/[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 77 by nigel, Sat Feb 24 21:40:45 2007 UTC revision 533 by ph10, Wed Jun 2 19:02:41 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 1423  auto_callout(uschar *code, const uschar Line 2316  auto_callout(uschar *code, const uschar
2316  {  {
2317  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2318  *code++ = 255;  *code++ = 255;
2319  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2320  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2321  return code + 2*LINK_SIZE;  return code + 2*LINK_SIZE;
2322  }  }
2323    
# Line 1449  Returns: nothing Line 2342  Returns: nothing
2342  static void  static void
2343  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)
2344  {  {
2345  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2346  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2347  }  }
2348    
# 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 1503  for (++c; c <= d; c++) Line 2392  for (++c; c <= d; c++)
2392    
2393  return TRUE;  return TRUE;
2394  }  }
2395    
2396    
2397    
2398    /*************************************************
2399    *        Check a character and a property        *
2400    *************************************************/
2401    
2402    /* This function is called by check_auto_possessive() when a property item
2403    is adjacent to a fixed character.
2404    
2405    Arguments:
2406      c            the character
2407      ptype        the property type
2408      pdata        the data for the type
2409      negated      TRUE if it's a negated property (\P or \p{^)
2410    
2411    Returns:       TRUE if auto-possessifying is OK
2412    */
2413    
2414    static BOOL
2415    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2416    {
2417    const ucd_record *prop = GET_UCD(c);
2418    switch(ptype)
2419      {
2420      case PT_LAMP:
2421      return (prop->chartype == ucp_Lu ||
2422              prop->chartype == ucp_Ll ||
2423              prop->chartype == ucp_Lt) == negated;
2424    
2425      case PT_GC:
2426      return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;
2427    
2428      case PT_PC:
2429      return (pdata == prop->chartype) == negated;
2430    
2431      case PT_SC:
2432      return (pdata == prop->script) == negated;
2433    
2434      /* These are specials */
2435    
2436      case PT_ALNUM:
2437      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2438              _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;
2439    
2440      case PT_SPACE:    /* Perl space */
2441      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2442              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2443              == negated;
2444    
2445      case PT_PXSPACE:  /* POSIX space */
2446      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2447              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2448              c == CHAR_FF || c == CHAR_CR)
2449              == negated;
2450    
2451      case PT_WORD:
2452      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2453              _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2454              c == CHAR_UNDERSCORE) == negated;
2455      }
2456    return FALSE;
2457    }
2458  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2459    
2460    
2461    
2462    /*************************************************
2463    *     Check if auto-possessifying is possible    *
2464    *************************************************/
2465    
2466    /* This function is called for unlimited repeats of certain items, to see
2467    whether the next thing could possibly match the repeated item. If not, it makes
2468    sense to automatically possessify the repeated item.
2469    
2470    Arguments:
2471      previous      pointer to the repeated opcode
2472      utf8          TRUE in UTF-8 mode
2473      ptr           next character in pattern
2474      options       options bits
2475      cd            contains pointers to tables etc.
2476    
2477    Returns:        TRUE if possessifying is wanted
2478    */
2479    
2480    static BOOL
2481    check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
2482      int options, compile_data *cd)
2483    {
2484    int c, next;
2485    int op_code = *previous++;
2486    
2487    /* Skip whitespace and comments in extended mode */
2488    
2489    if ((options & PCRE_EXTENDED) != 0)
2490      {
2491      for (;;)
2492        {
2493        while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2494        if (*ptr == CHAR_NUMBER_SIGN)
2495          {
2496          while (*(++ptr) != 0)
2497            if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2498          }
2499        else break;
2500        }
2501      }
2502    
2503    /* If the next item is one that we can handle, get its value. A non-negative
2504    value is a character, a negative value is an escape value. */
2505    
2506    if (*ptr == CHAR_BACKSLASH)
2507      {
2508      int temperrorcode = 0;
2509      next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
2510      if (temperrorcode != 0) return FALSE;
2511      ptr++;    /* Point after the escape sequence */
2512      }
2513    
2514    else if ((cd->ctypes[*ptr] & ctype_meta) == 0)
2515      {
2516    #ifdef SUPPORT_UTF8
2517      if (utf8) { GETCHARINC(next, ptr); } else
2518    #endif
2519      next = *ptr++;
2520      }
2521    
2522    else return FALSE;
2523    
2524    /* Skip whitespace and comments in extended mode */
2525    
2526    if ((options & PCRE_EXTENDED) != 0)
2527      {
2528      for (;;)
2529        {
2530        while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2531        if (*ptr == CHAR_NUMBER_SIGN)
2532          {
2533          while (*(++ptr) != 0)
2534            if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2535          }
2536        else break;
2537        }
2538      }
2539    
2540    /* If the next thing is itself optional, we have to give up. */
2541    
2542    if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2543      strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2544        return FALSE;
2545    
2546    /* Now compare the next item with the previous opcode. First, handle cases when
2547    the next item is a character. */
2548    
2549    if (next >= 0) switch(op_code)
2550      {
2551      case OP_CHAR:
2552    #ifdef SUPPORT_UTF8
2553      GETCHARTEST(c, previous);
2554    #else
2555      c = *previous;
2556    #endif
2557      return c != next;
2558    
2559      /* For CHARNC (caseless character) we must check the other case. If we have
2560      Unicode property support, we can use it to test the other case of
2561      high-valued characters. */
2562    
2563      case OP_CHARNC:
2564    #ifdef SUPPORT_UTF8
2565      GETCHARTEST(c, previous);
2566    #else
2567      c = *previous;
2568    #endif
2569      if (c == next) return FALSE;
2570    #ifdef SUPPORT_UTF8
2571      if (utf8)
2572        {
2573        unsigned int othercase;
2574        if (next < 128) othercase = cd->fcc[next]; else
2575    #ifdef SUPPORT_UCP
2576        othercase = UCD_OTHERCASE((unsigned int)next);
2577    #else
2578        othercase = NOTACHAR;
2579    #endif
2580        return (unsigned int)c != othercase;
2581        }
2582      else
2583    #endif  /* SUPPORT_UTF8 */
2584      return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2585    
