/[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 150 by ph10, Tue Apr 17 08:22:40 2007 UTC revision 576 by ph10, Sun Nov 21 18:45:10 2010 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2007 University of Cambridge             Copyright (c) 1997-2010 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 42  POSSIBILITY OF SUCH DAMAGE. Line 42  POSSIBILITY OF SUCH DAMAGE.
42  supporting internal functions that are not used by other modules. */  supporting internal functions that are not used by other modules. */
43    
44    
45    #ifdef HAVE_CONFIG_H
46    #include "config.h"
47    #endif
48    
49  #define NLBLOCK cd             /* Block containing newline information */  #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,      0,      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"
397    "inconsistent NEWLINE options",    "inconsistent NEWLINE options\0"
398    "\\g is not followed by 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  };    "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      ;
413    
414  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
415  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 262  For convenience, we use the same bit def Line 427  For convenience, we use the same bit def
427    
428  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
429    
430  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC
431    
432    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
433    UTF-8 mode. */
434    
435  static const unsigned char digitab[] =  static const unsigned char digitab[] =
436    {    {
437    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
# Line 298  static const unsigned char digitab[] = Line 467  static const unsigned char digitab[] =
467    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
468    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
469    
470  #else           /* This is the "abnormal" case, for EBCDIC systems */  #else
471    
472    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
473    
474  static const unsigned char digitab[] =  static const unsigned char digitab[] =
475    {    {
476    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
# Line 373  static const unsigned char ebcdic_charta Line 545  static const unsigned char ebcdic_charta
545  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
546    
547  static BOOL  static BOOL
548    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, int, int *,    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,
549      int *, branch_chain *, compile_data *, int *);      int *, int *, branch_chain *, compile_data *, int *);
550    
551    
552    
553  /*************************************************  /*************************************************
554    *            Find an error text                  *
555    *************************************************/
556    
557    /* The error texts are now all in one long string, to save on relocations. As
558    some of the text is of unknown length, we can't use a table of offsets.
559    Instead, just count through the strings. This is not a performance issue
560    because it happens only when there has been a compilation error.
561    
562    Argument:   the error number
563    Returns:    pointer to the error string
564    */
565    
566    static const char *
567    find_error_text(int n)
568    {
569    const char *s = error_texts;
570    for (; n > 0; n--)
571      {
572      while (*s++ != 0) {};
573      if (*s == 0) return "Error text not found (please report)";
574      }
575    return s;
576    }
577    
578    
579    /*************************************************
580  *            Handle escapes                      *  *            Handle escapes                      *
581  *************************************************/  *************************************************/
582    
# Line 399  Arguments: Line 597  Arguments:
597    
598  Returns:         zero or positive => a data character  Returns:         zero or positive => a data character
599                   negative => a special escape sequence                   negative => a special escape sequence
600                   on error, errorptr is set                   on error, errorcodeptr is set
601  */  */
602    
603  static int  static int
# Line 417  ptr--; /* Set Line 615  ptr--; /* Set
615    
616  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
617    
618  /* 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
619  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.
620  Otherwise further processing may be required. */  Otherwise further processing may be required. */
621    
622  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
623  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */
624  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0) c = i;
625    
626  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
627  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */
628  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
629  #endif  #endif
630    
# Line 442  else Line 640  else
640      /* 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
641      error. */      error. */
642    
643      case 'l':      case CHAR_l:
644      case 'L':      case CHAR_L:
645      case 'N':      case CHAR_u:
646      case 'u':      case CHAR_U:
     case 'U':  
647      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
648      break;      break;
649    
650      /* \g must be followed by a number, either plain or braced. If positive, it      /* \g must be followed by one of a number of specific things:
651      is an absolute backreference. If negative, it is a relative backreference.  
652      This is a Perl 5.10 feature. */      (1) A number, either plain or braced. If positive, it is an absolute
653        backreference. If negative, it is a relative backreference. This is a Perl
654        5.10 feature.
655    
656        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
657        is part of Perl's movement towards a unified syntax for back references. As
658        this is synonymous with \k{name}, we fudge it up by pretending it really
659        was \k.
660    
661        (3) For Oniguruma compatibility we also support \g followed by a name or a
662        number either in angle brackets or in single quotes. However, these are
663        (possibly recursive) subroutine calls, _not_ backreferences. Just return
664        the -ESC_g code (cf \k). */
665    
666        case CHAR_g:
667        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
668          {
669          c = -ESC_g;
670          break;
671          }
672    
673        /* Handle the Perl-compatible cases */
674    
675      case 'g':      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
     if (ptr[1] == '{')  
676        {        {
677          const uschar *p;
678          for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
679            if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
680          if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
681            {
682            c = -ESC_k;
683            break;
684            }
685        braced = TRUE;        braced = TRUE;
686        ptr++;        ptr++;
687        }        }
688      else braced = FALSE;      else braced = FALSE;
689    
690      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
691        {        {
692        negated = TRUE;        negated = TRUE;
693        ptr++;        ptr++;
# Line 471  else Line 696  else
696    
697      c = 0;      c = 0;
698      while ((digitab[ptr[1]] & ctype_digit) != 0)      while ((digitab[ptr[1]] & ctype_digit) != 0)
699        c = c * 10 + *(++ptr) - '0';        c = c * 10 + *(++ptr) - CHAR_0;
700    
701        if (c < 0)   /* Integer overflow */
702          {
703          *errorcodeptr = ERR61;
704          break;
705          }
706    
707      if (c == 0 || (braced && *(++ptr) != '}'))      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
708        {        {
709        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
710        return 0;        break;
711          }
712    
713        if (c == 0)
714          {
715          *errorcodeptr = ERR58;
716          break;
717        }        }
718    
719      if (negated)      if (negated)
# Line 484  else Line 721  else
721        if (c > bracount)        if (c > bracount)
722          {          {
723          *errorcodeptr = ERR15;          *errorcodeptr = ERR15;
724          return 0;          break;
725          }          }
726        c = bracount - (c - 1);        c = bracount - (c - 1);
727        }        }
# Line 504  else Line 741  else
741      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
742      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
743    
744      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:
745      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
746    
747      if (!isclass)      if (!isclass)
748        {        {
749        oldptr = ptr;        oldptr = ptr;
750        c -= '0';        c -= CHAR_0;
751        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
752          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - CHAR_0;
753          if (c < 0)    /* Integer overflow */
754            {
755            *errorcodeptr = ERR61;
756            break;
757            }
758        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
759          {          {
760          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 525  else Line 767  else
767      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.
768      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
769    
770      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
771        {        {
772        ptr--;        ptr--;
773        c = 0;        c = 0;
# Line 538  else Line 780  else
780      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
781      than 3 octal digits. */      than 3 octal digits. */
782    
783      case '0':      case CHAR_0:
784      c -= '0';      c -= CHAR_0;
785      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
786          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
787      if (!utf8 && c > 255) *errorcodeptr = ERR51;      if (!utf8 && c > 255) *errorcodeptr = ERR51;
788      break;      break;
789    
# Line 549  else Line 791  else
791      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
792      treated as a data character. */      treated as a data character. */
793    
794      case 'x':      case CHAR_x:
795      if (ptr[1] == '{')      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
796        {        {
797        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
798        int count = 0;        int count = 0;
# Line 559  else Line 801  else
801        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
802          {          {
803          register int cc = *pt++;          register int cc = *pt++;
804          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
805          count++;          count++;
806    
807  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
808          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
809          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
810  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
811          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
812          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
813  #endif  #endif
814          }          }
815    
816        if (*pt == '}')        if (*pt == CHAR_RIGHT_CURLY_BRACKET)
817          {          {
818          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
819          ptr = pt;          ptr = pt;
# Line 587  else Line 829  else
829      c = 0;      c = 0;
830      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
831        {        {
832        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
833        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
834  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
835        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
836        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
837  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
838        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
839        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
840  #endif  #endif
841        }        }
842      break;      break;
843    
844      /* 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.
845      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
846        coding is ASCII-specific, but then the whole concept of \cx is
847      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
848    
849      case 'c':      case CHAR_c:
850      c = *(++ptr);      c = *(++ptr);
851      if (c == 0)      if (c == 0)
852        {        {
853        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
854        return 0;        break;
855        }        }
856    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
857  #ifndef EBCDIC  /* ASCII coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
858      if (c >= 'a' && c <= 'z') c -= 32;        {
859          *errorcodeptr = ERR68;
860          break;
861          }
862        if (c >= CHAR_a && c <= CHAR_z) c -= 32;
863      c ^= 0x40;      c ^= 0x40;
864  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
865      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
866      c ^= 0xC0;      c ^= 0xC0;
867  #endif  #endif
868      break;      break;
869    
870      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
871      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
872      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
873      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
874      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
875    
876      default:      default:
877      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 637  else Line 884  else
884      }      }
885    }    }
886    
887    /* Perl supports \N{name} for character names, as well as plain \N for "not
888    newline". PCRE does not support \N{name}. */
889    
890    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET)
891      *errorcodeptr = ERR37;
892    
893    /* If PCRE_UCP is set, we change the values for \d etc. */
894    
895    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
896      c -= (ESC_DU - ESC_D);
897    
898    /* Set the pointer to the final character before returning. */
899    
900  *ptrptr = ptr;  *ptrptr = ptr;
901  return c;  return c;
902  }  }
# Line 677  if (c == 0) goto ERROR_RETURN; Line 937  if (c == 0) goto ERROR_RETURN;
937  /* \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
938  negation. */  negation. */
939    
940  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
941    {    {
942    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
943      {      {
944      *negptr = TRUE;      *negptr = TRUE;
945      ptr++;      ptr++;
946      }      }
947    for (i = 0; i < sizeof(name) - 1; i++)    for (i = 0; i < (int)sizeof(name) - 1; i++)
948      {      {
949      c = *(++ptr);      c = *(++ptr);
950      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
951      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
952      name[i] = c;      name[i] = c;
953      }      }
954    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
955    name[i] = 0;    name[i] = 0;
956    }    }
957    
# Line 713  top = _pcre_utt_size; Line 973  top = _pcre_utt_size;
973  while (bot < top)  while (bot < top)
974    {    {
975    i = (bot + top) >> 1;    i = (bot + top) >> 1;
976    c = strcmp(name, _pcre_utt[i].name);    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);
977    if (c == 0)    if (c == 0)
978      {      {
979      *dptr = _pcre_utt[i].value;      *dptr = _pcre_utt[i].value;
# Line 756  is_counted_repeat(const uschar *p) Line 1016  is_counted_repeat(const uschar *p)
1016  {  {
1017  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
1018  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
1019  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
1020    
1021  if (*p++ != ',') return FALSE;  if (*p++ != CHAR_COMMA) return FALSE;
1022  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
1023    
1024  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
1025  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
1026    
1027  return (*p == '}');  return (*p == CHAR_RIGHT_CURLY_BRACKET);
1028  }  }
1029    
1030    
# Line 797  int max = -1; Line 1057  int max = -1;
1057  /* 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
1058  an integer overflow. */  an integer overflow. */
1059    
1060  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;
1061  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1062    {    {
1063    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 807  if (min < 0 || min > 65535) Line 1067  if (min < 0 || min > 65535)
1067  /* 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.
1068  Also, max must not be less than min. */  Also, max must not be less than min. */
1069    
1070  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1071    {    {
1072    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1073      {      {
1074      max = 0;      max = 0;
1075      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;
1076      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1077        {        {
1078        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 837  return p; Line 1097  return p;
1097    
1098    
1099  /*************************************************  /*************************************************
1100  *       Find forward referenced subpattern       *  *  Subroutine for finding forward reference      *
1101  *************************************************/  *************************************************/
1102    
1103  /* This function scans along a pattern's text looking for capturing  /* This recursive function is called only from find_parens() below. The
1104    top-level call starts at the beginning of the pattern. All other calls must
1105    start at a parenthesis. It scans along a pattern's text looking for capturing
1106  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
1107  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
1108  returns when it reaches a given numbered subpattern. This is used for forward  returns when it reaches a given numbered subpattern. We know that if (?P< is
1109  references to subpatterns. We know that if (?P< is encountered, the name will  encountered, the name will be terminated by '>' because that is checked in the
1110  be terminated by '>' because that is checked in the first pass.  first pass. Recursion is used to keep track of subpatterns that reset the
1111    capturing group numbers - the (?| feature.
