/[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 226 by ph10, Tue Aug 21 11:46:08 2007 UTC revision 550 by ph10, Sun Oct 10 16:24:11 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 43  supporting internal functions that are n Line 43  supporting internal functions that are n
43    
44    
45  #ifdef HAVE_CONFIG_H  #ifdef HAVE_CONFIG_H
46  #include <config.h>  #include "config.h"
47  #endif  #endif
48    
49  #define NLBLOCK cd             /* Block containing newline information */  #define NLBLOCK cd             /* Block containing newline information */
# Line 53  supporting internal functions that are n Line 53  supporting internal functions that are n
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    
# Line 91  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  -ESC_H,      0,      0, -ESC_K,      0,      0,      0,      0,   /* H - O */       0,                       0,
116  -ESC_P, -ESC_Q, -ESC_R, -ESC_S,      0,      0, -ESC_V, -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  -ESC_h,      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, -ESC_v, -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 130  static const short int escapes[] = { Line 171  static const short int escapes[] = {
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 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
174  /*  D0 */   '}',     0, -ESC_K,       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,-ESC_V, -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,
# Line 140  static const short int escapes[] = { Line 181  static const short int escapes[] = {
181  #endif  #endif
182    
183    
184  /* Table of special "verbs" like (*PRUNE) */  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
185    searched linearly. Put all the names into a single string, in order to reduce
186    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    platforms. */
189    
190  typedef struct verbitem {  typedef struct verbitem {
191    const char *name;    int   len;                 /* Length of verb name */
192    int   len;    int   op;                  /* Op when no arg, or -1 if arg mandatory */
193    int   op;    int   op_arg;              /* Op when arg present, or -1 if not allowed */
194  } verbitem;  } verbitem;
195    
196  static verbitem verbs[] = {  static const char verbnames[] =
197    { "ACCEPT", 6, OP_ACCEPT },    "\0"                       /* Empty name is a shorthand for MARK */
198    { "COMMIT", 6, OP_COMMIT },    STRING_MARK0
199    { "F",      1, OP_FAIL },    STRING_ACCEPT0
200    { "FAIL",   4, OP_FAIL },    STRING_COMMIT0
201    { "PRUNE",  5, OP_PRUNE },    STRING_F0
202    { "SKIP",   4, OP_SKIP  },    STRING_FAIL0
203    { "THEN",   4, OP_THEN  }    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 int verbcount = sizeof(verbs)/sizeof(verbitem);  static const int verbcount = sizeof(verbs)/sizeof(verbitem);
   
220    
 /* Tables of names of POSIX character classes and their lengths. The list is  
 terminated by a zero length entry. The first three must be alpha, lower, upper,  
 as this is assumed for handling case independence. */  
221    
222  static const char *const posix_names[] = {  /* Tables of names of POSIX character classes and their lengths. The names are
223    "alpha", "lower", "upper",  now all in a single string, to reduce the number of relocations when a shared
224    "alnum", "ascii", "blank", "cntrl", "digit", "graph",  library is dynamically loaded. The list of lengths is terminated by a zero
225    "print", "punct", "space", "word",  "xdigit" };  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 200  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 207  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 is 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",    /** DEAD **/    "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 a braced name or an optionally braced non-zero number",    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
399    "(?+ or (?- or (?(+ or (?(- must be followed by a non-zero number",    "a numbered reference must not be zero\0"
400    "(*VERB) with an argument is not supported",    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
401    /* 60 */    /* 60 */
402    "(*VERB) not recognized",    "(*VERB) not recognized\0"
403    "number is too big"    "number is too big\0"
404  };    "subpattern name expected\0"
405      "digit expected after (?+\0"
406      "] is an invalid data character in JavaScript compatibility mode\0"
407      /* 65 */
408      "different names for subpatterns of the same number are not allowed\0"
409      "(*MARK) must have an argument\0"
410      "this version of PCRE is not compiled with PCRE_UCP support\0"
411      ;
412    
413  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
414  patterns. Note that the tables in chartables are dependent on the locale, and  patterns. Note that the tables in chartables are dependent on the locale, and
# Line 303  For convenience, we use the same bit def Line 426  For convenience, we use the same bit def
426    
427  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
428    
429  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC
430    
431    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
432    UTF-8 mode. */
433    
434  static const unsigned char digitab[] =  static const unsigned char digitab[] =
435    {    {
436    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
# Line 339  static const unsigned char digitab[] = Line 466  static const unsigned char digitab[] =
466    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
467    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
468    
469  #else           /* This is the "abnormal" case, for EBCDIC systems */  #else
470    
471    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
472    
473  static const unsigned char digitab[] =  static const unsigned char digitab[] =
474    {    {
475    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
# Line 420  static BOOL Line 550  static BOOL
550    
551    
552  /*************************************************  /*************************************************
553    *            Find an error text                  *
554    *************************************************/
555    
556    /* The error texts are now all in one long string, to save on relocations. As
557    some of the text is of unknown length, we can't use a table of offsets.
558    Instead, just count through the strings. This is not a performance issue
559    because it happens only when there has been a compilation error.
560    
561    Argument:   the error number
562    Returns:    pointer to the error string
563    */
564    
565    static const char *
566    find_error_text(int n)
567    {
568    const char *s = error_texts;
569    for (; n > 0; n--)
570      {
571      while (*s++ != 0) {};
572      if (*s == 0) return "Error text not found (please report)";
573      }
574    return s;
575    }
576    
577    
578    /*************************************************
579  *            Handle escapes                      *  *            Handle escapes                      *
580  *************************************************/  *************************************************/
581    
# Line 458  ptr--; /* Set Line 614  ptr--; /* Set
614    
615  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
616    
617  /* Non-alphamerics are literals. For digits or letters, do an initial lookup in  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
618  a table. A non-zero result is something that can be returned immediately.  in a table. A non-zero result is something that can be returned immediately.
619  Otherwise further processing may be required. */  Otherwise further processing may be required. */
620    
621  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
622  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */
623  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0) c = i;
624    
625  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
626  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */
627  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
628  #endif  #endif
629    
# Line 483  else Line 639  else
639      /* A number of Perl escapes are not handled by PCRE. We give an explicit      /* A number of Perl escapes are not handled by PCRE. We give an explicit
640      error. */      error. */
641    
642      case 'l':      case CHAR_l:
643      case 'L':      case CHAR_L:
644      case 'N':      case CHAR_u:
645      case 'u':      case CHAR_U:
     case 'U':  
646      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
647      break;      break;
648    
649      /* \g must be followed by a number, either plain or braced. If positive, it      /* \g must be followed by one of a number of specific things:
650      is an absolute backreference. If negative, it is a relative backreference.  
651      This is a Perl 5.10 feature. Perl 5.10 also supports \g{name} as a      (1) A number, either plain or braced. If positive, it is an absolute
652      reference to a named group. This is part of Perl's movement towards a      backreference. If negative, it is a relative backreference. This is a Perl
653      unified syntax for back references. As this is synonymous with \k{name}, we      5.10 feature.
654      fudge it up by pretending it really was \k. */  
655        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
656        is part of Perl's movement towards a unified syntax for back references. As
657        this is synonymous with \k{name}, we fudge it up by pretending it really
658        was \k.
659    
660        (3) For Oniguruma compatibility we also support \g followed by a name or a
661        number either in angle brackets or in single quotes. However, these are
662        (possibly recursive) subroutine calls, _not_ backreferences. Just return
663        the -ESC_g code (cf \k). */
664    
665        case CHAR_g:
666        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
667          {
668          c = -ESC_g;
669          break;
670          }
671    
672        /* Handle the Perl-compatible cases */
673    
674      case 'g':      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
     if (ptr[1] == '{')  
675        {        {
676        const uschar *p;        const uschar *p;
677        for (p = ptr+2; *p != 0 && *p != '}'; p++)        for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
678          if (*p != '-' && (digitab[*p] & ctype_digit) == 0) break;          if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
679        if (*p != 0 && *p != '}')        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
680          {          {
681          c = -ESC_k;          c = -ESC_k;
682          break;          break;
# Line 514  else Line 686  else
686        }        }
687      else braced = FALSE;      else braced = FALSE;
688    
689      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
690        {        {
691        negated = TRUE;        negated = TRUE;
692        ptr++;        ptr++;
# Line 523  else Line 695  else
695    
696      c = 0;      c = 0;
697      while ((digitab[ptr[1]] & ctype_digit) != 0)      while ((digitab[ptr[1]] & ctype_digit) != 0)
698        c = c * 10 + *(++ptr) - '0';        c = c * 10 + *(++ptr) - CHAR_0;
699    
700      if (c < 0)      if (c < 0)   /* Integer overflow */
701        {        {
702        *errorcodeptr = ERR61;        *errorcodeptr = ERR61;
703        break;        break;
704        }        }
705    
706      if (c == 0 || (braced && *(++ptr) != '}'))      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
707        {        {
708        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
709        break;        break;
710        }        }
711    
712        if (c == 0)
713          {
714          *errorcodeptr = ERR58;
715          break;
716          }
717    
718      if (negated)      if (negated)
719        {        {
720        if (c > bracount)        if (c > bracount)
# Line 562  else Line 740  else
740      value is greater than 377, the least significant 8 bits are taken. Inside a      value is greater than 377, the least significant 8 bits are taken. Inside a
741      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
742    
743      case '1': case '2': case '3': case '4': case '5':      case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:
744      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
745    
746      if (!isclass)      if (!isclass)
747        {        {
748        oldptr = ptr;        oldptr = ptr;
749        c -= '0';        c -= CHAR_0;
750        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
751          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - CHAR_0;
752        if (c < 0)        if (c < 0)    /* Integer overflow */
753          {          {
754          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
755          break;          break;
# Line 588  else Line 766  else
766      generates a binary zero byte and treats the digit as a following literal.      generates a binary zero byte and treats the digit as a following literal.
