/[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 172 by ph10, Tue Jun 5 10:40:13 2007 UTC revision 744 by zherczeg, Sun Nov 13 16:31:38 2011 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2007 University of Cambridge             Copyright (c) 1997-2011 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 42  POSSIBILITY OF SUCH DAMAGE. Line 42  POSSIBILITY OF SUCH DAMAGE.
42  supporting internal functions that are not used by other modules. */  supporting internal functions that are not used by other modules. */
43    
44    
45    #ifdef HAVE_CONFIG_H
46    #include "config.h"
47    #endif
48    
49  #define NLBLOCK cd             /* Block containing newline information */  #define NLBLOCK cd             /* Block containing newline information */
50  #define PSSTART start_pattern  /* Field containing processed string start */  #define PSSTART start_pattern  /* Field containing processed string start */
51  #define PSEND   end_pattern    /* Field containing processed string end */  #define PSEND   end_pattern    /* Field containing processed string end */
52    
   
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56  /* When DEBUG is defined, we need the pcre_printint() function, which is also  /* When PCRE_DEBUG is defined, we need the pcre_printint() function, which is
57  used by pcretest. DEBUG is not defined when building a production library. */  also used by pcretest. PCRE_DEBUG is not defined when building a production
58    library. */
59    
60  #ifdef DEBUG  #ifdef PCRE_DEBUG
61  #include "pcre_printint.src"  #include "pcre_printint.src"
62  #endif  #endif
63    
64    
65    /* Macro for setting individual bits in class bitmaps. */
66    
67    #define SETBIT(a,b) a[b/8] |= (1 << (b%8))
68    
69    /* Maximum length value to check against when making sure that the integer that
70    holds the compiled pattern length does not overflow. We make it a bit less than
71    INT_MAX to allow for adding in group terminating bytes, so that we don't have
72    to check them every time. */
73    
74    #define OFLOW_MAX (INT_MAX - 20)
75    
76    
77  /*************************************************  /*************************************************
78  *      Code parameters and static tables         *  *      Code parameters and static tables         *
79  *************************************************/  *************************************************/
# Line 76  is 4 there is plenty of room. */ Line 92  is 4 there is plenty of room. */
92    
93  #define COMPILE_WORK_SIZE (4096)  #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
102  are simple data values; negative values are for special things like \d and so  are simple data values; negative values are for special things like \d and so
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  #ifndef 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, -ESC_K,      0,      0,      0,      0,   /* H - O */       0,                       0,
116  -ESC_P, -ESC_Q, -ESC_R, -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, -ESC_k,      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 106  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, -ESC_k,     '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,-ESC_R,      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 125  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, lower, upper,  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 };
# Line 164  static const int posix_class_maps[] = { Line 261  static const int posix_class_maps[] = {
261    cbit_xdigit,-1,          0              /* 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    #ifdef SUPPORT_UCP
269    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  #define STRING(a)  # a
313  #define XSTRING(s) STRING(s)  #define XSTRING(s) STRING(s)
# Line 171  static const int posix_class_maps[] = { Line 315  static const int posix_class_maps[] = {
315  /* The texts of compile-time error messages. These are "char *" because they  /* The texts of compile-time error messages. These are "char *" because they
316  are passed to the outside world. Do not ever re-use any error number, because  are passed to the outside world. Do not ever re-use any error number, because
317  they are documented. Always add a new error instead. Messages marked DEAD below  they are documented. Always add a new error instead. Messages marked DEAD below
318  are no longer used. */  are no longer used. This used to be a table of strings, but in order to reduce
319    the number of relocations needed when a shared library is loaded dynamically,
320  static const char *error_texts[] = {  it is now one long string. We cannot use a table of offsets, because the
321    "no error",  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
322    "\\ at end of pattern",  simply count through to the one we want - this isn't a performance issue
323    "\\c at end of pattern",  because these strings are used only when there is a compilation error.
324    "unrecognized character follows \\",  
325    "numbers out of order in {} quantifier",  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",  /** DEAD **/    "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",  /** DEAD **/    "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 or name 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",  /** DEAD **/    "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 in subpattern name (missing terminator)",    "syntax error in subpattern name (missing terminator)\0"
381    "two named subpatterns 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)",    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
388    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")",    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
389    /* 50 */    /* 50 */
390    "repeated subpattern is too long",    "repeated subpattern is too long\0"    /** DEAD **/
391    "octal value is greater than \\377 (not in UTF-8 mode)",    "octal value is greater than \\377 (not in UTF-8 mode)\0"
392    "internal error: overran compiling workspace",    "internal error: overran compiling workspace\0"
393    "internal error: previously-checked referenced subpattern not found",    "internal error: previously-checked referenced subpattern not found\0"
394    "DEFINE group contains more than one branch",    "DEFINE group contains more than one branch\0"
395    /* 55 */    /* 55 */
396    "repeating a DEFINE group is not allowed",    "repeating a DEFINE group is not allowed\0"  /** DEAD **/
397    "inconsistent NEWLINE options",    "inconsistent NEWLINE options\0"
398    "\\g is not followed by a braced name or an optionally braced non-zero number",    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
399    "(?+ or (?- or (?(+ or (?(- must be followed by a non-zero number"    "a numbered reference must not be zero\0"
400  };    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
401      /* 60 */
402      "(*VERB) not recognized\0"
403      "number is too big\0"
404      "subpattern name expected\0"
405      "digit expected after (?+\0"
406      "] is an invalid data character in JavaScript compatibility mode\0"
407      /* 65 */
408      "different names for subpatterns of the same number are not allowed\0"
409      "(*MARK) must have an argument\0"
410      "this version of PCRE is not compiled with PCRE_UCP support\0"
411      "\\c must be followed by an ASCII character\0"
412      "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
413      ;
414    
415  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
416  patterns. Note that the tables in chartables are dependent on the locale, and  patterns. Note that the tables in chartables are dependent on the locale, and
# Line 263  For convenience, we use the same bit def Line 428  For convenience, we use the same bit def
428    
429  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
430    
431  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC
432    
433    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
434    UTF-8 mode. */
435    
436  static const unsigned char digitab[] =  static const unsigned char digitab[] =
437    {    {
438    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
# Line 299  static const unsigned char digitab[] = Line 468  static const unsigned char digitab[] =
468    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
469    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
470    
471  #else           /* This is the "abnormal" case, for EBCDIC systems */  #else
472    
473    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
474    
475  static const unsigned char digitab[] =  static const unsigned char digitab[] =
476    {    {
477    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
# Line 374  static const unsigned char ebcdic_charta Line 546  static const unsigned char ebcdic_charta
546  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
547    
548  static BOOL  static BOOL
549    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, int, int *,    compile_regex(int, uschar **, const uschar **, int *, BOOL, BOOL, int, int,
550      int *, branch_chain *, compile_data *, int *);      int *, int *, branch_chain *, compile_data *, int *);
551    
552    
553    
554    /*************************************************
555    *            Find an error text                  *
556    *************************************************/
557    
558    /* The error texts are now all in one long string, to save on relocations. As
559    some of the text is of unknown length, we can't use a table of offsets.
560    Instead, just count through the strings. This is not a performance issue
561    because it happens only when there has been a compilation error.
562    
563    Argument:   the error number
564    Returns:    pointer to the error string
565    */
566    
567    static const char *
568    find_error_text(int n)
569    {
570    const char *s = error_texts;
571    for (; n > 0; n--)
572      {
573      while (*s++ != 0) {};
574      if (*s == 0) return "Error text not found (please report)";
575      }
576    return s;
577    }
578    
579    
580    /*************************************************
581    *            Check for counted repeat            *
582    *************************************************/
583    
584    /* This function is called when a '{' is encountered in a place where it might
585    start a quantifier. It looks ahead to see if it really is a quantifier or not.
586    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
587    where the ddds are digits.
588    
589    Arguments:
590      p         pointer to the first char after '{'
591    
592    Returns:    TRUE or FALSE
593    */
594    
595    static BOOL
596    is_counted_repeat(const uschar *p)
597    {
598    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
599    while ((digitab[*p] & ctype_digit) != 0) p++;
600    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
601    
602    if (*p++ != CHAR_COMMA) return FALSE;
603    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
604    
605    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
606    while ((digitab[*p] & ctype_digit) != 0) p++;
607    
608    return (*p == CHAR_RIGHT_CURLY_BRACKET);
609    }
610    
611    
612    
# Line 400  Arguments: Line 631  Arguments:
631    
632  Returns:         zero or positive => a data character  Returns:         zero or positive => a data character
633                   negative => a special escape sequence                   negative => a special escape sequence
634                   on error, errorptr is set                   on error, errorcodeptr is set
635  */  */
636    
637  static int  static int
# Line 418  ptr--; /* Set Line 649  ptr--; /* Set
649    
650  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
651    
652  /* Non-alphamerics are literals. For digits or letters, do an initial lookup in  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
653  a table. A non-zero result is something that can be returned immediately.  in a table. A non-zero result is something that can be returned immediately.
654  Otherwise further processing may be required. */  Otherwise further processing may be required. */
655    
656  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
657  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */
658  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0) c = i;
659    
660  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
661  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */
662  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
663  #endif  #endif
664    
# Line 443  else Line 674  else
674      /* A number of Perl escapes are not handled by PCRE. We give an explicit      /* A number of Perl escapes are not handled by PCRE. We give an explicit
675      error. */      error. */
676    
677      case 'l':      case CHAR_l:
678      case 'L':      case CHAR_L:
     case 'N':  
     case 'u':  
     case 'U':  
679      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
680      break;      break;
681    
682      /* \g must be followed by a number, either plain or braced. If positive, it      case CHAR_u:
683      is an absolute backreference. If negative, it is a relative backreference.      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
684      This is a Perl 5.10 feature. Perl 5.10 also supports \g{name} as a        {
685      reference to a named group. This is part of Perl's movement towards a        /* In JavaScript, \u must be followed by four hexadecimal numbers.
686      unified syntax for back references. As this is synonymous with \k{name}, we        Otherwise it is a lowercase u letter. */
687      fudge it up by pretending it really was \k. */        if ((digitab[ptr[1]] & ctype_xdigit) != 0 && (digitab[ptr[2]] & ctype_xdigit) != 0
688               && (digitab[ptr[3]] & ctype_xdigit) != 0 && (digitab[ptr[4]] & ctype_xdigit) != 0)
689            {
690            int i;
691            c = 0;
692            for (i = 0; i < 4; ++i)
693              {
694              register int cc = *(++ptr);
695    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
696              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
697              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
698    #else           /* EBCDIC coding */
699              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
700              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
701    #endif
702              }
703            }
704          }
705        else
706          *errorcodeptr = ERR37;
707        break;
708    
709        case CHAR_U:
710        /* In JavaScript, \U is an uppercase U letter. */
711        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
712        break;
713    
714        /* In a character class, \g is just a literal "g". Outside a character
715        class, \g must be followed by one of a number of specific things:
716    
717        (1) A number, either plain or braced. If positive, it is an absolute
718        backreference. If negative, it is a relative backreference. This is a Perl
719        5.10 feature.
720    
721        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
722        is part of Perl's movement towards a unified syntax for back references. As
723        this is synonymous with \k{name}, we fudge it up by pretending it really
724        was \k.
725    
726        (3) For Oniguruma compatibility we also support \g followed by a name or a
727        number either in angle brackets or in single quotes. However, these are
728        (possibly recursive) subroutine calls, _not_ backreferences. Just return
729        the -ESC_g code (cf \k). */
730    
731        case CHAR_g:
732        if (isclass) break;
733        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
734          {
735          c = -ESC_g;
736          break;
737          }
738    
739        /* Handle the Perl-compatible cases */
740    
741      case 'g':      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
     if (ptr[1] == '{')  
742        {        {
743        const uschar *p;        const uschar *p;
744        for (p = ptr+2; *p != 0 && *p != '}'; p++)        for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
745          if (*p != '-' && (digitab[*p] & ctype_digit) == 0) break;          if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
746        if (*p != 0 && *p != '}')        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
747          {          {
748          c = -ESC_k;          c = -ESC_k;
749          break;          break;
# Line 474  else Line 753  else
753        }        }
754      else braced = FALSE;      else braced = FALSE;
755    
756      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
757        {        {
758        negated = TRUE;        negated = TRUE;
759        ptr++;        ptr++;
# Line 483  else Line 762  else
762    
763      c = 0;      c = 0;
764      while ((digitab[ptr[1]] & ctype_digit) != 0)      while ((digitab[ptr[1]] & ctype_digit) != 0)
765        c = c * 10 + *(++ptr) - '0';        c = c * 10 + *(++ptr) - CHAR_0;
766    
767      if (c == 0 || (braced && *(++ptr) != '}'))      if (c < 0)   /* Integer overflow */
768          {
769          *errorcodeptr = ERR61;
770          break;
771          }
772    
773        if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
774        {        {
775        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
776        return 0;        break;
777          }
778    
779        if (c == 0)
780          {
781          *errorcodeptr = ERR58;
782          break;
783        }        }
784    
785      if (negated)      if (negated)
# Line 496  else Line 787  else
787        if (c > bracount)        if (c > bracount)
788          {          {
789          *errorcodeptr = ERR15;          *errorcodeptr = ERR15;
790          return 0;          break;
791          }          }
792        c = bracount - (c - 1);        c = bracount - (c - 1);
793        }        }
# Line 516  else Line 807  else
807      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
808      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
809    
810      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:
811      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
812    
813      if (!isclass)      if (!isclass)
814        {        {
815        oldptr = ptr;        oldptr = ptr;
816        c -= '0';        c -= CHAR_0;
817        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
818          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - CHAR_0;
819          if (c < 0)    /* Integer overflow */
820            {
821            *errorcodeptr = ERR61;
822            break;
823            }
824        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
825          {          {
826          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 537  else Line 833  else
833      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.
834      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
835    
836      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
837        {        {
838        ptr--;        ptr--;
839        c = 0;        c = 0;
# Line 550  else Line 846  else
846      to do). Nowadays we allow for larger numbers in UTF-8 mode, but no more      to do). Nowadays we allow for larger numbers in UTF-8 mode, but no more
847      than 3 octal digits. */      than 3 octal digits. */
848    
849      case '0':      case CHAR_0:
850      c -= '0';      c -= CHAR_0;
851      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
852          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
853      if (!utf8 && c > 255) *errorcodeptr = ERR51;      if (!utf8 && c > 255) *errorcodeptr = ERR51;
854      break;      break;
855    
# Line 561  else Line 857  else
857      than 0xff in utf8 mode, but only if the ddd are hex digits. If not, { is      than 0xff in utf8 mode, but only if the ddd are hex digits. If not, { is
858      treated as a data character. */      treated as a data character. */
859    
860      case 'x':      case CHAR_x:
861      if (ptr[1] == '{')      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
862          {
863          /* In JavaScript, \x must be followed by two hexadecimal numbers.
864          Otherwise it is a lowercase x letter. */
865          if ((digitab[ptr[1]] & ctype_xdigit) != 0 && (digitab[ptr[2]] & ctype_xdigit) != 0)
866            {
867            int i;
868            c = 0;
869            for (i = 0; i < 2; ++i)
870              {
871              register int cc = *(++ptr);
872    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
873              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
874              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
875    #else           /* EBCDIC coding */
876              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
877              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
878    #endif
879              }
880            }
881          break;
882          }
883    
884        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
885        {        {
886        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
887        int count = 0;        int count = 0;
# Line 571  else Line 890  else
890        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
891          {          {
892          register int cc = *pt++;          register int cc = *pt++;
893          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
894          count++;          count++;
895    
896  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
897          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
898          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
899  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
900          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
901          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
902  #endif  #endif
903          }          }
904    
905        if (*pt == '}')        if (*pt == CHAR_RIGHT_CURLY_BRACKET)
906          {          {
907          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
908          ptr = pt;          ptr = pt;
# Line 599  else Line 918  else
918      c = 0;      c = 0;
919      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
920        {        {
921        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
922        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
923  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
924        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
925        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
926  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
927        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
928        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
929  #endif  #endif
930        }        }
931      break;      break;
932    
933      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
934      This coding is ASCII-specific, but then the whole concept of \cx is      An error is given if the byte following \c is not an ASCII character. This
935        coding is ASCII-specific, but then the whole concept of \cx is
936      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
937    
938      case 'c':      case CHAR_c:
939      c = *(++ptr);      c = *(++ptr);
940      if (c == 0)      if (c == 0)
941        {        {
942        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
943        return 0;        break;
944        }        }
945    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
946  #ifndef EBCDIC  /* ASCII coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
947      if (c >= 'a' && c <= 'z') c -= 32;        {
948          *errorcodeptr = ERR68;
949          break;
950          }
951        if (c >= CHAR_a && c <= CHAR_z) c -= 32;
952      c ^= 0x40;      c ^= 0x40;
953  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
954      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
955      c ^= 0xC0;      c ^= 0xC0;
956  #endif  #endif
957      break;      break;
958    
959      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
960      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
961      for Perl compatibility, it is a literal. This code looks a bit odd, but      otherwise, for Perl compatibility, it is a literal. This code looks a bit
962      there used to be some cases other than the default, and there may be again      odd, but there used to be some cases other than the default, and there may
963      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
964    
965      default:      default:
966      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 649  else Line 973  else
973      }      }
974    }    }
975    
976    /* Perl supports \N{name} for character names, as well as plain \N for "not
977    newline". PCRE does not support \N{name}. However, it does support
978    quantification such as \N{2,3}. */
979    
980    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
981         !is_counted_repeat(ptr+2))
982      *errorcodeptr = ERR37;
983    
984    /* If PCRE_UCP is set, we change the values for \d etc. */
985    
986    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
987      c -= (ESC_DU - ESC_D);
988    
989    /* Set the pointer to the final character before returning. */
990    
991  *ptrptr = ptr;  *ptrptr = ptr;
992  return c;  return c;
993  }  }
# Line 689  if (c == 0) goto ERROR_RETURN; Line 1028  if (c == 0) goto ERROR_RETURN;
1028  /* \P or \p can be followed by a name in {}, optionally preceded by ^ for  /* \P or \p can be followed by a name in {}, optionally preceded by ^ for
1029  negation. */  negation. */
1030    
1031  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
1032    {    {
1033    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1034      {      {
1035      *negptr = TRUE;      *negptr = TRUE;
1036      ptr++;      ptr++;
1037      }      }
1038    for (i = 0; i < sizeof(name) - 1; i++)    for (i = 0; i < (int)sizeof(name) - 1; i++)
1039      {      {
1040      c = *(++ptr);      c = *(++ptr);
1041      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
1042      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1043      name[i] = c;      name[i] = c;
1044      }      }
1045    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
1046    name[i] = 0;    name[i] = 0;
1047    }    }
1048    
# Line 725  top = _pcre_utt_size; Line 1064  top = _pcre_utt_size;
1064  while (bot < top)  while (bot < top)
1065    {    {
1066    i = (bot + top) >> 1;    i = (bot + top) >> 1;
1067    c = strcmp(name, _pcre_utt[i].name);    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);
1068    if (c == 0)    if (c == 0)
1069      {      {
1070      *dptr = _pcre_utt[i].value;      *dptr = _pcre_utt[i].value;
# Line 749  return -1; Line 1088  return -1;
1088    
1089    
1090  /*************************************************  /*************************************************
 *            Check for counted repeat            *  
 *************************************************/  
   
