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

Diff of /code/trunk/pcre_compile.c

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

revision 79 by nigel, Sat Feb 24 21:40:52 2007 UTC revision 528 by ph10, Sat May 29 16:40:22 2010 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2005 University of Cambridge             Copyright (c) 1997-2010 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 42  POSSIBILITY OF SUCH DAMAGE. Line 42  POSSIBILITY OF SUCH DAMAGE.
42  supporting internal functions that are not used by other modules. */  supporting internal functions that are not used by other modules. */
43    
44    
45    #ifdef HAVE_CONFIG_H
46    #include "config.h"
47    #endif
48    
49    #define NLBLOCK cd             /* Block containing newline information */
50    #define PSSTART start_pattern  /* Field containing processed string start */
51    #define PSEND   end_pattern    /* Field containing processed string end */
52    
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56    /* When PCRE_DEBUG is defined, we need the pcre_printint() function, which is
57    also used by pcretest. PCRE_DEBUG is not defined when building a production
58    library. */
59    
60    #ifdef PCRE_DEBUG
61    #include "pcre_printint.src"
62    #endif
63    
64    
65    /* Macro for setting individual bits in class bitmaps. */
66    
67    #define SETBIT(a,b) a[b/8] |= (1 << (b%8))
68    
69    /* Maximum length value to check against when making sure that the integer that
70    holds the compiled pattern length does not overflow. We make it a bit less than
71    INT_MAX to allow for adding in group terminating bytes, so that we don't have
72    to check them every time. */
73    
74    #define OFLOW_MAX (INT_MAX - 20)
75    
76    
77  /*************************************************  /*************************************************
78  *      Code parameters and static tables         *  *      Code parameters and static tables         *
79  *************************************************/  *************************************************/
80    
81  /* Maximum number of items on the nested bracket stacks at compile time. This  /* This value specifies the size of stack workspace that is used during the
82  applies to the nesting of all kinds of parentheses. It does not limit  first pre-compile phase that determines how much memory is required. The regex
83  un-nested, non-capturing parentheses. This number can be made bigger if  is partly compiled into this space, but the compiled parts are discarded as
84  necessary - it is used to dimension one int and one unsigned char vector at  soon as they can be, so that hopefully there will never be an overrun. The code
85  compile time. */  does, however, check for an overrun. The largest amount I've seen used is 218,
86    so this number is very generous.
87    
88    The same workspace is used during the second, actual compile phase for
89    remembering forward references to groups so that they can be filled in at the
90    end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE
91    is 4 there is plenty of room. */
92    
93  #define BRASTACK_SIZE 200  #define COMPILE_WORK_SIZE (4096)
94    
95    /* The overrun tests check for a slightly smaller size so that they detect the
96    overrun before it actually does run off the end of the data block. */
97    
98    #define WORK_SIZE_CHECK (COMPILE_WORK_SIZE - 100)
99    
100    
101  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
# Line 63  are simple data values; negative values Line 103  are simple data values; negative values
103  on. Zero means further processing is needed (for things like \x), or the escape  on. Zero means further processing is needed (for things like \x), or the escape
104  is invalid. */  is invalid. */
105    
106  #if !EBCDIC   /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC
107    
108    /* This is the "normal" table for ASCII systems or for EBCDIC systems running
109    in UTF-8 mode. */
110    
111  static const short int escapes[] = {  static const short int escapes[] = {
112       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,                       0,
113       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,                       0,
114     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */       0,                       0,
115       0,      0,      0,      0,      0,      0,      0,      0,   /* H - O */       0,                       0,
116  -ESC_P, -ESC_Q,      0, -ESC_S,      0,      0,      0, -ESC_W,   /* P - W */       0,                       0,
117  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */       CHAR_COLON,              CHAR_SEMICOLON,
118     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
119       0,      0,      0,      0,      0,      0,  ESC_n,      0,   /* h - o */       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
120  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0,      0, -ESC_w,   /* p - w */       CHAR_COMMERCIAL_AT,      -ESC_A,
121       0,      0, -ESC_z                                            /* x - z */       -ESC_B,                  -ESC_C,
122         -ESC_D,                  -ESC_E,
123         0,                       -ESC_G,
124         -ESC_H,                  0,
125         0,                       -ESC_K,
126         0,                       0,
127         -ESC_N,                  0,
128         -ESC_P,                  -ESC_Q,
129         -ESC_R,                  -ESC_S,
130         0,                       0,
131         -ESC_V,                  -ESC_W,
132         -ESC_X,                  0,
133         -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,
134         CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,
135         CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,
136         CHAR_GRAVE_ACCENT,       7,
137         -ESC_b,                  0,
138         -ESC_d,                  ESC_e,
139         ESC_f,                   0,
140         -ESC_h,                  0,
141         0,                       -ESC_k,
142         0,                       0,
143         ESC_n,                   0,
144         -ESC_p,                  0,
145         ESC_r,                   -ESC_s,
146         ESC_tee,                 0,
147         -ESC_v,                  -ESC_w,
148         0,                       0,
149         -ESC_z
150  };  };
151    
152  #else         /* This is the "abnormal" table for EBCDIC systems */  #else
153    
154    /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
155    
156  static const short int escapes[] = {  static const short int escapes[] = {
157  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
158  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 87  static const short int escapes[] = { Line 162  static const short int escapes[] = {
162  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,
163  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',
164  /*  80 */     0,     7, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,  /*  80 */     0,     7, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,
165  /*  88 */     0,     0,      0,     '{',      0,     0,      0,      0,  /*  88 */-ESC_h,     0,      0,     '{',      0,     0,      0,      0,
166  /*  90 */     0,     0,      0,     'l',      0, ESC_n,      0, -ESC_p,  /*  90 */     0,     0, -ESC_k,     'l',      0, ESC_n,      0, -ESC_p,
167  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,
168  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,     0, -ESC_w,      0,  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,-ESC_v, -ESC_w,      0,
169  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,
170  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,
171  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
172  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,
173  /*  C8 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
174  /*  D0 */   '}',     0,      0,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,-ESC_N,      0, -ESC_P,
175  /*  D8 */-ESC_Q,     0,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
176  /*  E0 */  '\\',     0, -ESC_S,       0,      0,     0, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
177  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
178  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,
179  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0
# Line 106  static const short int escapes[] = { Line 181  static const short int escapes[] = {
181  #endif  #endif
182    
183    
184  /* Tables of names of POSIX character classes and their lengths. The list is  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
185  terminated by a zero length entry. The first three must be alpha, upper, lower,  searched linearly. Put all the names into a single string, in order to reduce
186  as this is assumed for handling case independence. */  the number of relocations when a shared library is dynamically linked. The
187    string is built from string macros so that it works in UTF-8 mode on EBCDIC
188  static const char *const posix_names[] = {  platforms. */
189    "alpha", "lower", "upper",  
190    "alnum", "ascii", "blank", "cntrl", "digit", "graph",  typedef struct verbitem {
191    "print", "punct", "space", "word",  "xdigit" };    int   len;                 /* Length of verb name */
192      int   op;                  /* Op when no arg, or -1 if arg mandatory */
193      int   op_arg;              /* Op when arg present, or -1 if not allowed */
194    } verbitem;
195    
196    static const char verbnames[] =
197      "\0"                       /* Empty name is a shorthand for MARK */
198      STRING_MARK0
199      STRING_ACCEPT0
200      STRING_COMMIT0
201      STRING_F0
202      STRING_FAIL0
203      STRING_PRUNE0
204      STRING_SKIP0
205      STRING_THEN;
206    
207    static const verbitem verbs[] = {
208      { 0, -1,        OP_MARK },
209      { 4, -1,        OP_MARK },
210      { 6, OP_ACCEPT, -1 },
211      { 6, OP_COMMIT, -1 },
212      { 1, OP_FAIL,   -1 },
213      { 4, OP_FAIL,   -1 },
214      { 5, OP_PRUNE,  OP_PRUNE_ARG },
215      { 4, OP_SKIP,   OP_SKIP_ARG  },
216      { 4, OP_THEN,   OP_THEN_ARG  }
217    };
218    
219    static const int verbcount = sizeof(verbs)/sizeof(verbitem);
220    
221    
222    /* Tables of names of POSIX character classes and their lengths. The names are
223    now all in a single string, to reduce the number of relocations when a shared
224    library is dynamically loaded. The list of lengths is terminated by a zero
225    length entry. The first three must be alpha, lower, upper, as this is assumed
226    for handling case independence. */
227    
228    static const char posix_names[] =
229      STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
230      STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
231      STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
232      STRING_word0  STRING_xdigit;
233    
234  static const uschar posix_name_lengths[] = {  static const uschar posix_name_lengths[] = {
235    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
236    
237  /* Table of class bit maps for each POSIX class; up to three may be combined  /* Table of class bit maps for each POSIX class. Each class is formed from a
238  to form the class. The table for [:blank:] is dynamically modified to remove  base map, with an optional addition or removal of another map. Then, for some
239  the vertical space characters. */  classes, there is some additional tweaking: for [:blank:] the vertical space
240    characters are removed, and for [:alpha:] and [:alnum:] the underscore
241    character is removed. The triples in the table consist of the base map offset,
242    second map offset or -1 if no second map, and a non-negative value for map
243    addition or a negative value for map subtraction (if there are two maps). The
244    absolute value of the third field has these meanings: 0 => no tweaking, 1 =>
245    remove vertical space characters, 2 => remove underscore. */
246    
247  static const int posix_class_maps[] = {  static const int posix_class_maps[] = {
248    cbit_lower, cbit_upper, -1,             /* alpha */    cbit_word,  cbit_digit, -2,             /* alpha */
249    cbit_lower, -1,         -1,             /* lower */    cbit_lower, -1,          0,             /* lower */
250    cbit_upper, -1,         -1,             /* upper */    cbit_upper, -1,          0,             /* upper */
251    cbit_digit, cbit_lower, cbit_upper,     /* alnum */    cbit_word,  -1,          2,             /* alnum - word without underscore */
252    cbit_print, cbit_cntrl, -1,             /* ascii */    cbit_print, cbit_cntrl,  0,             /* ascii */
253    cbit_space, -1,         -1,             /* blank - a GNU extension */    cbit_space, -1,          1,             /* blank - a GNU extension */
254    cbit_cntrl, -1,         -1,             /* cntrl */    cbit_cntrl, -1,          0,             /* cntrl */
255    cbit_digit, -1,         -1,             /* digit */    cbit_digit, -1,          0,             /* digit */
256    cbit_graph, -1,         -1,             /* graph */    cbit_graph, -1,          0,             /* graph */
257    cbit_print, -1,         -1,             /* print */    cbit_print, -1,          0,             /* print */
258    cbit_punct, -1,         -1,             /* punct */    cbit_punct, -1,          0,             /* punct */
259    cbit_space, -1,         -1,             /* space */    cbit_space, -1,          0,             /* space */
260    cbit_word,  -1,         -1,             /* word - a Perl extension */    cbit_word,  -1,          0,             /* word - a Perl extension */
261    cbit_xdigit,-1,         -1              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
262  };  };
263    
264    /* Table of substitutes for \d etc when PCRE_UCP is set. The POSIX class
265    substitutes must be in the order of the names, defined above, and there are
266    both positive and negative cases. NULL means no substitute. */
267    
268  /* The texts of compile-time error messages. These are "char *" because they  #ifdef SUPPORT_UCP
269  are passed to the outside world. */  static const uschar *substitutes[] = {
270      (uschar *)"\\P{Nd}",    /* \D */
271      (uschar *)"\\p{Nd}",    /* \d */
272      (uschar *)"\\P{Xsp}",   /* \S */       /* NOTE: Xsp is Perl space */
273      (uschar *)"\\p{Xsp}",   /* \s */
274      (uschar *)"\\P{Xwd}",   /* \W */
275      (uschar *)"\\p{Xwd}"    /* \w */
276    };
277    
278    static const uschar *posix_substitutes[] = {
279      (uschar *)"\\p{L}",     /* alpha */
280      (uschar *)"\\p{Ll}",    /* lower */
281      (uschar *)"\\p{Lu}",    /* upper */
282      (uschar *)"\\p{Xan}",   /* alnum */
283      NULL,                   /* ascii */
284      (uschar *)"\\h",        /* blank */
285      NULL,                   /* cntrl */
286      (uschar *)"\\p{Nd}",    /* digit */
287      NULL,                   /* graph */
288      NULL,                   /* print */
289      NULL,                   /* punct */
290      (uschar *)"\\p{Xps}",   /* space */    /* NOTE: Xps is POSIX space */
291      (uschar *)"\\p{Xwd}",   /* word */
292      NULL,                   /* xdigit */
293      /* Negated cases */
294      (uschar *)"\\P{L}",     /* ^alpha */
295      (uschar *)"\\P{Ll}",    /* ^lower */
296      (uschar *)"\\P{Lu}",    /* ^upper */
297      (uschar *)"\\P{Xan}",   /* ^alnum */
298      NULL,                   /* ^ascii */
299      (uschar *)"\\H",        /* ^blank */
300      NULL,                   /* ^cntrl */
301      (uschar *)"\\P{Nd}",    /* ^digit */
302      NULL,                   /* ^graph */
303      NULL,                   /* ^print */
304      NULL,                   /* ^punct */
305      (uschar *)"\\P{Xps}",   /* ^space */   /* NOTE: Xps is POSIX space */
306      (uschar *)"\\P{Xwd}",   /* ^word */
307      NULL                    /* ^xdigit */
308    };
309    #define POSIX_SUBSIZE (sizeof(posix_substitutes)/sizeof(uschar *))
310    #endif
311    
312    #define STRING(a)  # a
313    #define XSTRING(s) STRING(s)
314    
315  static const char *error_texts[] = {  /* The texts of compile-time error messages. These are "char *" because they
316    "no error",  are passed to the outside world. Do not ever re-use any error number, because
317    "\\ at end of pattern",  they are documented. Always add a new error instead. Messages marked DEAD below
318    "\\c at end of pattern",  are no longer used. This used to be a table of strings, but in order to reduce
319    "unrecognized character follows \\",  the number of relocations needed when a shared library is loaded dynamically,
320    "numbers out of order in {} quantifier",  it is now one long string. We cannot use a table of offsets, because the
321    lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
322    simply count through to the one we want - this isn't a performance issue
323    because these strings are used only when there is a compilation error.
324    
325    Each substring ends with \0 to insert a null character. This includes the final
326    substring, so that the whole string ends with \0\0, which can be detected when
327    counting through. */
328    
329    static const char error_texts[] =
330      "no error\0"
331      "\\ at end of pattern\0"
332      "\\c at end of pattern\0"
333      "unrecognized character follows \\\0"
334      "numbers out of order in {} quantifier\0"
335    /* 5 */    /* 5 */
336    "number too big in {} quantifier",    "number too big in {} quantifier\0"
337    "missing terminating ] for character class",    "missing terminating ] for character class\0"
338    "invalid escape sequence in character class",    "invalid escape sequence in character class\0"
339    "range out of order in character class",    "range out of order in character class\0"
340    "nothing to repeat",    "nothing to repeat\0"
341    /* 10 */    /* 10 */
342    "operand of unlimited repeat could match the empty string",    "operand of unlimited repeat could match the empty string\0"  /** DEAD **/
343    "internal error: unexpected repeat",    "internal error: unexpected repeat\0"
344    "unrecognized character after (?",    "unrecognized character after (? or (?-\0"
345    "POSIX named classes are supported only within a class",    "POSIX named classes are supported only within a class\0"
346    "missing )",    "missing )\0"
347    /* 15 */    /* 15 */
348    "reference to non-existent subpattern",    "reference to non-existent subpattern\0"
349    "erroffset passed as NULL",    "erroffset passed as NULL\0"
350    "unknown option bit(s) set",    "unknown option bit(s) set\0"
351    "missing ) after comment",    "missing ) after comment\0"
352    "parentheses nested too deeply",    "parentheses nested too deeply\0"  /** DEAD **/
353    /* 20 */    /* 20 */
354    "regular expression too large",    "regular expression is too large\0"
355    "failed to get memory",    "failed to get memory\0"
356    "unmatched parentheses",    "unmatched parentheses\0"
357    "internal error: code overflow",    "internal error: code overflow\0"
358    "unrecognized character after (?<",    "unrecognized character after (?<\0"
359    /* 25 */    /* 25 */
360    "lookbehind assertion is not fixed length",    "lookbehind assertion is not fixed length\0"
361    "malformed number after (?(",    "malformed number or name after (?(\0"
362    "conditional group contains more than two branches",    "conditional group contains more than two branches\0"
363    "assertion expected after (?(",    "assertion expected after (?(\0"
364    "(?R or (?digits must be followed by )",    "(?R or (?[+-]digits must be followed by )\0"
365    /* 30 */    /* 30 */
366    "unknown POSIX class name",    "unknown POSIX class name\0"
367    "POSIX collating elements are not supported",    "POSIX collating elements are not supported\0"
368    "this version of PCRE is not compiled with PCRE_UTF8 support",    "this version of PCRE is not compiled with PCRE_UTF8 support\0"
369    "spare error",    "spare error\0"  /** DEAD **/
370    "character value in \\x{...} sequence is too large",    "character value in \\x{...} sequence is too large\0"
371    /* 35 */    /* 35 */
372    "invalid condition (?(0)",    "invalid condition (?(0)\0"
373    "\\C not allowed in lookbehind assertion",    "\\C not allowed in lookbehind assertion\0"
374    "PCRE does not support \\L, \\l, \\N, \\U, or \\u",    "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
375    "number after (?C is > 255",    "number after (?C is > 255\0"
376    "closing ) for (?C expected",    "closing ) for (?C expected\0"
377    /* 40 */    /* 40 */
378    "recursive call could loop indefinitely",    "recursive call could loop indefinitely\0"
379    "unrecognized character after (?P",    "unrecognized character after (?P\0"
380    "syntax error after (?P",    "syntax error in subpattern name (missing terminator)\0"
381    "two named groups have the same name",    "two named subpatterns have the same name\0"
382    "invalid UTF-8 string",    "invalid UTF-8 string\0"
383    /* 45 */    /* 45 */
384    "support for \\P, \\p, and \\X has not been compiled",    "support for \\P, \\p, and \\X has not been compiled\0"
385    "malformed \\P or \\p sequence",    "malformed \\P or \\p sequence\0"
386    "unknown property name after \\P or \\p"    "unknown property name after \\P or \\p\0"
387  };    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
388      "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
389      /* 50 */
390      "repeated subpattern is too long\0"    /** DEAD **/
391      "octal value is greater than \\377 (not in UTF-8 mode)\0"
392      "internal error: overran compiling workspace\0"
393      "internal error: previously-checked referenced subpattern not found\0"
394      "DEFINE group contains more than one branch\0"
395      /* 55 */
396      "repeating a DEFINE group is not allowed\0"
397      "inconsistent NEWLINE options\0"
398      "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
399      "a numbered reference must not be zero\0"
400      "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
401      /* 60 */
402      "(*VERB) not recognized\0"
403      "number is too big\0"
404      "subpattern name expected\0"
405      "digit expected after (?+\0"
406      "] is an invalid data character in JavaScript compatibility mode\0"
407      /* 65 */
408      "different names for subpatterns of the same number are not allowed\0"
409      "(*MARK) must have an argument\0"
410      "this version of PCRE is not compiled with PCRE_UCP support\0"
411      ;
412    
413  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
414  patterns. Note that the tables in chartables are dependent on the locale, and  patterns. Note that the tables in chartables are dependent on the locale, and
# Line 220  For convenience, we use the same bit def Line 426  For convenience, we use the same bit def
426    
427  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
428    
429  #if !EBCDIC    /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC
430    
431    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
432    UTF-8 mode. */
433    
434  static const unsigned char digitab[] =  static const unsigned char digitab[] =
435    {    {
436    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
# Line 256  static const unsigned char digitab[] = Line 466  static const unsigned char digitab[] =
466    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
467    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
468    
469  #else          /* This is the "abnormal" case, for EBCDIC systems */  #else
470    
471    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
472    
473  static const unsigned char digitab[] =  static const unsigned char digitab[] =
474    {    {
475    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
# Line 270  static const unsigned char digitab[] = Line 483  static const unsigned char digitab[] =
483    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 40 */
484    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  72- |     */
485    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 50 */
486    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88-     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  88- 95    */
487    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 60 */
488    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ?     */
489    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */
# Line 304  static const unsigned char ebcdic_charta Line 517  static const unsigned char ebcdic_charta
517    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */    0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*    - 71 */
518    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */    0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /*  72- |  */
519    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  & - 87 */
520    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88-  */    0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /*  88- 95 */
521    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*  - -103 */
522    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */    0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ?  */
523    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */
# Line 331  static const unsigned char ebcdic_charta Line 544  static const unsigned char ebcdic_charta
544  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
545    
546  static BOOL  static BOOL
547    compile_regex(int, int, int *, uschar **, const uschar **, int *, BOOL, int,    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,
548      int *, int *, branch_chain *, compile_data *);      int *, int *, branch_chain *, compile_data *, int *);
549    
550    
551    
552    /*************************************************
553    *            Find an error text                  *
554    *************************************************/
555    
556    /* The error texts are now all in one long string, to save on relocations. As
557    some of the text is of unknown length, we can't use a table of offsets.
558    Instead, just count through the strings. This is not a performance issue
559    because it happens only when there has been a compilation error.
560    
561    Argument:   the error number
562    Returns:    pointer to the error string
563    */
564    
565    static const char *
566    find_error_text(int n)
567    {
568    const char *s = error_texts;
569    for (; n > 0; n--)
570      {
571      while (*s++ != 0) {};
572      if (*s == 0) return "Error text not found (please report)";
573      }
574    return s;
575    }
576    
577    
578  /*************************************************  /*************************************************
# Line 342  static BOOL Line 581  static BOOL
581    
582  /* This function is called when a \ has been encountered. It either returns a  /* This function is called when a \ has been encountered. It either returns a
583  positive value for a simple escape such as \n, or a negative value which  positive value for a simple escape such as \n, or a negative value which
584  encodes one of the more complicated things such as \d. When UTF-8 is enabled,  encodes one of the more complicated things such as \d. A backreference to group
585  a positive value greater than 255 may be returned. On entry, ptr is pointing at  n is returned as -(ESC_REF + n); ESC_REF is the highest ESC_xxx macro. When
586  the \. On exit, it is on the final character of the escape sequence.  UTF-8 is enabled, a positive value greater than 255 may be returned. On entry,
587    ptr is pointing at the \. On exit, it is on the final character of the escape
588    sequence.
589    
590  Arguments:  Arguments:
591    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
# Line 355  Arguments: Line 596  Arguments:
596    
597  Returns:         zero or positive => a data character  Returns:         zero or positive => a data character
598                   negative => a special escape sequence                   negative => a special escape sequence
599                   on error, errorptr is set                   on error, errorcodeptr is set
600  */  */
601    
602  static int  static int
603  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,  check_escape(const uschar **ptrptr, int *errorcodeptr, int bracount,
604    int options, BOOL isclass)    int options, BOOL isclass)
605  {  {
606  const uschar *ptr = *ptrptr;  BOOL utf8 = (options & PCRE_UTF8) != 0;
607    const uschar *ptr = *ptrptr + 1;
608  int c, i;  int c, i;
609    
610    GETCHARINCTEST(c, ptr);           /* Get character value, increment pointer */
611    ptr--;                            /* Set pointer back to the last byte */
612    
613  /* If backslash is at the end of the pattern, it's an error. */  /* If backslash is at the end of the pattern, it's an error. */
614    
 c = *(++ptr);  
615  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
616    
617  /* Non-alphamerics are literals. For digits or letters, do an initial lookup in  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
618  a table. A non-zero result is something that can be returned immediately.  in a table. A non-zero result is something that can be returned immediately.
619  Otherwise further processing may be required. */  Otherwise further processing may be required. */
620    
621  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
622  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */
623  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0) c = i;
624    
625  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
626  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */
627  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
628  #endif  #endif
629    
# Line 388  else if ((i = escapes[c - 0x48]) != 0) Line 632  else if ((i = escapes[c - 0x48]) != 0)
632  else  else
633    {    {
634    const uschar *oldptr;    const uschar *oldptr;
635      BOOL braced, negated;
636    
637    switch (c)    switch (c)
638      {      {
639      /* A number of Perl escapes are not handled by PCRE. We give an explicit      /* A number of Perl escapes are not handled by PCRE. We give an explicit
640      error. */      error. */
641    
642      case 'l':      case CHAR_l:
643      case 'L':      case CHAR_L:
644      case 'N':      case CHAR_u:
645      case 'u':      case CHAR_U:
     case 'U':  
646      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
647      break;      break;
648    
649        /* \g must be followed by one of a number of specific things:
650    
651        (1) A number, either plain or braced. If positive, it is an absolute
652        backreference. If negative, it is a relative backreference. This is a Perl
653        5.10 feature.
654    
655        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
656        is part of Perl's movement towards a unified syntax for back references. As
657        this is synonymous with \k{name}, we fudge it up by pretending it really
658        was \k.
659    
660        (3) For Oniguruma compatibility we also support \g followed by a name or a
661        number either in angle brackets or in single quotes. However, these are
662        (possibly recursive) subroutine calls, _not_ backreferences. Just return
663        the -ESC_g code (cf \k). */
664    
665        case CHAR_g:
666        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
667          {
668          c = -ESC_g;
669          break;
670          }
671    
672        /* Handle the Perl-compatible cases */
673    
674        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
675          {
676          const uschar *p;
677          for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
678            if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
679          if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
680            {
681            c = -ESC_k;
682            break;
683            }
684          braced = TRUE;
685          ptr++;
686          }
687        else braced = FALSE;
688    
689        if (ptr[1] == CHAR_MINUS)
690          {
691          negated = TRUE;
692          ptr++;
693          }
694        else negated = FALSE;
695    
696        c = 0;
697        while ((digitab[ptr[1]] & ctype_digit) != 0)
698          c = c * 10 + *(++ptr) - CHAR_0;
699    
700        if (c < 0)   /* Integer overflow */
701          {
702          *errorcodeptr = ERR61;
703          break;
704          }
705    
706        if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
707          {
708          *errorcodeptr = ERR57;
709          break;
710          }
711    
712        if (c == 0)
713          {
714          *errorcodeptr = ERR58;
715          break;
716          }
717    
718        if (negated)
719          {
720          if (c > bracount)
721            {
722            *errorcodeptr = ERR15;
723            break;
724            }
725          c = bracount - (c - 1);
726          }
727    
728        c = -(ESC_REF + c);
729        break;
730    
731      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
732      starting with one that is not zero is not straightforward. By experiment,      starting with one that is not zero is not straightforward. By experiment,
733      the way Perl works seems to be as follows:      the way Perl works seems to be as follows:
# Line 413  else Line 740  else
740      value is greater than 377, the least significant 8 bits are taken. Inside a      value is greater than 377, the least significant 8 bits are taken. Inside a
741      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
742    
743      case '1': case '2': case '3': case '4': case '5':      case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:
744      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
745    
746      if (!isclass)      if (!isclass)
747        {        {
748        oldptr = ptr;        oldptr = ptr;
749        c -= '0';        c -= CHAR_0;
750        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
751          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - CHAR_0;
752          if (c < 0)    /* Integer overflow */
753            {
754            *errorcodeptr = ERR61;
755            break;
756            }
757        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
758          {          {
759          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 434  else Line 766  else
766      generates a binary zero byte and treats the digit as a following literal.      generates a binary zero byte and treats the digit as a following literal.
767      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
768    
769      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
770        {        {
771        ptr--;        ptr--;
772        c = 0;        c = 0;
# Line 442  else Line 774  else
774        }        }
775    
776      /* \0 always starts an octal number, but we may drop through to here with a      /* \0 always starts an octal number, but we may drop through to here with a
777      larger first octal digit. */      larger first octal digit. The original code used just to take the least
778        significant 8 bits of octal numbers (I think this is what early Perls used
779      case '0':      to do). Nowadays we allow for larger numbers in UTF-8 mode, but no more
780      c -= '0';      than 3 octal digits. */
781      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')  
782          c = c * 8 + *(++ptr) - '0';      case CHAR_0:
783      c &= 255;     /* Take least significant 8 bits */      c -= CHAR_0;
784        while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
785            c = c * 8 + *(++ptr) - CHAR_0;
786        if (!utf8 && c > 255) *errorcodeptr = ERR51;
787      break;      break;
788    
789      /* \x is complicated when UTF-8 is enabled. \x{ddd} is a character number      /* \x is complicated. \x{ddd} is a character number which can be greater
790      which can be greater than 0xff, but only if the ddd are hex digits. */      than 0xff in utf8 mode, but only if the ddd are hex digits. If not, { is
791        treated as a data character. */
792    
793      case 'x':      case CHAR_x:
794  #ifdef SUPPORT_UTF8      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
     if (ptr[1] == '{' && (options & PCRE_UTF8) != 0)  
795        {        {
796        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
797        register int count = 0;        int count = 0;
798    
799        c = 0;        c = 0;
800        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
801          {          {
802          int cc = *pt++;          register int cc = *pt++;
803            if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
804          count++;          count++;
805  #if !EBCDIC    /* ASCII coding */  
806          if (cc >= 'a') cc -= 32;               /* Convert to upper case */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
807          c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
808  #else          /* EBCDIC coding */          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
809          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */  #else           /* EBCDIC coding */
810          c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
811            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
812  #endif  #endif
813          }          }
814        if (*pt == '}')  
815          if (*pt == CHAR_RIGHT_CURLY_BRACKET)
816          {          {
817          if (c < 0 || count > 8) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
818          ptr = pt;          ptr = pt;
819          break;          break;
820          }          }
821    
822        /* If the sequence of hex digits does not end with '}', then we don't        /* If the sequence of hex digits does not end with '}', then we don't
823        recognize this construct; fall through to the normal \x handling. */        recognize this construct; fall through to the normal \x handling. */
824        }        }
 #endif  
825    
826      /* Read just a single hex char */      /* Read just a single-byte hex-defined char */
827    
828      c = 0;      c = 0;
829      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
830        {        {
831        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
832        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
833  #if !EBCDIC    /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
834        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
835        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
836  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
837        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
838        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
839  #endif  #endif
840        }        }
841      break;      break;
842    
843      /* Other special escapes not starting with a digit are straightforward */      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
844        This coding is ASCII-specific, but then the whole concept of \cx is
845        ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
846    
847      case 'c':      case CHAR_c:
848      c = *(++ptr);      c = *(++ptr);
849      if (c == 0)      if (c == 0)
850        {        {
851        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
852        return 0;        break;
853        }        }
854    
855      /* A letter is upper-cased; then the 0x40 bit is flipped. This coding  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
856      is ASCII-specific, but then the whole concept of \cx is ASCII-specific.      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
     (However, an EBCDIC equivalent has now been added.) */  
   