2586      /* For OP_NOT, its data is always a single-byte character. */
2587    
2588      case OP_NOT:
2589      if ((c = *previous) == next) return TRUE;
2590      if ((options & PCRE_CASELESS) == 0) return FALSE;
2591    #ifdef SUPPORT_UTF8
2592      if (utf8)
2593        {
2594        unsigned int othercase;
2595        if (next < 128) othercase = cd->fcc[next]; else
2596    #ifdef SUPPORT_UCP
2597        othercase = UCD_OTHERCASE(next);
2598    #else
2599        othercase = NOTACHAR;
2600    #endif
2601        return (unsigned int)c == othercase;
2602        }
2603      else
2604    #endif  /* SUPPORT_UTF8 */
2605      return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2606    
2607      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2608      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
2609    
2610      case OP_DIGIT:
2611      return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
2612    
2613      case OP_NOT_DIGIT:
2614      return next <= 127 && (cd->ctypes[next] & ctype_digit) != 0;
2615    
2616      case OP_WHITESPACE:
2617      return next > 127 || (cd->ctypes[next] & ctype_space) == 0;
2618    
2619      case OP_NOT_WHITESPACE:
2620      return next <= 127 && (cd->ctypes[next] & ctype_space) != 0;
2621    
2622      case OP_WORDCHAR:
2623      return next > 127 || (cd->ctypes[next] & ctype_word) == 0;
2624    
2625      case OP_NOT_WORDCHAR:
2626      return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;
2627    
2628      case OP_HSPACE:
2629      case OP_NOT_HSPACE:
2630      switch(next)
2631        {
2632        case 0x09:
2633        case 0x20:
2634        case 0xa0:
2635        case 0x1680:
2636        case 0x180e:
2637        case 0x2000:
2638        case 0x2001:
2639        case 0x2002:
2640        case 0x2003:
2641        case 0x2004:
2642        case 0x2005:
2643        case 0x2006:
2644        case 0x2007:
2645        case 0x2008:
2646        case 0x2009:
2647        case 0x200A:
2648        case 0x202f:
2649        case 0x205f:
2650        case 0x3000:
2651        return op_code == OP_NOT_HSPACE;
2652        default:
2653        return op_code != OP_NOT_HSPACE;
2654        }
2655    
2656      case OP_ANYNL:
2657      case OP_VSPACE:
2658      case OP_NOT_VSPACE:
2659      switch(next)
2660        {
2661        case 0x0a:
2662        case 0x0b:
2663        case 0x0c:
2664        case 0x0d:
2665        case 0x85:
2666        case 0x2028:
2667        case 0x2029:
2668        return op_code == OP_NOT_VSPACE;
2669        default:
2670        return op_code != OP_NOT_VSPACE;
2671        }
2672    
2673    #ifdef SUPPORT_UCP
2674      case OP_PROP:
2675      return check_char_prop(next, previous[0], previous[1], FALSE);
2676    
2677      case OP_NOTPROP:
2678      return check_char_prop(next, previous[0], previous[1], TRUE);
2679    #endif
2680    
2681      default:
2682      return FALSE;
2683      }
2684    
2685    
2686    /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP
2687    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
2688    generated only when PCRE_UCP is *not* set, that is, when only ASCII
2689    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
2690    replaced by OP_PROP codes when PCRE_UCP is set. */
2691    
2692    switch(op_code)
2693      {
2694      case OP_CHAR:
2695      case OP_CHARNC:
2696    #ifdef SUPPORT_UTF8
2697      GETCHARTEST(c, previous);
2698    #else
2699      c = *previous;
2700    #endif
2701      switch(-next)
2702        {
2703        case ESC_d:
2704        return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
2705    
2706        case ESC_D:
2707        return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
2708    
2709        case ESC_s:
2710        return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
2711    
2712        case ESC_S:
2713        return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
2714    
2715        case ESC_w:
2716        return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
2717    
2718        case ESC_W:
2719        return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
2720    
2721        case ESC_h:
2722        case ESC_H:
2723        switch(c)
2724          {
2725          case 0x09:
2726          case 0x20:
2727          case 0xa0:
2728          case 0x1680:
2729          case 0x180e:
2730          case 0x2000:
2731          case 0x2001:
2732          case 0x2002:
2733          case 0x2003:
2734          case 0x2004:
2735          case 0x2005:
2736          case 0x2006:
2737          case 0x2007:
2738          case 0x2008:
2739          case 0x2009:
2740          case 0x200A:
2741          case 0x202f:
2742          case 0x205f:
2743          case 0x3000:
2744          return -next != ESC_h;
2745          default:
2746          return -next == ESC_h;
2747          }
2748    
2749        case ESC_v:
2750        case ESC_V:
2751        switch(c)
2752          {
2753          case 0x0a:
2754          case 0x0b:
2755          case 0x0c:
2756          case 0x0d:
2757          case 0x85:
2758          case 0x2028:
2759          case 0x2029:
2760          return -next != ESC_v;
2761          default:
2762          return -next == ESC_v;
2763          }
2764    
2765        /* When PCRE_UCP is set, these values get generated for \d etc. Find
2766        their substitutions and process them. The result will always be either
2767        -ESC_p or -ESC_P. Then fall through to process those values. */
2768    
2769    #ifdef SUPPORT_UCP
2770        case ESC_du:
2771        case ESC_DU:
2772        case ESC_wu:
2773        case ESC_WU:
2774        case ESC_su:
2775        case ESC_SU:
2776          {
2777          int temperrorcode = 0;
2778          ptr = substitutes[-next - ESC_DU];
2779          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
2780          if (temperrorcode != 0) return FALSE;
2781          ptr++;    /* For compatibility */
2782          }
2783        /* Fall through */
2784    
2785        case ESC_p:
2786        case ESC_P:
2787          {
2788          int ptype, pdata, errorcodeptr;
2789          BOOL negated;
2790    
2791          ptr--;      /* Make ptr point at the p or P */
2792          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
2793          if (ptype < 0) return FALSE;
2794          ptr++;      /* Point past the final curly ket */
2795    
2796          /* If the property item is optional, we have to give up. (When generated
2797          from \d etc by PCRE_UCP, this test will have been applied much earlier,
2798          to the original \d etc. At this point, ptr will point to a zero byte. */
2799    
2800          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2801            strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2802              return FALSE;
2803    
2804          /* Do the property check. */
2805    
2806          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
2807          }
2808    #endif
2809    
2810        default:
2811        return FALSE;
2812        }
2813    
2814      /* In principle, support for Unicode properties should be integrated here as
2815      well. It means re-organizing the above code so as to get hold of the property
2816      values before switching on the op-code. However, I wonder how many patterns
2817      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
2818      these op-codes are never generated.) */