 /* This function is called when a '{' is encountered in a place where it might  
 start a quantifier. It looks ahead to see if it really is a quantifier or not.  
 It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}  
 where the ddds are digits.  
   
 Arguments:  
   p         pointer to the first char after '{'  
   
 Returns:    TRUE or FALSE  
 */  
   
 static BOOL  
 is_counted_repeat(const uschar *p)  
 {  
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
 if (*p == '}') return TRUE;  
   
 if (*p++ != ',') return FALSE;  
 if (*p == '}') return TRUE;  
   
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
   
 return (*p == '}');  
 }  
   
   
   
 /*************************************************  
1091  *         Read repeat counts                     *  *         Read repeat counts                     *
1092  *************************************************/  *************************************************/
1093    
# Line 809  int max = -1; Line 1115  int max = -1;
1115  /* Read the minimum value and do a paranoid check: a negative value indicates  /* Read the minimum value and do a paranoid check: a negative value indicates
1116  an integer overflow. */  an integer overflow. */
1117    
1118  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;
1119  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1120    {    {
1121    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 819  if (min < 0 || min > 65535) Line 1125  if (min < 0 || min > 65535)
1125  /* Read the maximum value if there is one, and again do a paranoid on its size.  /* Read the maximum value if there is one, and again do a paranoid on its size.
1126  Also, max must not be less than min. */  Also, max must not be less than min. */
1127    
1128  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1129    {    {
1130    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1131      {      {
1132      max = 0;      max = 0;
1133      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;
1134      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1135        {        {
1136        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 849  return p; Line 1155  return p;
1155    
1156    
1157  /*************************************************  /*************************************************
1158  *       Find forward referenced subpattern       *  *  Subroutine for finding forward reference      *
1159  *************************************************/  *************************************************/
1160    
1161  /* This function scans along a pattern's text looking for capturing  /* This recursive function is called only from find_parens() below. The
1162    top-level call starts at the beginning of the pattern. All other calls must
1163    start at a parenthesis. It scans along a pattern's text looking for capturing
1164  subpatterns, and counting them. If it finds a named pattern that matches the  subpatterns, and counting them. If it finds a named pattern that matches the
1165  name it is given, it returns its number. Alternatively, if the name is NULL, it  name it is given, it returns its number. Alternatively, if the name is NULL, it
1166  returns when it reaches a given numbered subpattern. This is used for forward  returns when it reaches a given numbered subpattern. Recursion is used to keep
1167  references to subpatterns. We know that if (?P< is encountered, the name will  track of subpatterns that reset the capturing group numbers - the (?| feature.
1168  be terminated by '>' because that is checked in the first pass.  
1169    This function was originally called only from the second pass, in which we know
1170    that if (?< or (?' or (?P< is encountered, the name will be correctly
1171    terminated because that is checked in the first pass. There is now one call to
1172    this function in the first pass, to check for a recursive back reference by
1173    name (so that we can make the whole group atomic). In this case, we need check
1174    only up to the current position in the pattern, and that is still OK because
1175    and previous occurrences will have been checked. To make this work, the test
1176    for "end of pattern" is a check against cd->end_pattern in the main loop,
1177    instead of looking for a binary zero. This means that the special first-pass
1178    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1179    processing items within the loop are OK, because afterwards the main loop will
1180    terminate.)
1181    
1182  Arguments:  Arguments:
1183    ptr          current position in the pattern    ptrptr       address of the current character pointer (updated)
1184    count        current count of capturing parens so far encountered    cd           compile background data
1185    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1186    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1187    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1188      utf8         TRUE if we are in UTF-8 mode
1189      count        pointer to the current capturing subpattern number (updated)
1190    
1191  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1192  */  */
1193    
1194  static int  static int
1195  find_parens(const uschar *ptr, int count, const uschar *name, int lorn,  find_parens_sub(uschar **ptrptr, compile_data *cd, const uschar *name, int lorn,
1196    BOOL xmode)    BOOL xmode, BOOL utf8, int *count)
1197  {  {
1198  const uschar *thisname;  uschar *ptr = *ptrptr;
1199    int start_count = *count;
1200    int hwm_count = start_count;
1201    BOOL dup_parens = FALSE;
1202    
1203  for (; *ptr != 0; ptr++)  /* If the first character is a parenthesis, check on the type of group we are
1204    dealing with. The very first call may not start with a parenthesis. */
1205    
1206    if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1207    {    {
1208    int term;    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1209    
1210      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1211    
1212      /* Handle a normal, unnamed capturing parenthesis. */
1213    
1214      else if (ptr[1] != CHAR_QUESTION_MARK)
1215        {
1216        *count += 1;
1217        if (name == NULL && *count == lorn) return *count;
1218        ptr++;
1219        }
1220    
1221      /* All cases now have (? at the start. Remember when we are in a group
1222      where the parenthesis numbers are duplicated. */
1223    
1224      else if (ptr[2] == CHAR_VERTICAL_LINE)
1225        {
1226        ptr += 3;
1227        dup_parens = TRUE;
1228        }
1229    
1230      /* Handle comments; all characters are allowed until a ket is reached. */
1231    
1232      else if (ptr[2] == CHAR_NUMBER_SIGN)
1233        {
1234        for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1235        goto FAIL_EXIT;
1236        }
1237    
1238      /* Handle a condition. If it is an assertion, just carry on so that it
1239      is processed as normal. If not, skip to the closing parenthesis of the
1240      condition (there can't be any nested parens). */
1241    
1242      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1243        {
1244        ptr += 2;
1245        if (ptr[1] != CHAR_QUESTION_MARK)
1246          {
1247          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1248          if (*ptr != 0) ptr++;
1249          }
1250        }
1251    
1252      /* Start with (? but not a condition. */
1253    
1254      else
1255        {
1256        ptr += 2;
1257        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1258    
1259        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1260    
1261        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1262            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1263          {
1264          int term;
1265          const uschar *thisname;
1266          *count += 1;
1267          if (name == NULL && *count == lorn) return *count;
1268          term = *ptr++;
1269          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1270          thisname = ptr;
1271          while (*ptr != term) ptr++;
1272          if (name != NULL && lorn == ptr - thisname &&
1273              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1274            return *count;
1275          term++;
1276          }
1277        }
1278      }
1279    
1280    /* Past any initial parenthesis handling, scan for parentheses or vertical
1281    bars. Stop if we get to cd->end_pattern. Note that this is important for the
1282    first-pass call when this value is temporarily adjusted to stop at the current
1283    position. So DO NOT change this to a test for binary zero. */
1284    
1285    for (; ptr < cd->end_pattern; ptr++)
1286      {
1287    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1288    
1289    if (*ptr == '\\')    if (*ptr == CHAR_BACKSLASH)
1290      {      {
1291      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1292      if (*ptr == 'Q') for (;;)      if (*ptr == CHAR_Q) for (;;)
1293        {        {
1294        while (*(++ptr) != 0 && *ptr != '\\');        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1295        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1296        if (*(++ptr) == 'E') break;        if (*(++ptr) == CHAR_E) break;
1297        }        }
1298      continue;      continue;
1299      }      }
1300    
1301    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1302      are handled for real. If the first character is '^', skip it. Also, if the
1303      first few characters (either before or after ^) are \Q\E or \E we skip them
1304      too. This makes for compatibility with Perl. Note the use of STR macros to
1305      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1306    
1307      if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1308        {
1309        BOOL negate_class = FALSE;
1310        for (;;)
1311          {
1312          if (ptr[1] == CHAR_BACKSLASH)
1313            {
1314            if (ptr[2] == CHAR_E)
1315              ptr+= 2;
1316            else if (strncmp((const char *)ptr+2,
1317                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1318              ptr += 4;
1319            else
1320              break;
1321            }
1322          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1323            {
1324            negate_class = TRUE;
1325            ptr++;
1326            }
1327          else break;
1328          }
1329    
1330    if (*ptr == '[')      /* If the next character is ']', it is a data character that must be
1331      {      skipped, except in JavaScript compatibility mode. */
1332      while (*(++ptr) != ']')  
1333        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1334            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1335          ptr++;
1336    
1337        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1338        {        {
1339        if (*ptr == '\\')        if (*ptr == 0) return -1;
1340          if (*ptr == CHAR_BACKSLASH)
1341          {          {
1342          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) goto FAIL_EXIT;
1343          if (*ptr == 'Q') for (;;)          if (*ptr == CHAR_Q) for (;;)
1344            {            {
1345            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1346            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1347            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1348            }            }
1349          continue;          continue;
1350          }          }
# Line 916  for (; *ptr != 0; ptr++) Line 1354  for (; *ptr != 0; ptr++)
1354    
1355    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1356    
1357    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1358      {      {
1359      while (*(++ptr) != 0 && *ptr != '\n');      ptr++;
1360      if (*ptr == 0) return -1;      while (*ptr != 0)
1361          {
1362          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1363          ptr++;
1364    #ifdef SUPPORT_UTF8
1365          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
1366    #endif
1367          }
1368        if (*ptr == 0) goto FAIL_EXIT;
1369      continue;      continue;
1370      }      }
1371    
1372    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1373    
1374    if (*ptr != '(') continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != '?')  
1375      {      {
1376      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, count);
1377      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1378      continue;      if (*ptr == 0) goto FAIL_EXIT;
1379        }
1380    
1381      else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1382        {
1383        if (dup_parens && *count < hwm_count) *count = hwm_count;
1384        goto FAIL_EXIT;
1385        }
1386    
1387      else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1388        {
1389        if (*count > hwm_count) hwm_count = *count;
1390        *count = start_count;
1391      }      }
1392      }
1393    
1394    ptr += 2;  FAIL_EXIT:
1395    if (*ptr == 'P') ptr++;                      /* Allow optional P */  *ptrptr = ptr;
1396    return -1;
1397    }
1398    
   /* We have to disambiguate (?<! and (?<= from (?<name> */  
1399    
   if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  
        *ptr != '\'')  
     continue;  
1400    
   count++;  
1401    
1402    if (name == NULL && count == lorn) return count;  /*************************************************
1403    term = *ptr++;  *       Find forward referenced subpattern       *
1404    if (term == '<') term = '>';  *************************************************/
1405    thisname = ptr;  
1406    while (*ptr != term) ptr++;  /* This function scans along a pattern's text looking for capturing
1407    if (name != NULL && lorn == ptr - thisname &&  subpatterns, and counting them. If it finds a named pattern that matches the
1408        strncmp((const char *)name, (const char *)thisname, lorn) == 0)  name it is given, it returns its number. Alternatively, if the name is NULL, it
1409      return count;  returns when it reaches a given numbered subpattern. This is used for forward
1410    references to subpatterns. We used to be able to start this scan from the
1411    current compiling point, using the current count value from cd->bracount, and
1412    do it all in a single loop, but the addition of the possibility of duplicate
1413    subpattern numbers means that we have to scan from the very start, in order to
1414    take account of such duplicates, and to use a recursive function to keep track
1415    of the different types of group.
1416    
1417    Arguments:
1418      cd           compile background data
1419      name         name to seek, or NULL if seeking a numbered subpattern
1420      lorn         name length, or subpattern number if name is NULL
1421      xmode        TRUE if we are in /x mode
1422      utf8         TRUE if we are in UTF-8 mode
1423    
1424    Returns:       the number of the found subpattern, or -1 if not found
1425    */
1426    
1427    static int
1428    find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode,
1429      BOOL utf8)
1430    {
1431    uschar *ptr = (uschar *)cd->start_pattern;
1432    int count = 0;
1433    int rc;
1434    
1435    /* If the pattern does not start with an opening parenthesis, the first call
1436    to find_parens_sub() will scan right to the end (if necessary). However, if it
1437    does start with a parenthesis, find_parens_sub() will return when it hits the
1438    matching closing parens. That is why we have to have a loop. */
1439    
1440    for (;;)
1441      {
1442      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, &count);
1443      if (rc > 0 || *ptr++ == 0) break;
1444    }    }
1445    
1446  return -1;  return rc;
1447  }  }
1448    
1449    
1450    
1451    
1452  /*************************************************  /*************************************************
1453  *      Find first significant op code            *  *      Find first significant op code            *
1454  *************************************************/  *************************************************/
1455    
1456  /* This is called by several functions that scan a compiled expression looking  /* This is called by several functions that scan a compiled expression looking
1457  for a fixed first character, or an anchoring op code etc. It skips over things  for a fixed first character, or an anchoring op code etc. It skips over things
1458  that do not influence this. For some calls, a change of option is important.  that do not influence this. For some calls, it makes sense to skip negative
1459  For some calls, it makes sense to skip negative forward and all backward  forward and all backward assertions, and also the \b assertion; for others it
1460  assertions, and also the \b assertion; for others it does not.  does not.
1461    
1462  Arguments:  Arguments:
1463    code         pointer to the start of the group    code         pointer to the start of the group
   options      pointer to external options  
   optbit       the option bit whose changing is significant, or  
                  zero if none are  
1464    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1465    
1466  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1467  */  */
1468    
1469  static const uschar*  static const uschar*
1470  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const uschar *code, BOOL skipassert)
   BOOL skipassert)  
1471  {  {
1472  for (;;)  for (;;)
1473    {    {
1474    switch ((int)*code)    switch ((int)*code)
1475      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1476      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1477      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1478      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
# Line 1008  for (;;) Line 1488  for (;;)
1488    
1489      case OP_CALLOUT:      case OP_CALLOUT:
1490      case OP_CREF:      case OP_CREF:
1491        case OP_NCREF:
1492      case OP_RREF:      case OP_RREF:
1493        case OP_NRREF:
1494      case OP_DEF:      case OP_DEF:
1495      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1496      break;      break;
# Line 1024  for (;;) Line 1506  for (;;)
1506    
1507    
1508  /*************************************************  /*************************************************
1509  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1510  *************************************************/  *************************************************/
1511    
1512  /* 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,
1513  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.
1514  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
1515    temporarily terminated with OP_END when this function is called.
1516    
1517    This function is called when a backward assertion is encountered, so that if it
1518    fails, the error message can point to the correct place in the pattern.
1519    However, we cannot do this when the assertion contains subroutine calls,
1520    because they can be forward references. We solve this by remembering this case
1521    and doing the check at the end; a flag specifies which mode we are running in.
1522    
1523  Arguments:  Arguments:
1524    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1525    options  the compiling options    utf8     TRUE in UTF-8 mode
1526      atend    TRUE if called when the pattern is complete
1527      cd       the "compile data" structure
1528    
1529  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1530                 or -1 if there is no fixed length,
1531               or -2 if \C was encountered               or -2 if \C was encountered
1532                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1533  */  */
1534    
1535  static int  static int
1536  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, BOOL utf8, BOOL atend, compile_data *cd)
1537  {  {
1538  int length = -1;  int length = -1;
1539    
# Line 1053  branch, check the length against that of Line 1546  branch, check the length against that of
1546  for (;;)  for (;;)
1547    {    {
1548    int d;    int d;
1549      uschar *ce, *cs;
1550    register int op = *cc;    register int op = *cc;
   