 #if !EBCDIC    /* ASCII coding */  
     if (c >= 'a' && c <= 'z') c -= 32;  
857      c ^= 0x40;      c ^= 0x40;
858  #else          /* EBCDIC coding */  #else           /* EBCDIC coding */
859      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
860      c ^= 0xC0;      c ^= 0xC0;
861  #endif  #endif
862      break;      break;
863    
864      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
865      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
866      for Perl compatibility, it is a literal. This code looks a bit odd, but      otherwise, for Perl compatibility, it is a literal. This code looks a bit
867      there used to be some cases other than the default, and there may be again      odd, but there used to be some cases other than the default, and there may
868      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
869    
870      default:      default:
871      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 541  else Line 878  else
878      }      }
879    }    }
880    
881    /* Perl supports \N{name} for character names, as well as plain \N for "not
882    newline". PCRE does not support \N{name}. */
883    
884    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET)
885      *errorcodeptr = ERR37;
886    
887    /* If PCRE_UCP is set, we change the values for \d etc. */
888    
889    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
890      c -= (ESC_DU - ESC_D);
891    
892    /* Set the pointer to the final character before returning. */
893    
894  *ptrptr = ptr;  *ptrptr = ptr;
895  return c;  return c;
896  }  }
# Line 560  escape sequence. Line 910  escape sequence.
910  Argument:  Argument:
911    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
912    negptr         points to a boolean that is set TRUE for negation else FALSE    negptr         points to a boolean that is set TRUE for negation else FALSE
913      dptr           points to an int that is set to the detailed property value
914    errorcodeptr   points to the error code variable    errorcodeptr   points to the error code variable
915    
916  Returns:     value from ucp_type_table, or -1 for an invalid type  Returns:         type value from ucp_type_table, or -1 for an invalid type
917  */  */
918    
919  static int  static int
920  get_ucp(const uschar **ptrptr, BOOL *negptr, int *errorcodeptr)  get_ucp(const uschar **ptrptr, BOOL *negptr, int *dptr, int *errorcodeptr)
921  {  {
922  int c, i, bot, top;  int c, i, bot, top;
923  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
924  char name[4];  char name[32];
925    
926  c = *(++ptr);  c = *(++ptr);
927  if (c == 0) goto ERROR_RETURN;  if (c == 0) goto ERROR_RETURN;
928    
929  *negptr = FALSE;  *negptr = FALSE;
930    
931  /* \P or \p can be followed by a one- or two-character name in {}, optionally  /* \P or \p can be followed by a name in {}, optionally preceded by ^ for
932  preceded by ^ for negation. */  negation. */
933    
934  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
935    {    {
936    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
937      {      {
938      *negptr = TRUE;      *negptr = TRUE;
939      ptr++;      ptr++;
940      }      }
941    for (i = 0; i <= 2; i++)    for (i = 0; i < (int)sizeof(name) - 1; i++)
942      {      {
943      c = *(++ptr);      c = *(++ptr);
944      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
945      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
946      name[i] = c;      name[i] = c;
947      }      }
948    if (c !='}')   /* Try to distinguish error cases */    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
     {  
     while (*(++ptr) != 0 && *ptr != '}');  
     if (*ptr == '}') goto UNKNOWN_RETURN; else goto ERROR_RETURN;  
     }  
949    name[i] = 0;    name[i] = 0;
950    }    }
951    
# Line 619  top = _pcre_utt_size; Line 966  top = _pcre_utt_size;
966    
967  while (bot < top)  while (bot < top)
968    {    {
969    i = (bot + top)/2;    i = (bot + top) >> 1;
970    c = strcmp(name, _pcre_utt[i].name);    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);
971    if (c == 0) return _pcre_utt[i].value;    if (c == 0)
972        {
973        *dptr = _pcre_utt[i].value;
974        return _pcre_utt[i].type;
975        }
976    if (c > 0) bot = i + 1; else top = i;    if (c > 0) bot = i + 1; else top = i;
977    }    }
978    
 UNKNOWN_RETURN:  
979  *errorcodeptr = ERR47;  *errorcodeptr = ERR47;
980  *ptrptr = ptr;  *ptrptr = ptr;
981  return -1;  return -1;
# Line 660  is_counted_repeat(const uschar *p) Line 1010  is_counted_repeat(const uschar *p)
1010  {  {
1011  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
1012  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
1013  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
1014    
1015  if (*p++ != ',') return FALSE;  if (*p++ != CHAR_COMMA) return FALSE;
1016  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
1017    
1018  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
1019  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
1020    
1021  return (*p == '}');  return (*p == CHAR_RIGHT_CURLY_BRACKET);
1022  }  }
1023    
1024    
# Line 698  read_repeat_counts(const uschar *p, int Line 1048  read_repeat_counts(const uschar *p, int
1048  int min = 0;  int min = 0;
1049  int max = -1;  int max = -1;
1050    
1051  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  /* Read the minimum value and do a paranoid check: a negative value indicates
1052    an integer overflow. */
1053    
1054    while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;
1055    if (min < 0 || min > 65535)
1056      {
1057      *errorcodeptr = ERR5;
1058      return p;
1059      }
1060    
1061    /* Read the maximum value if there is one, and again do a paranoid on its size.
1062    Also, max must not be less than min. */
1063    
1064  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1065    {    {
1066    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1067      {      {
1068      max = 0;      max = 0;
1069      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;
1070        if (max < 0 || max > 65535)
1071          {
1072          *errorcodeptr = ERR5;
1073          return p;
1074          }
1075      if (max < min)      if (max < min)
1076        {        {
1077        *errorcodeptr = ERR4;        *errorcodeptr = ERR4;
# Line 714  if (*p == '}') max = min; else Line 1080  if (*p == '}') max = min; else
1080      }      }
1081    }    }
1082    
1083  /* Do paranoid checks, then fill in the required variables, and pass back the  /* Fill in the required variables, and pass back the pointer to the terminating
1084  pointer to the terminating '}'. */  '}'. */
1085    
1086  if (min > 65535 || max > 65535)  *minp = min;
1087    *errorcodeptr = ERR5;  *maxp = max;
1088  else  return p;
1089    }
1090    
1091    
1092    
1093    /*************************************************
1094    *  Subroutine for finding forward reference      *
1095    *************************************************/
1096    
1097    /* This recursive function is called only from find_parens() below. The
1098    top-level call starts at the beginning of the pattern. All other calls must
1099    start at a parenthesis. It scans along a pattern's text looking for capturing
1100    subpatterns, and counting them. If it finds a named pattern that matches the
1101    name it is given, it returns its number. Alternatively, if the name is NULL, it
1102    returns when it reaches a given numbered subpattern. We know that if (?P< is
1103    encountered, the name will be terminated by '>' because that is checked in the
1104    first pass. Recursion is used to keep track of subpatterns that reset the
1105    capturing group numbers - the (?| feature.
1106    
1107    Arguments:
1108      ptrptr       address of the current character pointer (updated)
1109      cd           compile background data
1110      name         name to seek, or NULL if seeking a numbered subpattern
1111      lorn         name length, or subpattern number if name is NULL
1112      xmode        TRUE if we are in /x mode
1113      count        pointer to the current capturing subpattern number (updated)
1114    
1115    Returns:       the number of the named subpattern, or -1 if not found
1116    */
1117    
1118    static int
1119    find_parens_sub(uschar **ptrptr, compile_data *cd, const uschar *name, int lorn,
1120      BOOL xmode, int *count)
1121    {
1122    uschar *ptr = *ptrptr;
1123    int start_count = *count;
1124    int hwm_count = start_count;
1125    BOOL dup_parens = FALSE;
1126    
1127    /* If the first character is a parenthesis, check on the type of group we are
1128    dealing with. The very first call may not start with a parenthesis. */
1129    
1130    if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1131    {    {
1132    *minp = min;    if (ptr[1] == CHAR_QUESTION_MARK &&
1133    *maxp = max;        ptr[2] == CHAR_VERTICAL_LINE)
1134        {
1135        ptr += 3;
1136        dup_parens = TRUE;
1137        }
1138    
1139      /* Handle a normal, unnamed capturing parenthesis */
1140    
1141      else if (ptr[1] != CHAR_QUESTION_MARK && ptr[1] != CHAR_ASTERISK)
1142        {
1143        *count += 1;
1144        if (name == NULL && *count == lorn) return *count;
1145        ptr++;
1146        }
1147    
1148      /* Handle a condition. If it is an assertion, just carry on so that it
1149      is processed as normal. If not, skip to the closing parenthesis of the
1150      condition (there can't be any nested parens. */
1151    
1152      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1153        {
1154        ptr += 2;
1155        if (ptr[1] != CHAR_QUESTION_MARK)
1156          {
1157          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1158          if (*ptr != 0) ptr++;
1159          }
1160        }
1161    
1162      /* We have either (? or (* and not a condition */
1163    
1164      else
1165        {
1166        ptr += 2;
1167        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1168    
1169        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1170    
1171        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1172            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1173          {
1174          int term;
1175          const uschar *thisname;
1176          *count += 1;
1177          if (name == NULL && *count == lorn) return *count;
1178          term = *ptr++;
1179          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1180          thisname = ptr;
1181          while (*ptr != term) ptr++;
1182          if (name != NULL && lorn == ptr - thisname &&
1183              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1184            return *count;
1185          term++;
1186          }
1187        }
1188    }    }
1189  return p;  
1190    /* Past any initial parenthesis handling, scan for parentheses or vertical
1191    bars. */
1192    
1193    for (; *ptr != 0; ptr++)
1194      {
1195      /* Skip over backslashed characters and also entire \Q...\E */
1196    
1197      if (*ptr == CHAR_BACKSLASH)
1198        {
1199        if (*(++ptr) == 0) goto FAIL_EXIT;
1200        if (*ptr == CHAR_Q) for (;;)
1201          {
1202          while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1203          if (*ptr == 0) goto FAIL_EXIT;
1204          if (*(++ptr) == CHAR_E) break;
1205          }
1206        continue;
1207        }
1208    
1209      /* Skip over character classes; this logic must be similar to the way they
1210      are handled for real. If the first character is '^', skip it. Also, if the
1211      first few characters (either before or after ^) are \Q\E or \E we skip them
1212      too. This makes for compatibility with Perl. Note the use of STR macros to
1213      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1214    
1215      if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1216        {
1217        BOOL negate_class = FALSE;
1218        for (;;)
1219          {
1220          if (ptr[1] == CHAR_BACKSLASH)
1221            {
1222            if (ptr[2] == CHAR_E)
1223              ptr+= 2;
1224            else if (strncmp((const char *)ptr+2,
1225                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1226              ptr += 4;
1227            else
1228              break;
1229            }
1230          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1231            {
1232            negate_class = TRUE;
1233            ptr++;
1234            }
1235          else break;
1236          }
1237    
1238        /* If the next character is ']', it is a data character that must be
1239        skipped, except in JavaScript compatibility mode. */
1240    
1241        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1242            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1243          ptr++;
1244    
1245        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1246          {
1247          if (*ptr == 0) return -1;
1248          if (*ptr == CHAR_BACKSLASH)
1249            {
1250            if (*(++ptr) == 0) goto FAIL_EXIT;
1251            if (*ptr == CHAR_Q) for (;;)
1252              {
1253              while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1254              if (*ptr == 0) goto FAIL_EXIT;
1255              if (*(++ptr) == CHAR_E) break;
1256              }
1257            continue;
1258            }
1259          }
1260        continue;
1261        }
1262    
1263      /* Skip comments in /x mode */
1264    
1265      if (xmode && *ptr == CHAR_NUMBER_SIGN)
1266        {
1267        while (*(++ptr) != 0 && *ptr != CHAR_NL) {};
1268        if (*ptr == 0) goto FAIL_EXIT;
1269        continue;
1270        }
1271    
1272      /* Check for the special metacharacters */
1273    
1274      if (*ptr == CHAR_LEFT_PARENTHESIS)
1275        {
1276        int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, count);
1277        if (rc > 0) return rc;
1278        if (*ptr == 0) goto FAIL_EXIT;
1279        }
1280    
1281      else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1282        {
1283        if (dup_parens && *count < hwm_count) *count = hwm_count;
1284        *ptrptr = ptr;
1285        return -1;
1286        }
1287    
1288      else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1289        {
1290        if (*count > hwm_count) hwm_count = *count;
1291        *count = start_count;
1292        }
1293      }
1294    
1295    FAIL_EXIT:
1296    *ptrptr = ptr;
1297    return -1;
1298    }
1299    
1300    
1301    
1302    
1303    /*************************************************
1304    *       Find forward referenced subpattern       *
1305    *************************************************/
1306    
1307    /* This function scans along a pattern's text looking for capturing
1308    subpatterns, and counting them. If it finds a named pattern that matches the
1309    name it is given, it returns its number. Alternatively, if the name is NULL, it
1310    returns when it reaches a given numbered subpattern. This is used for forward
1311    references to subpatterns. We used to be able to start this scan from the
1312    current compiling point, using the current count value from cd->bracount, and
1313    do it all in a single loop, but the addition of the possibility of duplicate
1314    subpattern numbers means that we have to scan from the very start, in order to
1315    take account of such duplicates, and to use a recursive function to keep track
1316    of the different types of group.
1317    
1318    Arguments:
1319      cd           compile background data
1320      name         name to seek, or NULL if seeking a numbered subpattern
1321      lorn         name length, or subpattern number if name is NULL
1322      xmode        TRUE if we are in /x mode
1323    
1324    Returns:       the number of the found subpattern, or -1 if not found
1325    */
1326    
1327    static int
1328    find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode)
1329    {
1330    uschar *ptr = (uschar *)cd->start_pattern;
1331    int count = 0;
1332    int rc;
1333    
1334    /* If the pattern does not start with an opening parenthesis, the first call
1335    to find_parens_sub() will scan right to the end (if necessary). However, if it
1336    does start with a parenthesis, find_parens_sub() will return when it hits the
1337    matching closing parens. That is why we have to have a loop. */
1338    
1339    for (;;)
1340      {
1341      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, &count);
1342      if (rc > 0 || *ptr++ == 0) break;
1343      }
1344    
1345    return rc;
1346  }  }
1347    
1348    
1349    
1350    
1351  /*************************************************  /*************************************************
1352  *      Find first significant op code            *  *      Find first significant op code            *
1353  *************************************************/  *************************************************/
# Line 778  for (;;) Line 1397  for (;;)
1397    
1398      case OP_CALLOUT:      case OP_CALLOUT:
1399      case OP_CREF:      case OP_CREF:
1400      case OP_BRANUMBER:      case OP_NCREF:
1401        case OP_RREF:
1402        case OP_NRREF:
1403        case OP_DEF:
1404      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1405      break;      break;
1406    
# Line 793  for (;;) Line 1415  for (;;)
1415    
1416    
1417  /*************************************************  /*************************************************
1418  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1419  *************************************************/  *************************************************/
1420    
1421  /* Scan a pattern and compute the fixed length of subject that will match it,  /* Scan a branch and compute the fixed length of subject that will match it,
1422  if the length is fixed. This is needed for dealing with backward assertions.  if the length is fixed. This is needed for dealing with backward assertions.
1423  In UTF8 mode, the result is in characters rather than bytes.  In UTF8 mode, the result is in characters rather than bytes. The branch is
1424    temporarily terminated with OP_END when this function is called.
1425    
1426    This function is called when a backward assertion is encountered, so that if it
1427    fails, the error message can point to the correct place in the pattern.
1428    However, we cannot do this when the assertion contains subroutine calls,
1429    because they can be forward references. We solve this by remembering this case
1430    and doing the check at the end; a flag specifies which mode we are running in.
1431    
1432  Arguments:  Arguments:
1433    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1434    options  the compiling options    options  the compiling options
1435      atend    TRUE if called when the pattern is complete
1436      cd       the "compile data" structure
1437    
1438  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1439                 or -1 if there is no fixed length,
1440               or -2 if \C was encountered               or -2 if \C was encountered
1441                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1442  */  */
1443    
1444  static int  static int
1445  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)
1446  {  {
1447  int length = -1;  int length = -1;
1448    
# Line 822  branch, check the length against that of Line 1455  branch, check the length against that of
1455  for (;;)  for (;;)
1456    {    {
1457    int d;    int d;
1458      uschar *ce, *cs;
1459    register int op = *cc;    register int op = *cc;
   if (op >= OP_BRA) op = OP_BRA;  
   