1112    
1113  Arguments:  Arguments:
1114    ptr          current position in the pattern    ptrptr       address of the current character pointer (updated)
1115    count        current count of capturing parens so far encountered    cd           compile background data
1116    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1117    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1118    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1119      utf8         TRUE if we are in UTF-8 mode
1120      count        pointer to the current capturing subpattern number (updated)
1121    
1122  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1123  */  */
1124    
1125  static int  static int
1126  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,
1127    BOOL xmode)    BOOL xmode, BOOL utf8, int *count)
1128  {  {
1129  const uschar *thisname;  uschar *ptr = *ptrptr;
1130    int start_count = *count;
1131    int hwm_count = start_count;
1132    BOOL dup_parens = FALSE;
1133    
1134  for (; *ptr != 0; ptr++)  /* If the first character is a parenthesis, check on the type of group we are
1135    dealing with. The very first call may not start with a parenthesis. */
1136    
1137    if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1138    {    {
1139    int term;    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1140    
1141      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1142    
1143      /* Handle a normal, unnamed capturing parenthesis. */
1144    
1145      else if (ptr[1] != CHAR_QUESTION_MARK)
1146        {
1147        *count += 1;
1148        if (name == NULL && *count == lorn) return *count;
1149        ptr++;
1150        }
1151    
1152      /* All cases now have (? at the start. Remember when we are in a group
1153      where the parenthesis numbers are duplicated. */
1154    
1155      else if (ptr[2] == CHAR_VERTICAL_LINE)
1156        {
1157        ptr += 3;
1158        dup_parens = TRUE;
1159        }
1160    
1161      /* Handle comments; all characters are allowed until a ket is reached. */
1162    
1163      else if (ptr[2] == CHAR_NUMBER_SIGN)
1164        {
1165        for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1166        goto FAIL_EXIT;
1167        }
1168    
1169      /* Handle a condition. If it is an assertion, just carry on so that it
1170      is processed as normal. If not, skip to the closing parenthesis of the
1171      condition (there can't be any nested parens). */
1172    
1173      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1174        {
1175        ptr += 2;
1176        if (ptr[1] != CHAR_QUESTION_MARK)
1177          {
1178          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1179          if (*ptr != 0) ptr++;
1180          }
1181        }
1182    
1183      /* Start with (? but not a condition. */
1184    
1185      else
1186        {
1187        ptr += 2;
1188        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1189    
1190        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1191    
1192        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1193            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1194          {
1195          int term;
1196          const uschar *thisname;
1197          *count += 1;
1198          if (name == NULL && *count == lorn) return *count;
1199          term = *ptr++;
1200          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1201          thisname = ptr;
1202          while (*ptr != term) ptr++;
1203          if (name != NULL && lorn == ptr - thisname &&
1204              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1205            return *count;
1206          term++;
1207          }
1208        }
1209      }
1210    
1211    /* Past any initial parenthesis handling, scan for parentheses or vertical
1212    bars. */
1213    
1214    for (; *ptr != 0; ptr++)
1215      {
1216    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1217    
1218    if (*ptr == '\\')    if (*ptr == CHAR_BACKSLASH)
1219      {      {
1220      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1221      if (*ptr == 'Q') for (;;)      if (*ptr == CHAR_Q) for (;;)
1222        {        {
1223        while (*(++ptr) != 0 && *ptr != '\\');        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1224        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1225        if (*(++ptr) == 'E') break;        if (*(++ptr) == CHAR_E) break;
1226        }        }
1227      continue;      continue;
1228      }      }
1229    
1230    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1231      are handled for real. If the first character is '^', skip it. Also, if the
1232      first few characters (either before or after ^) are \Q\E or \E we skip them
1233      too. This makes for compatibility with Perl. Note the use of STR macros to
1234      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1235    
1236    if (*ptr == '[')    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1237      {      {
1238      while (*(++ptr) != ']')      BOOL negate_class = FALSE;
1239        for (;;)
1240        {        {
1241        if (*ptr == '\\')        if (ptr[1] == CHAR_BACKSLASH)
1242          {          {
1243          if (*(++ptr) == 0) return -1;          if (ptr[2] == CHAR_E)
1244          if (*ptr == 'Q') for (;;)            ptr+= 2;
1245            else if (strncmp((const char *)ptr+2,
1246                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1247              ptr += 4;
1248            else
1249              break;
1250            }
1251          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1252            {
1253            negate_class = TRUE;
1254            ptr++;
1255            }
1256          else break;
1257          }
1258    
1259        /* If the next character is ']', it is a data character that must be
1260        skipped, except in JavaScript compatibility mode. */
1261    
1262        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1263            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1264          ptr++;
1265    
1266        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1267          {
1268          if (*ptr == 0) return -1;
1269          if (*ptr == CHAR_BACKSLASH)
1270            {
1271            if (*(++ptr) == 0) goto FAIL_EXIT;
1272            if (*ptr == CHAR_Q) for (;;)
1273            {            {
1274            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1275            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1276            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1277            }            }
1278          continue;          continue;
1279          }          }
# Line 904  for (; *ptr != 0; ptr++) Line 1283  for (; *ptr != 0; ptr++)
1283    
1284    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1285    
1286    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1287      {      {
1288      while (*(++ptr) != 0 && *ptr != '\n');      ptr++;
1289      if (*ptr == 0) return -1;      while (*ptr != 0)
1290          {
1291          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1292          ptr++;
1293    #ifdef SUPPORT_UTF8
1294          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
1295    #endif
1296          }
1297        if (*ptr == 0) goto FAIL_EXIT;
1298      continue;      continue;
1299      }      }
1300    
1301    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1302    
1303    if (*ptr != '(') continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != '?')  
1304      {      {
1305      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, count);
1306      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1307      continue;      if (*ptr == 0) goto FAIL_EXIT;
1308      }      }
1309    
1310    ptr += 2;    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1311    if (*ptr == 'P') ptr++;                      /* Allow optional P */      {
1312        if (dup_parens && *count < hwm_count) *count = hwm_count;
1313        goto FAIL_EXIT;
1314        }
1315    
1316    /* We have to disambiguate (?<! and (?<= from (?<name> */    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1317        {
1318        if (*count > hwm_count) hwm_count = *count;
1319        *count = start_count;
1320        }
1321      }
1322    
1323    if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  FAIL_EXIT:
1324         *ptr != '\'')  *ptrptr = ptr;
1325      continue;  return -1;
1326    }
1327    
1328    
1329    
1330    
1331    /*************************************************
1332    *       Find forward referenced subpattern       *
1333    *************************************************/
1334    
1335    count++;  /* This function scans along a pattern's text looking for capturing
1336    subpatterns, and counting them. If it finds a named pattern that matches the
1337    name it is given, it returns its number. Alternatively, if the name is NULL, it
1338    returns when it reaches a given numbered subpattern. This is used for forward
1339    references to subpatterns. We used to be able to start this scan from the
1340    current compiling point, using the current count value from cd->bracount, and
1341    do it all in a single loop, but the addition of the possibility of duplicate
1342    subpattern numbers means that we have to scan from the very start, in order to
1343    take account of such duplicates, and to use a recursive function to keep track
1344    of the different types of group.
1345    
1346    Arguments:
1347      cd           compile background data
1348      name         name to seek, or NULL if seeking a numbered subpattern
1349      lorn         name length, or subpattern number if name is NULL
1350      xmode        TRUE if we are in /x mode
1351      utf8         TRUE if we are in UTF-8 mode
1352    
1353    if (name == NULL && count == lorn) return count;  Returns:       the number of the found subpattern, or -1 if not found
1354    term = *ptr++;  */
1355    if (term == '<') term = '>';  
1356    thisname = ptr;  static int
1357    while (*ptr != term) ptr++;  find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode,
1358    if (name != NULL && lorn == ptr - thisname &&    BOOL utf8)
1359        strncmp((const char *)name, (const char *)thisname, lorn) == 0)  {
1360      return count;  uschar *ptr = (uschar *)cd->start_pattern;
1361    int count = 0;
1362    int rc;
1363    
1364    /* If the pattern does not start with an opening parenthesis, the first call
1365    to find_parens_sub() will scan right to the end (if necessary). However, if it
1366    does start with a parenthesis, find_parens_sub() will return when it hits the
1367    matching closing parens. That is why we have to have a loop. */
1368    
1369    for (;;)
1370      {
1371      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, &count);
1372      if (rc > 0 || *ptr++ == 0) break;
1373    }    }
1374    
1375  return -1;  return rc;
1376  }  }
1377    
1378    
1379    
1380    
1381  /*************************************************  /*************************************************
1382  *      Find first significant op code            *  *      Find first significant op code            *
1383  *************************************************/  *************************************************/
# Line 996  for (;;) Line 1427  for (;;)
1427    
1428      case OP_CALLOUT:      case OP_CALLOUT:
1429      case OP_CREF:      case OP_CREF:
1430        case OP_NCREF:
1431      case OP_RREF:      case OP_RREF:
1432        case OP_NRREF:
1433      case OP_DEF:      case OP_DEF:
1434      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1435      break;      break;
# Line 1012  for (;;) Line 1445  for (;;)
1445    
1446    
1447  /*************************************************  /*************************************************
1448  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1449  *************************************************/  *************************************************/
1450    
1451  /* 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,
1452  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.
1453  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
1454    temporarily terminated with OP_END when this function is called.
1455    
1456    This function is called when a backward assertion is encountered, so that if it
1457    fails, the error message can point to the correct place in the pattern.
1458    However, we cannot do this when the assertion contains subroutine calls,
1459    because they can be forward references. We solve this by remembering this case
1460    and doing the check at the end; a flag specifies which mode we are running in.
1461    
1462  Arguments:  Arguments:
1463    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1464    options  the compiling options    options  the compiling options
1465      atend    TRUE if called when the pattern is complete
1466      cd       the "compile data" structure
1467    
1468  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1469                 or -1 if there is no fixed length,
1470               or -2 if \C was encountered               or -2 if \C was encountered
1471                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1472  */  */
1473    
1474  static int  static int
1475  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)
1476  {  {
1477  int length = -1;  int length = -1;
1478    
# Line 1041  branch, check the length against that of Line 1485  branch, check the length against that of
1485  for (;;)  for (;;)
1486    {    {
1487    int d;    int d;
1488      uschar *ce, *cs;
1489    register int op = *cc;    register int op = *cc;
   
1490    switch (op)    switch (op)
1491      {      {
1492      case OP_CBRA:      case OP_CBRA:
1493      case OP_BRA:      case OP_BRA:
1494      case OP_ONCE:      case OP_ONCE:
1495      case OP_COND:      case OP_COND:
1496      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);
1497      if (d < 0) return d;      if (d < 0) return d;
1498      branchlength += d;      branchlength += d;
1499      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1072  for (;;) Line 1516  for (;;)
1516      branchlength = 0;      branchlength = 0;
1517      break;      break;
1518    
1519        /* A true recursion implies not fixed length, but a subroutine call may
1520        be OK. If the subroutine is a forward reference, we can't deal with
1521        it until the end of the pattern, so return -3. */
1522    
1523        case OP_RECURSE:
1524        if (!atend) return -3;
1525        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1526        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1527        if (cc > cs && cc < ce) return -1;                /* Recursion */
1528        d = find_fixedlength(cs + 2, options, atend, cd);
1529        if (d < 0) return d;
1530        branchlength += d;
1531        cc += 1 + LINK_SIZE;
1532        break;
1533    
1534      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1535    
1536      case OP_ASSERT:      case OP_ASSERT:
# Line 1085  for (;;) Line 1544  for (;;)
1544    
1545      case OP_REVERSE:      case OP_REVERSE:
1546      case OP_CREF:      case OP_CREF:
1547        case OP_NCREF:
1548      case OP_RREF:      case OP_RREF:
1549        case OP_NRREF:
1550      case OP_DEF:      case OP_DEF:
1551      case OP_OPT:      case OP_OPT:
1552      case OP_CALLOUT:      case OP_CALLOUT:
1553      case OP_SOD:      case OP_SOD:
1554      case OP_SOM:      case OP_SOM:
1555        case OP_SET_SOM:
1556      case OP_EOD:      case OP_EOD:
1557      case OP_EODN:      case OP_EODN:
1558      case OP_CIRC:      case OP_CIRC:
# Line 1108  for (;;) Line 1570  for (;;)
1570      branchlength++;      branchlength++;
1571      cc += 2;      cc += 2;
1572  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1573      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1574        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1575  #endif  #endif
1576      break;      break;
1577    
# Line 1122  for (;;) Line 1582  for (;;)
1582      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1583      cc += 4;      cc += 4;
1584  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1585      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1586        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1587  #endif  #endif
1588      break;      break;
1589    
1590      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1591      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1592        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1593      cc += 4;      cc += 4;
1594      break;      break;
1595    
# Line 1148  for (;;) Line 1607  for (;;)
1607      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1608      case OP_WORDCHAR:      case OP_WORDCHAR:
1609      case OP_ANY:      case OP_ANY:
1610        case OP_ALLANY:
1611      branchlength++;      branchlength++;
1612      cc++;      cc++;
1613      break;      break;
# Line 1202  for (;;) Line 1662  for (;;)
1662    
1663    
1664  /*************************************************  /*************************************************
1665  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1666  *************************************************/  *************************************************/
1667    
1668  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1669  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1670    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1671    so that it can be called from pcre_study() when finding the minimum matching
1672    length.