1551    switch (op)    switch (op)
1552      {      {
1553        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1554        OP_BRA (normal non-capturing bracket) because the other variants of these
1555        opcodes are all concerned with unlimited repeated groups, which of course
1556        are not of fixed length. They will cause a -1 response from the default
1557        case of this switch. */
1558    
1559      case OP_CBRA:      case OP_CBRA:
1560      case OP_BRA:      case OP_BRA:
1561      case OP_ONCE:      case OP_ONCE:
1562        case OP_ONCE_NC:
1563      case OP_COND:      case OP_COND:
1564      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), utf8, atend, cd);
1565      if (d < 0) return d;      if (d < 0) return d;
1566      branchlength += d;      branchlength += d;
1567      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1070  for (;;) Line 1570  for (;;)
1570    
1571      /* Reached end of a branch; if it's a ket it is the end of a nested      /* Reached end of a branch; if it's a ket it is the end of a nested
1572      call. If it's ALT it is an alternation in a nested call. If it is      call. If it's ALT it is an alternation in a nested call. If it is
1573      END it's the end of the outer call. All can be handled by the same code. */      END it's the end of the outer call. All can be handled by the same code.
1574        Note that we must not include the OP_KETRxxx opcodes here, because they
1575        all imply an unlimited repeat. */
1576    
1577      case OP_ALT:      case OP_ALT:
1578      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1579      case OP_END:      case OP_END:
1580      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1581        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
# Line 1084  for (;;) Line 1584  for (;;)
1584      branchlength = 0;      branchlength = 0;
1585      break;      break;
1586    
1587        /* A true recursion implies not fixed length, but a subroutine call may
1588        be OK. If the subroutine is a forward reference, we can't deal with
1589        it until the end of the pattern, so return -3. */
1590    
1591        case OP_RECURSE:
1592        if (!atend) return -3;
1593        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1594        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1595        if (cc > cs && cc < ce) return -1;                /* Recursion */
1596        d = find_fixedlength(cs + 2, utf8, atend, cd);
1597        if (d < 0) return d;
1598        branchlength += d;
1599        cc += 1 + LINK_SIZE;
1600        break;
1601    
1602      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1603    
1604      case OP_ASSERT:      case OP_ASSERT:
# Line 1097  for (;;) Line 1612  for (;;)
1612    
1613      case OP_REVERSE:      case OP_REVERSE:
1614      case OP_CREF:      case OP_CREF:
1615        case OP_NCREF:
1616      case OP_RREF:      case OP_RREF:
1617        case OP_NRREF:
1618      case OP_DEF:      case OP_DEF:
     case OP_OPT:  
1619      case OP_CALLOUT:      case OP_CALLOUT:
1620      case OP_SOD:      case OP_SOD:
1621      case OP_SOM:      case OP_SOM:
1622        case OP_SET_SOM:
1623      case OP_EOD:      case OP_EOD:
1624      case OP_EODN:      case OP_EODN:
1625      case OP_CIRC:      case OP_CIRC:
1626        case OP_CIRCM:
1627      case OP_DOLL:      case OP_DOLL:
1628        case OP_DOLLM:
1629      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1630      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1631      cc += _pcre_OP_lengths[*cc];      cc += _pcre_OP_lengths[*cc];
# Line 1115  for (;;) Line 1634  for (;;)
1634      /* Handle literal characters */      /* Handle literal characters */
1635    
1636      case OP_CHAR:      case OP_CHAR:
1637      case OP_CHARNC:      case OP_CHARI:
1638      case OP_NOT:      case OP_NOT:
1639        case OP_NOTI:
1640      branchlength++;      branchlength++;
1641      cc += 2;      cc += 2;
1642  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1643      if ((options & PCRE_UTF8) != 0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       {  
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1644  #endif  #endif
1645      break;      break;
1646    
# Line 1134  for (;;) Line 1651  for (;;)
1651      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1652      cc += 4;      cc += 4;
1653  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1654      if ((options & PCRE_UTF8) != 0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       {  
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1655  #endif  #endif
1656      break;      break;
1657    
1658      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1659      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1660        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1661      cc += 4;      cc += 4;
1662      break;      break;
1663    
# Line 1160  for (;;) Line 1675  for (;;)
1675      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1676      case OP_WORDCHAR:      case OP_WORDCHAR:
1677      case OP_ANY:      case OP_ANY:
1678        case OP_ALLANY:
1679      branchlength++;      branchlength++;
1680      cc++;      cc++;
1681      break;      break;
# Line 1214  for (;;) Line 1730  for (;;)
1730    
1731    
1732  /*************************************************  /*************************************************
1733  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1734  *************************************************/  *************************************************/
1735    
1736  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1737  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1738    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1739    so that it can be called from pcre_study() when finding the minimum matching
1740    length.
1741    
1742  Arguments:  Arguments:
1743    code        points to start of expression    code        points to start of expression
1744    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1745    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1746    
1747  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
1748  */  */
1749    
1750  static const uschar *  const uschar *
1751  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1752  {  {
1753  for (;;)  for (;;)
1754    {    {
1755    register int c = *code;    register int c = *code;
1756    
1757    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1758    
1759    /* XCLASS is used for classes that cannot be represented just by a bit    /* XCLASS is used for classes that cannot be represented just by a bit
# Line 1242  for (;;) Line 1762  for (;;)
1762    
1763    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1764    
1765      /* Handle recursion */
1766    
1767      else if (c == OP_REVERSE)
1768        {
1769        if (number < 0) return (uschar *)code;
1770        code += _pcre_OP_lengths[c];
1771        }
1772    
1773    /* Handle capturing bracket */    /* Handle capturing bracket */
1774    
1775    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
1776               c == OP_CBRAPOS || c == OP_SCBRAPOS)
1777      {      {
1778      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
1779      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
1780      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1781      }      }
1782    
1783    /* In UTF-8 mode, opcodes that are followed by a character may be followed by    /* Otherwise, we can get the item's length from the table, except that for
1784    a multi-byte character. The length in the table is a minimum, so we have to    repeated character types, we have to test for \p and \P, which have an extra
1785    arrange to skip the extra bytes. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1786      must add in its length. */
1787    
1788    else    else
1789      {      {
1790        switch(c)
1791          {
1792          case OP_TYPESTAR:
1793          case OP_TYPEMINSTAR:
1794          case OP_TYPEPLUS:
1795          case OP_TYPEMINPLUS:
1796          case OP_TYPEQUERY:
1797          case OP_TYPEMINQUERY:
1798          case OP_TYPEPOSSTAR:
1799          case OP_TYPEPOSPLUS:
1800          case OP_TYPEPOSQUERY:
1801          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1802          break;
1803    
1804          case OP_TYPEUPTO:
1805          case OP_TYPEMINUPTO:
1806          case OP_TYPEEXACT:
1807          case OP_TYPEPOSUPTO:
1808          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1809          break;
1810    
1811          case OP_MARK:
1812          case OP_PRUNE_ARG:
1813          case OP_SKIP_ARG:
1814          code += code[1];
1815          break;
1816    
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];      code += _pcre_OP_lengths[c];
1825    
1826      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1827      a multi-byte character. The length in the table is a minimum, so we have to
1828      arrange to skip the extra bytes. */
1829    
1830  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1831      if (utf8) switch(c)      if (utf8) switch(c)
1832        {        {
1833        case OP_CHAR:        case OP_CHAR:
1834        case OP_CHARNC:        case OP_CHARI:
1835        case OP_EXACT:        case OP_EXACT:
1836          case OP_EXACTI:
1837        case OP_UPTO:        case OP_UPTO:
1838          case OP_UPTOI:
1839        case OP_MINUPTO:        case OP_MINUPTO:
1840          case OP_MINUPTOI:
1841        case OP_POSUPTO:        case OP_POSUPTO:
1842          case OP_POSUPTOI:
1843        case OP_STAR:        case OP_STAR:
1844          case OP_STARI:
1845        case OP_MINSTAR:        case OP_MINSTAR:
1846          case OP_MINSTARI:
1847        case OP_POSSTAR:        case OP_POSSTAR:
1848          case OP_POSSTARI:
1849        case OP_PLUS:        case OP_PLUS:
1850          case OP_PLUSI:
1851        case OP_MINPLUS:        case OP_MINPLUS:
1852          case OP_MINPLUSI:
1853        case OP_POSPLUS:        case OP_POSPLUS:
1854          case OP_POSPLUSI:
1855        case OP_QUERY:        case OP_QUERY:
1856          case OP_QUERYI:
1857        case OP_MINQUERY:        case OP_MINQUERY:
1858          case OP_MINQUERYI:
1859        case OP_POSQUERY:        case OP_POSQUERY:
1860          case OP_POSQUERYI:
1861        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1862        break;        break;
1863        }        }
1864    #else
1865        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1866  #endif  #endif
1867      }      }
1868    }    }
# Line 1315  for (;;) Line 1899  for (;;)
1899    
1900    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1901    
1902    /* Otherwise, we get the item's length from the table. In UTF-8 mode, opcodes    /* Otherwise, we can get the item's length from the table, except that for
1903    that are followed by a character may be followed by a multi-byte character.    repeated character types, we have to test for \p and \P, which have an extra
1904    The length in the table is a minimum, so we have to arrange to skip the extra    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1905    bytes. */    must add in its length. */
1906    
1907    else    else
1908      {      {
1909        switch(c)
1910          {
1911          case OP_TYPESTAR:
1912          case OP_TYPEMINSTAR:
1913          case OP_TYPEPLUS:
1914          case OP_TYPEMINPLUS:
1915          case OP_TYPEQUERY:
1916          case OP_TYPEMINQUERY:
1917          case OP_TYPEPOSSTAR:
1918          case OP_TYPEPOSPLUS:
1919          case OP_TYPEPOSQUERY:
1920          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1921          break;
1922    
1923          case OP_TYPEPOSUPTO:
1924          case OP_TYPEUPTO:
1925          case OP_TYPEMINUPTO:
1926          case OP_TYPEEXACT:
1927          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1928          break;
1929    
1930          case OP_MARK:
1931          case OP_PRUNE_ARG:
1932          case OP_SKIP_ARG:
1933          code += code[1];
1934          break;
1935    
1936          case OP_THEN_ARG:
1937          code += code[1];
1938          break;
1939          }
1940    
1941        /* Add in the fixed length from the table */
1942    
1943      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1944    
1945        /* In UTF-8 mode, opcodes that are followed by a character may be followed
1946        by a multi-byte character. The length in the table is a minimum, so we have
1947        to arrange to skip the extra bytes. */
1948    
1949  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1950      if (utf8) switch(c)      if (utf8) switch(c)
1951        {        {
1952        case OP_CHAR:        case OP_CHAR:
1953        case OP_CHARNC:        case OP_CHARI:
1954        case OP_EXACT:        case OP_EXACT:
1955          case OP_EXACTI:
1956        case OP_UPTO:        case OP_UPTO:
1957          case OP_UPTOI:
1958        case OP_MINUPTO:        case OP_MINUPTO:
1959          case OP_MINUPTOI:
1960        case OP_POSUPTO:        case OP_POSUPTO:
1961          case OP_POSUPTOI:
1962        case OP_STAR:        case OP_STAR:
1963          case OP_STARI:
1964        case OP_MINSTAR:        case OP_MINSTAR:
1965          case OP_MINSTARI:
1966        case OP_POSSTAR:        case OP_POSSTAR:
1967          case OP_POSSTARI:
1968        case OP_PLUS:        case OP_PLUS:
1969          case OP_PLUSI:
1970        case OP_MINPLUS:        case OP_MINPLUS:
1971          case OP_MINPLUSI:
1972        case OP_POSPLUS:        case OP_POSPLUS:
1973          case OP_POSPLUSI:
1974        case OP_QUERY:        case OP_QUERY:
1975          case OP_QUERYI:
1976        case OP_MINQUERY:        case OP_MINQUERY:
1977          case OP_MINQUERYI:
1978        case OP_POSQUERY:        case OP_POSQUERY:
1979          case OP_POSQUERYI:
1980        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1981        break;        break;
1982        }        }
1983    #else
1984        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1985  #endif  #endif
1986      }      }
1987    }    }
# Line 1359  for (;;) Line 1997  for (;;)
1997  can match the empty string or not. It is called from could_be_empty()  can match the empty string or not. It is called from could_be_empty()
1998  below and from compile_branch() when checking for an unlimited repeat of a  below and from compile_branch() when checking for an unlimited repeat of a
1999  group that can match nothing. Note that first_significant_code() skips over  group that can match nothing. Note that first_significant_code() skips over
2000  assertions. If we hit an unclosed bracket, we return "empty" - this means we've  backward and negative forward assertions when its final argument is TRUE. If we
2001  struck an inner bracket whose current branch will already have been scanned.  hit an unclosed bracket, we return "empty" - this means we've struck an inner
2002    bracket whose current branch will already have been scanned.
2003    
2004  Arguments:  Arguments:
2005    code        points to start of search    code        points to start of search
2006    endcode     points to where to stop    endcode     points to where to stop
2007    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
2008      cd          contains pointers to tables etc.
2009    
2010  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2011  */  */
2012    
2013  static BOOL  static BOOL
2014  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
2015      compile_data *cd)
2016  {  {
2017  register int c;  register int c;
2018  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], TRUE);
2019       code < endcode;       code < endcode;
2020       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], TRUE))
2021    {    {
2022    const uschar *ccode;    const uschar *ccode;
2023    
2024    c = *code;    c = *code;
2025    
2026    /* Groups with zero repeats can of course be empty; skip them. */    /* Skip over forward assertions; the other assertions are skipped by
2027      first_significant_code() with a TRUE final argument. */
2028    
2029    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_ASSERT)
2030      {      {
     code += _pcre_OP_lengths[c];  
2031      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
2032      c = *code;      c = *code;
2033      continue;      continue;
2034      }      }
2035    
2036    /* For other groups, scan the branches. */    /* For a recursion/subroutine call, if its end has been reached, which
2037      implies a backward reference subroutine call, we can scan it. If it's a
2038      forward reference subroutine call, we can't. To detect forward reference
2039      we have to scan up the list that is kept in the workspace. This function is
2040      called only when doing the real compile, not during the pre-compile that
2041      measures the size of the compiled pattern. */
2042    
2043    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE)    if (c == OP_RECURSE)
2044      {      {
2045        const uschar *scode;
2046      BOOL empty_branch;      BOOL empty_branch;
     if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */  
2047    
2048      /* Scan a closed bracket */      /* Test for forward reference */
2049    
2050        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
2051          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
2052    
2053        /* Not a forward reference, test for completed backward reference */
2054    
2055      empty_branch = FALSE;      empty_branch = FALSE;
2056        scode = cd->start_code + GET(code, 1);
2057        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2058    
2059        /* Completed backwards reference */
2060    
2061      do      do
2062        {        {
2063        if (!empty_branch && could_be_empty_branch(code, endcode, utf8))        if (could_be_empty_branch(scode, endcode, utf8, cd))
2064            {
2065          empty_branch = TRUE;          empty_branch = TRUE;
2066            break;
2067            }
2068          scode += GET(scode, 1);
2069          }
2070        while (*scode == OP_ALT);
2071    
2072        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2073        continue;
2074        }
2075    
2076      /* Groups with zero repeats can of course be empty; skip them. */
2077    
2078      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2079          c == OP_BRAPOSZERO)
2080        {
2081        code += _pcre_OP_lengths[c];
2082        do code += GET(code, 1); while (*code == OP_ALT);
2083        c = *code;
2084        continue;
2085        }
2086    
2087      /* A nested group that is already marked as "could be empty" can just be
2088      skipped. */
2089    
2090      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2091          c == OP_SCBRA || c == OP_SCBRAPOS)
2092        {
2093        do code += GET(code, 1); while (*code == OP_ALT);
2094        c = *code;
2095        continue;
2096        }
2097    
2098      /* For other groups, scan the branches. */
2099    
2100      if (c == OP_BRA  || c == OP_BRAPOS ||
2101          c == OP_CBRA || c == OP_CBRAPOS ||