1460    switch (op)    switch (op)
1461      {      {
1462        case OP_CBRA:
1463      case OP_BRA:      case OP_BRA:
1464      case OP_ONCE:      case OP_ONCE:
1465      case OP_COND:      case OP_COND:
1466      d = find_fixedlength(cc, options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);
1467      if (d < 0) return d;      if (d < 0) return d;
1468      branchlength += d;      branchlength += d;
1469      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 853  for (;;) Line 1486  for (;;)
1486      branchlength = 0;      branchlength = 0;
1487      break;      break;
1488    
1489        /* A true recursion implies not fixed length, but a subroutine call may
1490        be OK. If the subroutine is a forward reference, we can't deal with
1491        it until the end of the pattern, so return -3. */
1492    
1493        case OP_RECURSE:
1494        if (!atend) return -3;
1495        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1496        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1497        if (cc > cs && cc < ce) return -1;                /* Recursion */
1498        d = find_fixedlength(cs + 2, options, atend, cd);
1499        if (d < 0) return d;
1500        branchlength += d;
1501        cc += 1 + LINK_SIZE;
1502        break;
1503    
1504      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1505    
1506      case OP_ASSERT:      case OP_ASSERT:
# Line 865  for (;;) Line 1513  for (;;)
1513      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1514    
1515      case OP_REVERSE:      case OP_REVERSE:
     case OP_BRANUMBER:  
1516      case OP_CREF:      case OP_CREF:
1517        case OP_NCREF:
1518        case OP_RREF:
1519        case OP_NRREF:
1520        case OP_DEF:
1521      case OP_OPT:      case OP_OPT:
1522      case OP_CALLOUT:      case OP_CALLOUT:
1523      case OP_SOD:      case OP_SOD:
1524      case OP_SOM:      case OP_SOM:
1525        case OP_SET_SOM:
1526      case OP_EOD:      case OP_EOD:
1527      case OP_EODN:      case OP_EODN:
1528      case OP_CIRC:      case OP_CIRC:
# Line 884  for (;;) Line 1536  for (;;)
1536    
1537      case OP_CHAR:      case OP_CHAR:
1538      case OP_CHARNC:      case OP_CHARNC:
1539        case OP_NOT:
1540      branchlength++;      branchlength++;
1541      cc += 2;      cc += 2;
1542  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1543      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1544        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1545  #endif  #endif
1546      break;      break;
1547    
# Line 901  for (;;) Line 1552  for (;;)
1552      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1553      cc += 4;      cc += 4;
1554  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1555      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1556        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1557  #endif  #endif
1558      break;      break;
1559    
1560      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1561      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1562        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1563      cc += 4;      cc += 4;
1564      break;      break;
1565    
# Line 917  for (;;) Line 1567  for (;;)
1567    
1568      case OP_PROP:      case OP_PROP:
1569      case OP_NOTPROP:      case OP_NOTPROP:
1570      cc++;      cc += 2;
1571      /* Fall through */      /* Fall through */
1572    
1573      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
# Line 927  for (;;) Line 1577  for (;;)
1577      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1578      case OP_WORDCHAR:      case OP_WORDCHAR:
1579      case OP_ANY:      case OP_ANY:
1580        case OP_ALLANY:
1581      branchlength++;      branchlength++;
1582      cc++;      cc++;
1583      break;      break;
# Line 981  for (;;) Line 1632  for (;;)
1632    
1633    
1634  /*************************************************  /*************************************************
1635  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1636  *************************************************/  *************************************************/
1637    
1638  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1639  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1640    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1641    so that it can be called from pcre_study() when finding the minimum matching
1642    length.
1643    
1644  Arguments:  Arguments:
1645    code        points to start of expression    code        points to start of expression
1646    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1647    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1648    
1649  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
1650  */  */
1651    
1652  static const uschar *  const uschar *
1653  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1654  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
1655  for (;;)  for (;;)
1656    {    {
1657    register int c = *code;    register int c = *code;
1658    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1659    else if (c > OP_BRA)  
1660      /* XCLASS is used for classes that cannot be represented just by a bit
1661      map. This includes negated single high-valued characters. The length in
1662      the table is zero; the actual length is stored in the compiled code. */
1663    
1664      if (c == OP_XCLASS) code += GET(code, 1);
1665    
1666      /* Handle recursion */
1667    
1668      else if (c == OP_REVERSE)
1669        {
1670        if (number < 0) return (uschar *)code;
1671        code += _pcre_OP_lengths[c];
1672        }
1673    
1674      /* Handle capturing bracket */
1675    
1676      else if (c == OP_CBRA)
1677      {      {
1678      int n = c - OP_BRA;      int n = GET2(code, 1+LINK_SIZE);
     if (n > EXTRACT_BASIC_MAX) n = GET2(code, 2+LINK_SIZE);  
1679      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
1680      code += _pcre_OP_lengths[OP_BRA];      code += _pcre_OP_lengths[c];
1681      }      }
1682    
1683      /* Otherwise, we can get the item's length from the table, except that for
1684      repeated character types, we have to test for \p and \P, which have an extra
1685      two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1686      must add in its length. */
1687    
1688    else    else
1689      {      {
1690      code += _pcre_OP_lengths[c];      switch(c)
1691          {
1692          case OP_TYPESTAR:
1693          case OP_TYPEMINSTAR:
1694          case OP_TYPEPLUS:
1695          case OP_TYPEMINPLUS:
1696          case OP_TYPEQUERY:
1697          case OP_TYPEMINQUERY:
1698          case OP_TYPEPOSSTAR:
1699          case OP_TYPEPOSPLUS:
1700          case OP_TYPEPOSQUERY:
1701          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1702          break;
1703    
1704  #ifdef SUPPORT_UTF8        case OP_TYPEUPTO:
1705          case OP_TYPEMINUPTO:
1706          case OP_TYPEEXACT:
1707          case OP_TYPEPOSUPTO:
1708          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1709          break;
1710    
1711      /* In UTF-8 mode, opcodes that are followed by a character may be followed        case OP_MARK:
1712      by a multi-byte character. The length in the table is a minimum, so we have        case OP_PRUNE_ARG:
1713      to scan along to skip the extra bytes. All opcodes are less than 128, so we        case OP_SKIP_ARG:
1714      can use relatively efficient code. */        case OP_THEN_ARG:
1715          code += code[1];
1716          break;
1717          }
1718    
1719        /* Add in the fixed length from the table */
1720    
1721        code += _pcre_OP_lengths[c];
1722    
1723      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1724      a multi-byte character. The length in the table is a minimum, so we have to
1725      arrange to skip the extra bytes. */
1726    
1727    #ifdef SUPPORT_UTF8
1728      if (utf8) switch(c)      if (utf8) switch(c)
1729        {        {
1730        case OP_CHAR:        case OP_CHAR:
# Line 1031  for (;;) Line 1732  for (;;)
1732        case OP_EXACT:        case OP_EXACT:
1733        case OP_UPTO:        case OP_UPTO:
1734        case OP_MINUPTO:        case OP_MINUPTO:
1735          case OP_POSUPTO:
1736        case OP_STAR:        case OP_STAR:
1737        case OP_MINSTAR:        case OP_MINSTAR:
1738          case OP_POSSTAR:
1739        case OP_PLUS:        case OP_PLUS:
1740        case OP_MINPLUS:        case OP_MINPLUS:
1741          case OP_POSPLUS:
1742        case OP_QUERY:        case OP_QUERY:
1743        case OP_MINQUERY:        case OP_MINQUERY:
1744        while ((*code & 0xc0) == 0x80) code++;        case OP_POSQUERY:
1745        break;        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
   