1673    
1674  Arguments:  Arguments:
1675    code        points to start of expression    code        points to start of expression
1676    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1677    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1678    
1679  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
1680  */  */
1681    
1682  static const uschar *  const uschar *
1683  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1684  {  {
1685  for (;;)  for (;;)
1686    {    {
# Line 1230  for (;;) Line 1693  for (;;)
1693    
1694    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1695    
1696      /* Handle recursion */
1697    
1698      else if (c == OP_REVERSE)
1699        {
1700        if (number < 0) return (uschar *)code;
1701        code += _pcre_OP_lengths[c];
1702        }
1703    
1704    /* Handle capturing bracket */    /* Handle capturing bracket */
1705    
1706    else if (c == OP_CBRA)    else if (c == OP_CBRA)
# Line 1239  for (;;) Line 1710  for (;;)
1710      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1711      }      }
1712    
1713    /* 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
1714    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
1715    arrange to skip the extra bytes. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1716      must add in its length. */
1717    
1718    else    else
1719      {      {
1720        switch(c)
1721          {
1722          case OP_TYPESTAR:
1723          case OP_TYPEMINSTAR:
1724          case OP_TYPEPLUS:
1725          case OP_TYPEMINPLUS:
1726          case OP_TYPEQUERY:
1727          case OP_TYPEMINQUERY:
1728          case OP_TYPEPOSSTAR:
1729          case OP_TYPEPOSPLUS:
1730          case OP_TYPEPOSQUERY:
1731          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1732          break;
1733    
1734          case OP_TYPEUPTO:
1735          case OP_TYPEMINUPTO:
1736          case OP_TYPEEXACT:
1737          case OP_TYPEPOSUPTO:
1738          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1739          break;
1740    
1741          case OP_MARK:
1742          case OP_PRUNE_ARG:
1743          case OP_SKIP_ARG:
1744          code += code[1];
1745          break;
1746    
1747          case OP_THEN_ARG:
1748          code += code[1+LINK_SIZE];
1749          break;
1750          }
1751    
1752        /* Add in the fixed length from the table */
1753    
1754      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1755    
1756      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1757      a multi-byte character. The length in the table is a minimum, so we have to
1758      arrange to skip the extra bytes. */
1759    
1760  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1761      if (utf8) switch(c)      if (utf8) switch(c)
1762        {        {
# Line 1267  for (;;) Line 1778  for (;;)
1778        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1779        break;        break;
1780        }        }
1781    #else
1782        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1783  #endif  #endif
1784      }      }
1785    }    }
# Line 1303  for (;;) Line 1816  for (;;)
1816    
1817    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1818    
1819    /* 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
1820    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
1821    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
1822    bytes. */    must add in its length. */
1823    
1824    else    else
1825      {      {
1826        switch(c)
1827          {
1828          case OP_TYPESTAR:
1829          case OP_TYPEMINSTAR:
1830          case OP_TYPEPLUS:
1831          case OP_TYPEMINPLUS:
1832          case OP_TYPEQUERY:
1833          case OP_TYPEMINQUERY:
1834          case OP_TYPEPOSSTAR:
1835          case OP_TYPEPOSPLUS:
1836          case OP_TYPEPOSQUERY:
1837          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1838          break;
1839    
1840          case OP_TYPEPOSUPTO:
1841          case OP_TYPEUPTO:
1842          case OP_TYPEMINUPTO:
1843          case OP_TYPEEXACT:
1844          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1845          break;
1846    
1847          case OP_MARK:
1848          case OP_PRUNE_ARG:
1849          case OP_SKIP_ARG:
1850          code += code[1];
1851          break;
1852    
1853          case OP_THEN_ARG:
1854          code += code[1+LINK_SIZE];
1855          break;
1856          }
1857    
1858        /* Add in the fixed length from the table */
1859    
1860      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1861    
1862        /* In UTF-8 mode, opcodes that are followed by a character may be followed
1863        by a multi-byte character. The length in the table is a minimum, so we have
1864        to arrange to skip the extra bytes. */
1865    
1866  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1867      if (utf8) switch(c)      if (utf8) switch(c)
1868        {        {
# Line 1332  for (;;) Line 1884  for (;;)
1884        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1885        break;        break;
1886        }        }
1887    #else
1888        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1889  #endif  #endif
1890      }      }
1891    }    }
# Line 1347  for (;;) Line 1901  for (;;)
1901  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()
1902  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
1903  group that can match nothing. Note that first_significant_code() skips over  group that can match nothing. Note that first_significant_code() skips over
1904  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
1905  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
1906    bracket whose current branch will already have been scanned.
1907    
1908  Arguments:  Arguments:
1909    code        points to start of search    code        points to start of search
1910    endcode     points to where to stop    endcode     points to where to stop
1911    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
1912      cd          contains pointers to tables etc.
1913    
1914  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
1915  */  */
1916    
1917  static BOOL  static BOOL
1918  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
1919      compile_data *cd)
1920  {  {
1921  register int c;  register int c;
1922  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);
# Line 1370  for (code = first_significant_code(code Line 1927  for (code = first_significant_code(code
1927    
1928    c = *code;    c = *code;
1929    
1930    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE)    /* Skip over forward assertions; the other assertions are skipped by
1931      first_significant_code() with a TRUE final argument. */
1932    
1933      if (c == OP_ASSERT)
1934      {      {
1935      BOOL empty_branch;      do code += GET(code, 1); while (*code == OP_ALT);
1936      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      c = *code;
1937        continue;
1938        }
1939    
1940      /* Scan a closed bracket */    /* Groups with zero repeats can of course be empty; skip them. */
1941    
1942      empty_branch = FALSE;    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)
1943        {
1944        code += _pcre_OP_lengths[c];
1945        do code += GET(code, 1); while (*code == OP_ALT);
1946        c = *code;
1947        continue;
1948        }
1949    
1950      /* For a recursion/subroutine call, if its end has been reached, which
1951      implies a subroutine call, we can scan it. */
1952    
1953      if (c == OP_RECURSE)
1954        {
1955        BOOL empty_branch = FALSE;
1956        const uschar *scode = cd->start_code + GET(code, 1);
1957        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
1958      do      do
1959        {        {
1960        if (!empty_branch && could_be_empty_branch(code, endcode, utf8))        if (could_be_empty_branch(scode, endcode, utf8, cd))
1961            {
1962          empty_branch = TRUE;          empty_branch = TRUE;
1963        code += GET(code, 1);          break;
1964            }
1965          scode += GET(scode, 1);
1966        }        }
1967      while (*code == OP_ALT);      while (*scode == OP_ALT);
1968      if (!empty_branch) return FALSE;   /* All branches are non-empty */      if (!empty_branch) return FALSE;  /* All branches are non-empty */
1969        continue;
1970        }
1971    
1972      /* Move past the KET and fudge things so that the increment in the "for"    /* For other groups, scan the branches. */
     above has no effect. */  
1973    
1974      c = OP_END;    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)
1975      code += 1 + LINK_SIZE - _pcre_OP_lengths[c];      {
1976        BOOL empty_branch;
1977        if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
1978    
1979        /* If a conditional group has only one branch, there is a second, implied,
1980        empty branch, so just skip over the conditional, because it could be empty.
1981        Otherwise, scan the individual branches of the group. */
1982    
1983        if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
1984          code += GET(code, 1);
1985        else
1986          {
1987          empty_branch = FALSE;
1988          do
1989            {
1990            if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
1991              empty_branch = TRUE;
1992            code += GET(code, 1);
1993            }
1994          while (*code == OP_ALT);
1995          if (!empty_branch) return FALSE;   /* All branches are non-empty */
1996          }
1997    
1998        c = *code;
1999      continue;      continue;
2000      }      }
2001    
# Line 1399  for (code = first_significant_code(code Line 2003  for (code = first_significant_code(code
2003    
2004    switch (c)    switch (c)
2005      {      {
2006      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
2007        cannot be represented just by a bit map. This includes negated single
2008        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
2009        actual length is stored in the compiled code, so we must update "code"
2010        here. */
2011    
2012  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2013      case OP_XCLASS:      case OP_XCLASS:
2014      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
2015      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
2016  #endif  #endif
2017    
# Line 1447  for (code = first_significant_code(code Line 2055  for (code = first_significant_code(code
2055      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2056      case OP_WORDCHAR:      case OP_WORDCHAR:
2057      case OP_ANY:      case OP_ANY:
2058        case OP_ALLANY:
2059      case OP_ANYBYTE:      case OP_ANYBYTE:
2060      case OP_CHAR:      case OP_CHAR:
2061      case OP_CHARNC:      case OP_CHARNC:
# Line 1465  for (code = first_significant_code(code Line 2074  for (code = first_significant_code(code
2074      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2075      return FALSE;      return FALSE;
2076    
2077        /* These are going to continue, as they may be empty, but we have to
2078        fudge the length for the \p and \P cases. */
2079    
2080        case OP_TYPESTAR:
2081        case OP_TYPEMINSTAR:
2082        case OP_TYPEPOSSTAR:
2083        case OP_TYPEQUERY:
2084        case OP_TYPEMINQUERY:
2085        case OP_TYPEPOSQUERY:
2086        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2087        break;
2088    
2089        /* Same for these */
2090    
2091        case OP_TYPEUPTO:
2092        case OP_TYPEMINUPTO:
2093        case OP_TYPEPOSUPTO:
2094        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
2095        break;
2096    
2097      /* End of branch */      /* End of branch */
2098    
2099      case OP_KET:      case OP_KET:
# Line 1483  for (code = first_significant_code(code Line 2112  for (code = first_significant_code(code
2112      case OP_QUERY:      case OP_QUERY:
2113      case OP_MINQUERY:      case OP_MINQUERY:
2114      case OP_POSQUERY:      case OP_POSQUERY:
2115        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2116        break;
2117    
2118      case OP_UPTO:      case OP_UPTO:
2119      case OP_MINUPTO:      case OP_MINUPTO:
2120      case OP_POSUPTO:      case OP_POSUPTO:
2121      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2122        break;
2123    #endif
2124    
2125        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2126        string. */
2127    
2128        case OP_MARK:
2129        case OP_PRUNE_ARG:
2130        case OP_SKIP_ARG:
2131        code += code[1];
2132        break;
2133    
2134        case OP_THEN_ARG:
2135        code += code[1+LINK_SIZE];
2136        break;
2137    
2138        /* None of the remaining opcodes are required to match a character. */
2139    
2140        default:
2141      break;      break;
 #endif  
2142      }      }
2143    }    }
2144    
# Line 1511  Arguments: Line 2161  Arguments:
2161    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2162    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2163    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2164      cd          pointers to tables etc
2165    
2166  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2167  */  */
2168    
2169  static BOOL  static BOOL
2170  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2171    BOOL utf8)    BOOL utf8, compile_data *cd)
2172  {  {
2173  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2174    {    {
2175    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2176        return FALSE;
2177    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2178    }    }
2179  return TRUE;  return TRUE;
# Line 1534  return TRUE; Line 2186  return TRUE;
2186  *************************************************/  *************************************************/
2187    
2188  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2189  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
2190  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2191  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2192    
2193    Originally, this function only recognized a sequence of letters between the
2194    terminators, but it seems that Perl recognizes any sequence of characters,
2195    though of course unknown POSIX names are subsequently rejected. Perl gives an
2196    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2197    didn't consider this to be a POSIX class. Likewise for [:1234:].
2198    
2199    The problem in trying to be exactly like Perl is in the handling of escapes. We
2200    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2201    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2202    below handles the special case of \], but does not try to do any other escape
2203    processing. This makes it different from Perl for cases such as [:l\ower:]
2204    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2205    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2206    I think.