2102          c == OP_ONCE || c == OP_ONCE_NC ||
2103          c == OP_COND)
2104        {
2105        BOOL empty_branch;
2106        if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
2107    
2108        /* If a conditional group has only one branch, there is a second, implied,
2109        empty branch, so just skip over the conditional, because it could be empty.
2110        Otherwise, scan the individual branches of the group. */
2111    
2112        if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
2113        code += GET(code, 1);        code += GET(code, 1);
2114        else
2115          {
2116          empty_branch = FALSE;
2117          do
2118            {
2119            if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
2120              empty_branch = TRUE;
2121            code += GET(code, 1);
2122            }
2123          while (*code == OP_ALT);
2124          if (!empty_branch) return FALSE;   /* All branches are non-empty */
2125        }        }
2126      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
2127      c = *code;      c = *code;
2128      continue;      continue;
2129      }      }
# Line 1418  for (code = first_significant_code(code Line 2132  for (code = first_significant_code(code
2132    
2133    switch (c)    switch (c)
2134      {      {
2135      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
2136        cannot be represented just by a bit map. This includes negated single
2137        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
2138        actual length is stored in the compiled code, so we must update "code"
2139        here. */
2140    
2141  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2142      case OP_XCLASS:      case OP_XCLASS:
2143      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
2144      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
2145  #endif  #endif
2146    
# Line 1466  for (code = first_significant_code(code Line 2184  for (code = first_significant_code(code
2184      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2185      case OP_WORDCHAR:      case OP_WORDCHAR:
2186      case OP_ANY:      case OP_ANY:
2187        case OP_ALLANY:
2188      case OP_ANYBYTE:      case OP_ANYBYTE:
2189      case OP_CHAR:      case OP_CHAR:
2190      case OP_CHARNC:      case OP_CHARI:
2191      case OP_NOT:      case OP_NOT:
2192        case OP_NOTI:
2193      case OP_PLUS:      case OP_PLUS:
2194      case OP_MINPLUS:      case OP_MINPLUS:
2195      case OP_POSPLUS:      case OP_POSPLUS:
# Line 1484  for (code = first_significant_code(code Line 2204  for (code = first_significant_code(code
2204      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2205      return FALSE;      return FALSE;
2206    
2207        /* These are going to continue, as they may be empty, but we have to
2208        fudge the length for the \p and \P cases. */
2209    
2210        case OP_TYPESTAR:
2211        case OP_TYPEMINSTAR:
2212        case OP_TYPEPOSSTAR:
2213        case OP_TYPEQUERY:
2214        case OP_TYPEMINQUERY:
2215        case OP_TYPEPOSQUERY:
2216        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2217        break;
2218    
2219        /* Same for these */
2220    
2221        case OP_TYPEUPTO:
2222        case OP_TYPEMINUPTO:
2223        case OP_TYPEPOSUPTO:
2224        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
2225        break;
2226    
2227      /* End of branch */      /* End of branch */
2228    
2229      case OP_KET:      case OP_KET:
2230      case OP_KETRMAX:      case OP_KETRMAX:
2231      case OP_KETRMIN:      case OP_KETRMIN:
2232        case OP_KETRPOS:
2233      case OP_ALT:      case OP_ALT:
2234      return TRUE;      return TRUE;
2235    
# Line 1497  for (code = first_significant_code(code Line 2238  for (code = first_significant_code(code
2238    
2239  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2240      case OP_STAR:      case OP_STAR:
2241        case OP_STARI:
2242      case OP_MINSTAR:      case OP_MINSTAR:
2243        case OP_MINSTARI:
2244      case OP_POSSTAR:      case OP_POSSTAR:
2245        case OP_POSSTARI:
2246      case OP_QUERY:      case OP_QUERY:
2247        case OP_QUERYI:
2248      case OP_MINQUERY:      case OP_MINQUERY:
2249        case OP_MINQUERYI:
2250      case OP_POSQUERY:      case OP_POSQUERY:
2251        case OP_POSQUERYI:
2252        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2253        break;
2254    
2255      case OP_UPTO:      case OP_UPTO:
2256        case OP_UPTOI:
2257      case OP_MINUPTO:      case OP_MINUPTO:
2258        case OP_MINUPTOI:
2259      case OP_POSUPTO:      case OP_POSUPTO:
2260      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_POSUPTOI:
2261        if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2262        break;
2263    #endif
2264    
2265        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2266        string. */
2267    
2268        case OP_MARK:
2269        case OP_PRUNE_ARG:
2270        case OP_SKIP_ARG:
2271        code += code[1];
2272        break;
2273    
2274        case OP_THEN_ARG:
2275        code += code[1];
2276        break;
2277    
2278        /* None of the remaining opcodes are required to match a character. */
2279    
2280        default:
2281      break;      break;
 #endif  
2282      }      }
2283    }    }
2284    
# Line 1524  return TRUE; Line 2295  return TRUE;
2295  the current branch of the current pattern to see if it could match the empty  the current branch of the current pattern to see if it could match the empty
2296  string. If it could, we must look outwards for branches at other levels,  string. If it could, we must look outwards for branches at other levels,
2297  stopping when we pass beyond the bracket which is the subject of the recursion.  stopping when we pass beyond the bracket which is the subject of the recursion.
2298    This function is called only during the real compile, not during the
2299    pre-compile.
2300    
2301  Arguments:  Arguments:
2302    code        points to start of the recursion    code        points to start of the recursion
2303    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2304    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2305    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2306      cd          pointers to tables etc
2307    
2308  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2309  */  */
2310    
2311  static BOOL  static BOOL
2312  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2313    BOOL utf8)    BOOL utf8, compile_data *cd)
2314  {  {
2315  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2316    {    {
2317    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2318        return FALSE;
2319    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2320    }    }
2321  return TRUE;  return TRUE;
# Line 1553  return TRUE; Line 2328  return TRUE;
2328  *************************************************/  *************************************************/
2329    
2330  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2331  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
2332  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2333  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2334    
2335    Originally, this function only recognized a sequence of letters between the
2336    terminators, but it seems that Perl recognizes any sequence of characters,
2337    though of course unknown POSIX names are subsequently rejected. Perl gives an
2338    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2339    didn't consider this to be a POSIX class. Likewise for [:1234:].
2340    
2341    The problem in trying to be exactly like Perl is in the handling of escapes. We
2342    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2343    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2344    below handles the special case of \], but does not try to do any other escape
2345    processing. This makes it different from Perl for cases such as [:l\ower:]
2346    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2347    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2348    I think.
2349    
2350    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2351    It seems that the appearance of a nested POSIX class supersedes an apparent
2352    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2353    a digit.
2354    
2355    In Perl, unescaped square brackets may also appear as part of class names. For
2356    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2357    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2358    seem right at all. PCRE does not allow closing square brackets in POSIX class
2359    names.
2360    
2361  Argument:  Arguments:
2362    ptr      pointer to the initial [    ptr      pointer to the initial [
2363    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2364    
2365  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2366  */  */
2367    
2368  static BOOL  static BOOL
2369  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
2370  {  {
2371  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2372  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2373  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2374    {    {
2375    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2376    return TRUE;      ptr++;
2377      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2378      else
2379        {
2380        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2381          {
2382          *endptr = ptr;
2383          return TRUE;
2384          }
2385        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2386             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2387              ptr[1] == CHAR_EQUALS_SIGN) &&
2388            check_posix_syntax(ptr, endptr))
2389          return FALSE;
2390        }
2391    }    }
2392  return FALSE;  return FALSE;
2393  }  }
# Line 1600  Returns: a value representing the na Line 2412  Returns: a value representing the na
2412  static int  static int
2413  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
2414  {  {
2415    const char *pn = posix_names;
2416  register int yield = 0;  register int yield = 0;
2417  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2418    {    {
2419    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2420      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
2421      pn += posix_name_lengths[yield] + 1;
2422    yield++;    yield++;
2423    }    }
2424  return -1;  return -1;
# Line 1619  return -1; Line 2433  return -1;
2433  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2434  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2435  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
2436  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
2437  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
2438  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
2439  offsets adjusted. That one of the jobs of this function. Before it is called,  have their offsets adjusted. That one of the jobs of this function. Before it
2440  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2441    OP_END.
2442    
2443  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2444  recursions and subroutine calls. It must also check the list of such references  recursions and subroutine calls. It must also check the list of such references
# Line 1646  adjust_recurse(uschar *group, int adjust Line 2461  adjust_recurse(uschar *group, int adjust
2461    uschar *save_hwm)    uschar *save_hwm)
2462  {  {
2463  uschar *ptr = group;  uschar *ptr = group;
2464    
2465  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
2466    {    {
2467    int offset;    int offset;
# Line 1699  auto_callout(uschar *code, const uschar Line 2515  auto_callout(uschar *code, const uschar
2515  {  {
2516  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2517  *code++ = 255;  *code++ = 255;
2518  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2519  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2520  return code + 2*LINK_SIZE;  return code + 2*LINK_SIZE;
2521  }  }
2522    
# Line 1725  Returns: nothing Line 2541  Returns: nothing
2541  static void  static void
2542  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)
2543  {  {
2544  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2545  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2546  }  }
2547    
# Line 1757  get_othercase_range(unsigned int *cptr, Line 2573  get_othercase_range(unsigned int *cptr,
2573  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2574    
2575  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2576    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2577    
2578  if (c > d) return FALSE;  if (c > d) return FALSE;
2579    
# Line 1766  next = othercase + 1; Line 2582  next = othercase + 1;
2582    
2583  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2584    {    {
2585    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2586    next++;    next++;
2587    }    }
2588    
# Line 1775  for (++c; c <= d; c++) Line 2591  for (++c; c <= d; c++)
2591    
2592  return TRUE;  return TRUE;
2593  }  }
2594    
2595    
2596    
2597    /*************************************************
2598    *        Check a character and a property        *
2599    *************************************************/
2600    
2601    /* This function is called by check_auto_possessive() when a property item
2602    is adjacent to a fixed character.
2603    
2604    Arguments:
2605      c            the character
2606      ptype        the property type
2607      pdata        the data for the type
2608      negated      TRUE if it's a negated property (\P or \p{^)
2609    
2610    Returns:       TRUE if auto-possessifying is OK
2611    */
2612    
2613    static BOOL
2614    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2615    {
2616    const ucd_record *prop = GET_UCD(c);
2617    switch(ptype)
2618      {
2619      case PT_LAMP:
2620      return (prop->chartype == ucp_Lu ||
2621              prop->chartype == ucp_Ll ||
2622              prop->chartype == ucp_Lt) == negated;
2623    
2624      case PT_GC:
2625      return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;
2626    
2627      case PT_PC:
2628      return (pdata == prop->chartype) == negated;
2629    
2630      case PT_SC:
2631      return (pdata == prop->script) == negated;
2632    
2633      /* These are specials */
2634    
2635      case PT_ALNUM:
2636      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2637              _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;
2638    
2639      case PT_SPACE:    /* Perl space */
2640      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2641              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2642              == negated;
2643    
2644      case PT_PXSPACE:  /* POSIX space */
2645      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2646              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2647              c == CHAR_FF || c == CHAR_CR)
2648              == negated;
2649    
2650      case PT_WORD:
2651      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2652              _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2653              c == CHAR_UNDERSCORE) == negated;
2654      }
2655    return FALSE;
2656    }
2657  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2658    
2659    
# Line 1788  whether the next thing could possibly ma Line 2667  whether the next thing could possibly ma
2667  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
2668    
2669  Arguments:  Arguments:
2670    op_code       the repeated op code    previous      pointer to the repeated opcode
   this          data for this item, depends on the opcode  
2671    utf8          TRUE in UTF-8 mode    utf8          TRUE in UTF-8 mode
   utf8_char     used for utf8 character bytes, NULL if not relevant  
2672    ptr           next character in pattern    ptr           next character in pattern
2673    options       options bits    options       options bits
2674    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 1800  Returns: TRUE if possessifying is Line 2677  Returns: TRUE if possessifying is
2677  */  */
2678    
2679  static BOOL  static BOOL
2680  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
2681    const uschar *ptr, int options, compile_data *cd)    int options, compile_data *cd)
2682  {  {
2683  int next;  int c, next;
2684    int op_code = *previous++;
2685    
2686  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
2687    
# Line 1812  if ((options & PCRE_EXTENDED) != 0) Line 2690  if ((options & PCRE_EXTENDED) != 0)
2690    for (;;)    for (;;)
2691      {      {
2692      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2693      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2694        {        {
2695        while (*(++ptr) != 0)        ptr++;
2696          while (*ptr != 0)
2697            {
2698          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2699            ptr++;
2700    #ifdef SUPPORT_UTF8
2701            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2702    #endif
2703            }
2704        }        }
2705      else break;      else break;
2706      }      }
# Line 1824  if ((options & PCRE_EXTENDED) != 0) Line 2709  if ((options & PCRE_EXTENDED) != 0)
2709  /* If the next item is one that we can handle, get its value. A non-negative  /* If the next item is one that we can handle, get its value. A non-negative
2710  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2711    
2712  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2713    {    {
2714    int temperrorcode = 0;    int temperrorcode = 0;
2715    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
# Line 1849  if ((options & PCRE_EXTENDED) != 0) Line 2734  if ((options & PCRE_EXTENDED) != 0)
2734    for (;;)    for (;;)
2735      {      {
2736      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2737      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2738        {        {
2739        while (*(++ptr) != 0)        ptr++;
2740          while (*ptr != 0)
2741            {
2742          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2743            ptr++;
2744    #ifdef SUPPORT_UTF8
2745            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2746    #endif
2747            }
2748        }        }
2749      else break;      else break;
2750      }      }
# Line 1860  if ((options & PCRE_EXTENDED) != 0) Line 2752  if ((options & PCRE_EXTENDED) != 0)
2752    
2753  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
2754    
2755  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2756    return FALSE;    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2757        return FALSE;
 /* Now compare the next item with the previous opcode. If the previous is a  
 positive single character match, "item" either contains the character or, if  
 "item" is greater than 127 in utf8 mode, the character's bytes are in  
 utf8_char. */  
   