       /* XCLASS is used for classes that cannot be represented just by a bit  
       map. This includes negated single high-valued characters. The length in  
       the table is zero; the actual length is stored in the compiled code. */  
   
       case OP_XCLASS:  
       code += GET(code, 1) + 1;  
1746        break;        break;
1747        }        }
1748    #else
1749        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1750  #endif  #endif
1751      }      }
1752    }    }
# Line 1072  Returns:      pointer to the opcode for Line 1771  Returns:      pointer to the opcode for
1771  static const uschar *  static const uschar *
1772  find_recurse(const uschar *code, BOOL utf8)  find_recurse(const uschar *code, BOOL utf8)
1773  {  {
 #ifndef SUPPORT_UTF8  
 utf8 = utf8;               /* Stop pedantic compilers complaining */  
 #endif  
   
1774  for (;;)  for (;;)
1775    {    {
1776    register int c = *code;    register int c = *code;
1777    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1778    else if (c == OP_RECURSE) return code;    if (c == OP_RECURSE) return code;
1779    else if (c > OP_BRA)  
1780      {    /* XCLASS is used for classes that cannot be represented just by a bit
1781      code += _pcre_OP_lengths[OP_BRA];    map. This includes negated single high-valued characters. The length in
1782      }    the table is zero; the actual length is stored in the compiled code. */
1783    
1784      if (c == OP_XCLASS) code += GET(code, 1);
1785    
1786      /* Otherwise, we can get the item's length from the table, except that for
1787      repeated character types, we have to test for \p and \P, which have an extra
1788      two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1789      must add in its length. */
1790    
1791    else    else
1792      {      {
1793      code += _pcre_OP_lengths[c];      switch(c)
1794          {
1795          case OP_TYPESTAR:
1796          case OP_TYPEMINSTAR:
1797          case OP_TYPEPLUS:
1798          case OP_TYPEMINPLUS:
1799          case OP_TYPEQUERY:
1800          case OP_TYPEMINQUERY:
1801          case OP_TYPEPOSSTAR:
1802          case OP_TYPEPOSPLUS:
1803          case OP_TYPEPOSQUERY:
1804          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1805          break;
1806    
1807  #ifdef SUPPORT_UTF8        case OP_TYPEPOSUPTO:
1808          case OP_TYPEUPTO:
1809          case OP_TYPEMINUPTO:
1810          case OP_TYPEEXACT:
1811          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1812          break;
1813    
1814          case OP_MARK:
1815          case OP_PRUNE_ARG:
1816          case OP_SKIP_ARG:
1817          case OP_THEN_ARG:
1818          code += code[1];
1819          break;
1820          }
1821    
1822        /* Add in the fixed length from the table */
1823    
1824        code += _pcre_OP_lengths[c];
1825    
1826      /* In UTF-8 mode, opcodes that are followed by a character may be followed      /* In UTF-8 mode, opcodes that are followed by a character may be followed
1827      by a multi-byte character. The length in the table is a minimum, so we have      by a multi-byte character. The length in the table is a minimum, so we have
1828      to scan along to skip the extra bytes. All opcodes are less than 128, so we      to arrange to skip the extra bytes. */
     can use relatively efficient code. */  
1829    
1830    #ifdef SUPPORT_UTF8
1831      if (utf8) switch(c)      if (utf8) switch(c)
1832        {        {
1833        case OP_CHAR:        case OP_CHAR:
# Line 1103  for (;;) Line 1835  for (;;)
1835        case OP_EXACT:        case OP_EXACT:
1836        case OP_UPTO:        case OP_UPTO:
1837        case OP_MINUPTO:        case OP_MINUPTO:
1838          case OP_POSUPTO:
1839        case OP_STAR:        case OP_STAR:
1840        case OP_MINSTAR:        case OP_MINSTAR:
1841          case OP_POSSTAR:
1842        case OP_PLUS:        case OP_PLUS:
1843        case OP_MINPLUS:        case OP_MINPLUS:
1844          case OP_POSPLUS:
1845        case OP_QUERY:        case OP_QUERY:
1846        case OP_MINQUERY:        case OP_MINQUERY:
1847        while ((*code & 0xc0) == 0x80) code++;        case OP_POSQUERY:
1848        break;        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
   
       /* XCLASS is used for classes that cannot be represented just by a bit  
       map. This includes negated single high-valued characters. The length in  
       the table is zero; the actual length is stored in the compiled code. */  
   
