827        are not forgotten, as has happened in the past. */
1828    
1829        default:
1830        return -4;
1831      }      }
1832    }    }
1833  /* Control never gets here */  /* Control never gets here */
# Line 992  for (;;) Line 1837  for (;;)
1837    
1838    
1839  /*************************************************  /*************************************************
1840  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1841  *************************************************/  *************************************************/
1842    
1843  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1844  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1845    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1846    so that it can be called from pcre_study() when finding the minimum matching
1847    length.
1848    
1849  Arguments:  Arguments:
1850    code        points to start of expression    code        points to start of expression
1851    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1852    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1853    
1854  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
1855  */  */
1856    
1857  static const uschar *  const uschar *
1858  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1859  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
1860  for (;;)  for (;;)
1861    {    {
1862    register int c = *code;    register int c = *code;
1863    
1864    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1865    else if (c > OP_BRA)  
1866      /* XCLASS is used for classes that cannot be represented just by a bit
1867      map. This includes negated single high-valued characters. The length in
1868      the table is zero; the actual length is stored in the compiled code. */
1869    
1870      if (c == OP_XCLASS) code += GET(code, 1);
1871    
1872      /* Handle recursion */
1873    
1874      else if (c == OP_REVERSE)
1875        {
1876        if (number < 0) return (uschar *)code;
1877        code += _pcre_OP_lengths[c];
1878        }
1879    
1880      /* Handle capturing bracket */
1881    
1882      else if (c == OP_CBRA || c == OP_SCBRA ||
1883               c == OP_CBRAPOS || c == OP_SCBRAPOS)
1884      {      {
1885      int n = c - OP_BRA;      int n = GET2(code, 1+LINK_SIZE);
     if (n > EXTRACT_BASIC_MAX) n = GET2(code, 2+LINK_SIZE);  
1886      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
1887      code += _pcre_OP_lengths[OP_BRA];      code += _pcre_OP_lengths[c];
1888      }      }
1889    
1890      /* Otherwise, we can get the item's length from the table, except that for
1891      repeated character types, we have to test for \p and \P, which have an extra
1892      two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1893      must add in its length. */
1894    
1895    else    else
1896      {      {
1897      code += _pcre_OP_lengths[c];      switch(c)
1898          {
1899          case OP_TYPESTAR:
1900          case OP_TYPEMINSTAR:
1901          case OP_TYPEPLUS:
1902          case OP_TYPEMINPLUS:
1903          case OP_TYPEQUERY:
1904          case OP_TYPEMINQUERY:
1905          case OP_TYPEPOSSTAR:
1906          case OP_TYPEPOSPLUS:
1907          case OP_TYPEPOSQUERY:
1908          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1909          break;
1910    
1911  #ifdef SUPPORT_UTF8        case OP_TYPEUPTO:
1912          case OP_TYPEMINUPTO:
1913          case OP_TYPEEXACT:
1914          case OP_TYPEPOSUPTO:
1915          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1916          break;
1917    
1918      /* In UTF-8 mode, opcodes that are followed by a character may be followed        case OP_MARK:
1919      by a multi-byte character. The length in the table is a minimum, so we have        case OP_PRUNE_ARG:
1920      to scan along to skip the extra bytes. All opcodes are less than 128, so we        case OP_SKIP_ARG:
1921      can use relatively efficient code. */        code += code[1];
1922          break;
1923    
1924          case OP_THEN_ARG:
1925          code += code[1];
1926          break;
1927          }
1928    
1929        /* Add in the fixed length from the table */
1930    
1931        code += _pcre_OP_lengths[c];
1932    
1933      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1934      a multi-byte character. The length in the table is a minimum, so we have to
1935      arrange to skip the extra bytes. */
1936    
1937    #ifdef SUPPORT_UTF8
1938      if (utf8) switch(c)      if (utf8) switch(c)
1939        {        {
1940        case OP_CHAR:        case OP_CHAR:
1941        case OP_CHARNC:        case OP_CHARI:
1942        case OP_EXACT:        case OP_EXACT:
1943          case OP_EXACTI:
1944        case OP_UPTO:        case OP_UPTO:
1945          case OP_UPTOI:
1946        case OP_MINUPTO:        case OP_MINUPTO:
1947          case OP_MINUPTOI:
1948          case OP_POSUPTO:
1949          case OP_POSUPTOI:
1950        case OP_STAR:        case OP_STAR:
1951          case OP_STARI:
1952        case OP_MINSTAR:        case OP_MINSTAR:
1953          case OP_MINSTARI:
1954          case OP_POSSTAR:
1955          case OP_POSSTARI:
1956        case OP_PLUS:        case OP_PLUS:
1957          case OP_PLUSI:
1958        case OP_MINPLUS:        case OP_MINPLUS:
1959          case OP_MINPLUSI:
1960          case OP_POSPLUS:
1961          case OP_POSPLUSI:
1962        case OP_QUERY:        case OP_QUERY:
1963          case OP_QUERYI:
1964        case OP_MINQUERY:        case OP_MINQUERY:
1965        while ((*code & 0xc0) == 0x80) code++;        case OP_MINQUERYI:
1966        break;        case OP_POSQUERY:
1967          case OP_POSQUERYI:
1968        /* 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;  
1969        break;        break;
1970        }        }
1971    #else
1972        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1973  #endif  #endif
1974      }      }
1975    }    }
# Line 1083  Returns:      pointer to the opcode for Line 1994  Returns:      pointer to the opcode for
1994  static const uschar *  static const uschar *
1995  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const uschar *code, BOOL utf8)
1996  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
1997  for (;;)  for (;;)
1998    {    {
1999    register int c = *code;    register int c = *code;
2000    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
2001    else if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
2002    else if (c > OP_BRA)  
2003      {    /* XCLASS is used for classes that cannot be represented just by a bit
2004      code += _pcre_OP_lengths[OP_BRA];    map. This includes negated single high-valued characters. The length in
2005      }    the table is zero; the actual length is stored in the compiled code. */
2006    
2007      if (c == OP_XCLASS) code += GET(code, 1);
2008    
2009      /* Otherwise, we can get the item's length from the table, except that for
2010      repeated character types, we have to test for \p and \P, which have an extra
2011      two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2012      must add in its length. */
2013    
2014    else    else
2015      {      {
2016      code += _pcre_OP_lengths[c];      switch(c)
2017          {
2018          case OP_TYPESTAR:
2019          case OP_TYPEMINSTAR:
2020          case OP_TYPEPLUS:
2021          case OP_TYPEMINPLUS:
2022          case OP_TYPEQUERY:
2023          case OP_TYPEMINQUERY:
2024          case OP_TYPEPOSSTAR:
2025          case OP_TYPEPOSPLUS:
2026          case OP_TYPEPOSQUERY:
2027          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2028          break;
2029    
2030  #ifdef SUPPORT_UTF8        case OP_TYPEPOSUPTO:
2031          case OP_TYPEUPTO:
2032          case OP_TYPEMINUPTO:
2033          case OP_TYPEEXACT:
2034          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
2035          break;
2036    
2037      /* In UTF-8 mode, opcodes that are followed by a character may be followed        case OP_MARK:
2038      by a multi-byte character. The length in the table is a minimum, so we have        case OP_PRUNE_ARG:
2039      to scan along to skip the extra bytes. All opcodes are less than 128, so we        case OP_SKIP_ARG:
2040      can use relatively efficient code. */        code += code[1];
2041          break;
2042    
2043      if (utf8) switch(c)        case OP_THEN_ARG:
2044        {        code += code[1];
2045        case OP_CHAR:        break;
2046        case OP_CHARNC:        }
2047        case OP_EXACT:  
2048        /* Add in the fixed length from the table */
2049    
2050        code += _pcre_OP_lengths[c];
2051    
2052        /* In UTF-8 mode, opcodes that are followed by a character may be followed
2053        by a multi-byte character. The length in the table is a minimum, so we have
2054        to arrange to skip the extra bytes. */
2055    
2056    #ifdef SUPPORT_UTF8
2057        if (utf8) switch(c)
2058          {
2059          case OP_CHAR:
2060          case OP_CHARI:
2061          case OP_EXACT:
2062          case OP_EXACTI:
2063        case OP_UPTO:        case OP_UPTO:
2064          case OP_UPTOI:
2065        case OP_MINUPTO:        case OP_MINUPTO:
2066          case OP_MINUPTOI:
2067          case OP_POSUPTO:
2068          case OP_POSUPTOI:
2069        case OP_STAR:        case OP_STAR:
2070          case OP_STARI:
2071        case OP_MINSTAR:        case OP_MINSTAR:
2072          case OP_MINSTARI:
2073          case OP_POSSTAR:
2074          case OP_POSSTARI:
2075        case OP_PLUS:        case OP_PLUS:
2076          case OP_PLUSI:
2077        case OP_MINPLUS:        case OP_MINPLUS:
2078          case OP_MINPLUSI:
2079          case OP_POSPLUS:
2080          case OP_POSPLUSI:
2081        case OP_QUERY:        case OP_QUERY:
2082          case OP_QUERYI:
2083        case OP_MINQUERY:        case OP_MINQUERY:
2084        while ((*code & 0xc0) == 0x80) code++;        case OP_MINQUERYI:
2085        break;        case OP_POSQUERY:
2086          case OP_POSQUERYI:
2087        /* 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;  
2088        break;        break;
2089        }        }
2090    #else
2091        (void)(utf8);  /* Keep compiler happy by referencing function argument */
2092  #endif  #endif
2093      }      }
2094    }    }
# Line 1143  for (;;) Line 2101  for (;;)
2101  *************************************************/  *************************************************/
2102    
2103  /* 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
2104  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()
2105  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
2106  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
2107  whose current branch will already have been scanned.  backward and negative forward assertions when its final argument is TRUE. If we
2108    hit an unclosed bracket, we return "empty" - this means we've struck an inner
2109    bracket whose current branch will already have been scanned.