2758    
2759  /* Handle cases when the next item is a character. */  /* Now compare the next item with the previous opcode. First, handle cases when
2760    the next item is a character. */
2761    
2762  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
2763    {    {
2764    case OP_CHAR:    case OP_CHAR:
2765  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2766    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2767    #else
2768      c = *previous;
2769  #endif  #endif
2770    return item != next;    return c != next;
2771    
2772    /* For CHARNC (caseless character) we must check the other case. If we have    /* For CHARI (caseless character) we must check the other case. If we have
2773    Unicode property support, we can use it to test the other case of    Unicode property support, we can use it to test the other case of
2774    high-valued characters. */    high-valued characters. */
2775    
2776    case OP_CHARNC:    case OP_CHARI:
2777  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2778    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2779    #else
2780      c = *previous;
2781  #endif  #endif
2782    if (item == next) return FALSE;    if (c == next) return FALSE;
2783  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2784    if (utf8)    if (utf8)
2785      {      {
2786      unsigned int othercase;      unsigned int othercase;
2787      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2788  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2789      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
2790  #else  #else
2791      othercase = NOTACHAR;      othercase = NOTACHAR;
2792  #endif  #endif
2793      return (unsigned int)item != othercase;      return (unsigned int)c != othercase;
2794      }      }
2795    else    else
2796  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2797    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2798    
2799    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT and OP_NOTI, the data is always a single-byte character. These
2800      opcodes are not used for multi-byte characters, because they are coded using
2801      an XCLASS instead. */
2802    
2803    case OP_NOT:    case OP_NOT:
2804    if (next < 0) return FALSE;  /* Not a character */    return (c = *previous) == next;
2805    if (item == next) return TRUE;  
2806    if ((options & PCRE_CASELESS) == 0) return FALSE;    case OP_NOTI:
2807      if ((c = *previous) == next) return TRUE;
2808  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2809    if (utf8)    if (utf8)
2810      {      {
2811      unsigned int othercase;      unsigned int othercase;
2812      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2813  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2814      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
2815  #else  #else
2816      othercase = NOTACHAR;      othercase = NOTACHAR;
2817  #endif  #endif
2818      return (unsigned int)item == othercase;      return (unsigned int)c == othercase;
2819      }      }
2820    else    else
2821  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2822    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2823    
2824      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2825      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
2826    
2827    case OP_DIGIT:    case OP_DIGIT:
2828    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
# Line 1944  if (next >= 0) switch(op_code) Line 2842  if (next >= 0) switch(op_code)
2842    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
2843    return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;    return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;
2844    
2845      case OP_HSPACE:
2846      case OP_NOT_HSPACE:
2847      switch(next)
2848        {
2849        case 0x09:
2850        case 0x20:
2851        case 0xa0:
2852        case 0x1680:
2853        case 0x180e:
2854        case 0x2000:
2855        case 0x2001:
2856        case 0x2002:
2857        case 0x2003:
2858        case 0x2004:
2859        case 0x2005:
2860        case 0x2006:
2861        case 0x2007:
2862        case 0x2008:
2863        case 0x2009:
2864        case 0x200A:
2865        case 0x202f:
2866        case 0x205f:
2867        case 0x3000:
2868        return op_code == OP_NOT_HSPACE;
2869        default:
2870        return op_code != OP_NOT_HSPACE;
2871        }
2872    
2873      case OP_ANYNL:
2874      case OP_VSPACE:
2875      case OP_NOT_VSPACE:
2876      switch(next)
2877        {
2878        case 0x0a:
2879        case 0x0b:
2880        case 0x0c:
2881        case 0x0d:
2882        case 0x85:
2883        case 0x2028:
2884        case 0x2029:
2885        return op_code == OP_NOT_VSPACE;
2886        default:
2887        return op_code != OP_NOT_VSPACE;
2888        }
2889    
2890    #ifdef SUPPORT_UCP
2891      case OP_PROP:
2892      return check_char_prop(next, previous[0], previous[1], FALSE);
2893    
2894      case OP_NOTPROP:
2895      return check_char_prop(next, previous[0], previous[1], TRUE);
2896    #endif
2897    
2898    default:    default:
2899    return FALSE;    return FALSE;
2900    }    }
2901    
2902    
2903  /* Handle the case when the next item is \d, \s, etc. */  /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP
2904    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
2905    generated only when PCRE_UCP is *not* set, that is, when only ASCII
2906    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
2907    replaced by OP_PROP codes when PCRE_UCP is set. */
2908    
2909  switch(op_code)  switch(op_code)
2910    {    {
2911    case OP_CHAR:    case OP_CHAR:
2912    case OP_CHARNC:    case OP_CHARI:
2913  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2914    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2915    #else
2916      c = *previous;
2917  #endif  #endif
2918    switch(-next)    switch(-next)
2919      {      {
2920      case ESC_d:      case ESC_d:
2921      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
2922    
2923      case ESC_D:      case ESC_D:
2924      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
2925    
2926      case ESC_s:      case ESC_s:
2927      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
2928    
2929      case ESC_S:      case ESC_S:
2930      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
2931    
2932      case ESC_w:      case ESC_w:
2933      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
2934    
2935      case ESC_W:      case ESC_W:
2936      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
2937    
2938        case ESC_h:
2939        case ESC_H:
2940        switch(c)
2941          {
2942          case 0x09:
2943          case 0x20:
2944          case 0xa0:
2945          case 0x1680:
2946          case 0x180e:
2947          case 0x2000:
2948          case 0x2001:
2949          case 0x2002:
2950          case 0x2003:
2951          case 0x2004:
2952          case 0x2005:
2953          case 0x2006:
2954          case 0x2007:
2955          case 0x2008:
2956          case 0x2009:
2957          case 0x200A:
2958          case 0x202f:
2959          case 0x205f:
2960          case 0x3000:
2961          return -next != ESC_h;
2962          default:
2963          return -next == ESC_h;
2964          }
2965    
2966        case ESC_v:
2967        case ESC_V:
2968        switch(c)
2969          {
2970          case 0x0a:
2971          case 0x0b:
2972          case 0x0c:
2973          case 0x0d:
2974          case 0x85:
2975          case 0x2028:
2976          case 0x2029:
2977          return -next != ESC_v;
2978          default:
2979          return -next == ESC_v;
2980          }
2981    
2982        /* When PCRE_UCP is set, these values get generated for \d etc. Find
2983        their substitutions and process them. The result will always be either
2984        -ESC_p or -ESC_P. Then fall through to process those values. */
2985    
2986    #ifdef SUPPORT_UCP
2987        case ESC_du:
2988        case ESC_DU:
2989        case ESC_wu:
2990        case ESC_WU:
2991        case ESC_su:
2992        case ESC_SU:
2993          {
2994          int temperrorcode = 0;
2995          ptr = substitutes[-next - ESC_DU];
2996          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
2997          if (temperrorcode != 0) return FALSE;
2998          ptr++;    /* For compatibility */
2999          }
3000        /* Fall through */
3001    
3002        case ESC_p:
3003        case ESC_P:
3004          {
3005          int ptype, pdata, errorcodeptr;
3006          BOOL negated;
3007    
3008          ptr--;      /* Make ptr point at the p or P */
3009          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
3010          if (ptype < 0) return FALSE;
3011          ptr++;      /* Point past the final curly ket */
3012    
3013          /* If the property item is optional, we have to give up. (When generated
3014          from \d etc by PCRE_UCP, this test will have been applied much earlier,
3015          to the original \d etc. At this point, ptr will point to a zero byte. */
3016    
3017          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3018            strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3019              return FALSE;
3020    
3021          /* Do the property check. */
3022    
3023          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
3024          }
3025    #endif
3026    
3027      default:      default:
3028      return FALSE;      return FALSE;
3029      }      }
3030    
3031      /* In principle, support for Unicode properties should be integrated here as
3032      well. It means re-organizing the above code so as to get hold of the property
3033      values before switching on the op-code. However, I wonder how many patterns
3034      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
3035      these op-codes are never generated.) */
3036    
3037    case OP_DIGIT:    case OP_DIGIT:
3038    return next == -ESC_D || next == -ESC_s || next == -ESC_W;    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
3039             next == -ESC_h || next == -ESC_v || next == -ESC_R;
3040    
3041    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
3042    return next == -ESC_d;    return next == -ESC_d;
3043    
3044    case OP_WHITESPACE:    case OP_WHITESPACE:
3045    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
3046    
3047    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
3048    return next == -ESC_s;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
3049    
3050      case OP_HSPACE:
3051      return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
3052             next == -ESC_w || next == -ESC_v || next == -ESC_R;
3053    
3054      case OP_NOT_HSPACE:
3055      return next == -ESC_h;
3056    
3057      /* Can't have \S in here because VT matches \S (Perl anomaly) */
3058      case OP_ANYNL:
3059      case OP_VSPACE:
3060      return next == -ESC_V || next == -ESC_d || next == -ESC_w;
3061    
3062      case OP_NOT_VSPACE:
3063      return next == -ESC_v || next == -ESC_R;
3064    
3065    case OP_WORDCHAR:    case OP_WORDCHAR:
3066    return next == -ESC_W || next == -ESC_s;    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
3067             next == -ESC_v || next == -ESC_R;
3068    
3069    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
3070    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2027  Arguments: Line 3096  Arguments:
3096    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
3097    reqbyteptr     set to the last literal character required, else < 0    reqbyteptr     set to the last literal character required, else < 0
3098    bcptr          points to current branch chain    bcptr          points to current branch chain
3099      cond_depth     conditional nesting depth
3100    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
3101    lengthptr      NULL during the real compile phase    lengthptr      NULL during the real compile phase
3102                   points to length accumulator during pre-compile phase                   points to length accumulator during pre-compile phase
# Line 2038  Returns: TRUE on success Line 3108  Returns: TRUE on success
3108  static BOOL  static BOOL
3109  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
3110    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
3111    compile_data *cd, int *lengthptr)    int cond_depth, compile_data *cd, int *lengthptr)
3112  {  {
3113  int repeat_type, op_type;  int repeat_type, op_type;
3114  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
# Line 2047  int greedy_default, greedy_non_default; Line 3117  int greedy_default, greedy_non_default;
3117  int firstbyte, reqbyte;  int firstbyte, reqbyte;
3118  int zeroreqbyte, zerofirstbyte;  int zeroreqbyte, zerofirstbyte;
3119  int req_caseopt, reqvary, tempreqvary;  int req_caseopt, reqvary, tempreqvary;
3120  int options = *optionsptr;  int options = *optionsptr;               /* May change dynamically */
3121  int after_manual_callout = 0;  int after_manual_callout = 0;
3122  int length_prevgroup = 0;  int length_prevgroup = 0;
3123  register int c;  register int c;
# Line 2059  BOOL inescq = FALSE; Line 3129  BOOL inescq = FALSE;
3129  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
3130  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
3131  const uschar *tempptr;  const uschar *tempptr;
3132    const uschar *nestptr = NULL;
3133  uschar *previous = NULL;  uschar *previous = NULL;
3134  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
3135  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
3136  uschar classbits[32];  uschar classbits[32];
3137    
3138    /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3139    must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3140    dynamically as we process the pattern. */
3141    
3142  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3143  BOOL class_utf8;  BOOL class_utf8;
3144  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
3145  uschar *class_utf8data;  uschar *class_utf8data;
3146    uschar *class_utf8data_base;
3147  uschar utf8_char[6];  uschar utf8_char[6];
3148  #else  #else
3149  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
 uschar *utf8_char = NULL;  
3150  #endif  #endif
3151    
3152  #ifdef DEBUG  #ifdef PCRE_DEBUG
3153  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
3154  #endif  #endif
3155    
# Line 2107  req_caseopt = ((options & PCRE_CASELESS) Line 3182  req_caseopt = ((options & PCRE_CASELESS)
3182  for (;; ptr++)  for (;; ptr++)
3183    {    {
3184    BOOL negate_class;    BOOL negate_class;
3185      BOOL should_flip_negation;
3186    BOOL possessive_quantifier;    BOOL possessive_quantifier;
3187    BOOL is_quantifier;    BOOL is_quantifier;
3188    BOOL is_recurse;    BOOL is_recurse;
3189      BOOL reset_bracount;
3190    int class_charcount;    int class_charcount;
3191    int class_lastchar;    int class_lastchar;
3192    int newoptions;    int newoptions;
# Line 2120  for (;; ptr++) Line 3197  for (;; ptr++)
3197    int subfirstbyte;    int subfirstbyte;
3198    int terminator;    int terminator;
3199    int mclength;    int mclength;
3200      int tempbracount;
3201    uschar mcbuffer[8];    uschar mcbuffer[8];
3202    
3203    /* Get next byte in the pattern */    /* Get next byte in the pattern */
3204    
3205    c = *ptr;    c = *ptr;
3206    
3207      /* If we are at the end of a nested substitution, revert to the outer level
3208      string. Nesting only happens one level deep. */
3209    
3210      if (c == 0 && nestptr != NULL)
3211        {
3212        ptr = nestptr;
3213        nestptr = NULL;
3214        c = *ptr;
3215        }
3216    
3217    /* If we are in the pre-compile phase, accumulate the length used for the    /* If we are in the pre-compile phase, accumulate the length used for the
3218    previous cycle of this loop. */    previous cycle of this loop. */
3219    
3220    if (lengthptr != NULL)    if (lengthptr != NULL)
3221      {      {
3222  #ifdef DEBUG  #ifdef PCRE_DEBUG
3223      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3224  #endif  #endif
3225      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
3226        {        {
3227        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3228        goto FAILED;        goto FAILED;
# Line 2147  for (;; ptr++) Line 3235  for (;; ptr++)
3235      */      */
3236    
3237      if (code < last_code) code = last_code;      if (code < last_code) code = last_code;
3238      *lengthptr += code - last_code;  
3239        /* Paranoid check for integer overflow */
3240    
3241        if (OFLOW_MAX - *lengthptr < code - last_code)
3242          {
3243          *errorcodeptr = ERR20;
3244          goto FAILED;
3245          }
3246    
3247        *lengthptr += (int)(code - last_code);
3248      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));
3249    
3250      /* If "previous" is set and it is not at the start of the work space, move      /* If "previous" is set and it is not at the start of the work space, move
# Line 2174  for (;; ptr++) Line 3271  for (;; ptr++)
3271    /* In the real compile phase, just check the workspace used by the forward    /* In the real compile phase, just check the workspace used by the forward
3272    reference list. */    reference list. */
3273    
3274    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
3275      {      {
3276      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3277      goto FAILED;      goto FAILED;
# Line 2184  for (;; ptr++) Line 3281  for (;; ptr++)
3281    
3282    if (inescq && c != 0)    if (inescq && c != 0)
3283      {      {
3284      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3285        {        {
3286        inescq = FALSE;        inescq = FALSE;
3287        ptr++;        ptr++;
# Line 2210  for (;; ptr++) Line 3307  for (;; ptr++)
3307    /* 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
3308    a quantifier. */    a quantifier. */
3309    
3310    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
3311      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
3312        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
3313    
3314    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
3315         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2221  for (;; ptr++) Line 3319  for (;; ptr++)
3319      previous_callout = NULL;      previous_callout = NULL;
3320      }      }
3321    
3322    /* In extended mode, skip white space and comments */    /* In extended mode, skip white space and comments. */
3323    
3324    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3325      {      {
3326      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
3327      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
3328        {        {
3329        while (*(++ptr) != 0)        ptr++;
3330          while (*ptr != 0)
3331          {          {
3332          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3333            ptr++;
3334    #ifdef SUPPORT_UTF8
3335            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
3336    #endif
3337          }          }
3338        if (*ptr != 0) continue;        if (*ptr != 0) continue;
3339    
# Line 2251  for (;; ptr++) Line 3354  for (;; ptr++)
3354      {      {
3355      /* ===================================================================*/      /* ===================================================================*/
3356      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
3357      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
3358      case ')':      case CHAR_RIGHT_PARENTHESIS:
3359      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
3360      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
3361      *codeptr = code;      *codeptr = code;
3362      *ptrptr = ptr;      *ptrptr = ptr;
3363      if (lengthptr != NULL)      if (lengthptr != NULL)
3364        {        {
3365        *lengthptr += code - last_code;   /* To include callout length */        if (OFLOW_MAX - *lengthptr < code - last_code)
3366            {
3367            *errorcodeptr = ERR20;
3368            goto FAILED;
3369            }
3370          *lengthptr += (int)(code - last_code);   /* To include callout length */
3371        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
3372        }        }
3373      return TRUE;      return TRUE;
# Line 2269  for (;; ptr++) Line 3377  for (;; ptr++)
3377      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
3378      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3379    
3380      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
3381        previous = NULL;
3382      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3383        {        {
3384        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3385          *code++ = OP_CIRCM;
3386        }        }
3387      previous = NULL;      else *code++ = OP_CIRC;
     *code++ = OP_CIRC;  
3388      break;      break;
3389    
3390      case '$':      case CHAR_DOLLAR_SIGN:
3391      previous = NULL;      previous = NULL;
3392      *code++ = OP_DOLL;      *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
3393      break;      break;
3394    
3395      /* 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
3396      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
3397    
3398      case '.':      case CHAR_DOT:
3399      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3400      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
3401      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3402      previous = code;      previous = code;
3403      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
3404      break;      break;
3405    
3406    
# Line 2306  for (;; ptr++) Line 3415  for (;; ptr++)
3415      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,
3416      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
3417      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.
     */  
3418    
3419      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
3420        default (Perl) mode, it is treated as a data character. */
3421    
3422        case CHAR_RIGHT_SQUARE_BRACKET:
3423        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3424          {
3425          *errorcodeptr = ERR64;
3426          goto FAILED;
3427          }
3428        goto NORMAL_CHAR;
3429    
3430        case CHAR_LEFT_SQUARE_BRACKET:
3431      previous = code;      previous = code;
3432    
3433      /* 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
3434      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. */
3435    
3436      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3437          check_posix_syntax(ptr, &tempptr, cd))           ptr[1] == CHAR_EQUALS_SIGN) &&
3438            check_posix_syntax(ptr, &tempptr))
3439        {        {
3440        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
3441        goto FAILED;        goto FAILED;
3442        }        }
3443    
3444      /* 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,
3445        if the first few characters (either before or after ^) are \Q\E or \E we
3446        skip them too. This makes for compatibility with Perl. */
3447    
3448      if ((c = *(++ptr)) == '^')      negate_class = FALSE;
3449        for (;;)
3450        {        {
       negate_class = TRUE;  
3451        c = *(++ptr);        c = *(++ptr);
3452          if (c == CHAR_BACKSLASH)
3453            {
3454            if (ptr[1] == CHAR_E)
3455              ptr++;
3456            else if (strncmp((const char *)ptr+1,
3457                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
3458              ptr += 3;
3459            else
3460              break;
3461            }
3462          else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
3463            negate_class = TRUE;
3464          else break;
3465        }        }
3466      else  
3467        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
3468        an initial ']' is taken as a data character -- the code below handles
3469        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3470        [^] must match any character, so generate OP_ALLANY. */
3471    
3472        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3473            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3474        {        {
3475        negate_class = FALSE;        *code++ = negate_class? OP_ALLANY : OP_FAIL;
3476          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3477          zerofirstbyte = firstbyte;
3478          break;
3479        }        }
3480    
3481        /* If a class contains a negative special such as \S, we need to flip the
3482        negation flag at the end, so that support for characters > 255 works
3483        correctly (they are all included in the class). */
3484    
3485        should_flip_negation = FALSE;
3486    
3487      /* 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
3488      of just a single character (as long as it's < 256). However, For higher      of just a single character (as long as it's < 256). However, For higher
3489      valued UTF-8 characters, we don't yet do any optimization. */      valued UTF-8 characters, we don't yet do any optimization. */
# Line 2350  for (;; ptr++) Line 3501  for (;; ptr++)
3501  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3502      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
3503      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
3504        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
3505  #endif  #endif
3506    
3507      /* Process characters until ] is reached. By writing this as a "do" it      /* Process characters until ] is reached. By writing this as a "do" it
# Line 2365  for (;; ptr++) Line 3517  for (;; ptr++)
3517          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3518          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3519          }          }
3520    
3521          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
3522          data and reset the pointer. This is so that very large classes that
3523          contain a zillion UTF-8 characters no longer overwrite the work space
3524          (which is on the stack). */
3525    
3526          if (lengthptr != NULL)
3527            {
3528            *lengthptr += class_utf8data - class_utf8data_base;
3529            class_utf8data = class_utf8data_base;
3530            }
3531    
3532  #endif  #endif
3533    
3534        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
3535    
3536        if (inescq)        if (inescq)
3537          {          {
3538          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3539            {            {
3540            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
3541            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2386  for (;; ptr++) Line 3550  for (;; ptr++)
3550        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3551        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3552    
3553        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3554            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3555            check_posix_syntax(ptr, &tempptr, cd))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3556          {          {
3557          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3558          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3559          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3560          uschar pbits[32];          uschar pbits[32];
3561    
3562          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3563            {            {
3564            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3565            goto FAILED;            goto FAILED;
3566            }            }
3567    
3568          ptr += 2;          ptr += 2;
3569          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3570            {            {
3571            local_negate = TRUE;            local_negate = TRUE;
3572              should_flip_negation = TRUE;  /* Note negative special */
3573            ptr++;            ptr++;
3574            }            }
3575    
3576          posix_class = check_posix_name(ptr, tempptr - ptr);          posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3577          if (posix_class < 0)          if (posix_class < 0)
3578            {            {
3579            *errorcodeptr = ERR30;            *errorcodeptr = ERR30;
# Line 2422  for (;; ptr++) Line 3587  for (;; ptr++)
3587          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3588            posix_class = 0;            posix_class = 0;
3589    
3590          /* We build the bit map for the POSIX class in a chunk of local store          /* When PCRE_UCP is set, some of the POSIX classes are converted to
3591          because we may be adding and subtracting from it, and we don't want to          different escape sequences that use Unicode properties. */
3592          subtract bits that may be in the main map already. At the end we or the  
3593          result into the bit map that is being built. */  #ifdef SUPPORT_UCP
3594            if ((options & PCRE_UCP) != 0)
3595              {
3596              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3597              if (posix_substitutes[pc] != NULL)
3598                {
3599                nestptr = tempptr + 1;
3600                ptr = posix_substitutes[pc] - 1;
3601                continue;
3602                }
3603              }
3604    #endif
3605            /* In the non-UCP case, we build the bit map for the POSIX class in a
3606            chunk of local store because we may be adding and subtracting from it,
3607            and we don't want to subtract bits that may be in the main map already.
3608            At the end we or the result into the bit map that is being built. */
3609    
3610          posix_class *= 3;          posix_class *= 3;
3611    
# Line 2469  for (;; ptr++) Line 3649  for (;; ptr++)
3649    
3650        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3651        of the specials, which just set a flag. The sequence \b is a special        of the specials, which just set a flag. The sequence \b is a special
3652        case. Inside a class (and only there) it is treated as backspace.        case. Inside a class (and only there) it is treated as backspace. We
3653        Elsewhere it marks a word boundary. Other escapes have preset maps ready        assume that other escapes have more than one character in them, so set
3654        to or into the one we are building. We assume they have more than one        class_charcount bigger than one. Unrecognized escapes fall through and
3655        character in them, so set class_charcount bigger than one. */        are either treated as literal characters (by default), or are faulted if
3656          PCRE_EXTRA is set. */
3657    
3658        if (c == '\\')        if (c == CHAR_BACKSLASH)
3659          {          {
3660          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3661          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3662    
3663          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 */  
         else if (-c == ESC_R) c = 'R';   /* \R is literal R in a class */  
3664          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3665            {            {
3666            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3667              {              {
3668              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3669              }              }
3670            else inescq = TRUE;            else inescq = TRUE;
3671            continue;            continue;
3672            }            }
3673            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
3674    
3675          if (c < 0)          if (c < 0)
3676            {            {
3677            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
3678            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3679    
3680            /* Save time by not doing this in the pre-compile phase. */            switch (-c)
   