       case OP_XCLASS:  
       code += GET(code, 1) + 1;  
1849        break;        break;
1850        }        }
1851    #else
1852        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1853  #endif  #endif
1854      }      }
1855    }    }
# Line 1132  for (;;) Line 1862  for (;;)
1862  *************************************************/  *************************************************/
1863    
1864  /* 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
1865  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()
1866  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
1867  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
1868  whose current branch will already have been scanned.  backward and negative forward assertions when its final argument is TRUE. If we
1869    hit an unclosed bracket, we return "empty" - this means we've struck an inner
1870    bracket whose current branch will already have been scanned.
1871    
1872  Arguments:  Arguments:
1873    code        points to start of search    code        points to start of search
1874    endcode     points to where to stop    endcode     points to where to stop
1875    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
1876      cd          contains pointers to tables etc.
1877    
1878  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
1879  */  */
1880    
1881  static BOOL  static BOOL
1882  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
1883      compile_data *cd)
1884  {  {
1885  register int c;  register int c;
1886  for (code = first_significant_code(code + 1 + LINK_SIZE, NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);
1887       code < endcode;       code < endcode;
1888       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))
1889    {    {
# Line 1157  for (code = first_significant_code(code Line 1891  for (code = first_significant_code(code
1891    
1892    c = *code;    c = *code;
1893    
1894    if (c >= OP_BRA)    /* Skip over forward assertions; the other assertions are skipped by
1895      first_significant_code() with a TRUE final argument. */
1896    
1897      if (c == OP_ASSERT)
1898      {      {
1899      BOOL empty_branch;      do code += GET(code, 1); while (*code == OP_ALT);
1900      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      c = *code;
1901        continue;
1902        }
1903    
1904      /* Groups with zero repeats can of course be empty; skip them. */
1905    
1906      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)
1907        {
1908        code += _pcre_OP_lengths[c];
1909        do code += GET(code, 1); while (*code == OP_ALT);
1910        c = *code;
1911        continue;
1912        }
1913    
1914      /* Scan a closed bracket */    /* For a recursion/subroutine call, if its end has been reached, which
1915      implies a subroutine call, we can scan it. */
1916    
1917      empty_branch = FALSE;    if (c == OP_RECURSE)
1918        {
1919        BOOL empty_branch = FALSE;
1920        const uschar *scode = cd->start_code + GET(code, 1);
1921        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
1922      do      do
1923        {        {
1924        if (!empty_branch && could_be_empty_branch(code, endcode, utf8))        if (could_be_empty_branch(scode, endcode, utf8, cd))
1925            {
1926          empty_branch = TRUE;          empty_branch = TRUE;
1927            break;
1928            }
1929          scode += GET(scode, 1);
1930          }
1931        while (*scode == OP_ALT);
1932        if (!empty_branch) return FALSE;  /* All branches are non-empty */
1933        continue;
1934        }
1935    
1936      /* For other groups, scan the branches. */
1937    
1938      if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)
1939        {
1940        BOOL empty_branch;
1941        if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
1942    
1943        /* If a conditional group has only one branch, there is a second, implied,
1944        empty branch, so just skip over the conditional, because it could be empty.
1945        Otherwise, scan the individual branches of the group. */
1946    
1947        if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
1948        code += GET(code, 1);        code += GET(code, 1);
1949        else
1950          {
1951          empty_branch = FALSE;
1952          do
1953            {
1954            if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
1955              empty_branch = TRUE;
1956            code += GET(code, 1);
1957            }
1958          while (*code == OP_ALT);
1959          if (!empty_branch) return FALSE;   /* All branches are non-empty */
1960        }        }
1961      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
     code += 1 + LINK_SIZE;  
1962      c = *code;      c = *code;
1963        continue;
1964      }      }
1965    
1966    else switch (c)    /* Handle the other opcodes */
1967    
1968      switch (c)
1969      {      {
1970      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
1971        cannot be represented just by a bit map. This includes negated single
1972        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
1973        actual length is stored in the compiled code, so we must update "code"
1974        here. */
1975    
1976  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1977      case OP_XCLASS:      case OP_XCLASS:
1978      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
1979      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
1980  #endif  #endif
1981    
# Line 1227  for (code = first_significant_code(code Line 2019  for (code = first_significant_code(code
2019      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2020      case OP_WORDCHAR:      case OP_WORDCHAR:
2021      case OP_ANY:      case OP_ANY:
2022        case OP_ALLANY:
2023      case OP_ANYBYTE:      case OP_ANYBYTE:
2024      case OP_CHAR:      case OP_CHAR:
2025      case OP_CHARNC:      case OP_CHARNC:
2026      case OP_NOT:      case OP_NOT:
2027      case OP_PLUS:      case OP_PLUS:
2028      case OP_MINPLUS:      case OP_MINPLUS:
2029        case OP_POSPLUS:
2030      case OP_EXACT:      case OP_EXACT:
2031      case OP_NOTPLUS:      case OP_NOTPLUS:
2032      case OP_NOTMINPLUS:      case OP_NOTMINPLUS:
2033        case OP_NOTPOSPLUS:
2034      case OP_NOTEXACT:      case OP_NOTEXACT:
2035      case OP_TYPEPLUS:      case OP_TYPEPLUS:
2036      case OP_TYPEMINPLUS:      case OP_TYPEMINPLUS:
2037        case OP_TYPEPOSPLUS:
2038      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2039      return FALSE;      return FALSE;
2040    
2041        /* These are going to continue, as they may be empty, but we have to
2042        fudge the length for the \p and \P cases. */
2043    
2044        case OP_TYPESTAR:
2045        case OP_TYPEMINSTAR:
2046        case OP_TYPEPOSSTAR:
2047        case OP_TYPEQUERY:
2048        case OP_TYPEMINQUERY:
2049        case OP_TYPEPOSQUERY:
2050        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2051        break;
2052    
2053        /* Same for these */
2054    
2055        case OP_TYPEUPTO:
2056        case OP_TYPEMINUPTO:
2057        case OP_TYPEPOSUPTO:
2058        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
2059        break;
2060    
2061      /* End of branch */      /* End of branch */
2062    
2063      case OP_KET:      case OP_KET:
# Line 1250  for (code = first_significant_code(code Line 2066  for (code = first_significant_code(code
2066      case OP_ALT:      case OP_ALT:
2067      return TRUE;      return TRUE;
2068    
2069      /* In UTF-8 mode, STAR, MINSTAR, QUERY, MINQUERY, UPTO, and MINUPTO  may be      /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2070      followed by a multibyte character */      MINUPTO, and POSUPTO may be followed by a multibyte character */
2071    
2072  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2073      case OP_STAR:      case OP_STAR:
2074      case OP_MINSTAR:      case OP_MINSTAR:
2075        case OP_POSSTAR:
2076      case OP_QUERY:      case OP_QUERY:
2077      case OP_MINQUERY:      case OP_MINQUERY:
2078        case OP_POSQUERY:
2079        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2080        break;
2081    
2082      case OP_UPTO:      case OP_UPTO:
2083      case OP_MINUPTO:      case OP_MINUPTO:
2084      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_POSUPTO:
2085        if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2086      break;      break;
2087  #endif  #endif
2088    
2089        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2090        string. */
2091    
2092        case OP_MARK:
2093        case OP_PRUNE_ARG:
2094        case OP_SKIP_ARG:
2095        case OP_THEN_ARG:
2096        code += code[1];
2097        break;
2098    
2099        /* None of the remaining opcodes are required to match a character. */
2100    
2101        default:
2102        break;
2103      }      }
2104    }    }
2105    
# Line 1285  Arguments: Line 2122  Arguments:
2122    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2123    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2124    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2125      cd          pointers to tables etc
2126    
2127  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2128  */  */
2129    
2130  static BOOL  static BOOL
2131  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2132    BOOL utf8)    BOOL utf8, compile_data *cd)
2133  {  {
2134  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2135    {    {
2136    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2137        return FALSE;
2138    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2139    }    }
2140  return TRUE;  return TRUE;
# Line 1308  return TRUE; Line 2147  return TRUE;
2147  *************************************************/  *************************************************/
2148    
2149  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2150  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
2151  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2152  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2153    
2154    Originally, this function only recognized a sequence of letters between the
2155    terminators, but it seems that Perl recognizes any sequence of characters,
2156    though of course unknown POSIX names are subsequently rejected. Perl gives an
2157    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2158    didn't consider this to be a POSIX class. Likewise for [:1234:].
2159    
2160    The problem in trying to be exactly like Perl is in the handling of escapes. We
2161    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2162    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2163    below handles the special case of \], but does not try to do any other escape
2164    processing. This makes it different from Perl for cases such as [:l\ower:]
2165    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2166    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2167    I think.
2168    
2169  Argument:  Arguments:
2170    ptr      pointer to the initial [    ptr      pointer to the initial [
2171    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2172    
2173  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2174  */  */
2175    
2176  static BOOL  static BOOL
2177  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
2178  {  {
2179  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2180  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2181  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2182    {    {
2183    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else
2184    return TRUE;      {
2185        if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2186        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2187          {
2188          *endptr = ptr;
2189          return TRUE;
2190          }
2191        }
2192    }    }
2193  return FALSE;  return FALSE;
2194  }  }
# Line 1355  Returns:     a value representing the na Line 2213  Returns:     a value representing the na
2213  static int  static int
2214  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
2215  {  {
2216    const char *pn = posix_names;
2217  register int yield = 0;  register int yield = 0;
2218  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2219    {    {
2220    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2221      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
2222      pn += posix_name_lengths[yield] + 1;
2223    yield++;    yield++;
2224    }    }
2225  return -1;  return -1;
# Line 1374  return -1; Line 2234  return -1;
2234  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2235  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2236  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
2237  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
2238  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
2239  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
2240  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
2241  partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2242    OP_END.
2243    
2244    This function has been extended with the possibility of forward references for
2245    recursions and subroutine calls. It must also check the list of such references
2246    for the group we are dealing with. If it finds that one of the recursions in
2247    the current group is on this list, it adjusts the offset in the list, not the
2248    value in the reference (which is a group number).
2249    
2250  Arguments:  Arguments:
2251    group      points to the start of the group    group      points to the start of the group
2252    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
2253    utf8       TRUE in UTF-8 mode    utf8       TRUE in UTF-8 mode
2254    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
2255      save_hwm   the hwm forward reference pointer at the start of the group
2256    
2257  Returns:     nothing  Returns:     nothing
2258  */  */
2259    
2260  static void  static void
2261  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd)  adjust_recurse(uschar *group, int adjust, BOOL utf8, compile_data *cd,
2262      uschar *save_hwm)
2263  {  {
2264  uschar *ptr = group;  uschar *ptr = group;
2265    
2266  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
2267    {    {
2268    int offset = GET(ptr, 1);    int offset;
2269    if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);    uschar *hc;
2270    
2271      /* See if this recursion is on the forward reference list. If so, adjust the
2272      reference. */
2273    
2274      for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
2275        {
2276        offset = GET(hc, 0);
2277        if (cd->start_code + offset == ptr + 1)
2278          {
2279          PUT(hc, 0, offset + adjust);
2280          break;
2281          }
2282        }
2283    
2284      /* Otherwise, adjust the recursion offset if it's after the start of this
2285      group. */
2286    
2287      if (hc >= cd->hwm)
2288        {
2289        offset = GET(ptr, 1);
2290        if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
2291        }
2292    
2293    ptr += 1 + LINK_SIZE;    ptr += 1 + LINK_SIZE;
2294    }    }
2295  }  }
# Line 1475  Yield:        TRUE when range returned; Line 2368  Yield:        TRUE when range returned;
2368  */  */
2369    
2370  static BOOL  static BOOL
2371  get_othercase_range(int *cptr, int d, int *ocptr, int *odptr)  get_othercase_range(unsigned int *cptr, unsigned int d, unsigned int *ocptr,
2372      unsigned int *odptr)
2373  {  {
2374  int c, chartype, othercase, next;  unsigned int c, othercase, next;
2375    
2376  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2377    {    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
   if (_pcre_ucp_findchar(c, &chartype, &othercase) == ucp_L && othercase != 0)  
     break;  
   }  
2378    
2379  if (c > d) return FALSE;  if (c > d) return FALSE;
2380    
# Line 1492  next = othercase + 1; Line 2383  next = othercase + 1;
2383    
2384  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2385    {    {
2386    if (_pcre_ucp_findchar(c, &chartype, &othercase) != ucp_L ||    if (UCD_OTHERCASE(c) != next) break;
         othercase != next)  
     break;  
2387    next++;    next++;
2388    }    }
2389    
# Line 1506  return TRUE; Line 2395  return TRUE;
2395  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2396    
2397    
2398    
2399    /*************************************************
2400    *     Check if auto-possessifying is possible    *
2401    *************************************************/
2402    
2403    /* This function is called for unlimited repeats of certain items, to see
2404    whether the next thing could possibly match the repeated item. If not, it makes
2405    sense to automatically possessify the repeated item.
2406    
2407    Arguments:
2408      op_code       the repeated op code
2409      this          data for this item, depends on the opcode
2410      utf8          TRUE in UTF-8 mode
2411      utf8_char     used for utf8 character bytes, NULL if not relevant
2412      ptr           next character in pattern
2413      options       options bits
2414      cd            contains pointers to tables etc.
2415    
2416    Returns:        TRUE if possessifying is wanted
2417    */
2418    
2419    static BOOL
2420    check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,
2421      const uschar *ptr, int options, compile_data *cd)
2422    {
2423    int next;
2424    
2425    /* Skip whitespace and comments in extended mode */
2426    
2427    if ((options & PCRE_EXTENDED) != 0)
2428      {
2429      for (;;)
2430        {
2431        while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2432        if (*ptr == CHAR_NUMBER_SIGN)
2433          {
2434          while (*(++ptr) != 0)
2435            if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2436          }
2437        else break;
2438        }
2439      }
2440    
2441    /* If the next item is one that we can handle, get its value. A non-negative
2442    value is a character, a negative value is an escape value. */
2443    
2444    if (*ptr == CHAR_BACKSLASH)
2445      {
2446      int temperrorcode = 0;
2447      next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
2448      if (temperrorcode != 0) return FALSE;
2449      ptr++;    /* Point after the escape sequence */
2450      }
2451    
2452    else if ((cd->ctypes[*ptr] & ctype_meta) == 0)
2453      {
2454    #ifdef SUPPORT_UTF8
2455      if (utf8) { GETCHARINC(next, ptr); } else
2456    #endif
2457      next = *ptr++;
2458      }
2459    
2460    else return FALSE;
2461    
2462    /* Skip whitespace and comments in extended mode */
2463    
2464    if ((options & PCRE_EXTENDED) != 0)
2465      {
2466      for (;;)
2467        {
2468        while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2469        if (*ptr == CHAR_NUMBER_SIGN)
2470          {
2471          while (*(++ptr) != 0)
2472            if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2473          }
2474        else break;
2475        }
2476      }
2477    
2478    /* If the next thing is itself optional, we have to give up. */
2479    
2480    if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2481      strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2482        return FALSE;
2483    
2484    /* Now compare the next item with the previous opcode. If the previous is a
2485    positive single character match, "item" either contains the character or, if
2486    "item" is greater than 127 in utf8 mode, the character's bytes are in
2487    utf8_char. */
2488    
2489    
2490    /* Handle cases when the next item is a character. */
2491    
2492    if (next >= 0) switch(op_code)
2493      {
2494      case OP_CHAR:
2495    #ifdef SUPPORT_UTF8
2496      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2497    #else
2498      (void)(utf8_char);  /* Keep compiler happy by referencing function argument */
2499    #endif
2500      return item != next;
2501    
2502      /* For CHARNC (caseless character) we must check the other case. If we have
2503      Unicode property support, we can use it to test the other case of
2504      high-valued characters. */
2505    
2506      case OP_CHARNC:
2507    #ifdef SUPPORT_UTF8
2508      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2509    #endif
2510      if (item == next) return FALSE;
2511    #ifdef SUPPORT_UTF8
2512      if (utf8)
2513        {
2514        unsigned int othercase;
2515        if (next < 128) othercase = cd->fcc[next]; else
2516    #ifdef SUPPORT_UCP
2517        othercase = UCD_OTHERCASE((unsigned int)next);
2518    #else
2519        othercase = NOTACHAR;
2520    #endif
2521        return (unsigned int)item != othercase;
2522        }
2523      else
2524    #endif  /* SUPPORT_UTF8 */
2525      return (item != cd->fcc[next]);  /* Non-UTF-8 mode */
2526    
2527      /* For OP_NOT, "item" must be a single-byte character. */
2528    
2529      case OP_NOT:
2530      if (item == next) return TRUE;
2531      if ((options & PCRE_CASELESS) == 0) return FALSE;
2532    #ifdef SUPPORT_UTF8
2533      if (utf8)
2534        {
2535        unsigned int othercase;
2536        if (next < 128) othercase = cd->fcc[next]; else
2537    #ifdef SUPPORT_UCP
2538        othercase = UCD_OTHERCASE(next);
2539    #else
2540        othercase = NOTACHAR;
2541    #endif
2542        return (unsigned int)item == othercase;
2543        }
2544      else
2545    #endif  /* SUPPORT_UTF8 */
2546      return (item == cd->fcc[next]);  /* Non-UTF-8 mode */
2547    
2548      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2549      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
2550    
2551      case OP_DIGIT:
2552      return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
2553    
2554      case OP_NOT_DIGIT:
2555      return next <= 127 && (cd->ctypes[next] & ctype_digit) != 0;
2556    
2557      case OP_WHITESPACE:
2558      return next > 127 || (cd->ctypes[next] & ctype_space) == 0;
2559    
2560      case OP_NOT_WHITESPACE:
2561      return next <= 127 && (cd->ctypes[next] & ctype_space) != 0;
2562    
2563      case OP_WORDCHAR:
2564      return next > 127 || (cd->ctypes[next] & ctype_word) == 0;
2565    
2566      case OP_NOT_WORDCHAR:
2567      return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;
2568    
2569      case OP_HSPACE:
2570      case OP_NOT_HSPACE:
2571      switch(next)
2572        {
2573        case 0x09:
2574        case 0x20:
2575        case 0xa0:
2576        case 0x1680:
2577        case 0x180e:
2578        case 0x2000:
2579        case 0x2001:
2580        case 0x2002:
2581        case 0x2003:
2582        case 0x2004:
2583        case 0x2005:
2584        case 0x2006:
2585        case 0x2007:
2586        case 0x2008:
2587        case 0x2009:
2588        case 0x200A:
2589        case 0x202f:
2590        case 0x205f:
2591        case 0x3000:
2592        return op_code == OP_NOT_HSPACE;
2593        default:
2594        return op_code != OP_NOT_HSPACE;
2595        }
2596    
2597      case OP_ANYNL:
2598      case OP_VSPACE:
2599      case OP_NOT_VSPACE:
2600      switch(next)
2601        {
2602        case 0x0a:
2603        case 0x0b:
2604        case 0x0c:
2605        case 0x0d:
2606        case 0x85:
2607        case 0x2028:
2608        case 0x2029:
2609        return op_code == OP_NOT_VSPACE;
2610        default:
2611        return op_code != OP_NOT_VSPACE;
2612        }
2613    
2614      default:
2615      return FALSE;
2616      }
2617    
2618    
2619    /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP
2620    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
2621    generated only when PCRE_UCP is *not* set, that is, when only ASCII
2622    characteristics are recognized. */
2623    
2624    switch(op_code)
2625      {
2626      case OP_CHAR:
2627      case OP_CHARNC:
2628    #ifdef SUPPORT_UTF8
2629      if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2630    #endif
2631      switch(-next)
2632        {
2633        case ESC_d:
2634        return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;
2635    
2636        case ESC_D:
2637        return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;
2638    
2639        case ESC_s:
2640        return item > 127 || (cd->ctypes[item] & ctype_space) == 0;
2641    
2642        case ESC_S:
2643        return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;
2644    
2645        case ESC_w:
2646        return item > 127 || (cd->ctypes[item] & ctype_word) == 0;
2647    
2648        case ESC_W:
2649        return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;
2650    
2651        case ESC_h:
2652        case ESC_H:
2653        switch(item)
2654          {
2655          case 0x09:
2656          case 0x20:
2657          case 0xa0:
2658          case 0x1680:
2659          case 0x180e:
2660          case 0x2000:
2661          case 0x2001:
2662          case 0x2002:
2663          case 0x2003:
2664          case 0x2004:
2665          case 0x2005:
2666          case 0x2006:
2667          case 0x2007:
2668          case 0x2008:
2669          case 0x2009:
2670          case 0x200A:
2671          case 0x202f:
2672          case 0x205f:
2673          case 0x3000:
2674          return -next != ESC_h;
2675          default:
2676          return -next == ESC_h;
2677          }
2678    
2679        case ESC_v:
2680        case ESC_V:
2681        switch(item)
2682          {
2683          case 0x0a:
2684          case 0x0b:
2685          case 0x0c:
2686          case 0x0d:
2687          case 0x85:
2688          case 0x2028:
2689          case 0x2029:
2690          return -next != ESC_v;
2691          default:
2692          return -next == ESC_v;
2693          }
2694    
2695        default:
2696        return FALSE;
2697        }
2698    
2699      case OP_DIGIT:
2700      return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2701             next == -ESC_h || next == -ESC_v || next == -ESC_R;
2702    
2703      case OP_NOT_DIGIT:
2704      return next == -ESC_d;
2705    
2706      case OP_WHITESPACE:
2707      return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
2708    
2709      case OP_NOT_WHITESPACE:
2710      return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2711    
2712      case OP_HSPACE:
2713      return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
2714             next == -ESC_w || next == -ESC_v || next == -ESC_R;
2715    
2716      case OP_NOT_HSPACE:
2717      return next == -ESC_h;
2718    
2719      /* Can't have \S in here because VT matches \S (Perl anomaly) */
2720      case OP_ANYNL:
2721      case OP_VSPACE:
2722      return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2723    
2724      case OP_NOT_VSPACE:
2725      return next == -ESC_v || next == -ESC_R;
2726    
2727      case OP_WORDCHAR:
2728      return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
2729             next == -ESC_v || next == -ESC_R;
2730    
2731      case OP_NOT_WORDCHAR:
2732      return next == -ESC_w || next == -ESC_d;
2733    
2734      default:
2735      return FALSE;
2736      }
2737    
2738    /* Control does not reach here */
2739    }
2740    
2741    
2742    
2743  /*************************************************  /*************************************************
2744  *           Compile one branch                   *  *           Compile one branch                   *
2745  *************************************************/  *************************************************/
2746    
2747  /* Scan the pattern, compiling it into the code vector. If the options are  /* Scan the pattern, compiling it into the a vector. If the options are
2748  changed during the branch, the pointer is used to change the external options  changed during the branch, the pointer is used to change the external options
2749  bits.  bits. This function is used during the pre-compile phase when we are trying
2750    to find out the amount of memory needed, as well as during the real compile
2751    phase. The value of lengthptr distinguishes the two phases.
2752    
2753  Arguments:  Arguments:
2754    optionsptr     pointer to the option bits    optionsptr     pointer to the option bits
   brackets       points to number of extracting brackets used  
2755    codeptr        points to the pointer to the current code point    codeptr        points to the pointer to the current code point
2756    ptrptr         points to the current pattern pointer    ptrptr         points to the current pattern pointer
2757    errorcodeptr   points to error code variable    errorcodeptr   points to error code variable
# Line 1524  Arguments: Line 2759  Arguments:
2759    reqbyteptr     set to the last literal character required, else < 0    reqbyteptr     set to the last literal character required, else < 0
2760    bcptr          points to current branch chain    bcptr          points to current branch chain
2761    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
2762      lengthptr      NULL during the real compile phase
2763                     points to length accumulator during pre-compile phase
2764    
2765  Returns:         TRUE on success  Returns:         TRUE on success
2766                   FALSE, with *errorcodeptr set non-zero on error                   FALSE, with *errorcodeptr set non-zero on error
2767  */  */
2768    
2769  static BOOL  static BOOL
2770  compile_branch(int *optionsptr, int *brackets, uschar **codeptr,  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
2771    const uschar **ptrptr, int *errorcodeptr, int *firstbyteptr,    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
2772    int *reqbyteptr, branch_chain *bcptr, compile_data *cd)    compile_data *cd, int *lengthptr)
2773  {  {
2774  int repeat_type, op_type;  int repeat_type, op_type;
2775  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
# Line 1541  int greedy_default, greedy_non_default; Line 2778  int greedy_default, greedy_non_default;
2778  int firstbyte, reqbyte;  int firstbyte, reqbyte;
2779  int zeroreqbyte, zerofirstbyte;  int zeroreqbyte, zerofirstbyte;
2780  int req_caseopt, reqvary, tempreqvary;  int req_caseopt, reqvary, tempreqvary;
 int condcount = 0;  
2781  int options = *optionsptr;  int options = *optionsptr;
2782  int after_manual_callout = 0;  int after_manual_callout = 0;
2783    int length_prevgroup = 0;
2784  register int c;  register int c;
2785  register uschar *code = *codeptr;  register uschar *code = *codeptr;
2786    uschar *last_code = code;
2787    uschar *orig_code = code;
2788  uschar *tempcode;  uschar *tempcode;
2789  BOOL inescq = FALSE;  BOOL inescq = FALSE;
2790  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
2791  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
2792  const uschar *tempptr;  const uschar *tempptr;
2793    const uschar *nestptr = NULL;
2794  uschar *previous = NULL;  uschar *previous = NULL;
2795  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
2796    uschar *save_hwm = NULL;
2797  uschar classbits[32];  uschar classbits[32];
2798    
2799  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2800  BOOL class_utf8;  BOOL class_utf8;
2801  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
2802  uschar *class_utf8data;  uschar *class_utf8data;
2803    uschar *class_utf8data_base;
2804  uschar utf8_char[6];  uschar utf8_char[6];
2805  #else  #else
2806  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
2807    uschar *utf8_char = NULL;
2808    #endif
2809    
2810    #ifdef PCRE_DEBUG
2811    if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2812  #endif  #endif
2813    
2814  /* Set up the default and non-default settings for greediness */  /* Set up the default and non-default settings for greediness */
# Line 1593  req_caseopt = ((options & PCRE_CASELESS) Line 2840  req_caseopt = ((options & PCRE_CASELESS)
2840  for (;; ptr++)  for (;; ptr++)
2841    {    {
2842    BOOL negate_class;    BOOL negate_class;
2843      BOOL should_flip_negation;
2844    BOOL possessive_quantifier;    BOOL possessive_quantifier;
2845    BOOL is_quantifier;    BOOL is_quantifier;
2846      BOOL is_recurse;
2847      BOOL reset_bracount;
2848    int class_charcount;    int class_charcount;
2849    int class_lastchar;    int class_lastchar;
2850    int newoptions;    int newoptions;
2851    int recno;    int recno;
2852      int refsign;
2853    int skipbytes;    int skipbytes;
2854    int subreqbyte;    int subreqbyte;
2855    int subfirstbyte;    int subfirstbyte;
2856      int terminator;
2857    int mclength;    int mclength;
2858    uschar mcbuffer[8];    uschar mcbuffer[8];
2859    
2860    /* Next byte in the pattern */    /* Get next byte in the pattern */
2861    
2862    c = *ptr;    c = *ptr;
2863    
2864      /* If we are at the end of a nested substitution, revert to the outer level
2865      string. Nesting only happens one level deep. */
2866    
2867      if (c == 0 && nestptr != NULL)
2868        {
2869        ptr = nestptr;
2870        nestptr = NULL;
2871        c = *ptr;
2872        }
2873    
2874      /* If we are in the pre-compile phase, accumulate the length used for the
2875      previous cycle of this loop. */
2876    
2877      if (lengthptr != NULL)
2878        {
2879    #ifdef PCRE_DEBUG
2880        if (code > cd->hwm) cd->hwm = code;                 /* High water info */
2881    #endif
2882        if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
2883          {
2884          *errorcodeptr = ERR52;
2885          goto FAILED;
2886          }
2887    
2888        /* There is at least one situation where code goes backwards: this is the
2889        case of a zero quantifier after a class (e.g. [ab]{0}). At compile time,
2890        the class is simply eliminated. However, it is created first, so we have to
2891        allow memory for it. Therefore, don't ever reduce the length at this point.
2892        */
2893    
2894        if (code < last_code) code = last_code;
2895    
2896        /* Paranoid check for integer overflow */
2897    
2898        if (OFLOW_MAX - *lengthptr < code - last_code)
2899          {
2900          *errorcodeptr = ERR20;
2901          goto FAILED;
2902          }
2903    
2904        *lengthptr += code - last_code;
2905        DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));
2906    
2907        /* If "previous" is set and it is not at the start of the work space, move
2908        it back to there, in order to avoid filling up the work space. Otherwise,
2909        if "previous" is NULL, reset the current code pointer to the start. */
2910    
2911        if (previous != NULL)
2912          {
2913          if (previous > orig_code)
2914            {
2915            memmove(orig_code, previous, code - previous);
2916            code -= previous - orig_code;
2917            previous = orig_code;
2918            }
2919          }
2920        else code = orig_code;
2921    
2922        /* Remember where this code item starts so we can pick up the length
2923        next time round. */
2924    
2925        last_code = code;
2926        }
2927    
2928      /* In the real compile phase, just check the workspace used by the forward
2929      reference list. */
2930    
2931      else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
2932        {
2933        *errorcodeptr = ERR52;
2934        goto FAILED;
2935        }
2936    
2937    /* If in \Q...\E, check for the end; if not, we have a literal */    /* If in \Q...\E, check for the end; if not, we have a literal */
2938    
2939    if (inescq && c != 0)    if (inescq && c != 0)
2940      {      {
2941      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
2942        {        {
2943        inescq = FALSE;        inescq = FALSE;
2944        ptr++;        ptr++;
# Line 1623  for (;; ptr++) Line 2948  for (;; ptr++)
2948        {        {
2949        if (previous_callout != NULL)        if (previous_callout != NULL)
2950          {          {
2951          complete_callout(previous_callout, ptr, cd);          if (lengthptr == NULL)  /* Don't attempt in pre-compile phase */
2952              complete_callout(previous_callout, ptr, cd);
2953          previous_callout = NULL;          previous_callout = NULL;
2954          }          }
2955        if ((options & PCRE_AUTO_CALLOUT) != 0)        if ((options & PCRE_AUTO_CALLOUT) != 0)
# Line 1638  for (;; ptr++) Line 2964  for (;; ptr++)
2964    /* Fill in length of a previous callout, except when the next thing is    /* Fill in length of a previous callout, except when the next thing is
2965    a quantifier. */    a quantifier. */
2966    
2967    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
2968      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
2969        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
2970    
2971    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
2972         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
2973      {      {
2974      complete_callout(previous_callout, ptr, cd);      if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */
2975          complete_callout(previous_callout, ptr, cd);
2976      previous_callout = NULL;      previous_callout = NULL;
2977      }      }
2978    
# Line 1653  for (;; ptr++) Line 2981  for (;; ptr++)
2981    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
2982      {      {
2983      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
2984      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
2985        {        {
2986        /* The space before the ; is to avoid a warning on a silly compiler        while (*(++ptr) != 0)
2987        on the Macintosh. */          {
2988        while ((c = *(++ptr)) != 0 && c != NEWLINE) ;          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
2989        if (c != 0) continue;   /* Else fall through to handle end of string */          }
2990          if (*ptr != 0) continue;
2991    
2992          /* Else fall through to handle end of string */
2993          c = 0;
2994        }        }
2995      }      }
2996    
# Line 1672  for (;; ptr++) Line 3004  for (;; ptr++)
3004    
3005    switch(c)    switch(c)
3006      {      {
3007      /* The branch terminates at end of string, |, or ). */      /* ===================================================================*/
3008        case 0:                        /* The branch terminates at string end */
3009      case 0:      case CHAR_VERTICAL_LINE:       /* or | or ) */
3010      case '|':      case CHAR_RIGHT_PARENTHESIS:
     case ')':  
3011      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
3012      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
3013      *codeptr = code;      *codeptr = code;
3014      *ptrptr = ptr;      *ptrptr = ptr;
3015        if (lengthptr != NULL)
3016          {
3017          if (OFLOW_MAX - *lengthptr < code - last_code)
3018            {
3019            *errorcodeptr = ERR20;
3020            goto FAILED;
3021            }
3022          *lengthptr += code - last_code;   /* To include callout length */
3023          DPRINTF((">> end branch\n"));
3024          }
3025      return TRUE;      return TRUE;
3026    
3027    
3028        /* ===================================================================*/
3029      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
3030      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3031    
3032      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
3033      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3034        {        {
3035        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
# Line 1695  for (;; ptr++) Line 3038  for (;; ptr++)
3038      *code++ = OP_CIRC;      *code++ = OP_CIRC;
3039      break;      break;
3040    
3041      case '$':      case CHAR_DOLLAR_SIGN:
3042      previous = NULL;      previous = NULL;
3043      *code++ = OP_DOLL;      *code++ = OP_DOLL;
3044      break;      break;
# Line 1703  for (;; ptr++) Line 3046  for (;; ptr++)
3046      /* There can never be a first char if '.' is first, whatever happens about      /* There can never be a first char if '.' is first, whatever happens about
3047      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
3048    
3049      case '.':      case CHAR_DOT:
3050      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3051      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
3052      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3053      previous = code;      previous = code;
3054      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
3055      break;      break;
3056    
3057      /* Character classes. If the included characters are all < 255 in value, we  
3058      build a 32-byte bitmap of the permitted characters, except in the special      /* ===================================================================*/
3059      case where there is only one such character. For negated classes, we build      /* Character classes. If the included characters are all < 256, we build a
3060      the map as usual, then invert it at the end. However, we use a different      32-byte bitmap of the permitted characters, except in the special case
3061      opcode so that data characters > 255 can be handled correctly.      where there is only one such character. For negated classes, we build the
3062        map as usual, then invert it at the end. However, we use a different opcode
3063        so that data characters > 255 can be handled correctly.
3064    
3065      If the class contains characters outside the 0-255 range, a different      If the class contains characters outside the 0-255 range, a different
3066      opcode is compiled. It may optionally have a bit map for characters < 256,      opcode is compiled. It may optionally have a bit map for characters < 256,
3067      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
3068      whether the bitmap is present, and whether this is a negated class or not.      whether the bitmap is present, and whether this is a negated class or not.
     */  
3069    
3070      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
3071        default (Perl) mode, it is treated as a data character. */
3072    
3073        case CHAR_RIGHT_SQUARE_BRACKET:
3074        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3075          {
3076          *errorcodeptr = ERR64;
3077          goto FAILED;
3078          }
3079        goto NORMAL_CHAR;
3080    
3081        case CHAR_LEFT_SQUARE_BRACKET:
3082      previous = code;      previous = code;
3083    
3084      /* PCRE supports POSIX class stuff inside a class. Perl gives an error if      /* PCRE supports POSIX class stuff inside a class. Perl gives an error if
3085      they are encountered at the top level, so we'll do that too. */      they are encountered at the top level, so we'll do that too. */
3086    
3087      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3088          check_posix_syntax(ptr, &tempptr, cd))           ptr[1] == CHAR_EQUALS_SIGN) &&
3089            check_posix_syntax(ptr, &tempptr))
3090        {        {
3091        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
3092        goto FAILED;        goto FAILED;
3093        }        }
3094    
3095      /* If the first character is '^', set the negation flag and skip it. */      /* If the first character is '^', set the negation flag and skip it. Also,
3096        if the first few characters (either before or after ^) are \Q\E or \E we
3097        skip them too. This makes for compatibility with Perl. */
3098    
3099      if ((c = *(++ptr)) == '^')      negate_class = FALSE;
3100        for (;;)
3101        {        {
       negate_class = TRUE;  
3102        c = *(++ptr);        c = *(++ptr);
3103          if (c == CHAR_BACKSLASH)
3104            {
3105            if (ptr[1] == CHAR_E)
3106              ptr++;
3107            else if (strncmp((const char *)ptr+1,
3108                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
3109              ptr += 3;
3110            else
3111              break;
3112            }
3113          else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
3114            negate_class = TRUE;
3115          else break;
3116        }        }
3117      else  
3118        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
3119        an initial ']' is taken as a data character -- the code below handles
3120        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3121        [^] must match any character, so generate OP_ALLANY. */
3122    
3123        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3124            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3125        {        {
3126        negate_class = FALSE;        *code++ = negate_class? OP_ALLANY : OP_FAIL;
3127          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3128          zerofirstbyte = firstbyte;
3129          break;
3130        }        }
3131    
3132        /* If a class contains a negative special such as \S, we need to flip the
3133        negation flag at the end, so that support for characters > 255 works
3134        correctly (they are all included in the class). */
3135    
3136        should_flip_negation = FALSE;
3137    
3138      /* Keep a count of chars with values < 256 so that we can optimize the case      /* Keep a count of chars with values < 256 so that we can optimize the case
3139      of just a single character (as long as it's < 256). For higher valued UTF-8      of just a single character (as long as it's < 256). However, For higher
3140      characters, we don't yet do any optimization. */      valued UTF-8 characters, we don't yet do any optimization. */
3141    
3142      class_charcount = 0;      class_charcount = 0;
3143      class_lastchar = -1;      class_lastchar = -1;
3144    
3145        /* Initialize the 32-char bit map to all zeros. We build the map in a
3146        temporary bit of memory, in case the class contains only 1 character (less
3147        than 256), because in that case the compiled code doesn't use the bit map.
3148        */
3149    
3150        memset(classbits, 0, 32 * sizeof(uschar));
3151    
3152  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3153      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
3154      class_utf8data = code + LINK_SIZE + 34;   /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
3155        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
3156  #endif  #endif
3157    
     /* Initialize the 32-char bit map to all zeros. We have to build the  
     map in a temporary bit of store, in case the class contains only 1  
     character (< 256), because in that case the compiled code doesn't use the  
     bit map. */  
   