2110    
2111  Arguments:  Arguments:
2112    code        points to start of search    code        points to start of search
2113    endcode     points to where to stop    endcode     points to where to stop
2114    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
2115      cd          contains pointers to tables etc.
2116    
2117  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2118  */  */
2119    
2120  static BOOL  static BOOL
2121  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
2122      compile_data *cd)
2123  {  {
2124  register int c;  register int c;
2125  for (code = first_significant_code(code + 1 + LINK_SIZE, NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], TRUE);
2126       code < endcode;       code < endcode;
2127       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], TRUE))
2128    {    {
2129    const uschar *ccode;    const uschar *ccode;
2130    
2131    c = *code;    c = *code;
2132    
2133    if (c >= OP_BRA)    /* Skip over forward assertions; the other assertions are skipped by
2134      first_significant_code() with a TRUE final argument. */
2135    
2136      if (c == OP_ASSERT)
2137        {
2138        do code += GET(code, 1); while (*code == OP_ALT);
2139        c = *code;
2140        continue;
2141        }
2142    
2143      /* For a recursion/subroutine call, if its end has been reached, which
2144      implies a backward reference subroutine call, we can scan it. If it's a
2145      forward reference subroutine call, we can't. To detect forward reference
2146      we have to scan up the list that is kept in the workspace. This function is
2147      called only when doing the real compile, not during the pre-compile that
2148      measures the size of the compiled pattern. */
2149    
2150      if (c == OP_RECURSE)
2151      {      {
2152        const uschar *scode;
2153      BOOL empty_branch;      BOOL empty_branch;
     if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */  
2154    
2155      /* Scan a closed bracket */      /* Test for forward reference */
2156    
2157        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
2158          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
2159    
2160        /* Not a forward reference, test for completed backward reference */
2161    
2162      empty_branch = FALSE;      empty_branch = FALSE;
2163        scode = cd->start_code + GET(code, 1);
2164        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2165    
2166        /* Completed backwards reference */
2167    
2168      do      do
2169        {        {
2170        if (!empty_branch && could_be_empty_branch(code, endcode, utf8))        if (could_be_empty_branch(scode, endcode, utf8, cd))
2171            {
2172          empty_branch = TRUE;          empty_branch = TRUE;
2173            break;
2174            }
2175          scode += GET(scode, 1);
2176          }
2177        while (*scode == OP_ALT);
2178    
2179        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2180        continue;
2181        }
2182    
2183      /* Groups with zero repeats can of course be empty; skip them. */
2184    
2185      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2186          c == OP_BRAPOSZERO)
2187        {
2188        code += _pcre_OP_lengths[c];
2189        do code += GET(code, 1); while (*code == OP_ALT);
2190        c = *code;
2191        continue;
2192        }
2193    
2194      /* A nested group that is already marked as "could be empty" can just be
2195      skipped. */
2196    
2197      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2198          c == OP_SCBRA || c == OP_SCBRAPOS)
2199        {
2200        do code += GET(code, 1); while (*code == OP_ALT);
2201        c = *code;
2202        continue;
2203        }
2204    
2205      /* For other groups, scan the branches. */
2206    
2207      if (c == OP_BRA  || c == OP_BRAPOS ||
2208          c == OP_CBRA || c == OP_CBRAPOS ||
2209          c == OP_ONCE || c == OP_ONCE_NC ||
2210          c == OP_COND)
2211        {
2212        BOOL empty_branch;
2213        if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
2214    
2215        /* If a conditional group has only one branch, there is a second, implied,
2216        empty branch, so just skip over the conditional, because it could be empty.
2217        Otherwise, scan the individual branches of the group. */
2218    
2219        if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
2220        code += GET(code, 1);        code += GET(code, 1);
2221        else
2222          {
2223          empty_branch = FALSE;
2224          do
2225            {
2226            if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
2227              empty_branch = TRUE;
2228            code += GET(code, 1);
2229            }
2230          while (*code == OP_ALT);
2231          if (!empty_branch) return FALSE;   /* All branches are non-empty */
2232        }        }
2233      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
     code += 1 + LINK_SIZE;  
2234      c = *code;      c = *code;
2235        continue;
2236      }      }
2237    
2238    else switch (c)    /* Handle the other opcodes */
2239    
2240      switch (c)
2241      {      {
2242      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
2243        cannot be represented just by a bit map. This includes negated single
2244        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
2245        actual length is stored in the compiled code, so we must update "code"
2246        here. */
2247    
2248  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2249      case OP_XCLASS:      case OP_XCLASS:
2250      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
2251      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
2252  #endif  #endif
2253    
# Line 1238  for (code = first_significant_code(code Line 2291  for (code = first_significant_code(code
2291      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2292      case OP_WORDCHAR:      case OP_WORDCHAR:
2293      case OP_ANY:      case OP_ANY:
2294        case OP_ALLANY:
2295      case OP_ANYBYTE:      case OP_ANYBYTE:
2296      case OP_CHAR:      case OP_CHAR:
2297      case OP_CHARNC:      case OP_CHARI:
2298      case OP_NOT:      case OP_NOT:
2299        case OP_NOTI:
2300      case OP_PLUS:      case OP_PLUS:
2301      case OP_MINPLUS:      case OP_MINPLUS:
2302        case OP_POSPLUS:
2303      case OP_EXACT:      case OP_EXACT:
2304      case OP_NOTPLUS:      case OP_NOTPLUS:
2305      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
2306        case OP_NOTPOSPLUS:
2307      case OP_NOTEXACT:      case OP_NOTEXACT:
2308      case OP_TYPEPLUS:      case OP_TYPEPLUS:
2309      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
2310        case OP_TYPEPOSPLUS:
2311      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2312      return FALSE;      return FALSE;
2313    
2314        /* These are going to continue, as they may be empty, but we have to
2315        fudge the length for the \p and \P cases. */
2316    
2317        case OP_TYPESTAR:
2318        case OP_TYPEMINSTAR:
2319        case OP_TYPEPOSSTAR:
2320        case OP_TYPEQUERY:
2321        case OP_TYPEMINQUERY:
2322        case OP_TYPEPOSQUERY:
2323        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2324        break;
2325    
2326        /* Same for these */
2327    
2328        case OP_TYPEUPTO:
2329        case OP_TYPEMINUPTO:
2330        case OP_TYPEPOSUPTO:
2331        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
2332        break;
2333    
2334      /* End of branch */      /* End of branch */
2335    
2336      case OP_KET:      case OP_KET:
2337      case OP_KETRMAX:      case OP_KETRMAX:
2338      case OP_KETRMIN:      case OP_KETRMIN:
2339        case OP_KETRPOS:
2340      case OP_ALT:      case OP_ALT:
2341      return TRUE;      return TRUE;
2342    
2343      /* In UTF-8 mode, STAR, MINSTAR, QUERY, MINQUERY, UPTO, and MINUPTO  may be      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2344      followed by a multibyte character */      MINUPTO, and POSUPTO may be followed by a multibyte character */
2345    
2346  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2347      case OP_STAR:      case OP_STAR:
2348        case OP_STARI:
2349      case OP_MINSTAR:      case OP_MINSTAR:
2350        case OP_MINSTARI:
2351        case OP_POSSTAR:
2352        case OP_POSSTARI:
2353      case OP_QUERY:      case OP_QUERY:
2354        case OP_QUERYI:
2355      case OP_MINQUERY:      case OP_MINQUERY:
2356        case OP_MINQUERYI:
2357        case OP_POSQUERY:
2358        case OP_POSQUERYI:
2359        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2360        break;
2361    
2362      case OP_UPTO:      case OP_UPTO:
2363        case OP_UPTOI:
2364      case OP_MINUPTO:      case OP_MINUPTO:
2365      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_MINUPTOI:
2366        case OP_POSUPTO:
2367        case OP_POSUPTOI:
2368        if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2369      break;      break;
2370  #endif  #endif
2371    
2372        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2373        string. */
2374    
2375        case OP_MARK:
2376        case OP_PRUNE_ARG:
2377        case OP_SKIP_ARG:
2378        code += code[1];
2379        break;
2380    
2381        case OP_THEN_ARG:
2382        code += code[1];
2383        break;
2384    
2385        /* None of the remaining opcodes are required to match a character. */
2386    
2387        default:
2388        break;
2389      }      }
2390    }    }
2391    
# Line 1290  return TRUE; Line 2402  return TRUE;
2402  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
2403  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,
2404  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.