           if (lengthptr == NULL) switch (-c)  
3681              {              {
3682    #ifdef SUPPORT_UCP
3683                case ESC_du:     /* These are the values given for \d etc */
3684                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3685                case ESC_wu:     /* escape sequence with an appropriate \p */
3686                case ESC_WU:     /* or \P to test Unicode properties instead */
3687                case ESC_su:     /* of the default ASCII testing. */
3688                case ESC_SU:
3689                nestptr = ptr;
3690                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3691                class_charcount -= 2;                /* Undo! */
3692                continue;
3693    #endif
3694              case ESC_d:              case ESC_d:
3695              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3696              continue;              continue;
3697    
3698              case ESC_D:              case ESC_D:
3699                should_flip_negation = TRUE;
3700              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
3701              continue;              continue;
3702    
# Line 2514  for (;; ptr++) Line 3705  for (;; ptr++)
3705              continue;              continue;
3706    
3707              case ESC_W:              case ESC_W:
3708                should_flip_negation = TRUE;
3709              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3710              continue;              continue;
3711    
3712                /* Perl 5.004 onwards omits VT from \s, but we must preserve it
3713                if it was previously set by something earlier in the character
3714                class. */
3715    
3716              case ESC_s:              case ESC_s:
3717              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];              classbits[0] |= cbits[cbit_space];
3718              classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= cbits[cbit_space+1] & ~0x08;
3719                for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3720              continue;              continue;
3721    
3722              case ESC_S:              case ESC_S:
3723                should_flip_negation = TRUE;
3724              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
3725              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3726              continue;              continue;
3727    
3728              case ESC_E: /* Perl ignores an orphan \E */              case ESC_h:
3729                SETBIT(classbits, 0x09); /* VT */
3730                SETBIT(classbits, 0x20); /* SPACE */
3731                SETBIT(classbits, 0xa0); /* NSBP */
3732    #ifdef SUPPORT_UTF8
3733                if (utf8)
3734                  {
3735                  class_utf8 = TRUE;
3736                  *class_utf8data++ = XCL_SINGLE;
3737                  class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);
3738                  *class_utf8data++ = XCL_SINGLE;
3739                  class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);
3740                  *class_utf8data++ = XCL_RANGE;
3741                  class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);
3742                  class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);
3743                  *class_utf8data++ = XCL_SINGLE;
3744                  class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);
3745                  *class_utf8data++ = XCL_SINGLE;
3746                  class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);
3747                  *class_utf8data++ = XCL_SINGLE;
3748                  class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);
3749                  }
3750    #endif
3751              continue;              continue;
3752    
3753              default:    /* Not recognized; fall through */              case ESC_H:
3754              break;      /* Need "default" setting to stop compiler warning. */              for (c = 0; c < 32; c++)
3755              }                {
3756                  int x = 0xff;
3757                  switch (c)
3758                    {
3759                    case 0x09/8: x ^= 1 << (0x09%8); break;
3760                    case 0x20/8: x ^= 1 << (0x20%8); break;
3761                    case 0xa0/8: x ^= 1 << (0xa0%8); break;
3762                    default: break;
3763                    }
3764                  classbits[c] |= x;
3765                  }
3766    
3767    #ifdef SUPPORT_UTF8
3768                if (utf8)
3769                  {
3770                  class_utf8 = TRUE;
3771                  *class_utf8data++ = XCL_RANGE;
3772                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3773                  class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);
3774                  *class_utf8data++ = XCL_RANGE;
3775                  class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);
3776                  class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);
3777                  *class_utf8data++ = XCL_RANGE;
3778                  class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);
3779                  class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);
3780                  *class_utf8data++ = XCL_RANGE;
3781                  class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);
3782                  class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);
3783                  *class_utf8data++ = XCL_RANGE;
3784                  class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);
3785                  class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);
3786                  *class_utf8data++ = XCL_RANGE;
3787                  class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);
3788                  class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);
3789                  *class_utf8data++ = XCL_RANGE;
3790                  class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);
3791                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3792                  }
3793    #endif
3794                continue;
3795    
3796            /* In the pre-compile phase, just do the recognition. */              case ESC_v:
3797                SETBIT(classbits, 0x0a); /* LF */
3798                SETBIT(classbits, 0x0b); /* VT */
3799                SETBIT(classbits, 0x0c); /* FF */
3800                SETBIT(classbits, 0x0d); /* CR */
3801                SETBIT(classbits, 0x85); /* NEL */
3802    #ifdef SUPPORT_UTF8
3803                if (utf8)
3804                  {
3805                  class_utf8 = TRUE;
3806                  *class_utf8data++ = XCL_RANGE;
3807                  class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);
3808                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3809                  }
3810    #endif
3811                continue;
3812    
3813            else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||              case ESC_V:
3814                     c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;              for (c = 0; c < 32; c++)
3815                  {
3816                  int x = 0xff;
3817                  switch (c)
3818                    {
3819                    case 0x0a/8: x ^= 1 << (0x0a%8);
3820                                 x ^= 1 << (0x0b%8);
3821                                 x ^= 1 << (0x0c%8);
3822                                 x ^= 1 << (0x0d%8);
3823                                 break;
3824                    case 0x85/8: x ^= 1 << (0x85%8); break;
3825                    default: break;
3826                    }
3827                  classbits[c] |= x;
3828                  }
3829    
3830            /* We need to deal with \P and \p in both phases. */  #ifdef SUPPORT_UTF8
3831                if (utf8)
3832                  {
3833                  class_utf8 = TRUE;
3834                  *class_utf8data++ = XCL_RANGE;
3835                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3836                  class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);
3837                  *class_utf8data++ = XCL_RANGE;
3838                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3839                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3840                  }
3841    #endif
3842                continue;
3843    
3844  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3845            if (-c == ESC_p || -c == ESC_P)              case ESC_p:
3846              {              case ESC_P:
3847              BOOL negated;                {
3848              int pdata;                BOOL negated;
3849              int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                int pdata;
3850              if (ptype < 0) goto FAILED;                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3851              class_utf8 = TRUE;                if (ptype < 0) goto FAILED;
3852              *class_utf8data++ = ((-c == ESC_p) != negated)?                class_utf8 = TRUE;
3853                XCL_PROP : XCL_NOTPROP;                *class_utf8data++ = ((-c == ESC_p) != negated)?
3854              *class_utf8data++ = ptype;                  XCL_PROP : XCL_NOTPROP;
3855              *class_utf8data++ = pdata;                *class_utf8data++ = ptype;
3856              class_charcount -= 2;   /* Not a < 256 character */                *class_utf8data++ = pdata;
3857              continue;                class_charcount -= 2;   /* Not a < 256 character */
3858              }                continue;
3859                  }
3860  #endif  #endif
3861            /* Unrecognized escapes are faulted if PCRE is running in its              /* Unrecognized escapes are faulted if PCRE is running in its
3862            strict mode. By default, for compatibility with Perl, they are              strict mode. By default, for compatibility with Perl, they are
3863            treated as literals. */              treated as literals. */
3864    
3865            if ((options & PCRE_EXTRA) != 0)              default:
3866              {              if ((options & PCRE_EXTRA) != 0)
3867              *errorcodeptr = ERR7;                {
3868              goto FAILED;                *errorcodeptr = ERR7;
3869                  goto FAILED;
3870                  }
3871                class_charcount -= 2;  /* Undo the default count from above */
3872                c = *ptr;              /* Get the final character and fall through */
3873                break;
3874              }              }
   