2207    
2208  Argument:  Arguments:
2209    ptr      pointer to the initial [    ptr      pointer to the initial [
2210    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2211    
2212  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2213  */  */
2214    
2215  static BOOL  static BOOL
2216  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
2217  {  {
2218  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2219  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2220  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2221    {    {
2222    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else
2223    return TRUE;      {
2224        if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2225        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2226          {
2227          *endptr = ptr;
2228          return TRUE;
2229          }
2230        }
2231    }    }
2232  return FALSE;  return FALSE;
2233  }  }
# Line 1581  Returns: a value representing the na Line 2252  Returns: a value representing the na
2252  static int  static int
2253  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
2254  {  {
2255    const char *pn = posix_names;
2256  register int yield = 0;  register int yield = 0;
2257  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2258    {    {
2259    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2260      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
2261      pn += posix_name_lengths[yield] + 1;
2262    yield++;    yield++;
2263    }    }
2264  return -1;  return -1;
# Line 1600  return -1; Line 2273  return -1;
2273  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2274  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2275  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
2276  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
2277  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
2278  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
2279  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
2280  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2281    OP_END.
2282    
2283  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2284  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 1627  adjust_recurse(uschar *group, int adjust Line 2301  adjust_recurse(uschar *group, int adjust
2301    uschar *save_hwm)    uschar *save_hwm)
2302  {  {
2303  uschar *ptr = group;  uschar *ptr = group;
2304    
2305  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
2306    {    {
2307    int offset;    int offset;
# Line 1680  auto_callout(uschar *code, const uschar Line 2355  auto_callout(uschar *code, const uschar
2355  {  {
2356  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2357  *code++ = 255;  *code++ = 255;
2358  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2359  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2360  return code + 2*LINK_SIZE;  return code + 2*LINK_SIZE;
2361  }  }
2362    
# Line 1706  Returns: nothing Line 2381  Returns: nothing
2381  static void  static void
2382  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)
2383  {  {
2384  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2385  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2386  }  }
2387    
# Line 1738  get_othercase_range(unsigned int *cptr, Line 2413  get_othercase_range(unsigned int *cptr,
2413  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2414    
2415  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2416    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2417    
2418  if (c > d) return FALSE;  if (c > d) return FALSE;
2419    
# Line 1747  next = othercase + 1; Line 2422  next = othercase + 1;
2422    
2423  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2424    {    {
2425    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2426    next++;    next++;
2427    }    }
2428    
# Line 1756  for (++c; c <= d; c++) Line 2431  for (++c; c <= d; c++)
2431    
2432  return TRUE;  return TRUE;
2433  }  }
2434    
2435    
2436    
2437    /*************************************************
2438    *        Check a character and a property        *
2439    *************************************************/
2440    
2441    /* This function is called by check_auto_possessive() when a property item
2442    is adjacent to a fixed character.
2443    
2444    Arguments:
2445      c            the character
2446      ptype        the property type
2447      pdata        the data for the type
2448      negated      TRUE if it's a negated property (\P or \p{^)
2449    
2450    Returns:       TRUE if auto-possessifying is OK
2451    */
2452    
2453    static BOOL
2454    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2455    {
2456    const ucd_record *prop = GET_UCD(c);
2457    switch(ptype)
2458      {
2459      case PT_LAMP:
2460      return (prop->chartype == ucp_Lu ||
2461              prop->chartype == ucp_Ll ||
2462              prop->chartype == ucp_Lt) == negated;
2463    
2464      case PT_GC:
2465      return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;
2466    
2467      case PT_PC:
2468      return (pdata == prop->chartype) == negated;
2469    
2470      case PT_SC:
2471      return (pdata == prop->script) == negated;
2472    
2473      /* These are specials */
2474    
2475      case PT_ALNUM:
2476      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2477              _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;
2478    
2479      case PT_SPACE:    /* Perl space */
2480      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2481              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2482              == negated;
2483    
2484      case PT_PXSPACE:  /* POSIX space */
2485      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2486              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2487              c == CHAR_FF || c == CHAR_CR)
2488              == negated;
2489    
2490      case PT_WORD:
2491      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2492              _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2493              c == CHAR_UNDERSCORE) == negated;
2494      }
2495    return FALSE;
2496    }
2497  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2498    
2499    
# Line 1769  whether the next thing could possibly ma Line 2507  whether the next thing could possibly ma
2507  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
2508    
2509  Arguments:  Arguments:
2510    op_code       the repeated op code    previous      pointer to the repeated opcode
   this          data for this item, depends on the opcode  
2511    utf8          TRUE in UTF-8 mode    utf8          TRUE in UTF-8 mode
   utf8_char     used for utf8 character bytes, NULL if not relevant  
2512    ptr           next character in pattern    ptr           next character in pattern
2513    options       options bits    options       options bits
2514    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 1781  Returns: TRUE if possessifying is Line 2517  Returns: TRUE if possessifying is
2517  */  */
2518    
2519  static BOOL  static BOOL
2520  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
2521    const uschar *ptr, int options, compile_data *cd)    int options, compile_data *cd)
2522  {  {
2523  int next;  int c, next;
2524    int op_code = *previous++;
2525    
2526  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
2527    
# Line 1793  if ((options & PCRE_EXTENDED) != 0) Line 2530  if ((options & PCRE_EXTENDED) != 0)
2530    for (;;)    for (;;)
2531      {      {
2532      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2533      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2534        {        {
2535        while (*(++ptr) != 0)        ptr++;
2536          while (*ptr != 0)
2537            {
2538          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2539            ptr++;
2540    #ifdef SUPPORT_UTF8
2541            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2542    #endif
2543            }
2544        }        }
2545      else break;      else break;
2546      }      }
# Line 1805  if ((options & PCRE_EXTENDED) != 0) Line 2549  if ((options & PCRE_EXTENDED) != 0)
2549  /* 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
2550  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2551    
2552  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2553    {    {
2554    int temperrorcode = 0;    int temperrorcode = 0;
2555    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
# Line 1830  if ((options & PCRE_EXTENDED) != 0) Line 2574  if ((options & PCRE_EXTENDED) != 0)
2574    for (;;)    for (;;)
2575      {      {
2576      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2577      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2578        {        {
2579        while (*(++ptr) != 0)        ptr++;
2580          while (*ptr != 0)
2581            {
2582          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2583            ptr++;
2584    #ifdef SUPPORT_UTF8
2585            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2586    #endif
2587            }
2588        }        }
2589      else break;      else break;
2590      }      }
# Line 1841  if ((options & PCRE_EXTENDED) != 0) Line 2592  if ((options & PCRE_EXTENDED) != 0)
2592    
2593  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
2594    
2595  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2596    return FALSE;    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2597        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. */  
   
2598    
2599  /* Handle cases when the next item is a character. */  /* Now compare the next item with the previous opcode. First, handle cases when
2600    the next item is a character. */
2601    
2602  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
2603    {    {
2604    case OP_CHAR:    case OP_CHAR:
2605  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2606    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2607    #else
2608      c = *previous;
2609  #endif  #endif
2610    return item != next;    return c != next;
2611    
2612    /* For CHARNC (caseless character) we must check the other case. If we have    /* For CHARNC (caseless character) we must check the other case. If we have
2613    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
# Line 1866  if (next >= 0) switch(op_code) Line 2615  if (next >= 0) switch(op_code)
2615    
2616    case OP_CHARNC:    case OP_CHARNC:
2617  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2618    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2619    #else
2620      c = *previous;
2621  #endif  #endif
2622    if (item == next) return FALSE;    if (c == next) return FALSE;
2623  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2624    if (utf8)    if (utf8)
2625      {      {
2626      unsigned int othercase;      unsigned int othercase;
2627      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2628  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2629      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
2630  #else  #else
2631      othercase = NOTACHAR;      othercase = NOTACHAR;
2632  #endif  #endif
2633      return (unsigned int)item != othercase;      return (unsigned int)c != othercase;
2634      }      }
2635    else    else
2636  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2637    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2638    
2639    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT, its data is always a single-byte character. */
2640    
2641    case OP_NOT:    case OP_NOT:
2642    if (next < 0) return FALSE;  /* Not a character */    if ((c = *previous) == next) return TRUE;
   if (item == next) return TRUE;  
2643    if ((options & PCRE_CASELESS) == 0) return FALSE;    if ((options & PCRE_CASELESS) == 0) return FALSE;
2644  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2645    if (utf8)    if (utf8)
# Line 1897  if (next >= 0) switch(op_code) Line 2647  if (next >= 0) switch(op_code)
2647      unsigned int othercase;      unsigned int othercase;
2648      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2649  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2650      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
2651  #else  #else
2652      othercase = NOTACHAR;      othercase = NOTACHAR;
2653  #endif  #endif
2654      return (unsigned int)item == othercase;      return (unsigned int)c == othercase;
2655      }      }
2656    else    else
2657  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2658    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2659    
2660      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2661      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
2662    
2663    case OP_DIGIT:    case OP_DIGIT:
2664    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
# Line 1925  if (next >= 0) switch(op_code) Line 2678  if (next >= 0) switch(op_code)
2678    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
2679    return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;    return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;
2680    
2681      case OP_HSPACE:
2682      case OP_NOT_HSPACE:
2683      switch(next)
2684        {
2685        case 0x09:
2686        case 0x20:
2687        case 0xa0:
2688        case 0x1680:
2689        case 0x180e:
2690        case 0x2000:
2691        case 0x2001:
2692        case 0x2002:
2693        case 0x2003:
2694        case 0x2004:
2695        case 0x2005:
2696        case 0x2006:
2697        case 0x2007:
2698        case 0x2008:
2699        case 0x2009:
2700        case 0x200A:
2701        case 0x202f:
2702        case 0x205f:
2703        case 0x3000:
2704        return op_code == OP_NOT_HSPACE;
2705        default:
2706        return op_code != OP_NOT_HSPACE;
2707        }
2708    
2709      case OP_ANYNL:
2710      case OP_VSPACE:
2711      case OP_NOT_VSPACE:
2712      switch(next)
2713        {
2714        case 0x0a:
2715        case 0x0b:
2716        case 0x0c:
2717        case 0x0d:
2718        case 0x85:
2719        case 0x2028:
2720        case 0x2029:
2721        return op_code == OP_NOT_VSPACE;
2722        default:
2723        return op_code != OP_NOT_VSPACE;
2724        }
2725    
2726    #ifdef SUPPORT_UCP
2727      case OP_PROP:
2728      return check_char_prop(next, previous[0], previous[1], FALSE);
2729    
2730      case OP_NOTPROP:
2731      return check_char_prop(next, previous[0], previous[1], TRUE);
2732    #endif
2733    
2734    default:    default:
2735    return FALSE;    return FALSE;
2736    }    }
2737    
2738    
2739  /* 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
2740    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
2741    generated only when PCRE_UCP is *not* set, that is, when only ASCII
2742    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
2743    replaced by OP_PROP codes when PCRE_UCP is set. */
2744    
2745  switch(op_code)  switch(op_code)
2746    {    {
2747    case OP_CHAR:    case OP_CHAR:
2748    case OP_CHARNC:    case OP_CHARNC:
2749  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2750    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2751    #else
2752      c = *previous;
2753  #endif  #endif
2754    switch(-next)    switch(-next)
2755      {      {
2756      case ESC_d:      case ESC_d:
2757      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
2758    
2759      case ESC_D:      case ESC_D:
2760      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
2761    
2762      case ESC_s:      case ESC_s:
2763      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
2764    
2765      case ESC_S:      case ESC_S:
2766      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
2767    
2768      case ESC_w:      case ESC_w:
2769      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
2770    
2771      case ESC_W:      case ESC_W:
2772      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
2773    
2774        case ESC_h:
2775        case ESC_H:
2776        switch(c)
2777          {
2778          case 0x09:
2779          case 0x20:
2780          case 0xa0:
2781          case 0x1680:
2782          case 0x180e:
2783          case 0x2000:
2784          case 0x2001:
2785          case 0x2002:
2786          case 0x2003:
2787          case 0x2004:
2788          case 0x2005:
2789          case 0x2006:
2790          case 0x2007:
2791          case 0x2008:
2792          case 0x2009:
2793          case 0x200A:
2794          case 0x202f:
2795          case 0x205f:
2796          case 0x3000:
2797          return -next != ESC_h;
2798          default:
2799          return -next == ESC_h;
2800          }
2801    
2802        case ESC_v:
2803        case ESC_V:
2804        switch(c)
2805          {
2806          case 0x0a:
2807          case 0x0b:
2808          case 0x0c:
2809          case 0x0d:
2810          case 0x85:
2811          case 0x2028:
2812          case 0x2029:
2813          return -next != ESC_v;
2814          default:
2815          return -next == ESC_v;
2816          }
2817    
2818        /* When PCRE_UCP is set, these values get generated for \d etc. Find
2819        their substitutions and process them. The result will always be either
2820        -ESC_p or -ESC_P. Then fall through to process those values. */
2821    
2822    #ifdef SUPPORT_UCP
2823        case ESC_du:
2824        case ESC_DU:
2825        case ESC_wu:
2826        case ESC_WU:
2827        case ESC_su:
2828        case ESC_SU:
2829          {
2830          int temperrorcode = 0;
2831          ptr = substitutes[-next - ESC_DU];
2832          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
2833          if (temperrorcode != 0) return FALSE;
2834          ptr++;    /* For compatibility */
2835          }
2836        /* Fall through */
2837    
2838        case ESC_p:
2839        case ESC_P:
2840          {
2841          int ptype, pdata, errorcodeptr;
2842          BOOL negated;
2843    
2844          ptr--;      /* Make ptr point at the p or P */
2845          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
2846          if (ptype < 0) return FALSE;