2819    
2820      case OP_DIGIT:
2821      return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2822             next == -ESC_h || next == -ESC_v || next == -ESC_R;
2823    
2824      case OP_NOT_DIGIT:
2825      return next == -ESC_d;
2826    
2827      case OP_WHITESPACE:
2828      return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
2829    
2830      case OP_NOT_WHITESPACE:
2831      return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2832    
2833      case OP_HSPACE:
2834      return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
2835             next == -ESC_w || next == -ESC_v || next == -ESC_R;
2836    
2837      case OP_NOT_HSPACE:
2838      return next == -ESC_h;
2839    
2840      /* Can't have \S in here because VT matches \S (Perl anomaly) */
2841      case OP_ANYNL:
2842      case OP_VSPACE:
2843      return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2844    
2845      case OP_NOT_VSPACE:
2846      return next == -ESC_v || next == -ESC_R;
2847    
2848      case OP_WORDCHAR:
2849      return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
2850             next == -ESC_v || next == -ESC_R;
2851    
2852      case OP_NOT_WORDCHAR:
2853      return next == -ESC_w || next == -ESC_d;
2854    
2855      default:
2856      return FALSE;
2857      }
2858    
2859    /* Control does not reach here */
2860    }
2861    
2862    
2863    
2864  /*************************************************  /*************************************************
2865  *           Compile one branch                   *  *           Compile one branch                   *
2866  *************************************************/  *************************************************/
2867    
2868  /* 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
2869  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
2870  bits.  bits. This function is used during the pre-compile phase when we are trying
2871    to find out the amount of memory needed, as well as during the real compile
2872    phase. The value of lengthptr distinguishes the two phases.
2873    
2874  Arguments:  Arguments:
2875    optionsptr     pointer to the option bits    optionsptr     pointer to the option bits
   brackets       points to number of extracting brackets used  
2876    codeptr        points to the pointer to the current code point    codeptr        points to the pointer to the current code point
2877    ptrptr         points to the current pattern pointer    ptrptr         points to the current pattern pointer
2878    errorcodeptr   points to error code variable    errorcodeptr   points to error code variable
# Line 1524  Arguments: Line 2880  Arguments:
2880    reqbyteptr     set to the last literal character required, else < 0    reqbyteptr     set to the last literal character required, else < 0
2881    bcptr          points to current branch chain    bcptr          points to current branch chain
2882    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
2883      lengthptr      NULL during the real compile phase
2884                     points to length accumulator during pre-compile phase
2885    
2886  Returns:         TRUE on success  Returns:         TRUE on success
2887                   FALSE, with *errorcodeptr set non-zero on error                   FALSE, with *errorcodeptr set non-zero on error
2888  */  */
2889    
2890  static BOOL  static BOOL
2891  compile_branch(int *optionsptr, int *brackets, uschar **codeptr,  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
2892    const uschar **ptrptr, int *errorcodeptr, int *firstbyteptr,    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
2893    int *reqbyteptr, branch_chain *bcptr, compile_data *cd)    compile_data *cd, int *lengthptr)
2894  {  {
2895  int repeat_type, op_type;  int repeat_type, op_type;
2896  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 2899  int greedy_default, greedy_non_default;
2899  int firstbyte, reqbyte;  int firstbyte, reqbyte;
2900  int zeroreqbyte, zerofirstbyte;  int zeroreqbyte, zerofirstbyte;
2901  int req_caseopt, reqvary, tempreqvary;  int req_caseopt, reqvary, tempreqvary;
 int condcount = 0;  
2902  int options = *optionsptr;  int options = *optionsptr;
2903  int after_manual_callout = 0;  int after_manual_callout = 0;
2904    int length_prevgroup = 0;
2905  register int c;  register int c;
2906  register uschar *code = *codeptr;  register uschar *code = *codeptr;
2907    uschar *last_code = code;
2908    uschar *orig_code = code;
2909  uschar *tempcode;  uschar *tempcode;
2910  BOOL inescq = FALSE;  BOOL inescq = FALSE;
2911  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
2912  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
2913  const uschar *tempptr;  const uschar *tempptr;
2914    const uschar *nestptr = NULL;
2915  uschar *previous = NULL;  uschar *previous = NULL;
2916  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
2917    uschar *save_hwm = NULL;
2918  uschar classbits[32];  uschar classbits[32];
2919    
2920  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2921  BOOL class_utf8;  BOOL class_utf8;
2922  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
2923  uschar *class_utf8data;  uschar *class_utf8data;
2924    uschar *class_utf8data_base;
2925  uschar utf8_char[6];  uschar utf8_char[6];
2926  #else  #else
2927  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
2928    uschar *utf8_char = NULL;
2929    #endif
2930    
2931    #ifdef PCRE_DEBUG
2932    if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2933  #endif  #endif
2934    
2935  /* 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 2961  req_caseopt = ((options & PCRE_CASELESS)
2961  for (;; ptr++)  for (;; ptr++)
2962    {    {
2963    BOOL negate_class;    BOOL negate_class;
2964      BOOL should_flip_negation;
2965    BOOL possessive_quantifier;    BOOL possessive_quantifier;
2966    BOOL is_quantifier;    BOOL is_quantifier;
2967      BOOL is_recurse;
2968      BOOL reset_bracount;
2969    int class_charcount;    int class_charcount;
2970    int class_lastchar;    int class_lastchar;
2971    int newoptions;    int newoptions;
2972    int recno;    int recno;
2973      int refsign;
2974    int skipbytes;    int skipbytes;
2975    int subreqbyte;    int subreqbyte;
2976    int subfirstbyte;    int subfirstbyte;
2977      int terminator;
2978    int mclength;    int mclength;
2979    uschar mcbuffer[8];    uschar mcbuffer[8];
2980    
2981    /* Next byte in the pattern */    /* Get next byte in the pattern */
2982    
2983    c = *ptr;    c = *ptr;
2984    
2985      /* If we are at the end of a nested substitution, revert to the outer level
2986      string. Nesting only happens one level deep. */
2987    
2988      if (c == 0 && nestptr != NULL)
2989        {
2990        ptr = nestptr;
2991        nestptr = NULL;
2992        c = *ptr;
2993        }
2994    
2995      /* If we are in the pre-compile phase, accumulate the length used for the
2996      previous cycle of this loop. */
2997    
2998      if (lengthptr != NULL)
2999        {
3000    #ifdef PCRE_DEBUG
3001        if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3002    #endif
3003        if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
3004          {
3005          *errorcodeptr = ERR52;
3006          goto FAILED;
3007          }
3008    
3009        /* There is at least one situation where code goes backwards: this is the