           class_charcount -= 2;  /* Undo the default count from above */  
           c = *ptr;              /* Get the final character and fall through */  
3875            }            }
3876    
3877          /* 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
# Line 2582  for (;; ptr++) Line 3885  for (;; ptr++)
3885        entirely. The code for handling \Q and \E is messy. */        entirely. The code for handling \Q and \E is messy. */
3886    
3887        CHECK_RANGE:        CHECK_RANGE:
3888        while (ptr[1] == '\\' && ptr[2] == 'E')        while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3889          {          {
3890          inescq = FALSE;          inescq = FALSE;
3891          ptr += 2;          ptr += 2;
# Line 2590  for (;; ptr++) Line 3893  for (;; ptr++)
3893    
3894        oldptr = ptr;        oldptr = ptr;
3895    
3896        if (!inescq && ptr[1] == '-')        /* Remember \r or \n */
3897    
3898          if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3899    
3900          /* Check for range */
3901    
3902          if (!inescq && ptr[1] == CHAR_MINUS)
3903          {          {
3904          int d;          int d;
3905          ptr += 2;          ptr += 2;
3906          while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3907    
3908          /* If we hit \Q (not followed by \E) at this point, go into escaped          /* If we hit \Q (not followed by \E) at this point, go into escaped
3909          mode. */          mode. */
3910    
3911          while (*ptr == '\\' && ptr[1] == 'Q')          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3912            {            {
3913            ptr += 2;            ptr += 2;
3914            if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }            if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3915                { ptr += 2; continue; }
3916            inescq = TRUE;            inescq = TRUE;
3917            break;            break;
3918            }            }
3919    
3920          if (*ptr == 0 || (!inescq && *ptr == ']'))          if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3921            {            {
3922            ptr = oldptr;            ptr = oldptr;
3923            goto LONE_SINGLE_CHARACTER;            goto LONE_SINGLE_CHARACTER;
# Line 2626  for (;; ptr++) Line 3936  for (;; ptr++)
3936          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
3937          in such circumstances. */          in such circumstances. */
3938    
3939          if (!inescq && d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3940            {            {
3941            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3942            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3943    
3944            /* \b is backslash; \X is literal X; \R is literal R; any other            /* \b is backspace; any other special means the '-' was literal */
           special means the '-' was literal */  
3945    
3946            if (d < 0)            if (d < 0)
3947              {              {
3948              if (d == -ESC_b) d = '\b';              if (d == -ESC_b) d = CHAR_BS; else
             else if (d == -ESC_X) d = 'X';  
             else if (d == -ESC_R) d = 'R'; else  
3949                {                {
3950                ptr = oldptr;                ptr = oldptr;
3951                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 2657  for (;; ptr++) Line 3964  for (;; ptr++)
3964    
3965          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3966    
3967            /* Remember \r or \n */
3968    
3969            if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3970    
3971          /* 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
3972          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
3973          matching for characters > 127 is available only if UCP support is          matching for characters > 127 is available only if UCP support is
# Line 2679  for (;; ptr++) Line 3990  for (;; ptr++)
3990              unsigned int origd = d;              unsigned int origd = d;
3991              while (get_othercase_range(&cc, origd, &occ, &ocd))              while (get_othercase_range(&cc, origd, &occ, &ocd))
3992                {                {
3993                if (occ >= c && ocd <= d) continue;  /* Skip embedded ranges */                if (occ >= (unsigned int)c &&
3994                      ocd <= (unsigned int)d)
3995                    continue;                          /* Skip embedded ranges */
3996    
3997                if (occ < c  && ocd >= c - 1)        /* Extend the basic range */                if (occ < (unsigned int)c  &&
3998                      ocd >= (unsigned int)c - 1)      /* Extend the basic range */
3999                  {                                  /* if there is overlap,   */                  {                                  /* if there is overlap,   */
4000                  c = occ;                           /* noting that if occ < c */                  c = occ;                           /* noting that if occ < c */
4001                  continue;                          /* we can't have ocd > d  */                  continue;                          /* we can't have ocd > d  */
4002                  }                                  /* because a subrange is  */                  }                                  /* because a subrange is  */
4003                if (ocd > d && occ <= d + 1)         /* always shorter than    */                if (ocd > (unsigned int)d &&
4004                      occ <= (unsigned int)d + 1)      /* always shorter than    */
4005                  {                                  /* the basic range.       */                  {                                  /* the basic range.       */
4006                  d = ocd;                  d = ocd;
4007                  continue;                  continue;
# Line 2771  for (;; ptr++) Line 4086  for (;; ptr++)
4086          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
4087            {            {
4088            unsigned int othercase;            unsigned int othercase;
4089            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)            if ((othercase = UCD_OTHERCASE(c)) != c)
4090              {              {
4091              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
4092              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 2796  for (;; ptr++) Line 4111  for (;; ptr++)
4111          }          }
4112        }        }
4113    
4114      /* Loop until ']' reached. This "while" is the end of the "do" above. */      /* Loop until ']' reached. This "while" is the end of the "do" far above.
4115        If we are at the end of an internal nested string, revert to the outer
4116        string. */
4117    
4118        while (((c = *(++ptr)) != 0 ||
4119               (nestptr != NULL &&
4120                 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != 0)) &&
4121               (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
4122    
4123      while ((c = *(++ptr)) != 0 && (c != ']' || inescq));      /* Check for missing terminating ']' */
4124    
4125      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)
4126        {        {
4127        *errorcodeptr = ERR6;        *errorcodeptr = ERR6;
4128        goto FAILED;        goto FAILED;
4129        }        }
4130    
4131      /* If class_charcount is 1, we saw precisely one character whose value is      /* If class_charcount is 1, we saw precisely one character whose value is
4132      less than 256. In non-UTF-8 mode we can always optimize. In UTF-8 mode, we      less than 256. As long as there were no characters >= 128 and there was no
4133      can optimize the negative case only if there were no characters >= 128      use of \p or \P, in other words, no use of any XCLASS features, we can
4134      because OP_NOT and the related opcodes like OP_NOTSTAR operate on      optimize.
4135      single-bytes only. This is an historical hangover. Maybe one day we can  
4136      tidy these opcodes to handle multi-byte characters.      In UTF-8 mode, we can optimize the negative case only if there were no
4137        characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
4138        operate on single-bytes characters only. This is an historical hangover.
4139        Maybe one day we can tidy these opcodes to handle multi-byte characters.
4140    
4141      The optimization throws away the bit map. We turn the item into a      The optimization throws away the bit map. We turn the item into a
4142      1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note      1-character OP_CHAR[I] if it's positive, or OP_NOT[I] if it's negative.
4143      that OP_NOT does not support multibyte characters. In the positive case, it      Note that OP_NOT[I] does not support multibyte characters. In the positive
4144      can cause firstbyte to be set. Otherwise, there can be no first char if      case, it can cause firstbyte to be set. Otherwise, there can be no first
4145      this item is first, whatever repeat count may follow. In the case of      char if this item is first, whatever repeat count may follow. In the case
4146      reqbyte, save the previous value for reinstating. */      of reqbyte, save the previous value for reinstating. */
4147    
4148  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4149      if (class_charcount == 1 &&      if (class_charcount == 1 && !class_utf8 &&
4150            (!utf8 ||        (!utf8 || !negate_class || class_lastchar < 128))
           (!class_utf8 && (!negate_class || class_lastchar < 128))))  
   