2405    This function is called only during the real compile, not during the
2406    pre-compile.
2407    
2408  Arguments:  Arguments:
2409    code        points to start of the recursion    code        points to start of the recursion
2410    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2411    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2412    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2413      cd          pointers to tables etc
2414    
2415  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2416  */  */
2417    
2418  static BOOL  static BOOL
2419  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2420    BOOL utf8)    BOOL utf8, compile_data *cd)
2421  {  {
2422  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2423    {    {
2424    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2425        return FALSE;
2426    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2427    }    }
2428  return TRUE;  return TRUE;
# Line 1319  return TRUE; Line 2435  return TRUE;
2435  *************************************************/  *************************************************/
2436    
2437  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2438  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
2439  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2440  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2441    
2442    Originally, this function only recognized a sequence of letters between the
2443    terminators, but it seems that Perl recognizes any sequence of characters,
2444    though of course unknown POSIX names are subsequently rejected. Perl gives an
2445    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2446    didn't consider this to be a POSIX class. Likewise for [:1234:].
2447    
2448    The problem in trying to be exactly like Perl is in the handling of escapes. We
2449    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2450    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2451    below handles the special case of \], but does not try to do any other escape
2452    processing. This makes it different from Perl for cases such as [:l\ower:]
2453    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2454    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2455    I think.
2456    
2457    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2458    It seems that the appearance of a nested POSIX class supersedes an apparent
2459    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2460    a digit.
2461    
2462    In Perl, unescaped square brackets may also appear as part of class names. For
2463    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2464    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2465    seem right at all. PCRE does not allow closing square brackets in POSIX class
2466    names.
2467    
2468  Argument:  Arguments:
2469    ptr      pointer to the initial [    ptr      pointer to the initial [
2470    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2471    
2472  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2473  */  */
2474    
2475  static BOOL  static BOOL
2476  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
2477  {  {
2478  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2479  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2480  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2481    {    {
2482    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2483    return TRUE;      ptr++;
2484      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2485      else
2486        {
2487        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2488          {
2489          *endptr = ptr;
2490          return TRUE;
2491          }
2492        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2493             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2494              ptr[1] == CHAR_EQUALS_SIGN) &&
2495            check_posix_syntax(ptr, endptr))
2496          return FALSE;
2497        }
2498    }    }
2499  return FALSE;  return FALSE;
2500  }  }
# Line 1366  Returns:     a value representing the na Line 2519  Returns:     a value representing the na
2519  static int  static int
2520  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
2521  {  {
2522    const char *pn = posix_names;
2523  register int yield = 0;  register int yield = 0;
2524  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2525    {    {
2526    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2527      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
2528      pn += posix_name_lengths[yield] + 1;
2529    yield++;    yield++;
2530    }    }
2531  return -1;  return -1;
# Line 1385  return -1; Line 2540  return -1;
2540  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2541  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2542  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
2543  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
2544  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
2545  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
2546  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
2547  partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2548    OP_END.
2549    
2550    This function has been extended with the possibility of forward references for
2551    recursions and subroutine calls. It must also check the list of such references
2552    for the group we are dealing with. If it finds that one of the recursions in
2553    the current group is on this list, it adjusts the offset in the list, not the
2554    value in the reference (which is a group number).
2555    
2556  Arguments:  Arguments:
2557    group      points to the start of the group    group      points to the start of the group
2558    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
2559    utf8       TRUE in UTF-8 mode    utf8       TRUE in UTF-8 mode
2560    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
2561      save_hwm   the hwm forward reference pointer at the start of the group
2562    
2563  Returns:     nothing  Returns:     nothing
2564  */  */
2565    
2566  static void  static void
2567  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd)  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,
2568      uschar *save_hwm)
2569  {  {
2570  uschar *ptr = group;  uschar *ptr = group;
2571    
2572  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
2573    {    {
2574    int offset = GET(ptr, 1);    int offset;
2575    if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);    uschar *hc;
2576    
2577      /* See if this recursion is on the forward reference list. If so, adjust the
2578      reference. */
2579    
2580      for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
2581        {
2582        offset = GET(hc, 0);
2583        if (cd->start_code + offset == ptr + 1)
2584          {
2585          PUT(hc, 0, offset + adjust);
2586          break;
2587          }
2588        }
2589    
2590      /* Otherwise, adjust the recursion offset if it's after the start of this
2591      group. */
2592    
2593      if (hc >= cd->hwm)
2594        {
2595        offset = GET(ptr, 1);
2596        if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
2597        }
2598    
2599    ptr += 1 + LINK_SIZE;    ptr += 1 + LINK_SIZE;
2600    }    }
2601  }  }
# Line 1434  auto_callout(uschar *code, const uschar Line 2622  auto_callout(uschar *code, const uschar
2622  {  {
2623  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2624  *code++ = 255;  *code++ = 255;
2625  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2626  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2627  return code + 2*LINK_SIZE;  return code + 2*LINK_SIZE;
2628  }  }
2629    
# Line 1460  Returns:             nothing Line 2648  Returns:             nothing
2648  static void  static void
2649  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)
2650  {  {
2651  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2652  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2653  }  }
2654    
# Line 1486  Yield:        TRUE when range returned; Line 2674  Yield:        TRUE when range returned;
2674  */  */
2675    
2676  static BOOL  static BOOL
2677  get_othercase_range(int *cptr, int d, int *ocptr, int *odptr)  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,
2678      unsigned int *odptr)
2679  {  {
2680  int c, chartype, othercase, next;  unsigned int c, othercase, next;
2681    
2682  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2683    {    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
   if (_pcre_ucp_findchar(c, &chartype, &othercase) == ucp_L && othercase != 0)  
     break;  
   }  
2684    
2685  if (c > d) return FALSE;  if (c > d) return FALSE;
2686    
# Line 1503  next = othercase + 1; Line 2689  next = othercase + 1;
2689    
2690  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2691    {    {
2692    if (_pcre_ucp_findchar(c, &chartype, &othercase) != ucp_L ||    if (UCD_OTHERCASE(c) != next) break;
         othercase != next)  
     break;  
2693    next++;    next++;
2694    }    }
2695    
# Line 1514  for (++c; c <= d; c++) Line 2698  for (++c; c <= d; c++)
2698    
2699  return TRUE;  return TRUE;
2700  }  }
2701  #endif  /* SUPPORT_UCP */  
2702    
2703    
2704  /*************************************************  /*************************************************
2705  *           Compile one branch                   *  *        Check a character and a property        *
2706  *************************************************/  *************************************************/
2707    
2708  /* Scan the pattern, compiling it into the code vector. If the options are  /* This function is called by check_auto_possessive() when a property item
2709  changed during the branch, the pointer is used to change the external options  is adjacent to a fixed character.
 bits.  