3010        case of a zero quantifier after a class (e.g. [ab]{0}). At compile time,
3011        the class is simply eliminated. However, it is created first, so we have to
3012        allow memory for it. Therefore, don't ever reduce the length at this point.
3013        */
3014    
3015        if (code < last_code) code = last_code;
3016    
3017        /* Paranoid check for integer overflow */
3018    
3019        if (OFLOW_MAX - *lengthptr < code - last_code)
3020          {
3021          *errorcodeptr = ERR20;
3022          goto FAILED;
3023          }
3024    
3025        *lengthptr += (int)(code - last_code);
3026        DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));
3027    
3028        /* If "previous" is set and it is not at the start of the work space, move
3029        it back to there, in order to avoid filling up the work space. Otherwise,
3030        if "previous" is NULL, reset the current code pointer to the start. */
3031    
3032        if (previous != NULL)
3033          {
3034          if (previous > orig_code)
3035            {
3036            memmove(orig_code, previous, code - previous);
3037            code -= previous - orig_code;
3038            previous = orig_code;
3039            }
3040          }
3041        else code = orig_code;
3042    
3043        /* Remember where this code item starts so we can pick up the length
3044        next time round. */
3045    
3046        last_code = code;
3047        }
3048    
3049      /* In the real compile phase, just check the workspace used by the forward
3050      reference list. */
3051    
3052      else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
3053        {
3054        *errorcodeptr = ERR52;
3055        goto FAILED;
3056        }
3057    
3058    /* 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 */
3059    
3060    if (inescq && c != 0)    if (inescq && c != 0)
3061      {      {
3062      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3063        {        {
3064        inescq = FALSE;        inescq = FALSE;
3065        ptr++;        ptr++;
# Line 1623  for (;; ptr++) Line 3069  for (;; ptr++)
3069        {        {
3070        if (previous_callout != NULL)        if (previous_callout != NULL)
3071          {          {
3072          complete_callout(previous_callout, ptr, cd);          if (lengthptr == NULL)  /* Don't attempt in pre-compile phase */
3073              complete_callout(previous_callout, ptr, cd);
3074          previous_callout = NULL;          previous_callout = NULL;
3075          }          }
3076        if ((options & PCRE_AUTO_CALLOUT) != 0)        if ((options & PCRE_AUTO_CALLOUT) != 0)
# Line 1638  for (;; ptr++) Line 3085  for (;; ptr++)
3085    /* 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
3086    a quantifier. */    a quantifier. */
3087    
3088    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
3089      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
3090        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
3091    
3092    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
3093         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
3094      {      {
3095      complete_callout(previous_callout, ptr, cd);      if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */
3096          complete_callout(previous_callout, ptr, cd);
3097      previous_callout = NULL;      previous_callout = NULL;
3098      }      }
3099    
# Line 1653  for (;; ptr++) Line 3102  for (;; ptr++)
3102    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3103      {      {
3104      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
3105      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
3106        {        {
3107        /* The space before the ; is to avoid a warning on a silly compiler        while (*(++ptr) != 0)
3108        on the Macintosh. */          {
3109        while ((c = *(++ptr)) != 0 && c != NEWLINE) ;          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3110        if (c != 0) continue;   /* Else fall through to handle end of string */          }
3111          if (*ptr != 0) continue;
3112    
3113          /* Else fall through to handle end of string */
3114          c = 0;
3115        }        }
3116      }      }
3117    
# Line 1672  for (;; ptr++) Line 3125  for (;; ptr++)
3125    
3126    switch(c)    switch(c)
3127      {      {
3128      /* The branch terminates at end of string, |, or ). */      /* ===================================================================*/
3129        case 0:                        /* The branch terminates at string end */
3130      case 0:      case CHAR_VERTICAL_LINE:       /* or | or ) */
3131      case '|':      case CHAR_RIGHT_PARENTHESIS:
     case ')':  
3132      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
3133      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
3134      *codeptr = code;      *codeptr = code;
3135      *ptrptr = ptr;      *ptrptr = ptr;
3136        if (lengthptr != NULL)
3137          {
3138          if (OFLOW_MAX - *lengthptr < code - last_code)
3139            {
3140            *errorcodeptr = ERR20;
3141            goto FAILED;
3142            }
3143          *lengthptr += (int)(code - last_code);   /* To include callout length */
3144          DPRINTF((">> end branch\n"));
3145          }
3146      return TRUE;      return TRUE;
3147    
3148    
3149        /* ===================================================================*/
3150      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
3151      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3152    
3153      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
3154      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3155        {        {
3156        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
# Line 1695  for (;; ptr++) Line 3159  for (;; ptr++)
3159      *code++ = OP_CIRC;      *code++ = OP_CIRC;
3160      break;      break;
3161    
3162      case '$':      case CHAR_DOLLAR_SIGN:
3163      previous = NULL;      previous = NULL;
3164      *code++ = OP_DOLL;      *code++ = OP_DOLL;
3165      break;      break;
# Line 1703  for (;; ptr++) Line 3167  for (;; ptr++)
3167      /* 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
3168      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
3169    
3170      case '.':      case CHAR_DOT:
3171      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3172      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
3173      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3174      previous = code;      previous = code;
3175      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
3176      break;      break;
3177    
3178      /* Character classes. If the included characters are all < 255 in value, we  
3179      build a 32-byte bitmap of the permitted characters, except in the special      /* ===================================================================*/
3180      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
3181      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
3182      opcode so that data characters > 255 can be handled correctly.      where there is only one such character. For negated classes, we build the
3183        map as usual, then invert it at the end. However, we use a different opcode
3184        so that data characters > 255 can be handled correctly.