767      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
768    
769      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
770        {        {
771        ptr--;        ptr--;
772        c = 0;        c = 0;
# Line 601  else Line 779  else
779      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
780      than 3 octal digits. */      than 3 octal digits. */
781    
782      case '0':      case CHAR_0:
783      c -= '0';      c -= CHAR_0;
784      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
785          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
786      if (!utf8 && c > 255) *errorcodeptr = ERR51;      if (!utf8 && c > 255) *errorcodeptr = ERR51;
787      break;      break;
788    
# Line 612  else Line 790  else
790      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
791      treated as a data character. */      treated as a data character. */
792    
793      case 'x':      case CHAR_x:
794      if (ptr[1] == '{')      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
795        {        {
796        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
797        int count = 0;        int count = 0;
# Line 622  else Line 800  else
800        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
801          {          {
802          register int cc = *pt++;          register int cc = *pt++;
803          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
804          count++;          count++;
805    
806  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
807          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
808          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
809  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
810          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
811          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
812  #endif  #endif
813          }          }
814    
815        if (*pt == '}')        if (*pt == CHAR_RIGHT_CURLY_BRACKET)
816          {          {
817          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
818          ptr = pt;          ptr = pt;
# Line 650  else Line 828  else
828      c = 0;      c = 0;
829      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
830        {        {
831        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
832        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
833  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
834        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
835        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
836  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
837        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
838        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
839  #endif  #endif
840        }        }
841      break;      break;
# Line 666  else Line 844  else
844      This coding is ASCII-specific, but then the whole concept of \cx is      This coding is ASCII-specific, but then the whole concept of \cx is
845      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
846    
847      case 'c':      case CHAR_c:
848      c = *(++ptr);      c = *(++ptr);
849      if (c == 0)      if (c == 0)
850        {        {
# Line 674  else Line 852  else
852        break;        break;
853        }        }
854    
855  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
856      if (c >= 'a' && c <= 'z') c -= 32;      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
857      c ^= 0x40;      c ^= 0x40;
858  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
859      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
860      c ^= 0xC0;      c ^= 0xC0;
861  #endif  #endif
862      break;      break;
863    
864      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
865      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
866      for Perl compatibility, it is a literal. This code looks a bit odd, but      otherwise, for Perl compatibility, it is a literal. This code looks a bit
867      there used to be some cases other than the default, and there may be again      odd, but there used to be some cases other than the default, and there may
868      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
869    
870      default:      default:
871      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 700  else Line 878  else
878      }      }
879    }    }
880    
881    /* Perl supports \N{name} for character names, as well as plain \N for "not
882    newline". PCRE does not support \N{name}. */
883    
884    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET)
885      *errorcodeptr = ERR37;
886    
887    /* If PCRE_UCP is set, we change the values for \d etc. */
888    
889    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
890      c -= (ESC_DU - ESC_D);
891    
892    /* Set the pointer to the final character before returning. */
893    
894  *ptrptr = ptr;  *ptrptr = ptr;
895  return c;  return c;
896  }  }
# Line 740  if (c == 0) goto ERROR_RETURN; Line 931  if (c == 0) goto ERROR_RETURN;
931  /* \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
932  negation. */  negation. */
933    
934  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
935    {    {
936    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
937      {      {
938      *negptr = TRUE;      *negptr = TRUE;
939      ptr++;      ptr++;
# Line 751  if (c == '{') Line 942  if (c == '{')
942      {      {
943      c = *(++ptr);      c = *(++ptr);
944      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
945      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
946      name[i] = c;      name[i] = c;
947      }      }
948    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
949    name[i] = 0;    name[i] = 0;
950    }    }
951    
# Line 776  top = _pcre_utt_size; Line 967  top = _pcre_utt_size;
967  while (bot < top)  while (bot < top)
968    {    {
969    i = (bot + top) >> 1;    i = (bot + top) >> 1;
970    c = strcmp(name, _pcre_utt[i].name);    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);
971    if (c == 0)    if (c == 0)
972      {      {
973      *dptr = _pcre_utt[i].value;      *dptr = _pcre_utt[i].value;
# Line 819  is_counted_repeat(const uschar *p) Line 1010  is_counted_repeat(const uschar *p)
1010  {  {
1011  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
1012  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
1013  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
1014    
1015  if (*p++ != ',') return FALSE;  if (*p++ != CHAR_COMMA) return FALSE;
1016  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
1017    
1018  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
1019  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
1020    
1021  return (*p == '}');  return (*p == CHAR_RIGHT_CURLY_BRACKET);
1022  }  }
1023    
1024    
# Line 860  int max = -1; Line 1051  int max = -1;
1051  /* 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
1052  an integer overflow. */  an integer overflow. */
1053    
1054  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;
1055  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1056    {    {
1057    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 870  if (min < 0 || min > 65535) Line 1061  if (min < 0 || min > 65535)
1061  /* 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.
1062  Also, max must not be less than min. */  Also, max must not be less than min. */
1063    
1064  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1065    {    {
1066    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1067      {      {
1068      max = 0;      max = 0;
1069      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;
1070      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1071        {        {
1072        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 900  return p; Line 1091  return p;
1091    
1092    
1093  /*************************************************  /*************************************************
1094  *       Find forward referenced subpattern       *  *  Subroutine for finding forward reference      *
1095  *************************************************/  *************************************************/
1096    
1097  /* This function scans along a pattern's text looking for capturing  /* This recursive function is called only from find_parens() below. The
1098    top-level call starts at the beginning of the pattern. All other calls must
1099    start at a parenthesis. It scans along a pattern's text looking for capturing
1100  subpatterns, and counting them. If it finds a named pattern that matches the  subpatterns, and counting them. If it finds a named pattern that matches the
1101  name it is given, it returns its number. Alternatively, if the name is NULL, it  name it is given, it returns its number. Alternatively, if the name is NULL, it
1102  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
1103  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
1104  be terminated by '>' because that is checked in the first pass.  first pass. Recursion is used to keep track of subpatterns that reset the
1105    capturing group numbers - the (?| feature.
1106    
1107  Arguments:  Arguments:
1108    ptr          current position in the pattern    ptrptr       address of the current character pointer (updated)
1109    count        current count of capturing parens so far encountered    cd           compile background data
1110    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1111    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1112    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1113      count        pointer to the current capturing subpattern number (updated)
1114    
1115  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1116  */  */
1117    
1118  static int  static int
1119  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,
1120    BOOL xmode)    BOOL xmode, int *count)
1121  {  {
1122  const uschar *thisname;  uschar *ptr = *ptrptr;
1123    int start_count = *count;
1124    int hwm_count = start_count;
1125    BOOL dup_parens = FALSE;
1126    
1127  for (; *ptr != 0; ptr++)  /* If the first character is a parenthesis, check on the type of group we are
1128    dealing with. The very first call may not start with a parenthesis. */
1129    
1130    if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1131    {    {
1132    int term;    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1133    
1134      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1135    
1136      /* Handle a normal, unnamed capturing parenthesis. */
1137    
1138      else if (ptr[1] != CHAR_QUESTION_MARK)
1139        {
1140        *count += 1;
1141        if (name == NULL && *count == lorn) return *count;
1142        ptr++;
1143        }
1144    
1145      /* All cases now have (? at the start. Remember when we are in a group
1146      where the parenthesis numbers are duplicated. */
1147    
1148      else if (ptr[2] == CHAR_VERTICAL_LINE)
1149        {
1150        ptr += 3;
1151        dup_parens = TRUE;
1152        }
1153    
1154      /* Handle comments; all characters are allowed until a ket is reached. */
1155    
1156      else if (ptr[2] == CHAR_NUMBER_SIGN)
1157        {
1158        for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1159        goto FAIL_EXIT;
1160        }
1161    
1162      /* Handle a condition. If it is an assertion, just carry on so that it
1163      is processed as normal. If not, skip to the closing parenthesis of the
1164      condition (there can't be any nested parens). */
1165    
1166      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1167        {
1168        ptr += 2;
1169        if (ptr[1] != CHAR_QUESTION_MARK)
1170          {
1171          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1172          if (*ptr != 0) ptr++;
1173          }
1174        }
1175    
1176      /* Start with (? but not a condition. */
1177    
1178      else
1179        {
1180        ptr += 2;
1181        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1182    
1183        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1184    
1185        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1186            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1187          {
1188          int term;
1189          const uschar *thisname;
1190          *count += 1;
1191          if (name == NULL && *count == lorn) return *count;
1192          term = *ptr++;
1193          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1194          thisname = ptr;
1195          while (*ptr != term) ptr++;
1196          if (name != NULL && lorn == ptr - thisname &&
1197              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1198            return *count;
1199          term++;
1200          }
1201        }
1202      }
1203    
1204    /* Past any initial parenthesis handling, scan for parentheses or vertical
1205    bars. */
1206    
1207    for (; *ptr != 0; ptr++)
1208      {
1209    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1210    
1211    if (*ptr == '\\')    if (*ptr == CHAR_BACKSLASH)