2710    
2711  Arguments:  Arguments:
2712    optionsptr     pointer to the option bits    c            the character
2713    brackets       points to number of extracting brackets used    ptype        the property type
2714    codeptr        points to the pointer to the current code point    pdata        the data for the type
2715    ptrptr         points to the current pattern pointer    negated      TRUE if it's a negated property (\P or \p{^)
   errorcodeptr   points to error code variable  
   firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)  
   reqbyteptr     set to the last literal character required, else < 0  
   bcptr          points to current branch chain  
   cd             contains pointers to tables etc.  
2716    
2717  Returns:         TRUE on success  Returns:       TRUE if auto-possessifying is OK
                  FALSE, with *errorcodeptr set non-zero on error  
2718  */  */
2719    
2720  static BOOL  static BOOL
2721  compile_branch(int *optionsptr, int *brackets, uschar **codeptr,  check_char_prop(int c, int ptype, int pdata, BOOL negated)
   const uschar **ptrptr, int *errorcodeptr, int *firstbyteptr,  
   int *reqbyteptr, branch_chain *bcptr, compile_data *cd)  
2722  {  {
2723  int repeat_type, op_type;  const ucd_record *prop = GET_UCD(c);
2724  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  switch(ptype)
2725  int bravalue = 0;    {
2726  int greedy_default, greedy_non_default;    case PT_LAMP:
2727  int firstbyte, reqbyte;    return (prop->chartype == ucp_Lu ||
2728  int zeroreqbyte, zerofirstbyte;            prop->chartype == ucp_Ll ||
2729  int req_caseopt, reqvary, tempreqvary;            prop->chartype == ucp_Lt) == negated;
2730  int condcount = 0;  
2731  int options = *optionsptr;    case PT_GC:
2732  int after_manual_callout = 0;    return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;
2733  register int c;  
2734  register uschar *code = *codeptr;    case PT_PC:
2735  uschar *tempcode;    return (pdata == prop->chartype) == negated;
2736  BOOL inescq = FALSE;  
2737  BOOL groupsetfirstbyte = FALSE;    case PT_SC:
2738  const uschar *ptr = *ptrptr;    return (pdata == prop->script) == negated;
2739  const uschar *tempptr;  
2740  uschar *previous = NULL;    /* These are specials */
2741  uschar *previous_callout = NULL;  
2742  uschar classbits[32];    case PT_ALNUM:
2743      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2744  #ifdef SUPPORT_UTF8            _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;
2745  BOOL class_utf8;  
2746  BOOL utf8 = (options & PCRE_UTF8) != 0;    case PT_SPACE:    /* Perl space */
2747  uschar *class_utf8data;    return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2748  uschar utf8_char[6];            c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2749  #else            == negated;
2750  BOOL utf8 = FALSE;  
2751  #endif    case PT_PXSPACE:  /* POSIX space */
2752      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2753              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2754              c == CHAR_FF || c == CHAR_CR)
2755              == negated;
2756    
2757      case PT_WORD:
2758      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2759              _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2760              c == CHAR_UNDERSCORE) == negated;
2761      }
2762    return FALSE;
2763    }
2764    #endif  /* SUPPORT_UCP */
2765    
 /* Set up the default and non-default settings for greediness */  
2766    
 greedy_default = ((options & PCRE_UNGREEDY) != 0);  
 greedy_non_default = greedy_default ^ 1;  
2767    
2768  /* Initialize no first byte, no required byte. REQ_UNSET means "no char  /*************************************************
2769  matching encountered yet". It gets changed to REQ_NONE if we hit something that  *     Check if auto-possessifying is possible    *
2770  matches a non-fixed char first char; reqbyte just remains unset if we never  *************************************************/
 find one.  
2771    
2772  When we hit a repeat whose minimum is zero, we may have to adjust these values  /* This function is called for unlimited repeats of certain items, to see
2773  to take the zero repeat into account. This is implemented by setting them to  whether the next thing could possibly match the repeated item. If not, it makes
2774  zerofirstbyte and zeroreqbyte when such a repeat is encountered. The individual  sense to automatically possessify the repeated item.
 item types that can be repeated set these backoff variables appropriately. */  
2775    
2776  firstbyte = reqbyte = zerofirstbyte = zeroreqbyte = REQ_UNSET;  Arguments:
2777      previous      pointer to the repeated opcode
2778      utf8          TRUE in UTF-8 mode
2779      ptr           next character in pattern
2780      options       options bits
2781      cd            contains pointers to tables etc.
2782    
2783  /* The variable req_caseopt contains either the REQ_CASELESS value or zero,  Returns:        TRUE if possessifying is wanted
2784  according to the current setting of the caseless flag. REQ_CASELESS is a bit  */
 value > 255. It is added into the firstbyte or reqbyte variables to record the  
 case status of the value. This is used only for ASCII characters. */  
2785    
2786  req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;  static BOOL
2787    check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
2788      int options, compile_data *cd)
2789    {
2790    int c, next;
2791    int op_code = *previous++;
2792    
2793  /* Switch on next character until the end of the branch */  /* Skip whitespace and comments in extended mode */
2794    
2795  for (;; ptr++)  if ((options & PCRE_EXTENDED) != 0)
2796    {    {
2797    BOOL negate_class;    for (;;)
   BOOL possessive_quantifier;  
   BOOL is_quantifier;  
   int class_charcount;  
   int class_lastchar;  
   int newoptions;  
   int recno;  
   int skipbytes;  
   int subreqbyte;  
   int subfirstbyte;  
   int mclength;  
   uschar mcbuffer[8];  
   
   /* Next byte in the pattern */  
   
   c = *ptr;  
   
   /* If in \Q...\E, check for the end; if not, we have a literal */  
   
   if (inescq && c != 0)  
2798      {      {
2799      if (c == '\\' && ptr[1] == 'E')      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2800        if (*ptr == CHAR_NUMBER_SIGN)
2801        {        {
       inescq = FALSE;  
2802        ptr++;        ptr++;
2803        continue;        while (*ptr != 0)
       }  
     else  
       {  
       if (previous_callout != NULL)  
         {  
         complete_callout(previous_callout, ptr, cd);  
         previous_callout = NULL;  
         }  
       if ((options & PCRE_AUTO_CALLOUT) != 0)  
2804          {          {
2805          previous_callout = code;          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2806          code = auto_callout(code, ptr, cd);          ptr++;
2807    #ifdef SUPPORT_UTF8
2808            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2809    #endif
2810          }          }
       goto NORMAL_CHAR;  
2811        }        }
2812        else break;
2813      }      }
2814      }
2815    
2816    /* Fill in length of a previous callout, except when the next thing is  /* If the next item is one that we can handle, get its value. A non-negative
2817    a quantifier. */  value is a character, a negative value is an escape value. */
2818    
2819    if (*ptr == CHAR_BACKSLASH)
2820      {
2821      int temperrorcode = 0;
2822      next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
2823      if (temperrorcode != 0) return FALSE;
2824      ptr++;    /* Point after the escape sequence */
2825      }
2826    
2827    is_quantifier = c == '*' || c == '+' || c == '?' ||  else if ((cd->ctypes[*ptr] & ctype_meta) == 0)
2828      (c == '{' && is_counted_repeat(ptr+1));    {
2829    #ifdef SUPPORT_UTF8
2830      if (utf8) { GETCHARINC(next, ptr); } else
2831    #endif
2832      next = *ptr++;
2833      }
2834    
2835    if (!is_quantifier && previous_callout != NULL &&  else return FALSE;
        after_manual_callout-- <= 0)  
     {  
     complete_callout(previous_callout, ptr, cd);  
     previous_callout = NULL;  
     }  
2836    
2837    /* In extended mode, skip white space and comments */  /* Skip whitespace and comments in extended mode */
2838    
2839    if ((options & PCRE_EXTENDED) != 0)  if ((options & PCRE_EXTENDED) != 0)
2840      {
2841      for (;;)
2842      {      {
2843      if ((cd->ctypes[c] & ctype_space) != 0) continue;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2844      if (c == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2845        {        {
2846        /* The space before the ; is to avoid a warning on a silly compiler        ptr++;
2847        on the Macintosh. */        while (*ptr != 0)
2848        while ((c = *(++ptr)) != 0 && c != NEWLINE) ;          {
2849        if (c != 0) continue;   /* Else fall through to handle end of string */          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2850            ptr++;
2851    #ifdef SUPPORT_UTF8
2852            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2853    #endif
2854            }