3185    
3186      If the class contains characters outside the 0-255 range, a different      If the class contains characters outside the 0-255 range, a different
3187      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,
3188      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
3189      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.
     */  
3190    
3191      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
3192        default (Perl) mode, it is treated as a data character. */
3193    
3194        case CHAR_RIGHT_SQUARE_BRACKET:
3195        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3196          {
3197          *errorcodeptr = ERR64;
3198          goto FAILED;
3199          }
3200        goto NORMAL_CHAR;
3201    
3202        case CHAR_LEFT_SQUARE_BRACKET:
3203      previous = code;      previous = code;
3204    
3205      /* 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
3206      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. */
3207    
3208      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3209          check_posix_syntax(ptr, &tempptr, cd))           ptr[1] == CHAR_EQUALS_SIGN) &&
3210            check_posix_syntax(ptr, &tempptr))
3211        {        {
3212        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
3213        goto FAILED;        goto FAILED;
3214        }        }
3215    
3216      /* 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,
3217        if the first few characters (either before or after ^) are \Q\E or \E we
3218        skip them too. This makes for compatibility with Perl. */
3219    
3220      if ((c = *(++ptr)) == '^')      negate_class = FALSE;
3221        for (;;)
3222        {        {
       negate_class = TRUE;  
3223        c = *(++ptr);        c = *(++ptr);
3224          if (c == CHAR_BACKSLASH)
3225            {
3226            if (ptr[1] == CHAR_E)
3227              ptr++;
3228            else if (strncmp((const char *)ptr+1,
3229                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
3230              ptr += 3;
3231            else
3232              break;
3233            }
3234          else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
3235            negate_class = TRUE;
3236          else break;
3237        }        }
3238      else  
3239        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
3240        an initial ']' is taken as a data character -- the code below handles
3241        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3242        [^] must match any character, so generate OP_ALLANY. */
3243    
3244        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3245            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3246        {        {
3247        negate_class = FALSE;        *code++ = negate_class? OP_ALLANY : OP_FAIL;
3248          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3249          zerofirstbyte = firstbyte;
3250          break;
3251        }        }
3252    
3253        /* If a class contains a negative special such as \S, we need to flip the
3254        negation flag at the end, so that support for characters > 255 works
3255        correctly (they are all included in the class). */
3256    
3257        should_flip_negation = FALSE;
3258    
3259      /* 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
3260      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
3261      characters, we don't yet do any optimization. */      valued UTF-8 characters, we don't yet do any optimization. */
3262    
3263      class_charcount = 0;      class_charcount = 0;
3264      class_lastchar = -1;      class_lastchar = -1;
3265    
3266        /* Initialize the 32-char bit map to all zeros. We build the map in a
3267        temporary bit of memory, in case the class contains only 1 character (less
3268        than 256), because in that case the compiled code doesn't use the bit map.