1212      {      {
1213      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1214      if (*ptr == 'Q') for (;;)      if (*ptr == CHAR_Q) for (;;)
1215        {        {
1216        while (*(++ptr) != 0 && *ptr != '\\');        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1217        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1218        if (*(++ptr) == 'E') break;        if (*(++ptr) == CHAR_E) break;
1219        }        }
1220      continue;      continue;
1221      }      }
1222    
1223    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1224      are handled for real. If the first character is '^', skip it. Also, if the
1225      first few characters (either before or after ^) are \Q\E or \E we skip them
1226      too. This makes for compatibility with Perl. Note the use of STR macros to
1227      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1228    
1229    if (*ptr == '[')    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1230      {      {
1231      while (*(++ptr) != ']')      BOOL negate_class = FALSE;
1232        for (;;)
1233          {
1234          if (ptr[1] == CHAR_BACKSLASH)
1235            {
1236            if (ptr[2] == CHAR_E)
1237              ptr+= 2;
1238            else if (strncmp((const char *)ptr+2,
1239                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1240              ptr += 4;
1241            else
1242              break;
1243            }
1244          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1245            {
1246            negate_class = TRUE;
1247            ptr++;
1248            }
1249          else break;
1250          }
1251    
1252        /* If the next character is ']', it is a data character that must be
1253        skipped, except in JavaScript compatibility mode. */
1254    
1255        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1256            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1257          ptr++;
1258    
1259        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1260        {        {
1261        if (*ptr == 0) return -1;        if (*ptr == 0) return -1;
1262        if (*ptr == '\\')        if (*ptr == CHAR_BACKSLASH)
1263          {          {
1264          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) goto FAIL_EXIT;
1265          if (*ptr == 'Q') for (;;)          if (*ptr == CHAR_Q) for (;;)
1266            {            {
1267            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1268            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1269            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1270            }            }
1271          continue;          continue;
1272          }          }
# Line 968  for (; *ptr != 0; ptr++) Line 1276  for (; *ptr != 0; ptr++)
1276    
1277    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1278    
1279    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1280      {      {
1281      while (*(++ptr) != 0 && *ptr != '\n');      while (*(++ptr) != 0 && *ptr != CHAR_NL) {};
1282      if (*ptr == 0) return -1;      if (*ptr == 0) goto FAIL_EXIT;
1283      continue;      continue;
1284      }      }
1285    
1286    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1287    
1288    if (*ptr != '(') continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != '?' && ptr[1] != '*')  
1289      {      {
1290      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, count);
1291      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1292      continue;      if (*ptr == 0) goto FAIL_EXIT;
1293      }      }
1294    
1295    ptr += 2;    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1296    if (*ptr == 'P') ptr++;                      /* Allow optional P */      {
1297        if (dup_parens && *count < hwm_count) *count = hwm_count;
1298        goto FAIL_EXIT;
1299        }
1300    
1301    /* We have to disambiguate (?<! and (?<= from (?<name> */    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1302        {
1303        if (*count > hwm_count) hwm_count = *count;
1304        *count = start_count;
1305        }
1306      }
1307    
1308    if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  FAIL_EXIT:
1309         *ptr != '\'')  *ptrptr = ptr;
1310      continue;  return -1;
1311    }
1312    
1313    
1314    
1315    
1316    /*************************************************
1317    *       Find forward referenced subpattern       *
1318    *************************************************/
1319    
1320    /* This function scans along a pattern's text looking for capturing
1321    subpatterns, and counting them. If it finds a named pattern that matches the
1322    name it is given, it returns its number. Alternatively, if the name is NULL, it
1323    returns when it reaches a given numbered subpattern. This is used for forward
1324    references to subpatterns. We used to be able to start this scan from the
1325    current compiling point, using the current count value from cd->bracount, and
1326    do it all in a single loop, but the addition of the possibility of duplicate
1327    subpattern numbers means that we have to scan from the very start, in order to
1328    take account of such duplicates, and to use a recursive function to keep track
1329    of the different types of group.
1330    
1331    Arguments:
1332      cd           compile background data
1333      name         name to seek, or NULL if seeking a numbered subpattern
1334      lorn         name length, or subpattern number if name is NULL
1335      xmode        TRUE if we are in /x mode
1336    
1337    count++;  Returns:       the number of the found subpattern, or -1 if not found
1338    */
1339    
1340    if (name == NULL && count == lorn) return count;  static int
1341    term = *ptr++;  find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode)
1342    if (term == '<') term = '>';  {
1343    thisname = ptr;  uschar *ptr = (uschar *)cd->start_pattern;
1344    while (*ptr != term) ptr++;  int count = 0;
1345    if (name != NULL && lorn == ptr - thisname &&  int rc;
1346        strncmp((const char *)name, (const char *)thisname, lorn) == 0)  
1347      return count;  /* If the pattern does not start with an opening parenthesis, the first call
1348    to find_parens_sub() will scan right to the end (if necessary). However, if it
1349    does start with a parenthesis, find_parens_sub() will return when it hits the
1350    matching closing parens. That is why we have to have a loop. */
1351    
1352    for (;;)
1353      {
1354      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, &count);
1355      if (rc > 0 || *ptr++ == 0) break;
1356    }    }
1357    
1358  return -1;  return rc;
1359  }  }
1360    
1361    
1362    
1363    
1364  /*************************************************  /*************************************************
1365  *      Find first significant op code            *  *      Find first significant op code            *
1366  *************************************************/  *************************************************/
# Line 1060  for (;;) Line 1410  for (;;)
1410    
1411      case OP_CALLOUT:      case OP_CALLOUT:
1412      case OP_CREF:      case OP_CREF:
1413        case OP_NCREF:
1414      case OP_RREF:      case OP_RREF:
1415        case OP_NRREF:
1416      case OP_DEF:      case OP_DEF:
1417      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1418      break;      break;
# Line 1076  for (;;) Line 1428  for (;;)
1428    
1429    
1430  /*************************************************  /*************************************************
1431  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1432  *************************************************/  *************************************************/
1433    
1434  /* 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,
1435  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.
1436  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
1437    temporarily terminated with OP_END when this function is called.
1438    
1439    This function is called when a backward assertion is encountered, so that if it
1440    fails, the error message can point to the correct place in the pattern.
1441    However, we cannot do this when the assertion contains subroutine calls,
1442    because they can be forward references. We solve this by remembering this case
1443    and doing the check at the end; a flag specifies which mode we are running in.
1444    
1445  Arguments:  Arguments:
1446    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1447    options  the compiling options    options  the compiling options
1448      atend    TRUE if called when the pattern is complete
1449      cd       the "compile data" structure
1450    
1451  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1452                 or -1 if there is no fixed length,
1453               or -2 if \C was encountered               or -2 if \C was encountered
1454                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1455  */  */
1456    
1457  static int  static int
1458  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)
1459  {  {
1460  int length = -1;  int length = -1;
1461    
# Line 1105  branch, check the length against that of Line 1468  branch, check the length against that of
1468  for (;;)  for (;;)
1469    {    {
1470    int d;    int d;
1471      uschar *ce, *cs;
1472    register int op = *cc;    register int op = *cc;
1473    switch (op)    switch (op)
1474      {      {
# Line 1112  for (;;) Line 1476  for (;;)
1476      case OP_BRA:      case OP_BRA:
1477      case OP_ONCE:      case OP_ONCE:
1478      case OP_COND:      case OP_COND:
1479      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);
1480      if (d < 0) return d;      if (d < 0) return d;
1481      branchlength += d;      branchlength += d;
1482      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1135  for (;;) Line 1499  for (;;)
1499      branchlength = 0;      branchlength = 0;
1500      break;      break;
1501    
1502        /* A true recursion implies not fixed length, but a subroutine call may
1503        be OK. If the subroutine is a forward reference, we can't deal with
1504        it until the end of the pattern, so return -3. */
1505    
1506        case OP_RECURSE:
1507        if (!atend) return -3;
1508        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1509        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1510        if (cc > cs && cc < ce) return -1;                /* Recursion */
1511        d = find_fixedlength(cs + 2, options, atend, cd);
1512        if (d < 0) return d;
1513        branchlength += d;
1514        cc += 1 + LINK_SIZE;
1515        break;
1516    
1517      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1518    
1519      case OP_ASSERT:      case OP_ASSERT:
# Line 1148  for (;;) Line 1527  for (;;)
1527    
1528      case OP_REVERSE:      case OP_REVERSE:
1529      case OP_CREF:      case OP_CREF:
1530        case OP_NCREF:
1531      case OP_RREF:      case OP_RREF:
1532        case OP_NRREF:
1533      case OP_DEF:      case OP_DEF:
1534      case OP_OPT:      case OP_OPT:
1535      case OP_CALLOUT:      case OP_CALLOUT:
1536      case OP_SOD:      case OP_SOD:
1537      case OP_SOM:      case OP_SOM:
1538        case OP_SET_SOM:
1539      case OP_EOD:      case OP_EOD:
1540      case OP_EODN:      case OP_EODN:
1541      case OP_CIRC:      case OP_CIRC:
# Line 1171  for (;;) Line 1553  for (;;)
1553      branchlength++;      branchlength++;
1554      cc += 2;      cc += 2;
1555  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1556      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1557        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1558  #endif  #endif
1559      break;      break;
1560    
# Line 1185  for (;;) Line 1565  for (;;)
1565      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1566      cc += 4;      cc += 4;
1567  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1568      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1569        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1570  #endif  #endif
1571      break;      break;
1572    
# Line 1212  for (;;) Line 1590  for (;;)
1590      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1591      case OP_WORDCHAR:      case OP_WORDCHAR:
1592      case OP_ANY:      case OP_ANY:
1593        case OP_ALLANY:
1594      branchlength++;      branchlength++;
1595      cc++;      cc++;
1596      break;      break;
# Line 1266  for (;;) Line 1645  for (;;)
1645    
1646    
1647  /*************************************************  /*************************************************
1648  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1649  *************************************************/  *************************************************/
1650    
1651  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1652  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1653    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1654    so that it can be called from pcre_study() when finding the minimum matching
1655    length.