2855        }        }
2856        else break;
2857      }      }
2858      }
2859    
2860    /* No auto callout for quantifiers. */  /* If the next thing is itself optional, we have to give up. */
2861    
2862    if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2863      {    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2864      previous_callout = code;      return FALSE;
     code = auto_callout(code, ptr, cd);  
     }  
2865    
2866    switch(c)  /* Now compare the next item with the previous opcode. First, handle cases when
2867      {  the next item is a character. */
     /* The branch terminates at end of string, |, or ). */  
2868    
2869      case 0:  if (next >= 0) switch(op_code)
2870      case '|':    {
2871      case ')':    case OP_CHAR:
2872      *firstbyteptr = firstbyte;  #ifdef SUPPORT_UTF8
2873      *reqbyteptr = reqbyte;    GETCHARTEST(c, previous);
2874      *codeptr = code;  #else
2875      *ptrptr = ptr;    c = *previous;
2876      return TRUE;  #endif
2877      return c != next;
2878    
2879      /* Handle single-character metacharacters. In multiline mode, ^ disables    /* For CHARI (caseless character) we must check the other case. If we have
2880      the setting of any following char as a first character. */    Unicode property support, we can use it to test the other case of
2881      high-valued characters. */
2882    
2883      case '^':    case OP_CHARI:
2884      if ((options & PCRE_MULTILINE) != 0)  #ifdef SUPPORT_UTF8
2885        {    GETCHARTEST(c, previous);
2886        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;  #else
2887        }    c = *previous;
2888      previous = NULL;  #endif
2889      *code++ = OP_CIRC;    if (c == next) return FALSE;
2890      break;  #ifdef SUPPORT_UTF8
2891      if (utf8)
2892        {
2893        unsigned int othercase;
2894        if (next < 128) othercase = cd->fcc[next]; else
2895    #ifdef SUPPORT_UCP
2896        othercase = UCD_OTHERCASE((unsigned int)next);
2897    #else
2898        othercase = NOTACHAR;
2899    #endif
2900        return (unsigned int)c != othercase;
2901        }
2902      else
2903    #endif  /* SUPPORT_UTF8 */
2904      return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2905    
2906      case '$':    /* For OP_NOT and OP_NOTI, the data is always a single-byte character. These
2907      previous = NULL;    opcodes are not used for multi-byte characters, because they are coded using
2908      *code++ = OP_DOLL;    an XCLASS instead. */
     break;  
2909    
2910      /* There can never be a first char if '.' is first, whatever happens about    case OP_NOT:
2911      repeats. The value of reqbyte doesn't change either. */    return (c = *previous) == next;
2912    
2913      case '.':    case OP_NOTI:
2914      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;    if ((c = *previous) == next) return TRUE;
2915      zerofirstbyte = firstbyte;  #ifdef SUPPORT_UTF8
2916      zeroreqbyte = reqbyte;    if (utf8)
2917      previous = code;      {
2918      *code++ = OP_ANY;      unsigned int othercase;
2919      break;      if (next < 128) othercase = cd->fcc[next]; else
2920    #ifdef SUPPORT_UCP
2921        othercase = UCD_OTHERCASE(next);
2922    #else
2923        othercase = NOTACHAR;
2924    #endif
2925        return (unsigned int)c == othercase;
2926        }
2927      else
2928    #endif  /* SUPPORT_UTF8 */
2929      return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2930    
2931      /* Character classes. If the included characters are all < 255 in value, we    /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2932      build a 32-byte bitmap of the permitted characters, except in the special    When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
     case where there is only one such character. For negated classes, we build  
     the map as usual, then invert it at the end. However, we use a different  
     opcode so that data characters > 255 can be handled correctly.  
2933    
2934      If the class contains characters outside the 0-255 range, a different    case OP_DIGIT:
2935      opcode is compiled. It may optionally have a bit map for characters < 256,    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
     but those above are are explicitly listed afterwards. A flag byte tells  
     whether the bitmap is present, and whether this is a negated class or not.  
     */  
2936    
2937      case '[':    case OP_NOT_DIGIT:
2938      previous = code;    return next <= 127 && (cd->ctypes[next] & ctype_digit) != 0;
2939    
2940      /* PCRE supports POSIX class stuff inside a class. Perl gives an error if    case OP_WHITESPACE:
2941      they are encountered at the top level, so we'll do that too. */    return next > 127 || (cd->ctypes[next] & ctype_space) == 0;
2942    
2943      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&    case OP_NOT_WHITESPACE:
2944          check_posix_syntax(ptr, &tempptr, cd))    return next <= 127 && (cd->ctypes[next] & ctype_space) != 0;
       {  
       *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;  
       goto FAILED;  
       }  
2945    
2946      /* If the first character is '^', set the negation flag and skip it. */    case OP_WORDCHAR:
2947      return next > 127 || (cd->ctypes[next] & ctype_word) == 0;
2948    
2949      if ((c = *(++ptr)) == '^')    case OP_NOT_WORDCHAR:
2950        {    return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;
       negate_class = TRUE;  
       c = *(++ptr);  
       }  
     else  
       {  
       negate_class = FALSE;  
       }  
2951    
2952      /* Keep a count of chars with values < 256 so that we can optimize the case    case OP_HSPACE:
2953      of just a single character (as long as it's < 256). For higher valued UTF-8    case OP_NOT_HSPACE:
2954      characters, we don't yet do any optimization. */    switch(next)
2955        {
2956        case 0x09:
2957        case 0x20:
2958        case 0xa0:
2959        case 0x1680:
2960        case 0x180e:
2961        case 0x2000:
2962        case 0x2001:
2963        case 0x2002:
2964        case 0x2003:
2965        case 0x2004:
2966        case 0x2005:
2967        case 0x2006:
2968        case 0x2007:
2969        case 0x2008:
2970        case 0x2009:
2971        case 0x200A:
2972        case 0x202f:
2973        case 0x205f:
2974        case 0x3000:
2975        return op_code == OP_NOT_HSPACE;
2976        default:
2977        return op_code != OP_NOT_HSPACE;
2978        }
2979    
2980      class_charcount = 0;    case OP_ANYNL:
2981      class_lastchar = -1;    case OP_VSPACE:
2982      case OP_NOT_VSPACE:
2983      switch(next)
2984        {
2985        case 0x0a:
2986        case 0x0b:
2987        case 0x0c:
2988        case 0x0d:
2989        case 0x85:
2990        case 0x2028:
2991        case 0x2029:
2992        return op_code == OP_NOT_VSPACE;
2993        default:
2994        return op_code != OP_NOT_VSPACE;
2995        }
2996    
2997  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UCP
2998      class_utf8 = FALSE;                       /* No chars >= 256 */    case OP_PROP:
2999      class_utf8data = code + LINK_SIZE + 34;   /* For UTF-8 items */    return check_char_prop(next, previous[0], previous[1], FALSE);
3000    
3001      case OP_NOTPROP:
3002      return check_char_prop(next, previous[0], previous[1], TRUE);
3003  #endif  #endif
3004    
3005      /* Initialize the 32-char bit map to all zeros. We have to build the    default:
3006      map in a temporary bit of store, in case the class contains only 1    return FALSE;
3007      character (< 256), because in that case the compiled code doesn't use the    }
     bit map. */  
3008    
     memset(classbits, 0, 32 * sizeof(uschar));  
3009    
3010      /* Process characters until ] is reached. By writing this as a "do" it  /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP
3011      means that an initial ] is taken as a data character. The first pass  is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
3012      through the regex checked the overall syntax, so we don't need to be very  generated only when PCRE_UCP is *not* set, that is, when only ASCII
3013      strict here. At the start of the loop, c contains the first byte of the  characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
3014      character. */  replaced by OP_PROP codes when PCRE_UCP is set. */
3015    
3016      do  switch(op_code)
3017        {    {
3018      case OP_CHAR:
3019      case OP_CHARI:
3020  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3021        if (utf8 && c > 127)    GETCHARTEST(c, previous);
3022          {                           /* Braces are required because the */  #else
3023          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */    c = *previous;
         }  
3024  #endif  #endif
3025      switch(-next)
3026        {
3027        case ESC_d:
3028        return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
3029    
3030        /* Inside \Q...\E everything is literal except \E */      case ESC_D:
3031        return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
       if (inescq)  
         {  
         if (c == '\\' && ptr[1] == 'E')  
           {  
           inescq = FALSE;  