3269        */
3270    
3271        memset(classbits, 0, 32 * sizeof(uschar));
3272    
3273  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3274      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
3275      class_utf8data = code + LINK_SIZE + 34;   /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
3276        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
3277  #endif  #endif
3278    
     /* 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));  
   
3279      /* Process characters until ] is reached. By writing this as a "do" it      /* Process characters until ] is reached. By writing this as a "do" it
3280      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
3281      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. */  
3282    
3283      do      if (c != 0) do
3284        {        {
3285          const uschar *oldptr;
3286    
3287  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3288        if (utf8 && c > 127)        if (utf8 && c > 127)
3289          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3290          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3291          }          }
3292    
3293          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
3294          data and reset the pointer. This is so that very large classes that
3295          contain a zillion UTF-8 characters no longer overwrite the work space
3296          (which is on the stack). */
3297    
3298          if (lengthptr != NULL)
3299            {
3300            *lengthptr += class_utf8data - class_utf8data_base;
3301            class_utf8data = class_utf8data_base;
3302            }
3303    
3304  #endif  #endif
3305    
3306        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
3307    
3308        if (inescq)        if (inescq)
3309          {          {
3310          if (c == '\\' && ptr[1] == 'E')          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3311            {            {
3312            inescq = FALSE;            inescq = FALSE;                   /* Reset literal state */
3313            ptr++;            ptr++;                            /* Skip the 'E' */
3314            continue;            continue;                         /* Carry on with next */
3315            }            }
3316          else goto LONE_SINGLE_CHARACTER;          goto CHECK_RANGE;                   /* Could be range if \E follows */
3317          }          }
3318    
3319        /* 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 3322  for (;; ptr++)
3322        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3323        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3324    
3325        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3326            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3327            check_posix_syntax(ptr, &tempptr, cd))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3328          {          {
3329          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3330          int posix_class, i;          int posix_class, taboffset, tabopt;
3331          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3332            uschar pbits[32];
3333    
3334          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3335            {            {
3336            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3337            goto FAILED;            goto FAILED;
3338            }            }
3339    
3340          ptr += 2;          ptr += 2;
3341          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3342            {            {
3343            local_negate = TRUE;            local_negate = TRUE;
3344              should_flip_negation = TRUE;  /* Note negative special */
3345            ptr++;            ptr++;
3346            }            }
3347    
3348          posix_class = check_posix_name(ptr, tempptr - ptr);          posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3349          if (posix_class < 0)          if (posix_class < 0)
3350            {            {
3351            *errorcodeptr = ERR30;            *errorcodeptr = ERR30;
# Line 1835  for (;; ptr++) Line 3358  for (;; ptr++)
3358    
3359          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3360            posix_class = 0;            posix_class = 0;
3361    
3362          /* 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
3363          tables, or their negations. The [:blank:] class sets up the same          different escape sequences that use Unicode properties. */
3364          chars as the [:space:] class (all white space). We remove the vertical  
3365          white space chars afterwards. */  #ifdef SUPPORT_UCP
3366            if ((options & PCRE_UCP) != 0)
3367              {
3368              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3369              if (posix_substitutes[pc] != NULL)
3370                {
3371                nestptr = tempptr + 1;
3372                ptr = posix_substitutes[pc] - 1;
3373                continue;
3374                }
3375              }
3376    #endif
3377            /* In the non-UCP case, we build the bit map for the POSIX class in a
3378            chunk of local store because we may be adding and subtracting from it,
3379            and we don't want to subtract bits that may be in the main map already.
3380            At the end we or the result into the bit map that is being built. */
3381    
3382          posix_class *= 3;          posix_class *= 3;
3383          for (i = 0; i < 3; i++)  
3384            /* Copy in the first table (always present) */
3385    
3386            memcpy(pbits, cbits + posix_class_maps[posix_class],
3387              32 * sizeof(uschar));
3388    
3389            /* If there is a second table, add or remove it as required. */
3390    
3391            taboffset = posix_class_maps[posix_class + 1];
3392            tabopt = posix_class_maps[posix_class + 2];
3393    
3394            if (taboffset >= 0)
3395            {            {
3396            BOOL blankclass = strncmp((char *)ptr, "blank", 5) == 0;            if (tabopt >= 0)
3397            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;  
             }  
3398            else            else
3399              {              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;  
             }  
3400            }            }
3401    
3402            /* Not see if we need to remove any special characters. An option
3403            value of 1 removes vertical space and 2 removes underscore. */
3404    
3405            if (tabopt < 0) tabopt = -tabopt;
3406            if (tabopt == 1) pbits[1] &= ~0x3c;
3407              else if (tabopt == 2) pbits[11] &= 0x7f;
3408    
3409            /* Add the POSIX table or its complement into the main table that is
3410            being built and we are done. */
3411    
3412            if (local_negate)
3413              for (c = 0; c < 32; c++) classbits[c] |= ~pbits[c];
3414            else
3415              for (c = 0; c < 32; c++) classbits[c] |= pbits[c];
3416    
3417          ptr = tempptr + 1;          ptr = tempptr + 1;
3418          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */          class_charcount = 10;  /* Set > 1; assumes more than 1 per class */
3419          continue;    /* End of POSIX syntax handling */          continue;    /* End of POSIX syntax handling */
3420          }          }
3421    
3422        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3423        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
3424        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
3425        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
3426        it marks a word boundary. Other escapes have preset maps ready to        class_charcount bigger than one. Unrecognized escapes fall through and
3427        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
3428        character in them, so set class_charcount bigger than one. */        PCRE_EXTRA is set. */
3429    
3430        if (c == '\\')        if (c == CHAR_BACKSLASH)
3431          {          {
3432          c = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3433            if (*errorcodeptr != 0) goto FAILED;
3434    
3435          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 */  
3436          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3437            {            {
3438            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3439              {              {
3440              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3441              }              }