1656    
1657  Arguments:  Arguments:
1658    code        points to start of expression    code        points to start of expression
1659    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1660    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1661    
1662  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
1663  */  */
1664    
1665  static const uschar *  const uschar *
1666  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1667  {  {
1668  for (;;)  for (;;)
1669    {    {
# Line 1294  for (;;) Line 1676  for (;;)
1676    
1677    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1678    
1679      /* Handle recursion */
1680    
1681      else if (c == OP_REVERSE)
1682        {
1683        if (number < 0) return (uschar *)code;
1684        code += _pcre_OP_lengths[c];
1685        }
1686    
1687    /* Handle capturing bracket */    /* Handle capturing bracket */
1688    
1689    else if (c == OP_CBRA)    else if (c == OP_CBRA)
# Line 1305  for (;;) Line 1695  for (;;)
1695    
1696    /* Otherwise, we can get the item's length from the table, except that for    /* Otherwise, we can get the item's length from the table, except that for
1697    repeated character types, we have to test for \p and \P, which have an extra    repeated character types, we have to test for \p and \P, which have an extra
1698    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1699      must add in its length. */
1700    
1701    else    else
1702      {      {
# Line 1329  for (;;) Line 1720  for (;;)
1720        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
1721        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1722        break;        break;
1723    
1724          case OP_MARK:
1725          case OP_PRUNE_ARG:
1726          case OP_SKIP_ARG:
1727          code += code[1];
1728          break;
1729    
1730          case OP_THEN_ARG:
1731          code += code[1+LINK_SIZE];
1732          break;
1733        }        }
1734    
1735      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1360  for (;;) Line 1761  for (;;)
1761        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1762        break;        break;
1763        }        }
1764    #else
1765        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1766  #endif  #endif
1767      }      }
1768    }    }
# Line 1398  for (;;) Line 1801  for (;;)
1801    
1802    /* Otherwise, we can get the item's length from the table, except that for    /* Otherwise, we can get the item's length from the table, except that for
1803    repeated character types, we have to test for \p and \P, which have an extra    repeated character types, we have to test for \p and \P, which have an extra
1804    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1805      must add in its length. */
1806    
1807    else    else
1808      {      {
# Line 1422  for (;;) Line 1826  for (;;)
1826        case OP_TYPEEXACT:        case OP_TYPEEXACT:
1827        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1828        break;        break;
1829    
1830          case OP_MARK:
1831          case OP_PRUNE_ARG:
1832          case OP_SKIP_ARG:
1833          code += code[1];
1834          break;
1835    
1836          case OP_THEN_ARG:
1837          code += code[1+LINK_SIZE];
1838          break;
1839        }        }
1840    
1841      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1453  for (;;) Line 1867  for (;;)
1867        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1868        break;        break;
1869        }        }
1870    #else
1871        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1872  #endif  #endif
1873      }      }
1874    }    }
# Line 1468  for (;;) Line 1884  for (;;)
1884  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()
1885  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
1886  group that can match nothing. Note that first_significant_code() skips over  group that can match nothing. Note that first_significant_code() skips over
1887  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
1888  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
1889    bracket whose current branch will already have been scanned.
1890    
1891  Arguments:  Arguments:
1892    code        points to start of search    code        points to start of search
1893    endcode     points to where to stop    endcode     points to where to stop
1894    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
1895      cd          contains pointers to tables etc.
1896    
1897  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
1898  */  */
1899    
1900  static BOOL  static BOOL
1901  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
1902      compile_data *cd)
1903  {  {
1904  register int c;  register int c;
1905  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 1491  for (code = first_significant_code(code Line 1910  for (code = first_significant_code(code
1910    
1911    c = *code;    c = *code;
1912    
1913      /* Skip over forward assertions; the other assertions are skipped by
1914      first_significant_code() with a TRUE final argument. */
1915    
1916      if (c == OP_ASSERT)
1917        {
1918        do code += GET(code, 1); while (*code == OP_ALT);
1919        c = *code;
1920        continue;
1921        }
1922    
1923    /* Groups with zero repeats can of course be empty; skip them. */    /* Groups with zero repeats can of course be empty; skip them. */
1924    
1925    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)
1926      {      {
1927      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1928      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
# Line 1501  for (code = first_significant_code(code Line 1930  for (code = first_significant_code(code
1930      continue;      continue;
1931      }      }
1932    
1933      /* For a recursion/subroutine call, if its end has been reached, which
1934      implies a subroutine call, we can scan it. */
1935    
1936      if (c == OP_RECURSE)
1937        {
1938        BOOL empty_branch = FALSE;
1939        const uschar *scode = cd->start_code + GET(code, 1);
1940        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
1941        do
1942          {
1943          if (could_be_empty_branch(scode, endcode, utf8, cd))
1944            {
1945            empty_branch = TRUE;
1946            break;
1947            }
1948          scode += GET(scode, 1);
1949          }
1950        while (*scode == OP_ALT);
1951        if (!empty_branch) return FALSE;  /* All branches are non-empty */
1952        continue;
1953        }
1954    
1955    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
1956    
1957    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)
# Line 1508  for (code = first_significant_code(code Line 1959  for (code = first_significant_code(code
1959      BOOL empty_branch;      BOOL empty_branch;
1960      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
1961    
1962      /* Scan a closed bracket */      /* If a conditional group has only one branch, there is a second, implied,
1963        empty branch, so just skip over the conditional, because it could be empty.
1964        Otherwise, scan the individual branches of the group. */
1965    
1966      empty_branch = FALSE;      if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
     do  
       {  
       if (!empty_branch && could_be_empty_branch(code, endcode, utf8))  
         empty_branch = TRUE;  
1967        code += GET(code, 1);        code += GET(code, 1);
1968        else
1969          {
1970          empty_branch = FALSE;
1971          do
1972            {
1973            if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
1974              empty_branch = TRUE;
1975            code += GET(code, 1);
1976            }
1977          while (*code == OP_ALT);
1978          if (!empty_branch) return FALSE;   /* All branches are non-empty */
1979        }        }
1980      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
1981      c = *code;      c = *code;
1982      continue;      continue;
1983      }      }
# Line 1579  for (code = first_significant_code(code Line 2038  for (code = first_significant_code(code
2038      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2039      case OP_WORDCHAR:      case OP_WORDCHAR:
2040      case OP_ANY:      case OP_ANY:
2041        case OP_ALLANY:
2042      case OP_ANYBYTE:      case OP_ANYBYTE:
2043      case OP_CHAR:      case OP_CHAR:
2044      case OP_CHARNC:      case OP_CHARNC:
# Line 1596  for (code = first_significant_code(code Line 2056  for (code = first_significant_code(code
2056      case OP_TYPEPOSPLUS:      case OP_TYPEPOSPLUS:
2057      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2058      return FALSE;      return FALSE;
2059    
2060      /* These are going to continue, as they may be empty, but we have to      /* These are going to continue, as they may be empty, but we have to
2061      fudge the length for the \p and \P cases. */      fudge the length for the \p and \P cases. */
2062    
2063      case OP_TYPESTAR:      case OP_TYPESTAR:
2064      case OP_TYPEMINSTAR:      case OP_TYPEMINSTAR:
2065      case OP_TYPEPOSSTAR:      case OP_TYPEPOSSTAR:
# Line 1607  for (code = first_significant_code(code Line 2067  for (code = first_significant_code(code
2067      case OP_TYPEMINQUERY:      case OP_TYPEMINQUERY:
2068      case OP_TYPEPOSQUERY:      case OP_TYPEPOSQUERY:
2069      if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;      if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2070      break;      break;
2071    
2072      /* Same for these */      /* Same for these */
2073    
2074      case OP_TYPEUPTO:      case OP_TYPEUPTO:
2075      case OP_TYPEMINUPTO:      case OP_TYPEMINUPTO:
2076      case OP_TYPEPOSUPTO:      case OP_TYPEPOSUPTO:
# Line 1635  for (code = first_significant_code(code Line 2095  for (code = first_significant_code(code
2095      case OP_QUERY:      case OP_QUERY:
2096      case OP_MINQUERY:      case OP_MINQUERY:
2097      case OP_POSQUERY:      case OP_POSQUERY:
2098        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2099        break;
2100    
2101      case OP_UPTO:      case OP_UPTO:
2102      case OP_MINUPTO:      case OP_MINUPTO:
2103      case OP_POSUPTO:      case OP_POSUPTO:
2104      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2105      break;      break;
2106  #endif  #endif
2107    
2108        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2109        string. */
2110    
2111        case OP_MARK:
2112        case OP_PRUNE_ARG:
2113        case OP_SKIP_ARG:
2114        code += code[1];
2115        break;
2116    
2117        case OP_THEN_ARG:
2118        code += code[1+LINK_SIZE];
2119        break;
2120    
2121        /* None of the remaining opcodes are required to match a character. */
2122    
2123        default:
2124        break;
2125      }      }
2126    }    }
2127    
# Line 1663  Arguments: Line 2144  Arguments:
2144    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2145    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2146    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2147      cd          pointers to tables etc
2148    
2149  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2150  */  */
2151    
2152  static BOOL  static BOOL
2153  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2154    BOOL utf8)    BOOL utf8, compile_data *cd)
2155  {  {
2156  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2157    {    {
2158    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2159        return FALSE;
2160    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2161    }    }
2162  return TRUE;  return TRUE;
# Line 1686  return TRUE; Line 2169  return TRUE;
2169  *************************************************/  *************************************************/
2170    
2171  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2172  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
2173  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2174  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2175    
2176    Originally, this function only recognized a sequence of letters between the
2177    terminators, but it seems that Perl recognizes any sequence of characters,
2178    though of course unknown POSIX names are subsequently rejected. Perl gives an
2179    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2180    didn't consider this to be a POSIX class. Likewise for [:1234:].
2181    
2182    The problem in trying to be exactly like Perl is in the handling of escapes. We
2183    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2184    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2185    below handles the special case of \], but does not try to do any other escape
2186    processing. This makes it different from Perl for cases such as [:l\ower:]
2187    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2188    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2189    I think.