     memset(classbits, 0, 32 * sizeof(uschar));  
   
3158      /* Process characters until ] is reached. By writing this as a "do" it      /* Process characters until ] is reached. By writing this as a "do" it
3159      means that an initial ] is taken as a data character. The first pass      means that an initial ] is taken as a data character. At the start of the
3160      through the regex checked the overall syntax, so we don't need to be very      loop, c contains the first byte of the character. */
     strict here. At the start of the loop, c contains the first byte of the  
     character. */  
3161    
3162      do      if (c != 0) do
3163        {        {
3164          const uschar *oldptr;
3165    
3166  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3167        if (utf8 && c > 127)        if (utf8 && c > 127)
3168          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3169          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3170          }          }
3171    
3172          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
3173          data and reset the pointer. This is so that very large classes that
3174          contain a zillion UTF-8 characters no longer overwrite the work space
3175          (which is on the stack). */
3176    
3177          if (lengthptr != NULL)
3178            {
3179            *lengthptr += class_utf8data - class_utf8data_base;
3180            class_utf8data = class_utf8data_base;
3181            }
3182    
3183  #endif  #endif
3184    
3185        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
3186    
3187        if (inescq)        if (inescq)
3188          {          {
3189          if (c == '\\' && ptr[1] == 'E')          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3190            {            {
3191            inescq = FALSE;            inescq = FALSE;                   /* Reset literal state */
3192            ptr++;            ptr++;                            /* Skip the 'E' */
3193            continue;            continue;                         /* Carry on with next */
3194            }            }
3195          else goto LONE_SINGLE_CHARACTER;          goto CHECK_RANGE;                   /* Could be range if \E follows */
3196          }          }
3197    
3198        /* Handle POSIX class names. Perl allows a negation extension of the        /* Handle POSIX class names. Perl allows a negation extension of the
# Line 1801  for (;; ptr++) Line 3201  for (;; ptr++)
3201        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3202        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3203    
3204        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3205            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3206            check_posix_syntax(ptr, &tempptr, cd))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3207          {          {
3208          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3209          int posix_class, i;          int posix_class, taboffset, tabopt;
3210          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3211            uschar pbits[32];
3212    
3213          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3214            {            {
3215            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3216            goto FAILED;            goto FAILED;
3217            }            }
3218    
3219          ptr += 2;          ptr += 2;
3220          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3221            {            {
3222            local_negate = TRUE;            local_negate = TRUE;
3223              should_flip_negation = TRUE;  /* Note negative special */
3224            ptr++;            ptr++;
3225            }            }
3226    
# Line 1835  for (;; ptr++) Line 3237  for (;; ptr++)
3237    
3238          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3239            posix_class = 0;            posix_class = 0;
3240    
3241          /* Or into the map we are building up to 3 of the static class          /* When PCRE_UCP is set, some of the POSIX classes are converted to
3242          tables, or their negations. The [:blank:] class sets up the same          different escape sequences that use Unicode properties. */
3243          chars as the [:space:] class (all white space). We remove the vertical  
3244          white space chars afterwards. */  #ifdef SUPPORT_UCP
3245            if ((options & PCRE_UCP) != 0)
3246              {
3247              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3248              if (posix_substitutes[pc] != NULL)
3249                {
3250                nestptr = tempptr + 1;
3251                ptr = posix_substitutes[pc] - 1;
3252                continue;
3253                }
3254              }
3255    #endif
3256            /* In the non-UCP case, we build the bit map for the POSIX class in a
3257            chunk of local store because we may be adding and subtracting from it,
3258            and we don't want to subtract bits that may be in the main map already.
3259            At the end we or the result into the bit map that is being built. */
3260    
3261          posix_class *= 3;          posix_class *= 3;
3262          for (i = 0; i < 3; i++)  
3263            /* Copy in the first table (always present) */
3264    
3265            memcpy(pbits, cbits + posix_class_maps[posix_class],
3266              32 * sizeof(uschar));
3267    
3268            /* If there is a second table, add or remove it as required. */
3269    
3270            taboffset = posix_class_maps[posix_class + 1];
3271            tabopt = posix_class_maps[posix_class + 2];
3272    
3273            if (taboffset >= 0)
3274            {            {
3275            BOOL blankclass = strncmp((char *)ptr, "blank", 5) == 0;            if (tabopt >= 0)
3276            int taboffset = posix_class_maps[posix_class + i];              for (c = 0; c < 32; c++) pbits[c] |= cbits[c + taboffset];
           if (taboffset < 0) break;  
           if (local_negate)  
             {  
             if (i == 0)  
               for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+taboffset];  
             else  
               for (c = 0; c < 32; c++) classbits[c] &= ~cbits[c+taboffset];  
             if (blankclass) classbits[1] |= 0x3c;  
             }  
3277            else            else
3278              {              for (c = 0; c < 32; c++) pbits[c] &= ~cbits[c + taboffset];
             for (c = 0; c < 32; c++) classbits[c] |= cbits[c+taboffset];  
             if (blankclass) classbits[1] &= ~0x3c;  
             }  
3279            }            }
3280    
3281            /* Not see if we need to remove any special characters. An option
3282            value of 1 removes vertical space and 2 removes underscore. */
3283    
3284            if (tabopt < 0) tabopt = -tabopt;
3285            if (tabopt == 1) pbits[1] &= ~0x3c;
3286              else if (tabopt == 2) pbits[11] &= 0x7f;
3287    
3288            /* Add the POSIX table or its complement into the main table that is
3289            being built and we are done. */
3290    
3291            if (local_negate)
3292              for (c = 0; c < 32; c++) classbits[c] |= ~pbits[c];
3293            else
3294              for (c = 0; c < 32; c++) classbits[c] |= pbits[c];
3295    
3296          ptr = tempptr + 1;          ptr = tempptr + 1;
3297          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */
3298          continue;    /* End of POSIX syntax handling */          continue;    /* End of POSIX syntax handling */
3299          }          }
3300    
3301        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3302        of the specials, which just set a flag. Escaped items are checked for        of the specials, which just set a flag. The sequence \b is a special
3303        validity in the pre-compiling pass. The sequence \b is a special case.        case. Inside a class (and only there) it is treated as backspace. We
3304        Inside a class (and only there) it is treated as backspace. Elsewhere        assume that other escapes have more than one character in them, so set
3305        it marks a word boundary. Other escapes have preset maps ready to        class_charcount bigger than one. Unrecognized escapes fall through and
3306        or into the one we are building. We assume they have more than one        are either treated as literal characters (by default), or are faulted if
3307        character in them, so set class_charcount bigger than one. */        PCRE_EXTRA is set. */
3308    
3309        if (c == '\\')        if (c == CHAR_BACKSLASH)
3310          {          {
3311          c = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3312            if (*errorcodeptr != 0) goto FAILED;
3313    
3314          if (-c == ESC_b) c = '\b';       /* \b is backslash in a class */          if (-c == ESC_b) c = CHAR_BS;    /* \b is backspace in a class */
         else if (-c == ESC_X) c = 'X';   /* \X is literal X in a class */  
3315          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3316            {            {
3317            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3318              {              {
3319              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3320              }              }
3321            else inescq = TRUE;            else inescq = TRUE;
3322            continue;            continue;
3323            }            }
3324            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
3325    
3326          if (c < 0)          if (c < 0)
3327            {            {
3328            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
3329            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3330    
3331            switch (-c)            switch (-c)
3332              {              {
3333    #ifdef SUPPORT_UCP
3334                case ESC_du:     /* These are the values given for \d etc */
3335                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3336                case ESC_wu:     /* escape sequence with an appropriate \p */
3337                case ESC_WU:     /* or \P to test Unicode properties instead */
3338                case ESC_su:     /* of the default ASCII testing. */
3339                case ESC_SU:
3340                nestptr = ptr;
3341                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3342                class_charcount -= 2;                /* Undo! */
3343                continue;
3344    #endif
3345              case ESC_d:              case ESC_d:
3346              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3347              continue;              continue;
3348    
3349              case ESC_D:              case ESC_D:
3350                should_flip_negation = TRUE;
3351              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
3352              continue;              continue;
3353    
# Line 1910  for (;; ptr++) Line 3356  for (;; ptr++)
3356              continue;              continue;
3357    
3358              case ESC_W:              case ESC_W:
3359                should_flip_negation = TRUE;
3360              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3361              continue;              continue;
3362    
# Line 1919  for (;; ptr++) Line 3366  for (;; ptr++)
3366              continue;              continue;
3367    
3368              case ESC_S:              case ESC_S:
3369                should_flip_negation = TRUE;
3370              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
3371              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3372              continue;              continue;
3373    
3374                case ESC_h:
3375                SETBIT(classbits, 0x09); /* VT */
3376                SETBIT(classbits, 0x20); /* SPACE */
3377                SETBIT(classbits, 0xa0); /* NSBP */
3378    #ifdef SUPPORT_UTF8
3379                if (utf8)
3380                  {
3381                  class_utf8 = TRUE;
3382                  *class_utf8data++ = XCL_SINGLE;
3383                  class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);
3384                  *class_utf8data++ = XCL_SINGLE;
3385                  class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);
3386                  *class_utf8data++ = XCL_RANGE;
3387                  class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);
3388                  class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);
3389                  *class_utf8data++ = XCL_SINGLE;
3390                  class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);
3391                  *class_utf8data++ = XCL_SINGLE;
3392                  class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);
3393                  *class_utf8data++ = XCL_SINGLE;
3394                  class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);
3395                  }
3396    #endif
3397                continue;
3398    
3399                case ESC_H:
3400                for (c = 0; c < 32; c++)
3401                  {
3402                  int x = 0xff;
3403                  switch (c)
3404                    {
3405                    case 0x09/8: x ^= 1 << (0x09%8); break;
3406                    case 0x20/8: x ^= 1 << (0x20%8); break;
3407                    case 0xa0/8: x ^= 1 << (0xa0%8); break;
3408                    default: break;
3409                    }
3410                  classbits[c] |= x;
3411                  }
3412    
3413    #ifdef SUPPORT_UTF8
3414                if (utf8)
3415                  {
3416                  class_utf8 = TRUE;
3417                  *class_utf8data++ = XCL_RANGE;
3418                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3419                  class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);
3420                  *class_utf8data++ = XCL_RANGE;
3421                  class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);
3422                  class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);
3423                  *class_utf8data++ = XCL_RANGE;
3424                  class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);
3425                  class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);
3426                  *class_utf8data++ = XCL_RANGE;
3427                  class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);
3428                  class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);
3429                  *class_utf8data++ = XCL_RANGE;
3430                  class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);
3431                  class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);
3432                  *class_utf8data++ = XCL_RANGE;
3433                  class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);
3434                  class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);
3435                  *class_utf8data++ = XCL_RANGE;
3436                  class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);
3437                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3438                  }
3439    #endif
3440                continue;
3441    
3442                case ESC_v:
3443                SETBIT(classbits, 0x0a); /* LF */
3444                SETBIT(classbits, 0x0b); /* VT */
3445                SETBIT(classbits, 0x0c); /* FF */
3446                SETBIT(classbits, 0x0d); /* CR */
3447                SETBIT(classbits, 0x85); /* NEL */
3448    #ifdef SUPPORT_UTF8
3449                if (utf8)
3450                  {
3451                  class_utf8 = TRUE;
3452                  *class_utf8data++ = XCL_RANGE;
3453                  class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);
3454                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3455                  }
3456    #endif
3457                continue;
3458    
3459                case ESC_V:
3460                for (c = 0; c < 32; c++)
3461                  {
3462                  int x = 0xff;
3463                  switch (c)
3464                    {
3465                    case 0x0a/8: x ^= 1 << (0x0a%8);
3466                                 x ^= 1 << (0x0b%8);
3467                                 x ^= 1 << (0x0c%8);
3468                                 x ^= 1 << (0x0d%8);
3469                                 break;
3470                    case 0x85/8: x ^= 1 << (0x85%8); break;
3471                    default: break;
3472                    }
3473                  classbits[c] |= x;
3474                  }
3475    
3476    #ifdef SUPPORT_UTF8
3477                if (utf8)
3478                  {
3479                  class_utf8 = TRUE;
3480                  *class_utf8data++ = XCL_RANGE;
3481                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3482                  class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);
3483                  *class_utf8data++ = XCL_RANGE;
3484                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3485                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3486                  }
3487    #endif
3488                continue;
3489    
3490  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3491              case ESC_p:              case ESC_p:
3492              case ESC_P:              case ESC_P:
3493                {                {
3494                BOOL negated;                BOOL negated;
3495                int property = get_ucp(&ptr, &negated, errorcodeptr);                int pdata;
3496                if (property < 0) goto FAILED;                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3497                  if (ptype < 0) goto FAILED;
3498                class_utf8 = TRUE;                class_utf8 = TRUE;
3499                *class_utf8data++ = ((-c == ESC_p) != negated)?                *class_utf8data++ = ((-c == ESC_p) != negated)?
3500                  XCL_PROP : XCL_NOTPROP;                  XCL_PROP : XCL_NOTPROP;
3501                *class_utf8data++ = property;                *class_utf8data++ = ptype;
3502                  *class_utf8data++ = pdata;
3503                class_charcount -= 2;   /* Not a < 256 character */                class_charcount -= 2;   /* Not a < 256 character */
3504                  continue;
3505                }                }
             continue;  
3506  #endif  #endif
   
3507              /* Unrecognized escapes are faulted if PCRE is running in its              /* Unrecognized escapes are faulted if PCRE is running in its
3508              strict mode. By default, for compatibility with Perl, they are              strict mode. By default, for compatibility with Perl, they are
3509              treated as literals. */              treated as literals. */
# Line 1949  for (;; ptr++) Line 3514  for (;; ptr++)
3514                *errorcodeptr = ERR7;                *errorcodeptr = ERR7;
3515                goto FAILED;                goto FAILED;
3516                }                }
             c = *ptr;              /* The final character */  
3517              class_charcount -= 2;  /* Undo the default count from above */              class_charcount -= 2;  /* Undo the default count from above */
3518                c = *ptr;              /* Get the final character and fall through */
3519                break;
3520              }              }
3521            }            }
3522    
3523          /* Fall through if we have a single character (c >= 0). This may be          /* Fall through if we have a single character (c >= 0). This may be
3524          > 256 in UTF-8 mode. */          greater than 256 in UTF-8 mode. */
3525    
3526          }   /* End of backslash handling */          }   /* End of backslash handling */
3527    
3528        /* A single character may be followed by '-' to form a range. However,        /* A single character may be followed by '-' to form a range. However,
3529        Perl does not permit ']' to be the end of the range. A '-' character        Perl does not permit ']' to be the end of the range. A '-' character
3530        here is treated as a literal. */        at the end is treated as a literal. Perl ignores orphaned \E sequences
3531          entirely. The code for handling \Q and \E is messy. */
3532    
3533          CHECK_RANGE:
3534          while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3535            {
3536            inescq = FALSE;
3537            ptr += 2;
3538            }
3539    
3540          oldptr = ptr;
3541    
3542          /* Remember \r or \n */
3543    
3544          if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3545    
3546          /* Check for range */
3547    
3548        if (ptr[1] == '-' && ptr[2] != ']')        if (!inescq && ptr[1] == CHAR_MINUS)
3549          {          {
3550          int d;          int d;
3551          ptr += 2;          ptr += 2;
3552            while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3553    
3554            /* If we hit \Q (not followed by \E) at this point, go into escaped
3555            mode. */
3556    
3557            while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3558              {
3559              ptr += 2;
3560              if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3561                { ptr += 2; continue; }
3562              inescq = TRUE;
3563              break;
3564              }
3565    
3566            if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3567              {
3568              ptr = oldptr;
3569              goto LONE_SINGLE_CHARACTER;
3570              }
3571    
3572  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3573          if (utf8)          if (utf8)
# Line 1981  for (;; ptr++) Line 3582  for (;; ptr++)
3582          not any of the other escapes. Perl 5.6 treats a hyphen as a literal          not any of the other escapes. Perl 5.6 treats a hyphen as a literal
3583          in such circumstances. */          in such circumstances. */
3584    
3585          if (d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3586            {            {
3587            const uschar *oldptr = ptr;            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3588            d = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);            if (*errorcodeptr != 0) goto FAILED;
3589    
3590            /* \b is backslash; \X is literal X; any other special means the '-'            /* \b is backspace; any other special means the '-' was literal */
           was literal */  
3591    
3592            if (d < 0)            if (d < 0)
3593              {              {
3594              if (d == -ESC_b) d = '\b';              if (d == -ESC_b) d = CHAR_BS; else
             else if (d == -ESC_X) d = 'X'; else  
3595                {                {
3596                ptr = oldptr - 2;                ptr = oldptr;
3597                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3598                }                }
3599              }              }
3600            }            }
3601    
3602          /* The check that the two values are in the correct order happens in          /* Check that the two values are in the correct order. Optimize
3603          the pre-pass. Optimize one-character ranges */          one-character ranges */
3604    
3605            if (d < c)
3606              {
3607              *errorcodeptr = ERR8;
3608              goto FAILED;
3609              }
3610    
3611          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3612    
3613            /* Remember \r or \n */
3614    
3615            if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3616    
3617          /* In UTF-8 mode, if the upper limit is > 255, or > 127 for caseless          /* In UTF-8 mode, if the upper limit is > 255, or > 127 for caseless
3618          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
3619          matching for characters > 127 is available only if UCP support is          matching for characters > 127 is available only if UCP support is
# Line 2022  for (;; ptr++) Line 3631  for (;; ptr++)
3631  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3632            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
3633              {              {
3634              int occ, ocd;              unsigned int occ, ocd;
3635              int cc = c;              unsigned int cc = c;
3636              int origd = d;              unsigned int origd = d;
3637              while (get_othercase_range(&cc, origd, &occ, &ocd))              while (get_othercase_range(&cc, origd, &occ, &ocd))
3638                {                {
3639                if (occ >= c && ocd <= d) continue;  /* Skip embedded ranges */                if (occ >= (unsigned int)c &&
3640                      ocd <= (unsigned int)d)
3641                    continue;                          /* Skip embedded ranges */
3642    
3643                if (occ < c  && ocd >= c - 1)        /* Extend the basic range */                if (occ < (unsigned int)c  &&
3644                      ocd >= (unsigned int)c - 1)      /* Extend the basic range */
3645                  {                                  /* if there is overlap,   */                  {                                  /* if there is overlap,   */
3646                  c = occ;                           /* noting that if occ < c */                  c = occ;                           /* noting that if occ < c */
3647                  continue;                          /* we can't have ocd > d  */                  continue;                          /* we can't have ocd > d  */
3648                  }                                  /* because a subrange is  */                  }                                  /* because a subrange is  */
3649                if (ocd > d && occ <= d + 1)         /* always shorter than    */                if (ocd > (unsigned int)d &&
3650                      occ <= (unsigned int)d + 1)      /* always shorter than    */
3651                  {                                  /* the basic range.       */                  {                                  /* the basic range.       */
3652                  d = ocd;                  d = ocd;
3653                  continue;                  continue;
# Line 2082  for (;; ptr++) Line 3695  for (;; ptr++)
3695          ranges that lie entirely within 0-127 when there is UCP support; else          ranges that lie entirely within 0-127 when there is UCP support; else
3696          for partial ranges without UCP support. */          for partial ranges without UCP support. */
3697    
3698          for (; c <= d; c++)          class_charcount += d - c + 1;
3699            class_lastchar = d;
3700    
3701            /* We can save a bit of time by skipping this in the pre-compile. */
3702    
3703            if (lengthptr == NULL) for (; c <= d; c++)
3704            {            {
3705            classbits[c/8] |= (1 << (c&7));            classbits[c/8] |= (1 << (c&7));
3706            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
# Line 2090  for (;; ptr++) Line 3708  for (;; ptr++)
3708              int uc = cd->fcc[c];           /* flip case */              int uc = cd->fcc[c];           /* flip case */
3709              classbits[uc/8] |= (1 << (uc&7));              classbits[uc/8] |= (1 << (uc&7));
3710              }              }
           class_charcount++;                /* in case a one-char range */  
           class_lastchar = c;  
3711            }            }
3712    
3713          continue;   /* Go get the next char in the class */          continue;   /* Go get the next char in the class */
# Line 2115  for (;; ptr++) Line 3731  for (;; ptr++)
3731  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3732          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3733            {            {
3734            int chartype;            unsigned int othercase;
3735            int othercase;            if ((othercase = UCD_OTHERCASE(c)) != c)
           if (_pcre_ucp_findchar(c, &chartype, &othercase) >= 0 &&  
                othercase > 0)  
3736              {              {
3737              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3738              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 2143  for (;; ptr++) Line 3757  for (;; ptr++)
3757          }          }
3758        }        }
3759    
3760      /* Loop until ']' reached; the check for end of string happens inside the      /* Loop until ']' reached. This "while" is the end of the "do" far above.
3761      loop. This "while" is the end of the "do" above. */      If we are at the end of an internal nested string, revert to the outer
3762        string. */
3763    
3764        while (((c = *(++ptr)) != 0 ||
3765               (nestptr != NULL &&
3766                 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != 0)) &&
3767               (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
3768    
3769        /* Check for missing terminating ']' */
3770    
3771      while ((c = *(++ptr)) != ']' || inescq);      if (c == 0)
3772          {
3773          *errorcodeptr = ERR6;
3774          goto FAILED;
3775          }
3776    
3777      /* If class_charcount is 1, we saw precisely one character whose value is      /* If class_charcount is 1, we saw precisely one character whose value is
3778      less than 256. In non-UTF-8 mode we can always optimize. In UTF-8 mode, we      less than 256. As long as there were no characters >= 128 and there was no
3779      can optimize the negative case only if there were no characters >= 128      use of \p or \P, in other words, no use of any XCLASS features, we can
3780      because OP_NOT and the related opcodes like OP_NOTSTAR operate on      optimize.
3781      single-bytes only. This is an historical hangover. Maybe one day we can  
3782      tidy these opcodes to handle multi-byte characters.      In UTF-8 mode, we can optimize the negative case only if there were no
3783        characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
3784        operate on single-bytes only. This is an historical hangover. Maybe one day
3785        we can tidy these opcodes to handle multi-byte characters.
3786    
3787      The optimization throws away the bit map. We turn the item into a      The optimization throws away the bit map. We turn the item into a
3788      1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note      1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note
# Line 2161  for (;; ptr++) Line 3790  for (;; ptr++)
3790      can cause firstbyte to be set. Otherwise, there can be no first char if      can cause firstbyte to be set. Otherwise, there can be no first char if
3791      this item is first, whatever repeat count may follow. In the case of      this item is first, whatever repeat count may follow. In the case of
3792      reqbyte, save the previous value for reinstating. */      reqbyte, save the previous value for reinstating. */
3793    
3794  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3795      if (class_charcount == 1 &&      if (class_charcount == 1 && !class_utf8 &&
3796            (!utf8 ||        (!utf8 || !negate_class || class_lastchar < 128))
           (!class_utf8 && (!negate_class || class_lastchar < 128))))  
   