3442            else inescq = TRUE;            else inescq = TRUE;
3443            continue;            continue;
3444            }            }
3445            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
3446    
3447          if (c < 0)          if (c < 0)
3448            {            {
3449            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
3450            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3451    
3452            switch (-c)            switch (-c)
3453              {              {
3454    #ifdef SUPPORT_UCP
3455                case ESC_du:     /* These are the values given for \d etc */
3456                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3457                case ESC_wu:     /* escape sequence with an appropriate \p */
3458                case ESC_WU:     /* or \P to test Unicode properties instead */
3459                case ESC_su:     /* of the default ASCII testing. */
3460                case ESC_SU:
3461                nestptr = ptr;
3462                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3463                class_charcount -= 2;                /* Undo! */
3464                continue;
3465    #endif
3466              case ESC_d:              case ESC_d:
3467              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3468              continue;              continue;
3469    
3470              case ESC_D:              case ESC_D:
3471                should_flip_negation = TRUE;
3472              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
3473              continue;              continue;
3474    
# Line 1910  for (;; ptr++) Line 3477  for (;; ptr++)
3477              continue;              continue;
3478    
3479              case ESC_W:              case ESC_W:
3480                should_flip_negation = TRUE;
3481              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3482              continue;              continue;
3483    
# Line 1919  for (;; ptr++) Line 3487  for (;; ptr++)
3487              continue;              continue;
3488    
3489              case ESC_S:              case ESC_S:
3490                should_flip_negation = TRUE;
3491              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
3492              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3493              continue;              continue;
3494    
3495                case ESC_h:
3496                SETBIT(classbits, 0x09); /* VT */
3497                SETBIT(classbits, 0x20); /* SPACE */
3498                SETBIT(classbits, 0xa0); /* NSBP */
3499    #ifdef SUPPORT_UTF8
3500                if (utf8)
3501                  {
3502                  class_utf8 = TRUE;
3503                  *class_utf8data++ = XCL_SINGLE;
3504                  class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);
3505                  *class_utf8data++ = XCL_SINGLE;
3506                  class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);
3507                  *class_utf8data++ = XCL_RANGE;
3508                  class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);
3509                  class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);
3510                  *class_utf8data++ = XCL_SINGLE;
3511                  class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);
3512                  *class_utf8data++ = XCL_SINGLE;
3513                  class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);
3514                  *class_utf8data++ = XCL_SINGLE;
3515                  class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);
3516                  }
3517    #endif
3518                continue;
3519    
3520                case ESC_H:
3521                for (c = 0; c < 32; c++)
3522                  {
3523                  int x = 0xff;
3524                  switch (c)
3525                    {
3526                    case 0x09/8: x ^= 1 << (0x09%8); break;
3527                    case 0x20/8: x ^= 1 << (0x20%8); break;
3528                    case 0xa0/8: x ^= 1 << (0xa0%8); break;
3529                    default: break;
3530                    }
3531                  classbits[c] |= x;
3532                  }
3533    
3534    #ifdef SUPPORT_UTF8
3535                if (utf8)
3536                  {
3537                  class_utf8 = TRUE;
3538                  *class_utf8data++ = XCL_RANGE;
3539                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3540                  class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);
3541                  *class_utf8data++ = XCL_RANGE;
3542                  class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);
3543                  class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);
3544                  *class_utf8data++ = XCL_RANGE;
3545                  class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);
3546                  class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);
3547                  *class_utf8data++ = XCL_RANGE;
3548                  class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);
3549                  class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);
3550                  *class_utf8data++ = XCL_RANGE;
3551                  class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);
3552                  class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);
3553                  *class_utf8data++ = XCL_RANGE;
3554                  class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);
3555                  class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);
3556                  *class_utf8data++ = XCL_RANGE;
3557                  class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);
3558                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3559                  }
3560    #endif
3561                continue;
3562    
3563                case ESC_v:
3564                SETBIT(classbits, 0x0a); /* LF */
3565                SETBIT(classbits, 0x0b); /* VT */
3566                SETBIT(classbits, 0x0c); /* FF */
3567                SETBIT(classbits, 0x0d); /* CR */
3568                SETBIT(classbits, 0x85); /* NEL */
3569    #ifdef SUPPORT_UTF8
3570                if (utf8)
3571                  {
3572                  class_utf8 = TRUE;
3573                  *class_utf8data++ = XCL_RANGE;
3574                  class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);
3575                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3576                  }
3577    #endif
3578                continue;
3579    
3580                case ESC_V:
3581                for (c = 0; c < 32; c++)
3582                  {
3583                  int x = 0xff;
3584                  switch (c)
3585                    {
3586                    case 0x0a/8: x ^= 1 << (0x0a%8);
3587                                 x ^= 1 << (0x0b%8);
3588                                 x ^= 1 << (0x0c%8);
3589                                 x ^= 1 << (0x0d%8);
3590                                 break;
3591                    case 0x85/8: x ^= 1 << (0x85%8); break;
3592                    default: break;
3593                    }
3594                  classbits[c] |= x;
3595                  }
3596    
3597    #ifdef SUPPORT_UTF8
3598                if (utf8)
3599                  {
3600                  class_utf8 = TRUE;
3601                  *class_utf8data++ = XCL_RANGE;
3602                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3603                  class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);
3604                  *class_utf8data++ = XCL_RANGE;
3605                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3606                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3607                  }
3608    #endif
3609                continue;
3610    
3611  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3612              case ESC_p:              case ESC_p:
3613              case ESC_P:              case ESC_P:
3614                {                {
3615                BOOL negated;                BOOL negated;
3616                int property = get_ucp(&ptr, &negated, errorcodeptr);                int pdata;
3617                if (property < 0) goto FAILED;                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3618                  if (ptype < 0) goto FAILED;
3619                class_utf8 = TRUE;                class_utf8 = TRUE;
3620                *class_utf8data++ = ((-c == ESC_p) != negated)?                *class_utf8data++ = ((-c == ESC_p) != negated)?