           ptr++;  
           continue;  
           }  
         else goto LONE_SINGLE_CHARACTER;  
         }  
   
       /* Handle POSIX class names. Perl allows a negation extension of the  
       form [:^name:]. A square bracket that doesn't match the syntax is  
       treated as a literal. We also recognize the POSIX constructions  
       [.ch.] and [=ch=] ("collating elements") and fault them, as Perl  
       5.6 and 5.8 do. */  
   
       if (c == '[' &&  
           (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&  
           check_posix_syntax(ptr, &tempptr, cd))  
         {  
         BOOL local_negate = FALSE;  
         int posix_class, i;  
         register const uschar *cbits = cd->cbits;  
   
         if (ptr[1] != ':')  
           {  
           *errorcodeptr = ERR31;  
           goto FAILED;  
           }  
   
         ptr += 2;  
         if (*ptr == '^')  
           {  
           local_negate = TRUE;  
           ptr++;  
           }  
3032    
3033          posix_class = check_posix_name(ptr, tempptr - ptr);      case ESC_s:
3034          if (posix_class < 0)      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
           {  
           *errorcodeptr = ERR30;  
           goto FAILED;  
           }  
3035    
3036          /* If matching is caseless, upper and lower are converted to      case ESC_S:
3037          alpha. This relies on the fact that the class table starts with      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
3038          alpha, lower, upper as the first 3 entries. */  
3039        case ESC_w:
3040        return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
3041    
3042        case ESC_W:
3043        return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
3044    
3045        case ESC_h:
3046        case ESC_H:
3047        switch(c)
3048          {
3049          case 0x09:
3050          case 0x20:
3051          case 0xa0:
3052          case 0x1680:
3053          case 0x180e:
3054          case 0x2000:
3055          case 0x2001:
3056          case 0x2002:
3057          case 0x2003:
3058          case 0x2004:
3059          case 0x2005:
3060          case 0x2006:
3061          case 0x2007:
3062          case 0x2008:
3063          case 0x2009:
3064          case 0x200A:
3065          case 0x202f:
3066          case 0x205f:
3067          case 0x3000:
3068          return -next != ESC_h;
3069          default:
3070          return -next == ESC_h;
3071          }
3072    
3073          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)      case ESC_v:
3074            posix_class = 0;      case ESC_V:
3075        switch(c)
3076          {
3077          case 0x0a:
3078          case 0x0b:
3079          case 0x0c:
3080          case 0x0d:
3081          case 0x85:
3082          case 0x2028:
3083          case 0x2029:
3084          return -next != ESC_v;
3085          default:
3086          return -next == ESC_v;
3087          }
3088    
3089          /* Or into the map we are building up to 3 of the static class      /* When PCRE_UCP is set, these values get generated for \d etc. Find
3090          tables, or their negations. The [:blank:] class sets up the same      their substitutions and process them. The result will always be either
3091          chars as the [:space:] class (all white space). We remove the vertical      -ESC_p or -ESC_P. Then fall through to process those values. */
         white space chars afterwards. */  
3092    
3093          posix_class *= 3;  #ifdef SUPPORT_UCP
3094          for (i = 0; i < 3; i++)      case ESC_du:
3095            {      case ESC_DU:
3096            BOOL blankclass = strncmp((char *)ptr, "blank", 5) == 0;      case ESC_wu:
3097            int taboffset = posix_class_maps[posix_class + i];      case ESC_WU:
3098            if (taboffset < 0) break;      case ESC_su:
3099            if (local_negate)      case ESC_SU:
3100              {        {
3101              if (i == 0)        int temperrorcode = 0;
3102                for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+taboffset];        ptr = substitutes[-next - ESC_DU];
3103              else        next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
3104                for (c = 0; c < 32; c++) classbits[c] &= ~cbits[c+taboffset];        if (temperrorcode != 0) return FALSE;
3105              if (blankclass) classbits[1] |= 0x3c;        ptr++;    /* For compatibility */
3106              }        }
3107            else      /* Fall through */
             {  
             for (c = 0; c < 32; c++) classbits[c] |= cbits[c+taboffset];  
             if (blankclass) classbits[1] &= ~0x3c;  
             }  
           }  
3108    
3109          ptr = tempptr + 1;      case ESC_p:
3110          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */      case ESC_P:
3111          continue;    /* End of POSIX syntax handling */        {
3112          }        int ptype, pdata, errorcodeptr;
3113          BOOL negated;
3114    
3115        /* Backslash may introduce a single character, or it may introduce one        ptr--;      /* Make ptr point at the p or P */
3116        of the specials, which just set a flag. Escaped items are checked for        ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
3117        validity in the pre-compiling pass. The sequence \b is a special case.        if (ptype < 0) return FALSE;
3118        Inside a class (and only there) it is treated as backspace. Elsewhere        ptr++;      /* Point past the final curly ket */
3119        it marks a word boundary. Other escapes have preset maps ready to  
3120        or into the one we are building. We assume they have more than one        /* If the property item is optional, we have to give up. (When generated
3121        character in them, so set class_charcount bigger than one. */        from \d etc by PCRE_UCP, this test will have been applied much earlier,
3122          to the original \d etc. At this point, ptr will point to a zero byte. */
3123    
3124          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3125            strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3126              return FALSE;
3127    
3128        if (c == '\\')        /* Do the property check. */
         {  
         c = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);  
3129    
3130          if (-c == ESC_b) c = '\b';       /* \b is backslash in a class */        return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
3131          else if (-c == ESC_X) c = 'X';   /* \X is literal X in a class */        }
3132          else if (-c == ESC_Q)            /* Handle start of quoted string */  #endif
           {  
           if (ptr[1] == '\\' && ptr[2] == 'E')  
             {  
             ptr += 2; /* avoid empty string */  
             }  
           else inescq = TRUE;  
           continue;  
           }  
3133    
3134          if (c < 0)      default:
3135            {      return FALSE;
3136            register const uschar *cbits = cd->cbits;      }
           class_charcount += 2;     /* Greater than 1 is what matters */  
           switch (-c)  
             {  
             case ESC_d:  
             for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];  
             continue;  
3137    
3138              case ESC_D:    /* In principle, support for Unicode properties should be integrated here as
3139              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];    well. It means re-organizing the above code so as to get hold of the property
3140              continue;    values before switching on the op-code. However, I wonder how many patterns
3141      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
3142      these op-codes are never generated.) */
3143    
3144              case ESC_w:    case OP_DIGIT:
3145              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_word];    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
3146              continue;           next == -ESC_h || next == -ESC_v || next == -ESC_R;
3147    
3148              case ESC_W:    case OP_NOT_DIGIT:
3149              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];    return next == -ESC_d;
             continue;  
3150    
3151              case ESC_s:    case OP_WHITESPACE:
3152              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
             classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */  
             continue;  
3153    
3154              case ESC_S:    case OP_NOT_WHITESPACE:
3155              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
             classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */  
             continue;  
3156    
3157  #ifdef SUPPORT_UCP    case OP_HSPACE:
3158              case ESC_p:    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
3159              case ESC_P:           next == -ESC_w || next == -ESC_v || next == -ESC_R;
               {  
               BOOL negated;  
               int property = get_ucp(&ptr, &negated, errorcodeptr);  
               if (property < 0) goto FAILED;  
               class_utf8 = TRUE;  
               *class_utf8data++ = ((-c == ESC_p) != negated)?  