2190    
2191  Argument:  Arguments:
2192    ptr      pointer to the initial [    ptr      pointer to the initial [
2193    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2194    
2195  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2196  */  */
2197    
2198  static BOOL  static BOOL
2199  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
2200  {  {
2201  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2202  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2203  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2204    {    {
2205    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else
2206    return TRUE;      {
2207        if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2208        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2209          {
2210          *endptr = ptr;
2211          return TRUE;
2212          }
2213        }
2214    }    }
2215  return FALSE;  return FALSE;
2216  }  }
# Line 1733  Returns: a value representing the na Line 2235  Returns: a value representing the na
2235  static int  static int
2236  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
2237  {  {
2238    const char *pn = posix_names;
2239  register int yield = 0;  register int yield = 0;
2240  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2241    {    {
2242    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2243      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
2244      pn += posix_name_lengths[yield] + 1;
2245    yield++;    yield++;
2246    }    }
2247  return -1;  return -1;
# Line 1752  return -1; Line 2256  return -1;
2256  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2257  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2258  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
2259  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
2260  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
2261  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
2262  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
2263  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2264    OP_END.
2265    
2266  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2267  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 1833  auto_callout(uschar *code, const uschar Line 2338  auto_callout(uschar *code, const uschar
2338  {  {
2339  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2340  *code++ = 255;  *code++ = 255;
2341  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2342  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2343  return code + 2*LINK_SIZE;  return code + 2*LINK_SIZE;
2344  }  }
2345    
# Line 1859  Returns: nothing Line 2364  Returns: nothing
2364  static void  static void
2365  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)
2366  {  {
2367  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2368  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2369  }  }
2370    
# Line 1891  get_othercase_range(unsigned int *cptr, Line 2396  get_othercase_range(unsigned int *cptr,
2396  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2397    
2398  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2399    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2400    
2401  if (c > d) return FALSE;  if (c > d) return FALSE;
2402    
# Line 1900  next = othercase + 1; Line 2405  next = othercase + 1;
2405    
2406  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2407    {    {
2408    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2409    next++;    next++;
2410    }    }
2411    
# Line 1909  for (++c; c <= d; c++) Line 2414  for (++c; c <= d; c++)
2414    
2415  return TRUE;  return TRUE;
2416  }  }
2417    
2418    
2419    
2420    /*************************************************
2421    *        Check a character and a property        *
2422    *************************************************/
2423    
2424    /* This function is called by check_auto_possessive() when a property item
2425    is adjacent to a fixed character.
2426    
2427    Arguments:
2428      c            the character
2429      ptype        the property type
2430      pdata        the data for the type
2431      negated      TRUE if it's a negated property (\P or \p{^)
2432    
2433    Returns:       TRUE if auto-possessifying is OK
2434    */
2435    
2436    static BOOL
2437    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2438    {
2439    const ucd_record *prop = GET_UCD(c);
2440    switch(ptype)
2441      {
2442      case PT_LAMP:
2443      return (prop->chartype == ucp_Lu ||
2444              prop->chartype == ucp_Ll ||
2445              prop->chartype == ucp_Lt) == negated;
2446    
2447      case PT_GC:
2448      return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;
2449    
2450      case PT_PC:
2451      return (pdata == prop->chartype) == negated;
2452    
2453      case PT_SC:
2454      return (pdata == prop->script) == negated;
2455    
2456      /* These are specials */
2457    
2458      case PT_ALNUM:
2459      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2460              _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;
2461    
2462      case PT_SPACE:    /* Perl space */
2463      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2464              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2465              == negated;
2466    
2467      case PT_PXSPACE:  /* POSIX space */
2468      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2469              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2470              c == CHAR_FF || c == CHAR_CR)
2471              == negated;
2472    
2473      case PT_WORD:
2474      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2475              _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2476              c == CHAR_UNDERSCORE) == negated;
2477      }
2478    return FALSE;
2479    }
2480  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2481    
2482    
# Line 1922  whether the next thing could possibly ma Line 2490  whether the next thing could possibly ma
2490  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
2491    
2492  Arguments:  Arguments:
2493    op_code       the repeated op code    previous      pointer to the repeated opcode
   this          data for this item, depends on the opcode  
2494    utf8          TRUE in UTF-8 mode    utf8          TRUE in UTF-8 mode
   utf8_char     used for utf8 character bytes, NULL if not relevant  
2495    ptr           next character in pattern    ptr           next character in pattern
2496    options       options bits    options       options bits
2497    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 1934  Returns: TRUE if possessifying is Line 2500  Returns: TRUE if possessifying is
2500  */  */
2501    
2502  static BOOL  static BOOL
2503  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
2504    const uschar *ptr, int options, compile_data *cd)    int options, compile_data *cd)
2505  {  {
2506  int next;  int c, next;
2507    int op_code = *previous++;
2508    
2509  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
2510    
# Line 1946  if ((options & PCRE_EXTENDED) != 0) Line 2513  if ((options & PCRE_EXTENDED) != 0)
2513    for (;;)    for (;;)
2514      {      {
2515      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2516      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2517        {        {
2518        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2519          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
# Line 1958  if ((options & PCRE_EXTENDED) != 0) Line 2525  if ((options & PCRE_EXTENDED) != 0)
2525  /* 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
2526  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2527    
2528  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2529    {    {
2530    int temperrorcode = 0;    int temperrorcode = 0;
2531    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
# Line 1983  if ((options & PCRE_EXTENDED) != 0) Line 2550  if ((options & PCRE_EXTENDED) != 0)
2550    for (;;)    for (;;)
2551      {      {
2552      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2553      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2554        {        {
2555        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2556          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
# Line 1994  if ((options & PCRE_EXTENDED) != 0) Line 2561  if ((options & PCRE_EXTENDED) != 0)
2561    
2562  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
2563    
2564  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2565    return FALSE;    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2566        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. */  
   
2567    
2568  /* Handle cases when the next item is a character. */  /* Now compare the next item with the previous opcode. First, handle cases when
2569    the next item is a character. */
2570    
2571  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
2572    {    {
2573    case OP_CHAR:    case OP_CHAR:
2574  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2575    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2576    #else
2577      c = *previous;
2578  #endif  #endif
2579    return item != next;    return c != next;
2580    
2581    /* 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
2582    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 2019  if (next >= 0) switch(op_code) Line 2584  if (next >= 0) switch(op_code)
2584    
2585    case OP_CHARNC:    case OP_CHARNC:
2586  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2587    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2588    #else
2589      c = *previous;
2590  #endif  #endif
2591    if (item == next) return FALSE;    if (c == next) return FALSE;
2592  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2593    if (utf8)    if (utf8)
2594      {      {
2595      unsigned int othercase;      unsigned int othercase;
2596      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2597  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2598      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
2599  #else  #else
2600      othercase = NOTACHAR;      othercase = NOTACHAR;
2601  #endif  #endif
2602      return (unsigned int)item != othercase;      return (unsigned int)c != othercase;
2603      }      }
2604    else    else
2605  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2606    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2607    
2608    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT, its data is always a single-byte character. */
2609    
2610    case OP_NOT:    case OP_NOT:
2611    if (next < 0) return FALSE;  /* Not a character */    if ((c = *previous) == next) return TRUE;
   if (item == next) return TRUE;  
2612    if ((options & PCRE_CASELESS) == 0) return FALSE;    if ((options & PCRE_CASELESS) == 0) return FALSE;
2613  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2614    if (utf8)    if (utf8)
# Line 2050  if (next >= 0) switch(op_code) Line 2616  if (next >= 0) switch(op_code)
2616      unsigned int othercase;      unsigned int othercase;
2617      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2618  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2619      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
2620  #else  #else
2621      othercase = NOTACHAR;      othercase = NOTACHAR;
2622  #endif  #endif
2623      return (unsigned int)item == othercase;      return (unsigned int)c == othercase;
2624      }      }
2625    else    else
2626  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2627    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2628    
2629      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2630      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
2631    
2632    case OP_DIGIT:    case OP_DIGIT:
2633    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
# Line 2101  if (next >= 0) switch(op_code) Line 2670  if (next >= 0) switch(op_code)
2670      case 0x202f:      case 0x202f:
2671      case 0x205f:      case 0x205f:
2672      case 0x3000:      case 0x3000:
2673      return op_code != OP_HSPACE;      return op_code == OP_NOT_HSPACE;
2674      default:      default:
2675      return op_code == OP_HSPACE;      return op_code != OP_NOT_HSPACE;
2676      }      }
2677    
2678      case OP_ANYNL:
2679    case OP_VSPACE:    case OP_VSPACE:
2680    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
2681    switch(next)    switch(next)
# Line 2117  if (next >= 0) switch(op_code) Line 2687  if (next >= 0) switch(op_code)
2687      case 0x85:      case 0x85:
2688      case 0x2028:      case 0x2028:
2689      case 0x2029:      case 0x2029:
2690      return op_code != OP_VSPACE;      return op_code == OP_NOT_VSPACE;
2691      default:      default:
2692      return op_code == OP_VSPACE;      return op_code != OP_NOT_VSPACE;
2693      }      }
2694    
2695    #ifdef SUPPORT_UCP
2696      case OP_PROP:
2697      return check_char_prop(next, previous[0], previous[1], FALSE);
2698    
2699      case OP_NOTPROP:
2700      return check_char_prop(next, previous[0], previous[1], TRUE);
2701    #endif
2702    
2703    default:    default:
2704    return FALSE;    return FALSE;
2705    }    }
2706    
2707    
2708  /* 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
2709    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
2710    generated only when PCRE_UCP is *not* set, that is, when only ASCII
2711    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
2712    replaced by OP_PROP codes when PCRE_UCP is set. */
2713    
2714  switch(op_code)  switch(op_code)
2715    {    {
2716    case OP_CHAR:    case OP_CHAR:
2717    case OP_CHARNC:    case OP_CHARNC:
2718  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2719    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2720    #else
2721      c = *previous;
2722  #endif  #endif
2723    switch(-next)    switch(-next)
2724      {      {
2725      case ESC_d:      case ESC_d:
2726      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
2727    
2728      case ESC_D:      case ESC_D:
2729      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
2730    
2731      case ESC_s:      case ESC_s:
2732      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
2733    
2734      case ESC_S:      case ESC_S:
2735      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
2736    
2737      case ESC_w:      case ESC_w:
2738      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
2739    
2740      case ESC_W:      case ESC_W:
2741      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
2742    
2743      case ESC_h:      case ESC_h:
2744      case ESC_H:      case ESC_H:
2745      switch(item)      switch(c)
2746        {        {
2747        case 0x09:        case 0x09:
2748        case 0x20:        case 0x20:
# Line 2186  switch(op_code) Line 2770  switch(op_code)
2770    
2771      case ESC_v:      case ESC_v:
2772      case ESC_V:      case ESC_V:
2773      switch(item)      switch(c)
2774        {        {
2775        case 0x0a:        case 0x0a:
2776        case 0x0b:        case 0x0b:
# Line 2200  switch(op_code) Line 2784  switch(op_code)
2784        return -next == ESC_v;        return -next == ESC_v;
2785        }        }
2786    
2787        /* When PCRE_UCP is set, these values get generated for \d etc. Find
2788        their substitutions and process them. The result will always be either
2789        -ESC_p or -ESC_P. Then fall through to process those values. */
2790    
2791    #ifdef SUPPORT_UCP
2792        case ESC_du:
2793        case ESC_DU:
2794        case ESC_wu:
2795        case ESC_WU:
2796        case ESC_su:
2797        case ESC_SU:
2798          {
2799          int temperrorcode = 0;
2800          ptr = substitutes[-next - ESC_DU];
2801          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
2802          if (temperrorcode != 0) return FALSE;
2803          ptr++;    /* For compatibility */
2804          }
2805        /* Fall through */
2806    
2807        case ESC_p:
2808        case ESC_P:
2809          {
2810          int ptype, pdata, errorcodeptr;
2811          BOOL negated;
2812    
2813          ptr--;      /* Make ptr point at the p or P */
2814          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
2815          if (ptype < 0) return FALSE;
2816          ptr++;      /* Point past the final curly ket */
2817    
2818          /* If the property item is optional, we have to give up. (When generated
2819          from \d etc by PCRE_UCP, this test will have been applied much earlier,
2820          to the original \d etc. At this point, ptr will point to a zero byte. */
2821    
2822          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2823            strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2824              return FALSE;
2825    
2826          /* Do the property check. */
2827    
2828          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
2829          }
2830    #endif
2831    
2832      default:      default:
2833      return FALSE;      return FALSE;
2834      }      }
2835    
2836      /* In principle, support for Unicode properties should be integrated here as
2837      well. It means re-organizing the above code so as to get hold of the property
2838      values before switching on the op-code. However, I wonder how many patterns
2839      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
2840      these op-codes are never generated.) */
2841    
2842    case OP_DIGIT:    case OP_DIGIT:
2843    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2844           next == -ESC_h || next == -ESC_v;           next == -ESC_h || next == -ESC_v || next == -ESC_R;
2845    
2846    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
2847    return next == -ESC_d;    return next == -ESC_d;
2848    
2849    case OP_WHITESPACE:    case OP_WHITESPACE:
2850    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
2851    
2852    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
2853    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2854    
2855    case OP_HSPACE:    case OP_HSPACE:
2856    return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
2857             next == -ESC_w || next == -ESC_v || next == -ESC_R;
2858    
2859    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
2860    return next == -ESC_h;    return next == -ESC_h;
2861    
2862    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
2863      case OP_ANYNL:
2864    case OP_VSPACE:    case OP_VSPACE:
2865    return next == -ESC_V || next == -ESC_d || next == -ESC_w;    return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2866    
2867    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
2868    return next == -ESC_v;    return next == -ESC_v || next == -ESC_R;
2869    
2870    case OP_WORDCHAR:    case OP_WORDCHAR:
2871    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
2872             next == -ESC_v || next == -ESC_R;
2873    
2874    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
2875    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2295  BOOL inescq = FALSE; Line 2933  BOOL inescq = FALSE;
2933  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
2934  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
2935  const uschar *tempptr;  const uschar *tempptr;
2936    const uschar *nestptr = NULL;
2937  uschar *previous = NULL;  uschar *previous = NULL;
2938  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
2939  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
# Line 2304  uschar classbits[32]; Line 2943  uschar classbits[32];
2943  BOOL class_utf8;  BOOL class_utf8;
2944  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
2945  uschar *class_utf8data;  uschar *class_utf8data;
2946    uschar *class_utf8data_base;
2947  uschar utf8_char[6];  uschar utf8_char[6];
2948  #else  #else
2949  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
2950  uschar *utf8_char = NULL;  uschar *utf8_char = NULL;
2951  #endif  #endif
2952    
2953  #ifdef DEBUG  #ifdef PCRE_DEBUG
2954  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2955  #endif  #endif
2956    
# Line 2343  req_caseopt = ((options & PCRE_CASELESS) Line 2983  req_caseopt = ((options & PCRE_CASELESS)
2983  for (;; ptr++)  for (;; ptr++)
2984    {    {
2985    BOOL negate_class;    BOOL negate_class;
2986      BOOL should_flip_negation;
2987    BOOL possessive_quantifier;    BOOL possessive_quantifier;
2988    BOOL is_quantifier;    BOOL is_quantifier;
2989    BOOL is_recurse;    BOOL is_recurse;
# Line 2363  for (;; ptr++) Line 3004  for (;; ptr++)
3004    
3005    c = *ptr;    c = *ptr;
3006    
3007      /* If we are at the end of a nested substitution, revert to the outer level
3008      string. Nesting only happens one level deep. */
3009    
3010      if (c == 0 && nestptr != NULL)
3011        {
3012        ptr = nestptr;
3013        nestptr = NULL;
3014        c = *ptr;
3015        }
3016    
3017    /* 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
3018    previous cycle of this loop. */    previous cycle of this loop. */
3019    
3020    if (lengthptr != NULL)    if (lengthptr != NULL)
3021      {      {
3022  #ifdef DEBUG  #ifdef PCRE_DEBUG
3023      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3024  #endif  #endif
3025      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
3026        {        {
3027        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3028        goto FAILED;        goto FAILED;
# Line 2393  for (;; ptr++) Line 3044  for (;; ptr++)
3044        goto FAILED;        goto FAILED;
3045        }        }
3046    
3047      *lengthptr += code - last_code;      *lengthptr += (int)(code - last_code);
3048      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));
3049    
3050      /* 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 2420  for (;; ptr++) Line 3071  for (;; ptr++)
3071    /* 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
3072    reference list. */    reference list. */
3073    
3074    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
3075      {      {
3076      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3077      goto FAILED;      goto FAILED;
# Line 2430  for (;; ptr++) Line 3081  for (;; ptr++)
3081    
3082    if (inescq && c != 0)    if (inescq && c != 0)
3083      {      {
3084      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3085        {        {
3086        inescq = FALSE;        inescq = FALSE;
3087        ptr++;        ptr++;
# Line 2456  for (;; ptr++) Line 3107  for (;; ptr++)
3107    /* 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
3108    a quantifier. */    a quantifier. */
3109    
3110    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
3111      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
3112        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
3113    
3114    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
3115         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2472  for (;; ptr++) Line 3124  for (;; ptr++)
3124    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3125      {      {
3126      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
3127      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
3128        {        {
3129        while (*(++ptr) != 0)        while (*(++ptr) != 0)
3130          {          {
# Line 2497  for (;; ptr++) Line 3149  for (;; ptr++)
3149      {      {
3150      /* ===================================================================*/      /* ===================================================================*/
3151      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
3152      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
3153      case ')':      case CHAR_RIGHT_PARENTHESIS:
3154      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
3155      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
3156      *codeptr = code;      *codeptr = code;
# Line 2510  for (;; ptr++) Line 3162  for (;; ptr++)
3162          *errorcodeptr = ERR20;          *errorcodeptr = ERR20;
3163          goto FAILED;          goto FAILED;
3164          }          }
3165        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += (int)(code - last_code);   /* To include callout length */
3166        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
3167        }        }
3168      return TRUE;      return TRUE;
# Line 2520  for (;; ptr++) Line 3172  for (;; ptr++)
3172      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
3173      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3174    
3175      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
3176      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3177        {        {
3178        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
# Line 2529  for (;; ptr++) Line 3181  for (;; ptr++)
3181      *code++ = OP_CIRC;      *code++ = OP_CIRC;
3182      break;      break;
3183    
3184      case '$':      case CHAR_DOLLAR_SIGN:
3185      previous = NULL;      previous = NULL;
3186      *code++ = OP_DOLL;      *code++ = OP_DOLL;
3187      break;      break;
# Line 2537  for (;; ptr++) Line 3189  for (;; ptr++)
3189      /* 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
3190      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
3191    
3192      case '.':      case CHAR_DOT:
3193      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3194      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
3195      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3196      previous = code;      previous = code;
3197      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
3198      break;      break;
3199    
3200    
# Line 2557  for (;; ptr++) Line 3209  for (;; ptr++)
3209      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,
3210      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
3211      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.