2847          ptr++;      /* Point past the final curly ket */
2848    
2849          /* If the property item is optional, we have to give up. (When generated
2850          from \d etc by PCRE_UCP, this test will have been applied much earlier,
2851          to the original \d etc. At this point, ptr will point to a zero byte. */
2852    
2853          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2854            strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2855              return FALSE;
2856    
2857          /* Do the property check. */
2858    
2859          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
2860          }
2861    #endif
2862    
2863      default:      default:
2864      return FALSE;      return FALSE;
2865      }      }
2866    
2867      /* In principle, support for Unicode properties should be integrated here as
2868      well. It means re-organizing the above code so as to get hold of the property
2869      values before switching on the op-code. However, I wonder how many patterns
2870      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
2871      these op-codes are never generated.) */
2872    
2873    case OP_DIGIT:    case OP_DIGIT:
2874    return next == -ESC_D || next == -ESC_s || next == -ESC_W;    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2875             next == -ESC_h || next == -ESC_v || next == -ESC_R;
2876    
2877    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
2878    return next == -ESC_d;    return next == -ESC_d;
2879    
2880    case OP_WHITESPACE:    case OP_WHITESPACE:
2881    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
2882    
2883    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
2884    return next == -ESC_s;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2885    
2886      case OP_HSPACE:
2887      return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
2888             next == -ESC_w || next == -ESC_v || next == -ESC_R;
2889    
2890      case OP_NOT_HSPACE:
2891      return next == -ESC_h;
2892    
2893      /* Can't have \S in here because VT matches \S (Perl anomaly) */
2894      case OP_ANYNL:
2895      case OP_VSPACE:
2896      return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2897    
2898      case OP_NOT_VSPACE:
2899      return next == -ESC_v || next == -ESC_R;
2900    
2901    case OP_WORDCHAR:    case OP_WORDCHAR:
2902    return next == -ESC_W || next == -ESC_s;    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
2903             next == -ESC_v || next == -ESC_R;
2904    
2905    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
2906    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2040  BOOL inescq = FALSE; Line 2964  BOOL inescq = FALSE;
2964  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
2965  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
2966  const uschar *tempptr;  const uschar *tempptr;
2967    const uschar *nestptr = NULL;
2968  uschar *previous = NULL;  uschar *previous = NULL;
2969  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
2970  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
# Line 2049  uschar classbits[32]; Line 2974  uschar classbits[32];
2974  BOOL class_utf8;  BOOL class_utf8;
2975  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
2976  uschar *class_utf8data;  uschar *class_utf8data;
2977    uschar *class_utf8data_base;
2978  uschar utf8_char[6];  uschar utf8_char[6];
2979  #else  #else
2980  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
2981  uschar *utf8_char = NULL;  uschar *utf8_char = NULL;
2982  #endif  #endif
2983    
2984  #ifdef DEBUG  #ifdef PCRE_DEBUG
2985  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2986  #endif  #endif
2987    
# Line 2088  req_caseopt = ((options & PCRE_CASELESS) Line 3014  req_caseopt = ((options & PCRE_CASELESS)
3014  for (;; ptr++)  for (;; ptr++)
3015    {    {
3016    BOOL negate_class;    BOOL negate_class;
3017      BOOL should_flip_negation;
3018    BOOL possessive_quantifier;    BOOL possessive_quantifier;
3019    BOOL is_quantifier;    BOOL is_quantifier;
3020    BOOL is_recurse;    BOOL is_recurse;
3021      BOOL reset_bracount;
3022    int class_charcount;    int class_charcount;
3023    int class_lastchar;    int class_lastchar;
3024    int newoptions;    int newoptions;
3025    int recno;    int recno;
3026      int refsign;
3027    int skipbytes;    int skipbytes;
3028    int subreqbyte;    int subreqbyte;
3029    int subfirstbyte;    int subfirstbyte;
# Line 2106  for (;; ptr++) Line 3035  for (;; ptr++)
3035    
3036    c = *ptr;    c = *ptr;
3037    
3038      /* If we are at the end of a nested substitution, revert to the outer level
3039      string. Nesting only happens one level deep. */
3040    
3041      if (c == 0 && nestptr != NULL)
3042        {
3043        ptr = nestptr;
3044        nestptr = NULL;
3045        c = *ptr;
3046        }
3047    
3048    /* 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
3049    previous cycle of this loop. */    previous cycle of this loop. */
3050    
3051    if (lengthptr != NULL)    if (lengthptr != NULL)
3052      {      {
3053  #ifdef DEBUG  #ifdef PCRE_DEBUG
3054      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3055  #endif  #endif
3056      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
3057        {        {
3058        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3059        goto FAILED;        goto FAILED;
# Line 2127  for (;; ptr++) Line 3066  for (;; ptr++)
3066      */      */
3067    
3068      if (code < last_code) code = last_code;      if (code < last_code) code = last_code;
3069      *lengthptr += code - last_code;  
3070        /* Paranoid check for integer overflow */
3071    
3072        if (OFLOW_MAX - *lengthptr < code - last_code)
3073          {
3074          *errorcodeptr = ERR20;
3075          goto FAILED;
3076          }
3077    
3078        *lengthptr += (int)(code - last_code);
3079      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));
3080    
3081      /* 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 2154  for (;; ptr++) Line 3102  for (;; ptr++)
3102    /* 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
3103    reference list. */    reference list. */
3104    
3105    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
3106      {      {
3107      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3108      goto FAILED;      goto FAILED;
# Line 2164  for (;; ptr++) Line 3112  for (;; ptr++)
3112    
3113    if (inescq && c != 0)    if (inescq && c != 0)
3114      {      {
3115      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3116        {        {
3117        inescq = FALSE;        inescq = FALSE;
3118        ptr++;        ptr++;
# Line 2190  for (;; ptr++) Line 3138  for (;; ptr++)
3138    /* 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
3139    a quantifier. */    a quantifier. */
3140    
3141    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
3142      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
3143        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
3144    
3145    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
3146         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2206  for (;; ptr++) Line 3155  for (;; ptr++)
3155    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3156      {      {
3157      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
3158      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
3159        {        {
3160        while (*(++ptr) != 0)        ptr++;
3161          while (*ptr != 0)
3162          {          {
3163          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3164            ptr++;
3165    #ifdef SUPPORT_UTF8
3166            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
3167    #endif
3168          }          }
3169        if (*ptr != 0) continue;        if (*ptr != 0) continue;
3170    
# Line 2231  for (;; ptr++) Line 3185  for (;; ptr++)
3185      {      {
3186      /* ===================================================================*/      /* ===================================================================*/
3187      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
3188      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
3189      case ')':      case CHAR_RIGHT_PARENTHESIS:
3190      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
3191      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
3192      *codeptr = code;      *codeptr = code;
3193      *ptrptr = ptr;      *ptrptr = ptr;
3194      if (lengthptr != NULL)      if (lengthptr != NULL)
3195        {        {
3196        *lengthptr += code - last_code;   /* To include callout length */        if (OFLOW_MAX - *lengthptr < code - last_code)
3197            {
3198            *errorcodeptr = ERR20;
3199            goto FAILED;
3200            }
3201          *lengthptr += (int)(code - last_code);   /* To include callout length */
3202        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
3203        }        }
3204      return TRUE;      return TRUE;
# Line 2249  for (;; ptr++) Line 3208  for (;; ptr++)
3208      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
3209      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3210    
3211      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
3212      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3213        {        {
3214        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
# Line 2258  for (;; ptr++) Line 3217  for (;; ptr++)
3217      *code++ = OP_CIRC;      *code++ = OP_CIRC;
3218      break;      break;
3219    
3220      case '$':      case CHAR_DOLLAR_SIGN:
3221      previous = NULL;      previous = NULL;
3222      *code++ = OP_DOLL;      *code++ = OP_DOLL;
3223      break;      break;
# Line 2266  for (;; ptr++) Line 3225  for (;; ptr++)
3225      /* 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
3226      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
3227    
3228      case '.':      case CHAR_DOT:
3229      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3230      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
3231      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3232      previous = code;      previous = code;
3233      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
3234      break;      break;
3235    
3236    
# Line 2286  for (;; ptr++) Line 3245  for (;; ptr++)
3245      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,
3246      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
3247      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.
     */  
3248    
3249      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
3250        default (Perl) mode, it is treated as a data character. */
3251    
3252        case CHAR_RIGHT_SQUARE_BRACKET:
3253        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3254          {
3255          *errorcodeptr = ERR64;
3256          goto FAILED;
3257          }
3258        goto NORMAL_CHAR;
3259    
3260        case CHAR_LEFT_SQUARE_BRACKET:
3261      previous = code;      previous = code;
3262    
3263      /* 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
3264      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. */
3265    
3266      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3267          check_posix_syntax(ptr, &tempptr, cd))           ptr[1] == CHAR_EQUALS_SIGN) &&
3268            check_posix_syntax(ptr, &tempptr))
3269        {        {
3270        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
3271        goto FAILED;        goto FAILED;
3272        }        }
3273    
3274      /* 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,
3275        if the first few characters (either before or after ^) are \Q\E or \E we
3276        skip them too. This makes for compatibility with Perl. */
3277    
3278      if ((c = *(++ptr)) == '^')      negate_class = FALSE;
3279        for (;;)
3280        {        {
       negate_class = TRUE;  
3281        c = *(++ptr);        c = *(++ptr);
3282          if (c == CHAR_BACKSLASH)
3283            {
3284            if (ptr[1] == CHAR_E)
3285              ptr++;
3286            else if (strncmp((const char *)ptr+1,
3287                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
3288              ptr += 3;
3289            else
3290              break;
3291            }
3292          else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
3293            negate_class = TRUE;
3294          else break;
3295        }        }
3296      else  
3297        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
3298        an initial ']' is taken as a data character -- the code below handles
3299        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3300        [^] must match any character, so generate OP_ALLANY. */
3301    
3302        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3303            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3304        {        {
3305        negate_class = FALSE;        *code++ = negate_class? OP_ALLANY : OP_FAIL;
3306          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3307          zerofirstbyte = firstbyte;
3308          break;
3309        }        }
3310    
3311        /* If a class contains a negative special such as \S, we need to flip the
3312        negation flag at the end, so that support for characters > 255 works
3313        correctly (they are all included in the class). */
3314    
3315        should_flip_negation = FALSE;
3316    
3317      /* 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
3318      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
3319      valued UTF-8 characters, we don't yet do any optimization. */      valued UTF-8 characters, we don't yet do any optimization. */
# Line 2330  for (;; ptr++) Line 3331  for (;; ptr++)
3331  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3332      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
3333      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
3334        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
3335  #endif  #endif
3336    
3337      /* 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 2345  for (;; ptr++) Line 3347  for (;; ptr++)
3347          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3348          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3349          }          }
3350    
3351          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
3352          data and reset the pointer. This is so that very large classes that
3353          contain a zillion UTF-8 characters no longer overwrite the work space
3354          (which is on the stack). */
3355    
3356          if (lengthptr != NULL)
3357            {
3358            *lengthptr += class_utf8data - class_utf8data_base;
3359            class_utf8data = class_utf8data_base;
3360            }
3361    
3362  #endif  #endif
3363    
3364        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
3365    
3366        if (inescq)        if (inescq)
3367          {          {
3368          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3369            {            {
3370            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
3371            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2366  for (;; ptr++) Line 3380  for (;; ptr++)
3380        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3381        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3382    
3383        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3384            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3385            check_posix_syntax(ptr, &tempptr, cd))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3386          {          {
3387          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3388          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3389          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3390          uschar pbits[32];          uschar pbits[32];
3391    
3392          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3393            {            {
3394            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3395            goto FAILED;            goto FAILED;
3396            }            }
3397    
3398          ptr += 2;          ptr += 2;
3399          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3400            {            {
3401            local_negate = TRUE;            local_negate = TRUE;
3402              should_flip_negation = TRUE;  /* Note negative special */
3403            ptr++;            ptr++;
3404            }            }
3405    
3406          posix_class = check_posix_name(ptr, tempptr - ptr);          posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3407          if (posix_class < 0)          if (posix_class < 0)
3408            {            {
3409            *errorcodeptr = ERR30;            *errorcodeptr = ERR30;
# Line 2402  for (;; ptr++) Line 3417  for (;; ptr++)
3417          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3418            posix_class = 0;            posix_class = 0;
3419    
3420          /* 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
3421          because we may be adding and subtracting from it, and we don't want to          different escape sequences that use Unicode properties. */
3422          subtract bits that may be in the main map already. At the end we or the  
3423          result into the bit map that is being built. */  #ifdef SUPPORT_UCP
3424            if ((options & PCRE_UCP) != 0)
3425              {
3426              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3427              if (posix_substitutes[pc] != NULL)
3428                {
3429                nestptr = tempptr + 1;
3430                ptr = posix_substitutes[pc] - 1;
3431                continue;
3432                }
3433              }
3434    #endif
3435            /* In the non-UCP case, we build the bit map for the POSIX class in a
3436            chunk of local store because we may be adding and subtracting from it,
3437            and we don't want to subtract bits that may be in the main map already.