3621                  XCL_PROP : XCL_NOTPROP;                  XCL_PROP : XCL_NOTPROP;
3622                *class_utf8data++ = property;                *class_utf8data++ = ptype;
3623                  *class_utf8data++ = pdata;
3624                class_charcount -= 2;   /* Not a < 256 character */                class_charcount -= 2;   /* Not a < 256 character */
3625                  continue;
3626                }                }
             continue;  
3627  #endif  #endif
   
3628              /* Unrecognized escapes are faulted if PCRE is running in its              /* Unrecognized escapes are faulted if PCRE is running in its
3629              strict mode. By default, for compatibility with Perl, they are              strict mode. By default, for compatibility with Perl, they are
3630              treated as literals. */              treated as literals. */
# Line 1949  for (;; ptr++) Line 3635  for (;; ptr++)
3635                *errorcodeptr = ERR7;                *errorcodeptr = ERR7;
3636                goto FAILED;                goto FAILED;
3637                }                }
             c = *ptr;              /* The final character */  
3638              class_charcount -= 2;  /* Undo the default count from above */              class_charcount -= 2;  /* Undo the default count from above */
3639                c = *ptr;              /* Get the final character and fall through */
3640                break;
3641              }              }
3642            }            }
3643    
3644          /* 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
3645          > 256 in UTF-8 mode. */          greater than 256 in UTF-8 mode. */
3646    
3647          }   /* End of backslash handling */          }   /* End of backslash handling */
3648    
3649        /* A single character may be followed by '-' to form a range. However,        /* A single character may be followed by '-' to form a range. However,
3650        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
3651        here is treated as a literal. */        at the end is treated as a literal. Perl ignores orphaned \E sequences
3652          entirely. The code for handling \Q and \E is messy. */
3653    
3654          CHECK_RANGE:
3655          while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3656            {
3657            inescq = FALSE;
3658            ptr += 2;
3659            }
3660    
3661          oldptr = ptr;
3662    
3663          /* Remember \r or \n */
3664    
3665        if (ptr[1] == '-' && ptr[2] != ']')        if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3666    
3667          /* Check for range */
3668    
3669          if (!inescq && ptr[1] == CHAR_MINUS)
3670          {          {
3671          int d;          int d;
3672          ptr += 2;          ptr += 2;
3673            while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3674    
3675            /* If we hit \Q (not followed by \E) at this point, go into escaped
3676            mode. */
3677    
3678            while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3679              {
3680              ptr += 2;
3681              if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3682                { ptr += 2; continue; }
3683              inescq = TRUE;
3684              break;
3685              }
3686    
3687            if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3688              {
3689              ptr = oldptr;
3690              goto LONE_SINGLE_CHARACTER;
3691              }
3692    
3693  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3694          if (utf8)          if (utf8)
# Line 1981  for (;; ptr++) Line 3703  for (;; ptr++)
3703          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
3704          in such circumstances. */          in such circumstances. */
3705    
3706          if (d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3707            {            {
3708            const uschar *oldptr = ptr;            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3709            d = check_escape(&ptr, errorcodeptr, *brackets, options, TRUE);            if (*errorcodeptr != 0) goto FAILED;
3710    
3711            /* \b is backslash; \X is literal X; any other special means the '-'            /* \b is backspace; any other special means the '-' was literal */
           was literal */  
3712    
3713            if (d < 0)            if (d < 0)
3714              {              {
3715              if (d == -ESC_b) d = '\b';              if (d == -ESC_b) d = CHAR_BS; else
             else if (d == -ESC_X) d = 'X'; else  
3716                {                {
3717                ptr = oldptr - 2;                ptr = oldptr;
3718                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3719                }                }
3720              }              }
3721            }            }
3722    
3723          /* The check that the two values are in the correct order happens in          /* Check that the two values are in the correct order. Optimize
3724          the pre-pass. Optimize one-character ranges */          one-character ranges */
3725    
3726            if (d < c)
3727              {
3728              *errorcodeptr = ERR8;
3729              goto FAILED;
3730              }
3731    
3732          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3733    
3734            /* Remember \r or \n */
3735    
3736            if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3737    
3738          /* 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
3739          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
3740          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 3752  for (;; ptr++)
3752  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3753            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
3754              {              {
3755              int occ, ocd;              unsigned int occ, ocd;
3756              int cc = c;              unsigned int cc = c;
3757              int origd = d;              unsigned int origd = d;
3758              while (get_othercase_range(&cc, origd, &occ, &ocd))              while (get_othercase_range(&cc, origd, &occ, &ocd))
3759                {                {
3760                if (occ >= c && ocd <= d) continue;  /* Skip embedded ranges */                if (occ >= (unsigned int)c &&
3761                      ocd <= (unsigned int)d)
3762                    continue;                          /* Skip embedded ranges */
3763    
3764                if (occ < c  && ocd >= c - 1)        /* Extend the basic range */                if (occ < (unsigned int)c  &&
3765                      ocd >= (unsigned int)c - 1)      /* Extend the basic range */
3766                  {                                  /* if there is overlap,   */                  {                                  /* if there is overlap,   */
3767                  c = occ;                           /* noting that if occ < c */                  c = occ;                           /* noting that if occ < c */
3768                  continue;                          /* we can't have ocd > d  */                  continue;                          /* we can't have ocd > d  */
3769                  }                                  /* because a subrange is  */                  }                                  /* because a subrange is  */
3770                if (ocd > d && occ <= d + 1)         /* always shorter than    */                if (ocd > (unsigned int)d &&
3771                      occ <= (unsigned int)d + 1)      /* always shorter than    */
3772                  {                                  /* the basic range.       */                  {                                  /* the basic range.       */
3773                  d = ocd;                  d = ocd;
3774                  continue;                  continue;
# Line 2082  for (;; ptr++) Line 3816  for (;; ptr++)
3816          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
3817          for partial ranges without UCP support. */          for partial ranges without UCP support. */
3818    
3819          for (; c <= d; c++)          class_charcount += d - c + 1;
3820            class_lastchar = d;
3821    
3822            /* We can save a bit of time by skipping this in the pre-compile. */
3823    
3824            if (lengthptr == NULL) for (; c <= d; c++)
3825            {            {
3826            classbits[c/8] |= (1 << (c&7));            classbits[c/8] |= (1 << (c&7));
3827            if ((options & PCRE_CASELESS) != 0)            if ((options & PCRE_CASELESS) != 0)
# Line 2090  for (;; ptr++) Line 3829  for (;; ptr++)
3829              int uc = cd->fcc[c];           /* flip case */              int uc = cd->fcc[c];           /* flip case */
3830              classbits[uc/8] |= (1 << (uc&7));              classbits[uc/8] |= (1 << (uc&7));
3831              }              }
           class_charcount++;                /* in case a one-char range */  
           class_lastchar = c;  
3832            }            }
3833    
3834          continue;   /* Go get the next char in the class */          continue;   /* Go get the next char in the class */
# Line 2115  for (;; ptr++) Line 3852  for (;; ptr++)
3852  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3853          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3854            {            {
3855            int chartype;            unsigned int othercase;
3856            int othercase;            if ((othercase = UCD_OTHERCASE(c)) != c)
           if (_pcre_ucp_findchar(c, &chartype, &othercase) >= 0 &&  
                othercase > 0)  
3857              {              {
3858              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3859              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 2143  for (;; ptr++) Line 3878  for (;; ptr++)
3878          }          }
3879        }        }
3880    
3881      /* 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.
3882      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
3883        string. */
3884    
3885        while (((c = *(++ptr)) != 0 ||
3886               (nestptr != NULL &&