     */  
3212    
3213      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
3214        default (Perl) mode, it is treated as a data character. */
3215    
3216        case CHAR_RIGHT_SQUARE_BRACKET:
3217        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3218          {
3219          *errorcodeptr = ERR64;
3220          goto FAILED;
3221          }
3222        goto NORMAL_CHAR;
3223    
3224        case CHAR_LEFT_SQUARE_BRACKET:
3225      previous = code;      previous = code;
3226    
3227      /* 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
3228      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. */
3229    
3230      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3231          check_posix_syntax(ptr, &tempptr, cd))           ptr[1] == CHAR_EQUALS_SIGN) &&
3232            check_posix_syntax(ptr, &tempptr))
3233        {        {
3234        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
3235        goto FAILED;        goto FAILED;
3236        }        }
3237    
# Line 2580  for (;; ptr++) Line 3243  for (;; ptr++)
3243      for (;;)      for (;;)
3244        {        {
3245        c = *(++ptr);        c = *(++ptr);
3246        if (c == '\\')        if (c == CHAR_BACKSLASH)
3247          {          {
3248          if (ptr[1] == 'E') ptr++;          if (ptr[1] == CHAR_E)
3249            else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;            ptr++;
3250              else break;          else if (strncmp((const char *)ptr+1,
3251                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
3252              ptr += 3;
3253            else
3254              break;
3255          }          }
3256        else if (!negate_class && c == '^')        else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
3257          negate_class = TRUE;          negate_class = TRUE;
3258        else break;        else break;
3259        }        }
3260    
3261        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
3262        an initial ']' is taken as a data character -- the code below handles
3263        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3264        [^] must match any character, so generate OP_ALLANY. */
3265    
3266        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3267            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3268          {
3269          *code++ = negate_class? OP_ALLANY : OP_FAIL;
3270          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3271          zerofirstbyte = firstbyte;
3272          break;
3273          }
3274    
3275        /* If a class contains a negative special such as \S, we need to flip the
3276        negation flag at the end, so that support for characters > 255 works
3277        correctly (they are all included in the class). */
3278    
3279        should_flip_negation = FALSE;
3280    
3281      /* 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
3282      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
3283      valued UTF-8 characters, we don't yet do any optimization. */      valued UTF-8 characters, we don't yet do any optimization. */
# Line 2608  for (;; ptr++) Line 3295  for (;; ptr++)
3295  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3296      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
3297      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
3298        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
3299  #endif  #endif
3300    
3301      /* 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 2623  for (;; ptr++) Line 3311  for (;; ptr++)
3311          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3312          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3313          }          }
3314    
3315          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
3316          data and reset the pointer. This is so that very large classes that
3317          contain a zillion UTF-8 characters no longer overwrite the work space
3318          (which is on the stack). */
3319    
3320          if (lengthptr != NULL)
3321            {
3322            *lengthptr += class_utf8data - class_utf8data_base;
3323            class_utf8data = class_utf8data_base;
3324            }
3325    
3326  #endif  #endif
3327    
3328        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
3329    
3330        if (inescq)        if (inescq)
3331          {          {
3332          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3333            {            {
3334            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
3335            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2644  for (;; ptr++) Line 3344  for (;; ptr++)
3344        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3345        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3346    
3347        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3348            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3349            check_posix_syntax(ptr, &tempptr, cd))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3350          {          {
3351          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3352          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3353          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3354          uschar pbits[32];          uschar pbits[32];
3355    
3356          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3357            {            {
3358            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3359            goto FAILED;            goto FAILED;
3360            }            }
3361    
3362          ptr += 2;          ptr += 2;
3363          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3364            {            {
3365            local_negate = TRUE;            local_negate = TRUE;
3366              should_flip_negation = TRUE;  /* Note negative special */
3367            ptr++;            ptr++;
3368            }            }
3369    
3370          posix_class = check_posix_name(ptr, tempptr - ptr);          posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3371          if (posix_class < 0)          if (posix_class < 0)
3372            {            {
3373            *errorcodeptr = ERR30;            *errorcodeptr = ERR30;
# Line 2680  for (;; ptr++) Line 3381  for (;; ptr++)
3381          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3382            posix_class = 0;            posix_class = 0;
3383    
3384          /* 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
3385          because we may be adding and subtracting from it, and we don't want to          different escape sequences that use Unicode properties. */
3386          subtract bits that may be in the main map already. At the end we or the  
3387          result into the bit map that is being built. */  #ifdef SUPPORT_UCP
3388            if ((options & PCRE_UCP) != 0)
3389              {
3390              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3391              if (posix_substitutes[pc] != NULL)
3392                {
3393                nestptr = tempptr + 1;
3394                ptr = posix_substitutes[pc] - 1;
3395                continue;
3396                }
3397              }
3398    #endif
3399            /* In the non-UCP case, we build the bit map for the POSIX class in a
3400            chunk of local store because we may be adding and subtracting from it,
3401            and we don't want to subtract bits that may be in the main map already.
3402            At the end we or the result into the bit map that is being built. */
3403    
3404          posix_class *= 3;          posix_class *= 3;
3405    
# Line 2727  for (;; ptr++) Line 3443  for (;; ptr++)
3443    
3444        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3445        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
3446        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
3447        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
3448        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
3449        character in them, so set class_charcount bigger than one. */        are either treated as literal characters (by default), or are faulted if
3450          PCRE_EXTRA is set. */
3451    
3452        if (c == '\\')        if (c == CHAR_BACKSLASH)
3453          {          {
3454          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3455          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3456    
3457          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 */  
3458          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3459            {            {
3460            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3461              {              {
3462              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3463              }              }
# Line 2756  for (;; ptr++) Line 3471  for (;; ptr++)
3471            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
3472            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3473    
3474            /* Save time by not doing this in the pre-compile phase. */            switch (-c)
   
           if (lengthptr == NULL) switch (-c)  
3475              {              {
3476    #ifdef SUPPORT_UCP
3477                case ESC_du:     /* These are the values given for \d etc */
3478                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3479                case ESC_wu:     /* escape sequence with an appropriate \p */
3480                case ESC_WU:     /* or \P to test Unicode properties instead */
3481                case ESC_su:     /* of the default ASCII testing. */
3482                case ESC_SU:
3483                nestptr = ptr;
3484                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3485                class_charcount -= 2;                /* Undo! */
3486                continue;
3487    #endif
3488              case ESC_d:              case ESC_d:
3489              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3490              continue;              continue;
3491    
3492              case ESC_D:              case ESC_D:
3493                should_flip_negation = TRUE;
3494              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
3495              continue;              continue;
3496    
# Line 2773  for (;; ptr++) Line 3499  for (;; ptr++)
3499              continue;              continue;
3500    
3501              case ESC_W:              case ESC_W:
3502                should_flip_negation = TRUE;
3503              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3504              continue;              continue;
3505    
# Line 2782  for (;; ptr++) Line 3509  for (;; ptr++)
3509              continue;              continue;
3510    
3511              case ESC_S:              case ESC_S:
3512                should_flip_negation = TRUE;
3513              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
3514              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3515              continue;              continue;
3516    
3517              case ESC_E: /* Perl ignores an orphan \E */              case ESC_h:
             continue;  
   
             default:    /* Not recognized; fall through */  
             break;      /* Need "default" setting to stop compiler warning. */  
             }  
   
           /* In the pre-compile phase, just do the recognition. */  
   
           else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||  
                    c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;  
   
           /* We need to deal with \H, \h, \V, and \v in both phases because  
           they use extra memory. */  
   
           if (-c == ESC_h)  
             {  
3518              SETBIT(classbits, 0x09); /* VT */              SETBIT(classbits, 0x09); /* VT */
3519              SETBIT(classbits, 0x20); /* SPACE */              SETBIT(classbits, 0x20); /* SPACE */
3520              SETBIT(classbits, 0xa0); /* NSBP */              SETBIT(classbits, 0xa0); /* NSBP */
# Line 2826  for (;; ptr++) Line 3538  for (;; ptr++)
3538                }                }
3539  #endif  #endif
3540              continue;              continue;
             }  
3541    
3542            if (-c == ESC_H)              case ESC_H:
             {  
3543              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3544                {                {
3545                int x = 0xff;                int x = 0xff;
# Line 2871  for (;; ptr++) Line 3581  for (;; ptr++)
3581                }                }
3582  #endif  #endif
3583              continue;              continue;
             }  
3584    
3585            if (-c == ESC_v)              case ESC_v:
             {  
3586              SETBIT(classbits, 0x0a); /* LF */              SETBIT(classbits, 0x0a); /* LF */
3587              SETBIT(classbits, 0x0b); /* VT */              SETBIT(classbits, 0x0b); /* VT */
3588              SETBIT(classbits, 0x0c); /* FF */              SETBIT(classbits, 0x0c); /* FF */
# Line 2890  for (;; ptr++) Line 3598  for (;; ptr++)
3598                }                }
3599  #endif  #endif
3600              continue;              continue;
             }  
3601    
3602            if (-c == ESC_V)              case ESC_V:
             {  
3603              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3604                {                {
3605                int x = 0xff;                int x = 0xff;
# Line 2923  for (;; ptr++) Line 3629  for (;; ptr++)
3629                }                }
3630  #endif  #endif
3631              continue;              continue;
             }  
   
           /* We need to deal with \P and \p in both phases. */  
3632    
3633  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3634            if (-c == ESC_p || -c == ESC_P)              case ESC_p:
3635              {              case ESC_P:
3636              BOOL negated;                {
3637              int pdata;                BOOL negated;
3638              int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                int pdata;
3639              if (ptype < 0) goto FAILED;                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3640              class_utf8 = TRUE;                if (ptype < 0) goto FAILED;
3641              *class_utf8data++ = ((-c == ESC_p) != negated)?                class_utf8 = TRUE;