3438            At the end we or the result into the bit map that is being built. */
3439    
3440          posix_class *= 3;          posix_class *= 3;
3441    
# Line 2449  for (;; ptr++) Line 3479  for (;; ptr++)
3479    
3480        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3481        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
3482        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
3483        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
3484        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
3485        character in them, so set class_charcount bigger than one. */        are either treated as literal characters (by default), or are faulted if
3486          PCRE_EXTRA is set. */
3487    
3488        if (c == '\\')        if (c == CHAR_BACKSLASH)
3489          {          {
3490          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3491          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3492    
3493          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 */  
3494          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3495            {            {
3496            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3497              {              {
3498              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3499              }              }
3500            else inescq = TRUE;            else inescq = TRUE;
3501            continue;            continue;
3502            }            }
3503            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
3504    
3505          if (c < 0)          if (c < 0)
3506            {            {
3507            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
3508            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3509    
3510            /* Save time by not doing this in the pre-compile phase. */            switch (-c)
   
           if (lengthptr == NULL) switch (-c)  
3511              {              {
3512    #ifdef SUPPORT_UCP
3513                case ESC_du:     /* These are the values given for \d etc */
3514                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3515                case ESC_wu:     /* escape sequence with an appropriate \p */
3516                case ESC_WU:     /* or \P to test Unicode properties instead */
3517                case ESC_su:     /* of the default ASCII testing. */
3518                case ESC_SU:
3519                nestptr = ptr;
3520                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3521                class_charcount -= 2;                /* Undo! */
3522                continue;
3523    #endif
3524              case ESC_d:              case ESC_d:
3525              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3526              continue;              continue;
3527    
3528              case ESC_D:              case ESC_D:
3529                should_flip_negation = TRUE;
3530              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
3531              continue;              continue;
3532    
# Line 2494  for (;; ptr++) Line 3535  for (;; ptr++)
3535              continue;              continue;
3536    
3537              case ESC_W:              case ESC_W:
3538                should_flip_negation = TRUE;
3539              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3540              continue;              continue;
3541    
3542                /* Perl 5.004 onwards omits VT from \s, but we must preserve it
3543                if it was previously set by something earlier in the character
3544                class. */
3545    
3546              case ESC_s:              case ESC_s:
3547              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];              classbits[0] |= cbits[cbit_space];
3548              classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= cbits[cbit_space+1] & ~0x08;
3549                for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3550              continue;              continue;
3551    
3552              case ESC_S:              case ESC_S:
3553                should_flip_negation = TRUE;
3554              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
3555              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3556              continue;              continue;
3557    
3558              case ESC_E: /* Perl ignores an orphan \E */              case ESC_h:
3559                SETBIT(classbits, 0x09); /* VT */
3560                SETBIT(classbits, 0x20); /* SPACE */
3561                SETBIT(classbits, 0xa0); /* NSBP */
3562    #ifdef SUPPORT_UTF8
3563                if (utf8)
3564                  {
3565                  class_utf8 = TRUE;
3566                  *class_utf8data++ = XCL_SINGLE;
3567                  class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);
3568                  *class_utf8data++ = XCL_SINGLE;
3569                  class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);
3570                  *class_utf8data++ = XCL_RANGE;
3571                  class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);
3572                  class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);
3573                  *class_utf8data++ = XCL_SINGLE;
3574                  class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);
3575                  *class_utf8data++ = XCL_SINGLE;
3576                  class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);
3577                  *class_utf8data++ = XCL_SINGLE;
3578                  class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);
3579                  }
3580    #endif
3581              continue;              continue;
3582    
3583              default:    /* Not recognized; fall through */              case ESC_H:
3584              break;      /* Need "default" setting to stop compiler warning. */              for (c = 0; c < 32; c++)
3585              }                {
3586                  int x = 0xff;
3587                  switch (c)
3588                    {
3589                    case 0x09/8: x ^= 1 << (0x09%8); break;
3590                    case 0x20/8: x ^= 1 << (0x20%8); break;
3591                    case 0xa0/8: x ^= 1 << (0xa0%8); break;
3592                    default: break;
3593                    }
3594                  classbits[c] |= x;
3595                  }
3596    
3597            /* In the pre-compile phase, just do the recognition. */  #ifdef SUPPORT_UTF8
3598                if (utf8)
3599                  {
3600                  class_utf8 = TRUE;
3601                  *class_utf8data++ = XCL_RANGE;
3602                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3603                  class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);
3604                  *class_utf8data++ = XCL_RANGE;
3605                  class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);
3606                  class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);
3607                  *class_utf8data++ = XCL_RANGE;
3608                  class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);
3609                  class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);
3610                  *class_utf8data++ = XCL_RANGE;
3611                  class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);
3612                  class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);
3613                  *class_utf8data++ = XCL_RANGE;
3614                  class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);
3615                  class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);
3616                  *class_utf8data++ = XCL_RANGE;
3617                  class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);
3618                  class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);
3619                  *class_utf8data++ = XCL_RANGE;
3620                  class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);
3621                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3622                  }
3623    #endif
3624                continue;
3625    
3626            else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||              case ESC_v:
3627                     c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;              SETBIT(classbits, 0x0a); /* LF */
3628                SETBIT(classbits, 0x0b); /* VT */
3629                SETBIT(classbits, 0x0c); /* FF */
3630                SETBIT(classbits, 0x0d); /* CR */
3631                SETBIT(classbits, 0x85); /* NEL */
3632    #ifdef SUPPORT_UTF8
3633                if (utf8)
3634                  {
3635                  class_utf8 = TRUE;
3636                  *class_utf8data++ = XCL_RANGE;
3637                  class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);
3638                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3639                  }
3640    #endif
3641                continue;
3642    
3643            /* We need to deal with \P and \p in both phases. */              case ESC_V:
3644                for (c = 0; c < 32; c++)
3645                  {
3646                  int x = 0xff;
3647                  switch (c)
3648                    {
3649                    case 0x0a/8: x ^= 1 << (0x0a%8);
3650                                 x ^= 1 << (0x0b%8);
3651                                 x ^= 1 << (0x0c%8);
3652                                 x ^= 1 << (0x0d%8);
3653                                 break;
3654                    case 0x85/8: x ^= 1 << (0x85%8); break;
3655                    default: break;
3656                    }
3657                  classbits[c] |= x;
3658                  }
3659    
3660  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UTF8
3661            if (-c == ESC_p || -c == ESC_P)              if (utf8)
3662              {                {
3663              BOOL negated;                class_utf8 = TRUE;
3664              int pdata;                *class_utf8data++ = XCL_RANGE;
3665              int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3666              if (ptype < 0) goto FAILED;                class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);
3667              class_utf8 = TRUE;                *class_utf8data++ = XCL_RANGE;
3668              *class_utf8data++ = ((-c == ESC_p) != negated)?                class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3669                XCL_PROP : XCL_NOTPROP;                class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3670              *class_utf8data++ = ptype;                }
3671              *class_utf8data++ = pdata;  #endif
             class_charcount -= 2;   /* Not a < 256 character */  
3672              continue;              continue;
3673              }  
3674    #ifdef SUPPORT_UCP
3675                case ESC_p:
3676                case ESC_P:
3677                  {
3678                  BOOL negated;
3679                  int pdata;
3680                  int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3681                  if (ptype < 0) goto FAILED;
3682                  class_utf8 = TRUE;
3683                  *class_utf8data++ = ((-c == ESC_p) != negated)?
3684                    XCL_PROP : XCL_NOTPROP;
3685                  *class_utf8data++ = ptype;
3686                  *class_utf8data++ = pdata;
3687                  class_charcount -= 2;   /* Not a < 256 character */
3688                  continue;
3689                  }
3690  #endif  #endif
3691            /* Unrecognized escapes are faulted if PCRE is running in its              /* Unrecognized escapes are faulted if PCRE is running in its
3692            strict mode. By default, for compatibility with Perl, they are              strict mode. By default, for compatibility with Perl, they are
3693            treated as literals. */              treated as literals. */
3694    
3695            if ((options & PCRE_EXTRA) != 0)              default:
3696              {              if ((options & PCRE_EXTRA) != 0)
3697              *errorcodeptr = ERR7;                {
3698              goto FAILED;                *errorcodeptr = ERR7;
3699                  goto FAILED;
3700                  }
3701                class_charcount -= 2;  /* Undo the default count from above */
3702                c = *ptr;              /* Get the final character and fall through */
3703                break;
3704              }              }
   
           class_charcount -= 2;  /* Undo the default count from above */  
           c = *ptr;              /* Get the final character and fall through */  
3705            }            }
3706    
3707          /* 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 2562  for (;; ptr++) Line 3715  for (;; ptr++)
3715        entirely. The code for handling \Q and \E is messy. */        entirely. The code for handling \Q and \E is messy. */
3716    
3717        CHECK_RANGE:        CHECK_RANGE:
3718        while (ptr[1] == '\\' && ptr[2] == 'E')        while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3719          {          {
3720          inescq = FALSE;          inescq = FALSE;
3721          ptr += 2;          ptr += 2;
# Line 2570  for (;; ptr++) Line 3723  for (;; ptr++)
3723    
3724        oldptr = ptr;        oldptr = ptr;
3725    
3726        if (!inescq && ptr[1] == '-')        /* Remember \r or \n */
3727    
3728          if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3729    
3730          /* Check for range */
3731    
3732          if (!inescq && ptr[1] == CHAR_MINUS)
3733          {          {
3734          int d;          int d;
3735          ptr += 2;          ptr += 2;
3736          while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3737    
3738          /* 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
3739          mode. */          mode. */
3740    
3741          while (*ptr == '\\' && ptr[1] == 'Q')          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3742            {            {
3743            ptr += 2;            ptr += 2;
3744            if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }            if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3745                { ptr += 2; continue; }
3746            inescq = TRUE;            inescq = TRUE;
3747            break;            break;
3748            }            }
3749    
3750          if (*ptr == 0 || (!inescq && *ptr == ']'))          if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3751            {            {
3752            ptr = oldptr;            ptr = oldptr;
3753            goto LONE_SINGLE_CHARACTER;            goto LONE_SINGLE_CHARACTER;
# Line 2606  for (;; ptr++) Line 3766  for (;; ptr++)
3766          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
3767          in such circumstances. */          in such circumstances. */
3768    
3769          if (!inescq && d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3770            {            {
3771            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3772            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3773    
3774            /* \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 */  
3775    
3776            if (d < 0)            if (d < 0)
3777              {              {
3778              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  
3779                {                {
3780                ptr = oldptr;                ptr = oldptr;
3781                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 2637  for (;; ptr++) Line 3794  for (;; ptr++)
3794    
3795          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3796    
3797            /* Remember \r or \n */
3798    
3799            if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3800    
3801          /* 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
3802          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
3803          matching for characters > 127 is available only if UCP support is          matching for characters > 127 is available only if UCP support is
# Line 2659  for (;; ptr++) Line 3820  for (;; ptr++)
3820              unsigned int origd = d;              unsigned int origd = d;
3821              while (get_othercase_range(&cc, origd, &occ, &ocd))              while (get_othercase_range(&cc, origd, &occ, &ocd))
3822                {                {
3823                if (occ >= c && ocd <= d) continue;  /* Skip embedded ranges */                if (occ >= (unsigned int)c &&
3824                      ocd <= (unsigned int)d)
3825                    continue;                          /* Skip embedded ranges */
3826    
3827                if (occ < c  && ocd >= c - 1)        /* Extend the basic range */                if (occ < (unsigned int)c  &&
3828                      ocd >= (unsigned int)c - 1)      /* Extend the basic range */
3829                  {                                  /* if there is overlap,   */                  {                                  /* if there is overlap,   */
3830                  c = occ;                           /* noting that if occ < c */                  c = occ;                           /* noting that if occ < c */
3831                  continue;                          /* we can't have ocd > d  */                  continue;                          /* we can't have ocd > d  */
3832                  }                                  /* because a subrange is  */                  }                                  /* because a subrange is  */
3833                if (ocd > d && occ <= d + 1)         /* always shorter than    */                if (ocd > (unsigned int)d &&
3834                      occ <= (unsigned int)d + 1)      /* always shorter than    */
3835                  {                                  /* the basic range.       */                  {                                  /* the basic range.       */
3836                  d = ocd;                  d = ocd;
3837                  continue;                  continue;
# Line 2751  for (;; ptr++) Line 3916  for (;; ptr++)
3916          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3917            {            {
3918            unsigned int othercase;            unsigned int othercase;
3919            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)            if ((othercase = UCD_OTHERCASE(c)) != c)
3920              {              {
3921              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3922              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 2776  for (;; ptr++) Line 3941  for (;; ptr++)
3941          }          }
3942        }        }
3943    
3944      /* 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.