                 XCL_PROP : XCL_NOTPROP;  
               *class_utf8data++ = property;  
               class_charcount -= 2;   /* Not a < 256 character */  
               }  
             continue;  
 #endif  
3160    
3161              /* Unrecognized escapes are faulted if PCRE is running in its    case OP_NOT_HSPACE:
3162              strict mode. By default, for compatibility with Perl, they are    return next == -ESC_h;
             treated as literals. */  
3163    
3164              default:    /* Can't have \S in here because VT matches \S (Perl anomaly) */
3165              if ((options & PCRE_EXTRA) != 0)    case OP_ANYNL:
3166                {    case OP_VSPACE:
3167                *errorcodeptr = ERR7;    return next == -ESC_V || next == -ESC_d || next == -ESC_w;
               goto FAILED;  
               }  
             c = *ptr;              /* The final character */  
             class_charcount -= 2;  /* Undo the default count from above */  
             }  
           }  
3168    
3169          /* Fall through if we have a single character (c >= 0). This may be    case OP_NOT_VSPACE:
3170          > 256 in UTF-8 mode. */    return next == -ESC_v || next == -ESC_R;
3171    
3172          }   /* End of backslash handling */    case OP_WORDCHAR:
3173      return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
3174             next == -ESC_v || next == -ESC_R;
3175    
3176        /* A single character may be followed by '-' to form a range. However,    case OP_NOT_WORDCHAR:
3177        Perl does not permit ']' to be the end of the range. A '-' character    return next == -ESC_w || next == -ESC_d;
       here is treated as a literal. */  
3178    
3179        if (ptr[1] == '-' && ptr[2] != ']')    default:
3180          {    return FALSE;
3181          int d;    }
         ptr += 2;  
3182    
3183  #ifdef SUPPORT_UTF8  /* Control does not reach here */
3184          if (utf8)  }
           {                           /* Braces are required because the */  
           GETCHARLEN(d, ptr, ptr);    /* macro generates multiple statements */  
           }  
         else  
 #endif  
         d = *ptr;  /* Not UTF-8 mode */  
3185    
         /* The second part of a range can be a single-character escape, but  
         not any of the other escapes. Perl 5.6 treats a hyphen as a literal  
         in such circumstances. */  
3186    
         if (d == '\\')  
           {  
           const uschar *oldptr = ptr;  
           d = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);  
3187    
3188            /* \b is backslash; \X is literal X; any other special means the '-'  /*************************************************
3189            was literal */  *           Compile one branch                   *
3190    *************************************************/
3191    
3192            if (d < 0)  /* Scan the pattern, compiling it into the a vector. If the options are
3193              {  changed during the branch, the pointer is used to change the external options
3194              if (d == -ESC_b) d = '\b';  bits. This function is used during the pre-compile phase when we are trying
3195              else if (d == -ESC_X) d = 'X'; else  to find out the amount of memory needed, as well as during the real compile
3196                {  phase. The value of lengthptr distinguishes the two phases.
               ptr = oldptr - 2;  
               goto LONE_SINGLE_CHARACTER;  /* A few lines below */  
               }  
             }  
           }  
3197    
3198          /* The check that the two values are in the correct order happens in  Arguments:
3199          the pre-pass. Optimize one-character ranges */    optionsptr     pointer to the option bits
3200      codeptr        points to the pointer to the current code point
3201      ptrptr         points to the current pattern pointer
3202      errorcodeptr   points to error code variable
3203      firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
3204      reqbyteptr     set to the last literal character required, else < 0
3205      bcptr          points to current branch chain
3206      cond_depth     conditional nesting depth
3207      cd             contains pointers to tables etc.
3208      lengthptr      NULL during the real compile phase
3209                     points to length accumulator during pre-compile phase
3210    
3211          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */  Returns:         TRUE on success
3212                     FALSE, with *errorcodeptr set non-zero on error
3213    */
3214    
3215          /* In UTF-8 mode, if the upper limit is > 255, or > 127 for caseless  static BOOL
3216          matching, we have to use an XCLASS with extra data items. Caseless  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
3217          matching for characters > 127 is available only if UCP support is    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
3218      int cond_depth, compile_data *cd, int *lengthptr)
3219    {
3220    int repeat_type, op_type;
3221    int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
3222    int bravalue = 0;
3223    int greedy_default, greedy_non_default;
3224    int firstbyte, reqbyte;
3225    int zeroreqbyte, zerofirstbyte;
3226    int req_caseopt, reqvary, tempreqvary;
3227    int options = *optionsptr;               /* May change dynamically */
3228    int after_manual_callout = 0;
3229    int length_prevgroup = 0;
3230    register int c;
3231    register uschar *code = *codeptr;
3232    uschar *last_code = code;
3233    uschar *orig_code = code;
3234    uschar *tempcode;
3235    BOOL inescq = FALSE;
3236    BOOL groupsetfirstbyte = FALSE;
3237    const uschar *ptr = *ptrptr;
3238    const uschar *tempptr;
3239    const uschar *nestptr = NULL;
3240    uschar *previous = NULL;
3241    uschar *previous_callout = NULL;
3242    uschar *save_hwm = NULL;
3243    uschar classbits[32];
3244    
3245    /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3246    must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3247    dynamically as we process the pattern. */
3248    
3249    #ifdef SUPPORT_UTF8
3250    BOOL class_utf8;
3251    BOOL utf8 = (options & PCRE_UTF8) != 0;
3252    uschar *class_utf8data;
3253    uschar *class_utf8data_base;
3254    uschar utf8_char[6];
3255    #else
3256    BOOL utf8 = FALSE;
3257    #endif
3258    
3259    #ifdef PCRE_DEBUG
3260    if (lengthptr != NULL) DPRINTF((">> start branch\n"));
3261    #endif
3262    
3263    /* Set up the default and non-default settings for greediness */
3264    
3265    greedy_default = ((options & PCRE_UNGREEDY) != 0);
3266    greedy_non_default = greedy_default ^ 1;
3267    
3268    /* Initialize no first byte, no required byte. REQ_UNSET means "no char
3269    matching encountered yet". It gets changed to REQ_NONE if we hit something that
3270    matches a non-fixed char first char; reqbyte just remains unset if we never
3271    find one.
3272    
3273    When we hit a repeat whose minimum is zero, we may have to adjust these values
3274    to take the zero repeat into account. This is implemented by setting them to
3275    zerofirstbyte and zeroreqbyte when such a repeat is encountered. The individual
3276    item types that can be repeated set these backoff variables appropriately. */
3277    
3278    firstbyte = reqbyte = zerofirstbyte = zeroreqbyte = REQ_UNSET;
3279    
3280    /* The variable req_caseopt contains either the REQ_CASELESS value or zero,
3281    according to the current setting of the caseless flag. REQ_CASELESS is a bit
3282    value > 255. It is added into the firstbyte or reqbyte variables to record the
3283    case status of the value. This is used only for ASCII characters. */
3284    
3285    req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
3286    
3287    /* Switch on next character until the end of the branch */
3288    
3289    for (;; ptr++)
3290      {
3291      BOOL negate_class;
3292      BOOL should_flip_negation;
3293      BOOL possessive_quantifier;
3294      BOOL is_quantifier;
3295      BOOL is_recurse;
3296      BOOL reset_bracount;
3297      int class_charcount;
3298      int class_lastchar;
3299      int newoptions;
3300      int recno;
3301      int refsign;
3302      int skipbytes;
3303      int subreqbyte;
3304      int subfirstbyte;
3305      int terminator;
3306      int mclength;
3307      int tempbracount;
3308      uschar mcbuffer[8];
3309    
3310      /* Get next byte in the pattern */
3311    
3312      c = *ptr;
3313    
3314      /* If we are at the end of a nested substitution, revert to the outer level
3315      string. Nesting only happens one level deep. */
3316    
3317      if (c == 0 && nestptr != NULL)
3318        {
3319        ptr = nestptr;
3320        nestptr = NULL;
3321        c = *ptr;
3322        }
3323    
3324      /* If we are in the pre-compile phase, accumulate the length used for the
3325      previous cycle of this loop. */
3326    
3327      if (lengthptr != NULL)
3328        {
3329    #ifdef PCRE_DEBUG
3330        if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3331    #endif
3332        if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
3333          {
3334          *errorcodeptr = ERR52;
3335          goto FAILED;
3336          }
3337    
3338        /* There is at least one situation where code goes backwards: this is the
3339        case of a zero quantifier after a class (e.g. [ab]{0}). At compile time,
3340        the class is simply eliminated. However, it is created first, so we have to
3341        allow memory for it. Therefore, don't ever reduce the length at this point.