3797  #else  #else
3798      if (class_charcount == 1)      if (class_charcount == 1)
3799  #endif  #endif
# Line 2209  for (;; ptr++) Line 3836  for (;; ptr++)
3836      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3837    
3838      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
3839      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode, unless there was a negated special
3840      we can omit the bitmap. */      such as \S in the class, and PCRE_UCP is not set, because in that case all
3841        characters > 255 are in the class, so any that were explicitly given as
3842        well can be ignored. If (when there are explicit characters > 255 that must
3843        be listed) there are no characters < 256, we can omit the bitmap in the
3844        actual compiled code. */
3845    
3846  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3847      if (class_utf8)      if (class_utf8 && (!should_flip_negation || (options & PCRE_UCP) != 0))
3848        {        {
3849        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
3850        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
3851        code += LINK_SIZE;        code += LINK_SIZE;
3852        *code = negate_class? XCL_NOT : 0;        *code = negate_class? XCL_NOT : 0;
3853    
3854        /* If the map is required, install it, and move on to the end of        /* If the map is required, move up the extra data to make room for it;
3855        the extra data */        otherwise just move the code pointer to the end of the extra data. */
3856    
3857        if (class_charcount > 0)        if (class_charcount > 0)
3858          {          {
3859          *code++ |= XCL_MAP;          *code++ |= XCL_MAP;
3860            memmove(code + 32, code, class_utf8data - code);
3861          memcpy(code, classbits, 32);          memcpy(code, classbits, 32);
3862          code = class_utf8data;          code = class_utf8data + 32;
         }  
   
       /* If the map is not required, slide down the extra data. */  
   
       else  
         {  
         int len = class_utf8data - (code + 33);  
         memmove(code + 1, code + 33, len);  
         code += len + 1;  
3863          }          }
3864          else code = class_utf8data;
3865    
3866        /* Now fill in the complete length of the item */        /* Now fill in the complete length of the item */
3867    
# Line 2246  for (;; ptr++) Line 3870  for (;; ptr++)
3870        }        }
3871  #endif  #endif
3872    
3873      /* If there are no characters > 255, negate the 32-byte map if necessary,      /* If there are no characters > 255, or they are all to be included or
3874      and copy it into the code vector. If this is the first thing in the branch,      excluded, set the opcode to OP_CLASS or OP_NCLASS, depending on whether the
3875      there can be no first char setting, whatever the repeat count. Any reqbyte      whole class was negated and whether there were negative specials such as \S
3876      setting must remain unchanged after any kind of repeat. */      (non-UCP) in the class. Then copy the 32-byte map into the code vector,
3877        negating it if necessary. */
3878    
3879        *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
3880      if (negate_class)      if (negate_class)
3881        {        {
3882        *code++ = OP_NCLASS;        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
3883        for (c = 0; c < 32; c++) code[c] = ~classbits[c];          for (c = 0; c < 32; c++) code[c] = ~classbits[c];
3884        }        }
3885      else      else
3886        {        {
       *code++ = OP_CLASS;  
3887        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
3888        }        }
3889      code += 32;      code += 32;
3890      break;      break;
3891    
3892    
3893        /* ===================================================================*/
3894      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
3895      has been tested above. */      has been tested above. */
3896    
3897      case '{':      case CHAR_LEFT_CURLY_BRACKET:
3898      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
3899      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
3900      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
3901      goto REPEAT;      goto REPEAT;
3902    
3903      case '*':      case CHAR_ASTERISK:
3904      repeat_min = 0;      repeat_min = 0;
3905      repeat_max = -1;      repeat_max = -1;
3906      goto REPEAT;      goto REPEAT;
3907    
3908      case '+':      case CHAR_PLUS:
3909      repeat_min = 1;      repeat_min = 1;
3910      repeat_max = -1;      repeat_max = -1;
3911      goto REPEAT;      goto REPEAT;
3912    
3913      case '?':      case CHAR_QUESTION_MARK:
3914      repeat_min = 0;      repeat_min = 0;
3915      repeat_max = 1;      repeat_max = 1;
3916    
# Line 2318  for (;; ptr++) Line 3945  for (;; ptr++)
3945      but if PCRE_UNGREEDY is set, it works the other way round. We change the      but if PCRE_UNGREEDY is set, it works the other way round. We change the
3946      repeat type to the non-default. */      repeat type to the non-default. */
3947    
3948      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
3949        {        {
3950        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
3951        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
3952        ptr++;        ptr++;
3953        }        }
3954      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
3955        {        {
3956        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
3957        ptr++;        ptr++;
3958        }        }
3959      else repeat_type = greedy_default;      else repeat_type = greedy_default;
3960    
     /* If previous was a recursion, we need to wrap it inside brackets so that  
     it can be replicated if necessary. */  
   
     if (*previous == OP_RECURSE)  
       {  
       memmove(previous + 1 + LINK_SIZE, previous, 1 + LINK_SIZE);  
       code += 1 + LINK_SIZE;  
       *previous = OP_BRA;  
       PUT(previous, 1, code - previous);  
       *code = OP_KET;  
       PUT(code, 1, code - previous);  
       code += 1 + LINK_SIZE;  
       }  
   
3961      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
3962      repeat item instead. If a char item has a minumum of more than one, ensure      repeat item instead. If a char item has a minumum of more than one, ensure
3963      that it is set in reqbyte - it might not be if a sequence such as x{3} is      that it is set in reqbyte - it might not be if a sequence such as x{3} is
# Line 2378  for (;; ptr++) Line 3991  for (;; ptr++)
3991          if (repeat_min > 1) reqbyte = c | req_caseopt | cd->req_varyopt;          if (repeat_min > 1) reqbyte = c | req_caseopt | cd->req_varyopt;
3992          }          }
3993    
3994          /* If the repetition is unlimited, it pays to see if the next thing on
3995          the line is something that cannot possibly match this character. If so,
3996          automatically possessifying this item gains some performance in the case
3997          where the match fails. */
3998    
3999          if (!possessive_quantifier &&
4000              repeat_max < 0 &&
4001              check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,
4002                options, cd))
4003            {
4004            repeat_type = 0;    /* Force greedy */
4005            possessive_quantifier = TRUE;
4006            }
4007    
4008        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */        goto OUTPUT_SINGLE_REPEAT;   /* Code shared with single character types */
4009        }        }
4010    
4011      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
4012      one of the special opcodes, replacing it. The code is shared with single-      one of the special opcodes, replacing it. The code is shared with single-
4013      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
4014      repeat_type. OP_NOT is currently used only for single-byte chars. */      repeat_type. We can also test for auto-possessification. OP_NOT is
4015        currently used only for single-byte chars. */
4016    
4017      else if (*previous == OP_NOT)      else if (*previous == OP_NOT)
4018        {        {
4019        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */
4020        c = previous[1];        c = previous[1];
4021          if (!possessive_quantifier &&
4022              repeat_max < 0 &&
4023              check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))
4024            {
4025            repeat_type = 0;    /* Force greedy */
4026            possessive_quantifier = TRUE;
4027            }
4028        goto OUTPUT_SINGLE_REPEAT;        goto OUTPUT_SINGLE_REPEAT;
4029        }        }
4030    
# Line 2403  for (;; ptr++) Line 4038  for (;; ptr++)
4038      else if (*previous < OP_EODN)      else if (*previous < OP_EODN)
4039        {        {
4040        uschar *oldcode;        uschar *oldcode;
4041        int prop_type;        int prop_type, prop_value;
4042        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */        op_type = OP_TYPESTAR - OP_STAR;  /* Use type opcodes */
4043        c = *previous;        c = *previous;
4044    
4045          if (!possessive_quantifier &&
4046              repeat_max < 0 &&
4047              check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))
4048            {
4049            repeat_type = 0;    /* Force greedy */
4050            possessive_quantifier = TRUE;
4051            }
4052    
4053        OUTPUT_SINGLE_REPEAT:        OUTPUT_SINGLE_REPEAT:
4054        prop_type = (*previous == OP_PROP || *previous == OP_NOTPROP)?        if (*previous == OP_PROP || *previous == OP_NOTPROP)
4055          previous[1] : -1;          {
4056            prop_type = previous[1];
4057            prop_value = previous[2];
4058            }
4059          else prop_type = prop_value = -1;
4060    
4061        oldcode = code;        oldcode = code;
4062        code = previous;                  /* Usually overwrite previous item */        code = previous;                  /* Usually overwrite previous item */
# Line 2419  for (;; ptr++) Line 4066  for (;; ptr++)
4066    
4067        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
4068    
4069          /*--------------------------------------------------------------------*/
4070          /* This code is obsolete from release 8.00; the restriction was finally
4071          removed: */
4072    
4073        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4074        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4075    
4076        if (repeat_max != 1) cd->nopartial = TRUE;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4077          /*--------------------------------------------------------------------*/
4078    
4079        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
4080    
# Line 2443  for (;; ptr++) Line 4095  for (;; ptr++)
4095          }          }
4096    
4097        /* A repeat minimum of 1 is optimized into some special cases. If the        /* A repeat minimum of 1 is optimized into some special cases. If the
4098        maximum is unlimited, we use OP_PLUS. Otherwise, the original item it        maximum is unlimited, we use OP_PLUS. Otherwise, the original item is
4099        left in place and, if the maximum is greater than 1, we use OP_UPTO with        left in place and, if the maximum is greater than 1, we use OP_UPTO with
4100        one less than the maximum. */        one less than the maximum. */
4101    
# Line 2470  for (;; ptr++) Line 4122  for (;; ptr++)
4122    
4123          /* If the maximum is unlimited, insert an OP_STAR. Before doing so,          /* If the maximum is unlimited, insert an OP_STAR. Before doing so,
4124          we have to insert the character for the previous code. For a repeated          we have to insert the character for the previous code. For a repeated
4125          Unicode property match, there is an extra byte that defines the          Unicode property match, there are two extra bytes that define the
4126          required property. In UTF-8 mode, long characters have their length in          required property. In UTF-8 mode, long characters have their length in
4127          c, with the 0x80 bit as a flag. */          c, with the 0x80 bit as a flag. */
4128    
# Line 2486  for (;; ptr++) Line 4138  for (;; ptr++)
4138  #endif  #endif
4139              {              {
4140              *code++ = c;              *code++ = c;
4141              if (prop_type >= 0) *code++ = prop_type;              if (prop_type >= 0)
4142                  {
4143                  *code++ = prop_type;
4144                  *code++ = prop_value;
4145                  }
4146              }              }
4147            *code++ = OP_STAR + repeat_type;            *code++ = OP_STAR + repeat_type;
4148            }            }
4149    
4150          /* Else insert an UPTO if the max is greater than the min, again          /* Else insert an UPTO if the max is greater than the min, again
4151          preceded by the character, for the previously inserted code. */          preceded by the character, for the previously inserted code. If the
4152            UPTO is just for 1 instance, we can use QUERY instead. */
4153    
4154          else if (repeat_max != repeat_min)          else if (repeat_max != repeat_min)
4155            {            {
# Line 2505  for (;; ptr++) Line 4162  for (;; ptr++)
4162            else            else
4163  #endif  #endif
4164            *code++ = c;            *code++ = c;
4165            if (prop_type >= 0) *code++ = prop_type;            if (prop_type >= 0)
4166                {
4167                *code++ = prop_type;
4168                *code++ = prop_value;
4169                }
4170            repeat_max -= repeat_min;            repeat_max -= repeat_min;
4171            *code++ = OP_UPTO + repeat_type;  
4172            PUT2INC(code, 0, repeat_max);            if (repeat_max == 1)
4173                {
4174                *code++ = OP_QUERY + repeat_type;
4175                }
4176              else
4177                {
4178                *code++ = OP_UPTO + repeat_type;
4179                PUT2INC(code, 0, repeat_max);
4180                }
4181            }            }
4182          }          }
4183    
# Line 2524  for (;; ptr++) Line 4193  for (;; ptr++)
4193  #endif  #endif
4194        *code++ = c;        *code++ = c;
4195    
4196        /* For a repeated Unicode property match, there is an extra byte that        /* For a repeated Unicode property match, there are two extra bytes that
4197        defines the required property. */        define the required property. */
4198    
4199  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4200        if (prop_type >= 0) *code++ = prop_type;        if (prop_type >= 0)
4201            {
4202            *code++ = prop_type;
4203            *code++ = prop_value;
4204            }
4205  #endif  #endif
4206        }        }
4207    
# Line 2548  for (;; ptr++) Line 4221  for (;; ptr++)
4221          goto END_REPEAT;          goto END_REPEAT;
4222          }          }
4223    
4224          /*--------------------------------------------------------------------*/
4225          /* This code is obsolete from release 8.00; the restriction was finally
4226          removed: */
4227    
4228        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4229        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4230    
4231        if (repeat_max != 1) cd->nopartial = TRUE;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4232          /*--------------------------------------------------------------------*/
4233    
4234        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4235          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 2571  for (;; ptr++) Line 4249  for (;; ptr++)
4249      /* If previous was a bracket group, we may have to replicate it in certain      /* If previous was a bracket group, we may have to replicate it in certain
4250      cases. */      cases. */
4251    
4252      else if (*previous >= OP_BRA || *previous == OP_ONCE ||      else if (*previous == OP_BRA  || *previous == OP_CBRA ||
4253               *previous == OP_COND)               *previous == OP_ONCE || *previous == OP_COND)
4254        {        {
4255        register int i;        register int i;
4256        int ketoffset = 0;        int ketoffset = 0;
4257        int len = code - previous;        int len = code - previous;
4258        uschar *bralink = NULL;        uschar *bralink = NULL;
4259    
4260          /* Repeating a DEFINE group is pointless */
4261    
4262          if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
4263            {
4264            *errorcodeptr = ERR55;
4265            goto FAILED;
4266            }
4267    
4268        /* If the maximum repeat count is unlimited, find the end of the bracket        /* If the maximum repeat count is unlimited, find the end of the bracket
4269        by scanning through from the start, and compute the offset back to it        by scanning through from the start, and compute the offset back to it
4270        from the current code pointer. There may be an OP_OPT setting following        from the current code pointer. There may be an OP_OPT setting following
# Line 2601  for (;; ptr++) Line 4287  for (;; ptr++)
4287    
4288        if (repeat_min == 0)        if (repeat_min == 0)
4289          {          {
4290          /* If the maximum is also zero, we just omit the group from the output          /* If the maximum is also zero, we used to just omit the group from the
4291          altogether. */          output altogether, like this:
   
         if (repeat_max == 0)  
           {  
           code = previous;