4151  #else  #else
4152      if (class_charcount == 1)      if (class_charcount == 1)
4153  #endif  #endif
4154        {        {
4155        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
4156    
4157        /* The OP_NOT opcode works on one-byte characters only. */        /* The OP_NOT[I] opcodes work on one-byte characters only. */
4158    
4159        if (negate_class)        if (negate_class)
4160          {          {
4161          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
4162          zerofirstbyte = firstbyte;          zerofirstbyte = firstbyte;
4163          *code++ = OP_NOT;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
4164          *code++ = class_lastchar;          *code++ = class_lastchar;
4165          break;          break;
4166          }          }
# Line 2867  for (;; ptr++) Line 4190  for (;; ptr++)
4190      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
4191    
4192      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
4193      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode, unless there was a negated special
4194      we can omit the bitmap in the actual compiled code. */      such as \S in the class, and PCRE_UCP is not set, because in that case all
4195        characters > 255 are in the class, so any that were explicitly given as
4196        well can be ignored. If (when there are explicit characters > 255 that must
4197        be listed) there are no characters < 256, we can omit the bitmap in the
4198        actual compiled code. */
4199    
4200  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4201      if (class_utf8)      if (class_utf8 && (!should_flip_negation || (options & PCRE_UCP) != 0))
4202        {        {
4203        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
4204        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 2897  for (;; ptr++) Line 4224  for (;; ptr++)
4224        }        }
4225  #endif  #endif
4226    
4227      /* If there are no characters > 255, negate the 32-byte map if necessary,      /* If there are no characters > 255, or they are all to be included or
4228      and copy it into the code vector. If this is the first thing in the branch,      excluded, set the opcode to OP_CLASS or OP_NCLASS, depending on whether the
4229      there can be no first char setting, whatever the repeat count. Any reqbyte      whole class was negated and whether there were negative specials such as \S
4230      setting must remain unchanged after any kind of repeat. */      (non-UCP) in the class. Then copy the 32-byte map into the code vector,
4231        negating it if necessary. */
4232    
4233        *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
4234      if (negate_class)      if (negate_class)
4235        {        {
       *code++ = OP_NCLASS;  
4236        if (lengthptr == NULL)    /* Save time in the pre-compile phase */        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
4237          for (c = 0; c < 32; c++) code[c] = ~classbits[c];          for (c = 0; c < 32; c++) code[c] = ~classbits[c];
4238        }        }
4239      else      else
4240        {        {
       *code++ = OP_CLASS;  
4241        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
4242        }        }
4243      code += 32;      code += 32;
# Line 2921  for (;; ptr++) Line 4248  for (;; ptr++)
4248      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
4249      has been tested above. */      has been tested above. */
425