3945        If we are at the end of an internal nested string, revert to the outer
3946        string. */
3947    
3948        while (((c = *(++ptr)) != 0 ||
3949               (nestptr != NULL &&
3950                 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != 0)) &&
3951               (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
3952    
3953      while ((c = *(++ptr)) != 0 && (c != ']' || inescq));      /* Check for missing terminating ']' */
3954    
3955      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)
3956        {        {
3957        *errorcodeptr = ERR6;        *errorcodeptr = ERR6;
3958        goto FAILED;        goto FAILED;
3959        }        }
3960    
3961      /* 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
3962      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
3963      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
3964      because OP_NOT and the related opcodes like OP_NOTSTAR operate on      optimize.
3965      single-bytes only. This is an historical hangover. Maybe one day we can  
3966      tidy these opcodes to handle multi-byte characters.      In UTF-8 mode, we can optimize the negative case only if there were no
3967        characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
3968        operate on single-bytes only. This is an historical hangover. Maybe one day
3969        we can tidy these opcodes to handle multi-byte characters.
3970    
3971      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
3972      1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note      1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note
# Line 2801  for (;; ptr++) Line 3976  for (;; ptr++)
3976      reqbyte, save the previous value for reinstating. */      reqbyte, save the previous value for reinstating. */
3977    
3978  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3979      if (class_charcount == 1 &&      if (class_charcount == 1 && !class_utf8 &&
3980            (!utf8 ||        (!utf8 || !negate_class || class_lastchar < 128))
           (!class_utf8 && (!negate_class || class_lastchar < 128))))  
   
3981  #else  #else
3982      if (class_charcount == 1)      if (class_charcount == 1)
3983  #endif  #endif
# Line 2847  for (;; ptr++) Line 4020  for (;; ptr++)
4020      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
4021    
4022      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
4023      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode, unless there was a negated special
4024      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
4025        characters > 255 are in the class, so any that were explicitly given as
4026        well can be ignored. If (when there are explicit characters > 255 that must
4027        be listed) there are no characters < 256, we can omit the bitmap in the
4028        actual compiled code. */
4029    
4030  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4031      if (class_utf8)      if (class_utf8 && (!should_flip_negation || (options & PCRE_UCP) != 0))
4032        {        {
4033        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
4034        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 2877  for (;; ptr++) Line 4054  for (;; ptr++)
4054        }        }
4055  #endif  #endif
4056    
4057      /* 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
4058      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
4059      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
4060      setting must remain unchanged after any kind of repeat. */      (non-UCP) in the class. Then copy the 32-byte map into the code vector,
4061        negating it if necessary. */
4062    
4063        *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
4064      if (negate_class)      if (negate_class)
4065        {        {
       *code++ = OP_NCLASS;  
4066        if (lengthptr == NULL)    /* Save time in the pre-compile phase */        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
4067          for (c = 0; c < 32; c++) code[c] = ~classbits[c];          for (c = 0; c < 32; c++) code[c] = ~classbits[c];
4068        }        }
4069      else      else
4070        {        {
       *code++ = OP_CLASS;  
4071        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
4072        }        }
4073      code += 32;      code += 32;
# Line 2901  for (;; ptr++) Line 4078  for (;; ptr++)
4078      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
4079      has been tested above. */      has been tested above. */
4080    
4081      case '{':      case CHAR_LEFT_CURLY_BRACKET:
4082      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
4083      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
4084      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
4085      goto REPEAT;      goto REPEAT;
4086    
4087      case '*':      case CHAR_ASTERISK:
4088      repeat_min = 0;      repeat_min = 0;
4089      repeat_max = -1;      repeat_max = -1;
4090      goto REPEAT;      goto REPEAT;
4091    
4092      case '+':      case CHAR_PLUS:
4093      repeat_min = 1;      repeat_min = 1;
4094      repeat_max = -1;      repeat_max = -1;
4095      goto REPEAT;      goto REPEAT;
4096    
4097      case '?':      case CHAR_QUESTION_MARK:
4098      repeat_min = 0;      repeat_min = 0;
4099      repeat_max = 1;      repeat_max = 1;
4100    
# Line 2952  for (;; ptr++) Line 4129  for (;; ptr++)
4129      but if PCRE_UNGREEDY is set, it works the other way round. We change the      but if PCRE_UNGREEDY is set, it works the other way round. We change the
4130      repeat type to the non-default. */      repeat type to the non-default. */
4131    
4132      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
4133        {        {
4134        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
4135        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
4136        ptr++;        ptr++;
4137        }        }
4138      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
4139        {        {
4140        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
4141        ptr++;        ptr++;
# Line 3005  for (;; ptr++) Line 4182  for (;; ptr++)
4182    
4183        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4184            repeat_max < 0 &&            repeat_max < 0 &&
4185            check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
             options, cd))  
4186          {          {
4187          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4188          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3027  for (;; ptr++) Line 4203  for (;; ptr++)
4203        c = previous[1];        c = previous[1];
4204        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4205            repeat_max < 0 &&            repeat_max < 0 &&
4206            check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4207          {          {
4208          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4209          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3051  for (;; ptr++) Line 4227  for (;; ptr++)
4227    
4228        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4229            repeat_max < 0 &&            repeat_max < 0 &&
4230            check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4231          {          {
4232          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4233          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3073  for (;; ptr++) Line 4249  for (;; ptr++)
4249    
4250        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
4251    
4252          /*--------------------------------------------------------------------*/
4253          /* This code is obsolete from release 8.00; the restriction was finally
4254          removed: */
4255    
4256        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4257        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4258    
4259        if (repeat_max != 1) cd->nopartial = TRUE;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4260          /*--------------------------------------------------------------------*/
4261    
4262        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
4263    
# Line 3223  for (;; ptr++) Line 4404  for (;; ptr++)
4404          goto END_REPEAT;          goto END_REPEAT;
4405          }          }
4406    
4407          /*--------------------------------------------------------------------*/
4408          /* This code is obsolete from release 8.00; the restriction was finally
4409          removed: */
4410    
4411        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4412        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4413    
4414        if (repeat_max != 1) cd->nopartial = TRUE;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4415          /*--------------------------------------------------------------------*/
4416    
4417        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4418          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 3251  for (;; ptr++) Line 4437  for (;; ptr++)
4437        {        {
4438        register int i;        register int i;
4439        int ketoffset = 0;        int ketoffset = 0;
4440        int len = code - previous;        int len = (int)(code - previous);
4441        uschar *bralink = NULL;        uschar *bralink = NULL;
4442    
4443        /* Repeating a DEFINE group is pointless */        /* Repeating a DEFINE group is pointless */
# Line 3262  for (;; ptr++) Line 4448  for (;; ptr++)
4448          goto FAILED;          goto FAILED;
4449          }          }
4450    
       /* This is a paranoid check to stop integer overflow later on */  
   
       if (len > MAX_DUPLENGTH)  
         {  
         *errorcodeptr = ERR50;  
         goto FAILED;  
         }  
   
4451        /* If the maximum repeat count is unlimited, find the end of the bracket        /* If the maximum repeat count is unlimited, find the end of the bracket
4452        by scanning through from the start, and compute the offset back to it        by scanning through from the start, and compute the offset back to it
4453        from the current code pointer. There may be an OP_OPT setting following        from the current code pointer. There may be an OP_OPT setting following
# Line 3280  for (;; ptr++) Line 4458  for (;; ptr++)
4458          {          {
4459          register uschar *ket = previous;          register uschar *ket = previous;
4460          do ket += GET(ket, 1); while (*ket != OP_KET);          do ket += GET(ket, 1); while (*ket != OP_KET);
4461          ketoffset = code - ket;          ketoffset = (int)(code - ket);
4462          }          }
4463    
4464        /* The case of a zero minimum is special because of the need to stick        /* The case of a zero minimum is special because of the need to stick
# Line 3292  for (;; ptr++) Line 4470  for (;; ptr++)
4470    
4471        if (repeat_min == 0)        if (repeat_min == 0)
4472          {          {
4473          /* If the maximum is also zero, we just omit the group from the output          /* If the maximum is also zero, we used to just omit the group from the
4474          altogether. */          output altogether, like this:
4475    
4476          if (repeat_max == 0)          ** if (repeat_max == 0)
4477            {          **   {
4478            code = previous;          **   code = previous;
4479            goto END_REPEAT;          **   goto END_REPEAT;
4480            }          **   }
4481    
4482            However, that fails when a group is referenced as a subroutine from
4483            elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it
4484            so that it is skipped on execution. As we don't have a list of which
4485            groups are referenced, we cannot do this selectively.