3342        */
3343    
3344        if (code < last_code) code = last_code;
3345    
3346        /* Paranoid check for integer overflow */
3347    
3348        if (OFLOW_MAX - *lengthptr < code - last_code)
3349          {
3350          *errorcodeptr = ERR20;
3351          goto FAILED;
3352          }
3353    
3354        *lengthptr += (int)(code - last_code);
3355        DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));
3356    
3357        /* If "previous" is set and it is not at the start of the work space, move
3358        it back to there, in order to avoid filling up the work space. Otherwise,
3359        if "previous" is NULL, reset the current code pointer to the start. */
3360    
3361        if (previous != NULL)
3362          {
3363          if (previous > orig_code)
3364            {
3365            memmove(orig_code, previous, code - previous);
3366            code -= previous - orig_code;
3367            previous = orig_code;
3368            }
3369          }
3370        else code = orig_code;
3371    
3372        /* Remember where this code item starts so we can pick up the length
3373        next time round. */
3374    
3375        last_code = code;
3376        }
3377    
3378      /* In the real compile phase, just check the workspace used by the forward
3379      reference list. */
3380    
3381      else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
3382        {
3383        *errorcodeptr = ERR52;
3384        goto FAILED;
3385        }
3386    
3387      /* If in \Q...\E, check for the end; if not, we have a literal */
3388    
3389      if (inescq && c != 0)
3390        {
3391        if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3392          {
3393          inescq = FALSE;
3394          ptr++;
3395          continue;
3396          }
3397        else
3398          {
3399          if (previous_callout != NULL)
3400            {
3401            if (lengthptr == NULL)  /* Don't attempt in pre-compile phase */
3402              complete_callout(previous_callout, ptr, cd);
3403            previous_callout = NULL;
3404            }
3405          if ((options & PCRE_AUTO_CALLOUT) != 0)
3406            {
3407            previous_callout = code;
3408            code = auto_callout(code, ptr, cd);
3409            }
3410          goto NORMAL_CHAR;
3411          }
3412        }
3413    
3414      /* Fill in length of a previous callout, except when the next thing is
3415      a quantifier. */
3416    
3417      is_quantifier =
3418        c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
3419        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
3420    
3421      if (!is_quantifier && previous_callout != NULL &&
3422           after_manual_callout-- <= 0)
3423        {
3424        if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */
3425          complete_callout(previous_callout, ptr, cd);
3426        previous_callout = NULL;
3427        }
3428    
3429      /* In extended mode, skip white space and comments. */
3430    
3431      if ((options & PCRE_EXTENDED) != 0)
3432        {
3433        if ((cd->ctypes[c] & ctype_space) != 0) continue;
3434        if (c == CHAR_NUMBER_SIGN)
3435          {
3436          ptr++;
3437          while (*ptr != 0)
3438            {
3439            if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3440            ptr++;
3441    #ifdef SUPPORT_UTF8
3442            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
3443    #endif
3444            }
3445          if (*ptr != 0) continue;
3446    
3447          /* Else fall through to handle end of string */
3448          c = 0;
3449          }
3450        }
3451    
3452      /* No auto callout for quantifiers. */
3453    
3454      if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier)
3455        {
3456        previous_callout = code;
3457        code = auto_callout(code, ptr, cd);
3458        }
3459    
3460      switch(c)
3461        {
3462        /* ===================================================================*/
3463        case 0:                        /* The branch terminates at string end */
3464        case CHAR_VERTICAL_LINE:       /* or | or ) */
3465        case CHAR_RIGHT_PARENTHESIS:
3466        *firstbyteptr = firstbyte;
3467        *reqbyteptr = reqbyte;
3468        *codeptr = code;
3469        *ptrptr = ptr;
3470        if (lengthptr != NULL)
3471          {
3472          if (OFLOW_MAX - *lengthptr < code - last_code)
3473            {
3474            *errorcodeptr = ERR20;
3475            goto FAILED;
3476            }
3477          *lengthptr += (int)(code - last_code);   /* To include callout length */
3478          DPRINTF((">> end branch\n"));
3479          }
3480        return TRUE;
3481    
3482    
3483        /* ===================================================================*/
3484        /* Handle single-character metacharacters. In multiline mode, ^ disables
3485        the setting of any following char as a first character. */
3486    
3487        case CHAR_CIRCUMFLEX_ACCENT:
3488        previous = NULL;
3489        if ((options & PCRE_MULTILINE) != 0)
3490          {
3491          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3492          *code++ = OP_CIRCM;
3493          }
3494        else *code++ = OP_CIRC;
3495        break;
3496    
3497        case CHAR_DOLLAR_SIGN:
3498        previous = NULL;
3499        *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
3500        break;
3501    
3502        /* There can never be a first char if '.' is first, whatever happens about
3503        repeats. The value of reqbyte doesn't change either. */
3504    
3505        case CHAR_DOT:
3506        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3507        zerofirstbyte = firstbyte;
3508        zeroreqbyte = reqbyte;
3509        previous = code;
3510        *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
3511        break;
3512    
3513    
3514        /* ===================================================================*/
3515        /* Character classes. If the included characters are all < 256, we build a
3516        32-byte bitmap of the permitted characters, except in the special case
3517        where there is only one such character. For negated classes, we build the
3518        map as usual, then invert it at the end. However, we use a different opcode
3519        so that data characters > 255 can be handled correctly.
3520    
3521        If the class contains characters outside the 0-255 range, a different
3522        opcode is compiled. It may optionally have a bit map for characters < 256,
3523        but those above are are explicitly listed afterwards. A flag byte tells
3524        whether the bitmap is present, and whether this is a negated class or not.
3525    
3526        In JavaScript compatibility mode, an isolated ']' causes an error. In
3527        default (Perl) mode, it is treated as a data character. */
3528    
3529        case CHAR_RIGHT_SQUARE_BRACKET:
3530        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3531          {
3532          *errorcodeptr = ERR64;
3533          goto FAILED;
3534          }
3535        goto NORMAL_CHAR;
3536    
3537        case CHAR_LEFT_SQUARE_BRACKET:
3538        previous = code;
3539    
3540        /* PCRE supports POSIX class stuff inside a class. Perl gives an error if
3541        they are encountered at the top level, so we'll do that too. */
3542    
3543        if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3544             ptr[1] == CHAR_EQUALS_SIGN) &&
3545            check_posix_syntax(ptr, &tempptr))
3546          {
3547          *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
3548          goto FAILED;
3549          }
3550    
3551        /* If the first character is '^', set the negation flag and skip it. Also,
3552        if the first few characters (either before or after ^) are \Q\E or \E we
3553        skip them too. This makes for compatibility with Perl. */
3554    
3555        negate_class = FALSE;
3556        for (;;)
3557          {
3558          c = *(++ptr);
3559          if (c == CHAR_BACKSLASH)
3560            {
3561            if (ptr[1] == CHAR_E)
3562              ptr++;
3563            else if (strncmp((const char *)ptr+1,
3564                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
3565              ptr += 3;
3566            else
3567              break;
3568            }
3569          else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
3570            negate_class = TRUE;
3571          else break;
3572          }
3573    
3574        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
3575        an initial ']' is taken as a data character -- the code below handles
3576        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3577        [^] must match any character, so generate OP_ALLANY. */
3578    
3579        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3580            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3581          {
3582          *code++ = negate_class? OP_ALLANY : OP_FAIL;
3583          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3584          zerofirstbyte = firstbyte;
3585          break;
3586          }
3587    
3588        /* If a class contains a negative special such as \S, we need to flip the
3589        negation flag at the end, so that support for characters > 255 works