4486    
4487            If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4488            and do no more at this point. However, we do need to adjust any
4489            OP_RECURSE calls inside the group that refer to the group itself or any
4490            internal or forward referenced group, because the offset is from the
4491            start of the whole regex. Temporarily terminate the pattern while doing
4492            this. */
4493    
4494          /* If the maximum is 1 or unlimited, we just have to stick in the          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
         BRAZERO and do no more at this point. However, we do need to adjust  
         any OP_RECURSE calls inside the group that refer to the group itself or  
         any internal or forward referenced group, because the offset is from  
         the start of the whole regex. Temporarily terminate the pattern while  
         doing this. */  
   
         if (repeat_max <= 1)  
4495            {            {
4496            *code = OP_END;            *code = OP_END;
4497            adjust_recurse(previous, 1, utf8, cd, save_hwm);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
4498            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
4499            code++;            code++;
4500              if (repeat_max == 0)
4501                {
4502                *previous++ = OP_SKIPZERO;
4503                goto END_REPEAT;
4504                }
4505            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4506            }            }
4507    
# Line 3338  for (;; ptr++) Line 4526  for (;; ptr++)
4526            /* We chain together the bracket offset fields that have to be            /* We chain together the bracket offset fields that have to be
4527            filled in later when the ends of the brackets are reached. */            filled in later when the ends of the brackets are reached. */
4528    
4529            offset = (bralink == NULL)? 0 : previous - bralink;            offset = (bralink == NULL)? 0 : (int)(previous - bralink);
4530            bralink = previous;            bralink = previous;
4531            PUTINC(previous, 0, offset);            PUTINC(previous, 0, offset);
4532            }            }
# Line 3358  for (;; ptr++) Line 4546  for (;; ptr++)
4546          if (repeat_min > 1)          if (repeat_min > 1)
4547            {            {
4548            /* In the pre-compile phase, we don't actually do the replication. We            /* In the pre-compile phase, we don't actually do the replication. We
4549            just adjust the length as if we had. */            just adjust the length as if we had. Do some paranoid checks for
4550              potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
4551              integer type when available, otherwise double. */
4552    
4553            if (lengthptr != NULL)            if (lengthptr != NULL)
4554              *lengthptr += (repeat_min - 1)*length_prevgroup;              {
4555                int delta = (repeat_min - 1)*length_prevgroup;
4556                if ((INT64_OR_DOUBLE)(repeat_min - 1)*
4557                      (INT64_OR_DOUBLE)length_prevgroup >
4558                        (INT64_OR_DOUBLE)INT_MAX ||
4559                    OFLOW_MAX - *lengthptr < delta)
4560                  {
4561                  *errorcodeptr = ERR20;
4562                  goto FAILED;
4563                  }
4564                *lengthptr += delta;
4565                }
4566    
4567            /* This is compiling for real */            /* This is compiling for real */
4568    
# Line 3399  for (;; ptr++) Line 4600  for (;; ptr++)
4600          /* In the pre-compile phase, we don't actually do the replication. We          /* In the pre-compile phase, we don't actually do the replication. We
4601          just adjust the length as if we had. For each repetition we must add 1          just adjust the length as if we had. For each repetition we must add 1
4602          to the length for BRAZERO and for all but the last repetition we must          to the length for BRAZERO and for all but the last repetition we must
4603          add 2 + 2*LINKSIZE to allow for the nesting that occurs. */          add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
4604            paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
4605            a 64-bit integer type when available, otherwise double. */
4606    
4607          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4608            *lengthptr += repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            {
4609              2 - 2*LINK_SIZE;  /* Last one doesn't nest */            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4610                          2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4611              if ((INT64_OR_DOUBLE)repeat_max *
4612                    (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4613                      > (INT64_OR_DOUBLE)INT_MAX ||
4614                  OFLOW_MAX - *lengthptr < delta)
4615                {
4616                *errorcodeptr = ERR20;
4617                goto FAILED;
4618                }
4619              *lengthptr += delta;
4620              }
4621    
4622          /* This is compiling for real */          /* This is compiling for real */
4623    
# Line 3421  for (;; ptr++) Line 4635  for (;; ptr++)
4635              {              {
4636              int offset;              int offset;
4637              *code++ = OP_BRA;              *code++ = OP_BRA;
4638              offset = (bralink == NULL)? 0 : code - bralink;              offset = (bralink == NULL)? 0 : (int)(code - bralink);
4639              bralink = code;              bralink = code;
4640              PUTINC(code, 0, offset);              PUTINC(code, 0, offset);
4641              }              }
# Line 3442  for (;; ptr++) Line 4656  for (;; ptr++)
4656          while (bralink != NULL)          while (bralink != NULL)
4657            {            {
4658            int oldlinkoffset;            int oldlinkoffset;
4659            int offset = code - bralink + 1;            int offset = (int)(code - bralink + 1);
4660            uschar *bra = code - offset;            uschar *bra = code - offset;
4661            oldlinkoffset = GET(bra, 1);            oldlinkoffset = GET(bra, 1);
4662            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;
# Line 3473  for (;; ptr++) Line 4687  for (;; ptr++)
4687            uschar *scode = bracode;            uschar *scode = bracode;
4688            do            do
4689              {              {
4690              if (could_be_empty_branch(scode, ketcode, utf8))              if (could_be_empty_branch(scode, ketcode, utf8, cd))
4691                {                {
4692                *bracode += OP_SBRA - OP_BRA;                *bracode += OP_SBRA - OP_BRA;
4693                break;                break;
# Line 3485  for (;; ptr++) Line 4699  for (;; ptr++)
4699          }          }
4700        }        }
4701    
4702        /* If previous is OP_FAIL, it was generated by an empty class [] in
4703        JavaScript mode. The other ways in which OP_FAIL can be generated, that is
4704        by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
4705        error above. We can just ignore the repeat in JS case. */
4706    
4707        else if (*previous == OP_FAIL) goto END_REPEAT;
4708    
4709      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
4710    
4711      else      else
# Line 3509  for (;; ptr++) Line 4730  for (;; ptr++)
4730      if (possessive_quantifier)      if (possessive_quantifier)
4731        {        {
4732        int len;        int len;
4733        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||  
4734            *tempcode == OP_NOTEXACT)        if (*tempcode == OP_TYPEEXACT)
4735            tempcode += _pcre_OP_lengths[*tempcode] +
4736              ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
4737    
4738          else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
4739            {
4740          tempcode += _pcre_OP_lengths[*tempcode];          tempcode += _pcre_OP_lengths[*tempcode];
4741        len = code - tempcode;  #ifdef SUPPORT_UTF8
4742            if (utf8 && tempcode[-1] >= 0xc0)
4743              tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];
4744    #endif
4745            }
4746    
4747          len = (int)(code - tempcode);
4748        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4749          {          {
4750          case OP_STAR:  *tempcode = OP_POSSTAR; break;          case OP_STAR:  *tempcode = OP_POSSTAR; break;
# Line 3530  for (;; ptr++) Line 4762  for (;; ptr++)
4762          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4763          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4764    
4765            /* Because we are moving code along, we must ensure that any
4766            pending recursive references are updated. */
4767    
4768          default:          default:
4769            *code = OP_END;
4770            adjust_recurse(tempcode, 1 + LINK_SIZE, utf8, cd, save_hwm);
4771          memmove(tempcode + 1+LINK_SIZE, tempcode, len);          memmove(tempcode + 1+LINK_SIZE, tempcode, len);
4772          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
4773          len += 1 + LINK_SIZE;          len += 1 + LINK_SIZE;
# Line 3542  for (;; ptr++) Line 4779  for (;; ptr++)
4779          }          }
4780        }        }
4781    
4782      /* In all case we no longer have a previous item. We also set the      /* In all case we no longer have a previous item. We also set the
4783      "follows varying string" flag for subsequently encountered reqbytes if      "follows varying string" flag for subsequently encountered reqbytes if
4784      it isn't already set and we have just passed a varying length item. */      it isn't already set and we have just passed a varying length item. */
4785    
4786        END_REPEAT:
4787        previous = NULL;
4788        cd->req_varyopt |= reqvary;
4789        break;
4790    
4791    
4792        /* ===================================================================*/
4793        /* Start of nested parenthesized sub-expression, or comment or lookahead or
4794        lookbehind or option setting or condition or all the other extended
4795        parenthesis forms.  */
4796    
4797        case CHAR_LEFT_PARENTHESIS:
4798        newoptions = options;
4799        skipbytes = 0;
4800        bravalue = OP_CBRA;
4801        save_hwm = cd->hwm;
4802        reset_bracount = FALSE;
4803    
4804        /* First deal with various "verbs" that can be introduced by '*'. */
4805    
4806        if (*(++ptr) == CHAR_ASTERISK &&
4807             ((cd->ctypes[ptr[1]] & ctype_letter) != 0 || ptr[1] == ':'))
4808          {
4809          int i, namelen;
4810          int arglen = 0;
4811          const char *vn = verbnames;
4812          const uschar *name = ptr + 1;
4813          const uschar *arg = NULL;
4814          previous = NULL;
4815          while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
4816          namelen = (int)(ptr - name);
4817    
4818          if (*ptr == CHAR_COLON)
4819            {
4820            arg = ++ptr;
4821            while ((cd->ctypes[*ptr] & (ctype_letter|ctype_digit)) != 0
4822              || *ptr == '_') ptr++;
4823            arglen = (int)(ptr - arg);
4824            }
4825    
4826          if (*ptr != CHAR_RIGHT_PARENTHESIS)
4827            {
4828            *errorcodeptr = ERR60;
4829            goto FAILED;
4830            }
4831    
4832          /* Scan the table of verb names */
4833    
4834          for (i = 0; i < verbcount; i++)
4835            {
4836            if (namelen == verbs[i].len &&
4837                strncmp((char *)name, vn, namelen) == 0)
4838              {
4839              /* Check for open captures before ACCEPT */
4840    
4841              if (verbs[i].op == OP_ACCEPT)
4842                {
4843                open_capitem *oc;
4844                cd->had_accept = TRUE;
4845                for (oc = cd->open_caps; oc != NULL; oc = oc->next)
4846                  {
4847                  *code++ = OP_CLOSE;
4848                  PUT2INC(code, 0, oc->number);
4849                  }
4850                }
4851    
4852              /* Handle the cases with/without an argument */
4853    
4854              if (arglen == 0)
4855                {
4856                if (verbs[i].op < 0)   /* Argument is mandatory */
4857                  {
4858                  *errorcodeptr = ERR66;
4859                  goto FAILED;
4860                  }
4861                *code = verbs[i].op;
4862                if (*code++ == OP_THEN)
4863                  {
4864                  PUT(code, 0, code - bcptr->current_branch - 1);
4865                  code += LINK_SIZE;
4866                  }
4867                }
4868    
4869              else
4870                {
4871                if (verbs[i].op_arg < 0)   /* Argument is forbidden */
4872                  {
4873                  *errorcodeptr = ERR59;
4874                  goto FAILED;
4875                  }
4876                *code = verbs[i].op_arg;
4877                if (*code++ == OP_THEN_ARG)
4878                  {
4879                  PUT(code, 0, code - bcptr->current_branch - 1);
4880                  code += LINK_SIZE;
4881                  }
4882                *code++ = arglen;
4883                memcpy(code, arg, arglen);
4884                code += arglen;
4885                *code++ = 0;
4886                }
4887    
4888      END_REPEAT:            break;  /* Found verb, exit loop */
4889      previous = NULL;            }
     cd->req_varyopt |= reqvary;  
     break;  
4890    
4891            vn += verbs[i].len + 1;
4892            }
4893    
4894      /* ===================================================================*/        if (i < verbcount) continue;    /* Successfully handled a verb */
4895      /* Start of nested parenthesized sub-expression, or comment or lookahead or        *errorcodeptr = ERR60;          /* Verb not recognized */
4896      lookbehind or option setting or condition or all the other extended        goto FAILED;
4897      parenthesis forms. First deal with the specials; all are introduced by ?,        }
     and the appearance of any of them means that this is not a capturing  
     group. */  
4898    
4899      case '(':      /* Deal with the extended parentheses; all are introduced by '?', and the
4900      newoptions = options;      appearance of any of them means that this is not a capturing group. */
     skipbytes = 0;  
     bravalue = OP_CBRA;  
     save_hwm = cd->hwm;  
4901    
4902      if (*(++ptr) == '?')      else if (*ptr == CHAR_QUESTION_MARK)
4903        {        {
4904        int i, set, unset, namelen;        int i, set, unset, namelen;
4905        int *optset;        int *optset;
# Line 3574  for (;; ptr++) Line 4908  for (;; ptr++)
4908    
4909        switch (*(++ptr))        switch (*(++ptr))
4910          {          {
4911          case '#':                 /* Comment; skip to ket */          case CHAR_NUMBER_SIGN:                 /* Comment; skip to ket */
4912          ptr++;          ptr++;
4913          while (*ptr != 0 && *ptr != ')') ptr++;          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
4914          if (*ptr == 0)          if (*ptr == 0)
4915            {            {
4916            *errorcodeptr = ERR18;            *errorcodeptr = ERR18;
# Line 3586  for (;; ptr++) Line 4920  for (;; ptr++)
4920    
4921    
4922          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4923          case ':':                 /* Non-capturing bracket */          case CHAR_VERTICAL_LINE:  /* Reset capture count for each branch */
4924            reset_bracount = TRUE;
4925            /* Fall through */
4926    
4927            /* ------------------------------------------------------------ */
4928            case CHAR_COLON:          /* Non-capturing bracket */
4929          bravalue = OP_BRA;          bravalue = OP_BRA;
4930          ptr++;          ptr++;
4931          break;          break;
4932    
4933    
4934          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4935          case '(':          case CHAR_LEFT_PARENTHESIS:
4936          bravalue = OP_COND;       /* Conditional group */          bravalue = OP_COND;       /* Conditional group */
4937    
4938          /* A condition can be an assertion, a number (referring to a numbered          /* A condition can be an assertion, a number (referring to a numbered
# Line 3613  for (;; ptr++) Line 4952  for (;; ptr++)
4952          the switch. This will take control down to where bracketed groups,          the switch. This will take control down to where bracketed groups,
4953          including assertions, are processed. */          including assertions, are processed. */
4954    
4955          if (ptr[1] == '?' && (ptr[2] == '=' || ptr[2] == '!' || ptr[2] == '<'))          if (ptr[1] == CHAR_QUESTION_MARK && (ptr[2] == CHAR_EQUALS_SIGN ||
4956                ptr[2] == CHAR_EXCLAMATION_MARK || ptr[2] == CHAR_LESS_THAN_SIGN))
4957            break;            break;
4958    
4959          /* Most other conditions use OP_CREF (a couple change to OP_RREF          /* Most other conditions use OP_CREF (a couple change to OP_RREF
# Line 3621  for (;; ptr++) Line 4961  for (;; ptr++)
4961    
4962          code[1+LINK_SIZE] = OP_CREF;          code[1+LINK_SIZE] = OP_CREF;
4963          skipbytes = 3;          skipbytes = 3;
4964            refsign = -1;
4965    
4966          /* Check for a test for recursion in a named group. */          /* Check for a test for recursion in a named group. */
4967    
4968          if (ptr[1] == 'R' && ptr[2] == '&')          if (ptr[1] == CHAR_R && ptr[2] == CHAR_AMPERSAND)