3642                XCL_PROP : XCL_NOTPROP;                *class_utf8data++ = ((-c == ESC_p) != negated)?
3643              *class_utf8data++ = ptype;                  XCL_PROP : XCL_NOTPROP;
3644              *class_utf8data++ = pdata;                *class_utf8data++ = ptype;
3645              class_charcount -= 2;   /* Not a < 256 character */                *class_utf8data++ = pdata;
3646              continue;                class_charcount -= 2;   /* Not a < 256 character */
3647              }                continue;
3648                  }
3649  #endif  #endif
3650            /* Unrecognized escapes are faulted if PCRE is running in its              /* Unrecognized escapes are faulted if PCRE is running in its
3651            strict mode. By default, for compatibility with Perl, they are              strict mode. By default, for compatibility with Perl, they are
3652            treated as literals. */              treated as literals. */
3653    
3654            if ((options & PCRE_EXTRA) != 0)              default:
3655              {              if ((options & PCRE_EXTRA) != 0)
3656              *errorcodeptr = ERR7;                {
3657              goto FAILED;                *errorcodeptr = ERR7;
3658                  goto FAILED;
3659                  }
3660                class_charcount -= 2;  /* Undo the default count from above */
3661                c = *ptr;              /* Get the final character and fall through */
3662                break;
3663              }              }
   
           class_charcount -= 2;  /* Undo the default count from above */  
           c = *ptr;              /* Get the final character and fall through */  
3664            }            }
3665    
3666          /* 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 2968  for (;; ptr++) Line 3674  for (;; ptr++)
3674        entirely. The code for handling \Q and \E is messy. */        entirely. The code for handling \Q and \E is messy. */
3675    
3676        CHECK_RANGE:        CHECK_RANGE:
3677        while (ptr[1] == '\\' && ptr[2] == 'E')        while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3678          {          {
3679          inescq = FALSE;          inescq = FALSE;
3680          ptr += 2;          ptr += 2;
# Line 2976  for (;; ptr++) Line 3682  for (;; ptr++)
3682    
3683        oldptr = ptr;        oldptr = ptr;
3684    
3685        if (!inescq && ptr[1] == '-')        /* Remember \r or \n */
3686    
3687          if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3688    
3689          /* Check for range */
3690    
3691          if (!inescq && ptr[1] == CHAR_MINUS)
3692          {          {
3693          int d;          int d;
3694          ptr += 2;          ptr += 2;
3695          while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3696    
3697          /* 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
3698          mode. */          mode. */
3699    
3700          while (*ptr == '\\' && ptr[1] == 'Q')          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3701            {            {
3702            ptr += 2;            ptr += 2;
3703            if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }            if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3704                { ptr += 2; continue; }
3705            inescq = TRUE;            inescq = TRUE;
3706            break;            break;
3707            }            }
3708    
3709          if (*ptr == 0 || (!inescq && *ptr == ']'))          if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3710            {            {
3711            ptr = oldptr;            ptr = oldptr;
3712            goto LONE_SINGLE_CHARACTER;            goto LONE_SINGLE_CHARACTER;
# Line 3012  for (;; ptr++) Line 3725  for (;; ptr++)
3725          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
3726          in such circumstances. */          in such circumstances. */
3727    
3728          if (!inescq && d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3729            {            {
3730            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3731            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3732    
3733            /* \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 */  
3734    
3735            if (d < 0)            if (d < 0)
3736              {              {
3737              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  
3738                {                {
3739                ptr = oldptr;                ptr = oldptr;
3740                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 3043  for (;; ptr++) Line 3753  for (;; ptr++)
3753    
3754          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3755    
3756            /* Remember \r or \n */
3757    
3758            if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3759    
3760          /* 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
3761          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
3762          matching for characters > 127 is available only if UCP support is          matching for characters > 127 is available only if UCP support is
# Line 3161  for (;; ptr++) Line 3875  for (;; ptr++)
3875          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3876            {            {
3877            unsigned int othercase;            unsigned int othercase;
3878            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)            if ((othercase = UCD_OTHERCASE(c)) != c)
3879              {              {
3880              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3881              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 3186  for (;; ptr++) Line 3900  for (;; ptr++)
3900          }          }
3901        }        }
3902    
3903      /* 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.
3904        If we are at the end of an internal nested string, revert to the outer
3905        string. */
3906    
3907        while (((c = *(++ptr)) != 0 ||
3908               (nestptr != NULL &&
3909                 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != 0)) &&
3910               (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
3911    
3912      while ((c = *(++ptr)) != 0 && (c != ']' || inescq));      /* Check for missing terminating ']' */
3913    
3914      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)
3915        {        {
3916        *errorcodeptr = ERR6;        *errorcodeptr = ERR6;
3917        goto FAILED;        goto FAILED;
3918        }        }
3919    
     /* Remember whether \r or \n are in this class */  
   
     if (negate_class)  
       {  
       if ((classbits[1] & 0x24) != 0x24) cd->external_options |= PCRE_HASCRORLF;  
       }  
     else  
       {  
       if ((classbits[1] & 0x24) != 0) cd->external_options |= PCRE_HASCRORLF;  
       }  
   
3920      /* 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
3921      less than 256. As long as there were no characters >= 128 and there was no      less than 256. As long as there were no characters >= 128 and there was no
3922      use of \p or \P, in other words, no use of any XCLASS features, we can      use of \p or \P, in other words, no use of any XCLASS features, we can
3923      optimize.      optimize.
3924    
3925      In UTF-8 mode, we can optimize the negative case only if there were no      In UTF-8 mode, we can optimize the negative case only if there were no
3926      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
3927      operate on single-bytes only. This is an historical hangover. Maybe one day      operate on single-bytes only. This is an historical hangover. Maybe one day
# Line 3225  for (;; ptr++) Line 3935  for (;; ptr++)
3935      reqbyte, save the previous value for reinstating. */      reqbyte, save the previous value for reinstating. */
3936    
3937  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3938      if (class_charcount == 1 && !class_utf8 &&      if (class_charcount == 1 && !class_utf8 &&
3939        (!utf8 || !negate_class || class_lastchar < 128))        (!utf8 || !negate_class || class_lastchar < 128))
3940  #else  #else
3941      if (class_charcount == 1)      if (class_charcount == 1)
# Line 3269  for (;; ptr++) Line 3979  for (;; ptr++)
3979      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3980    
3981      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
3982      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode, unless there was a negated special
3983      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
3984        characters > 255 are in the class, so any that were explicitly given as
3985        well can be ignored. If (when there are explicit characters > 255 that must
3986        be listed) there are no characters < 256, we can omit the bitmap in the
3987        actual compiled code. */
3988    
3989  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3990      if (class_utf8)      if (class_utf8 && (!should_flip_negation || (options & PCRE_UCP) != 0))
3991        {        {
3992        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
3993        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 3299  for (;; ptr++) Line 4013  for (;; ptr++)
4013        }        }
4014  #endif  #endif
4015    
4016      /* 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
4017      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
4018      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
4019      setting must remain unchanged after any kind of repeat. */      (non-UCP) in the class. Then copy the 32-byte map into the code vector,
4020        negating it if necessary. */
4021    
4022        *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
4023      if (negate_class)      if (negate_class)
4024        {        {
       *code++ = OP_NCLASS;  
4025        if (lengthptr == NULL)    /* Save time in the pre-compile phase */        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
4026          for (c = 0; c < 32; c++) code[c] = ~classbits[c];          for (c = 0; c < 32; c++) code[c] = ~classbits[c];
4027        }        }
4028      else      else
4029        {        {
       *code++ = OP_CLASS;  
4030        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
4031        }        }
4032      code += 32;      code += 32;
# Line 3323  for (;; ptr++) Line 4037  for (;; ptr++)
4037      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
4038      has been tested above. */      has been tested above. */
4039    
4040      case '{':      case CHAR_LEFT_CURLY_BRACKET:
4041      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
4042      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
4043      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
4044      goto REPEAT;      goto REPEAT;
4045    
4046      case '*':      case CHAR_ASTERISK:
4047      repeat_min = 0;      repeat_min = 0;
4048      repeat_max = -1;      repeat_max = -1;
4049      goto REPEAT;      goto REPEAT;
4050    
4051      case '+':      case CHAR_PLUS:
4052      repeat_min = 1;      repeat_min = 1;
4053      repeat_max = -1;      repeat_max = -1;
4054      goto REPEAT;      goto REPEAT;
4055    
4056      case '?':      case CHAR_QUESTION_MARK:
4057      repeat_min = 0;      repeat_min = 0;
4058      repeat_max = 1;      repeat_max = 1;
4059    
# Line 3374  for (;; ptr++) Line 4088  for (;; ptr++)
4088      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
4089      repeat type to the non-default. */      repeat type to the non-default. */
4090    
4091      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
4092        {        {
4093        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
4094        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
4095        ptr++;        ptr++;
4096        }        }
4097      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
4098        {        {
4099        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
4100        ptr++;        ptr++;
# Line 3427  for (;; ptr++) Line 4141  for (;; ptr++)
4141    
4142        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4143            repeat_max < 0 &&            repeat_max < 0 &&
4144            check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
             options, cd))  
4145          {          {
4146          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4147          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3449  for (;; ptr++) Line 4162  for (;; ptr++)
4162        c = previous[1];        c = previous[1];
4163        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4164            repeat_max < 0 &&            repeat_max < 0 &&
4165            check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4166          {          {
4167          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4168          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3473  for (;; ptr++) Line 4186  for (;; ptr++)
4186    
4187        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4188            repeat_max < 0 &&            repeat_max < 0 &&
4189            check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4190          {          {
4191          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4192          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3495  for (;; ptr++) Line 4208  for (;; ptr++)
4208    
4209        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
4210    
4211          /*--------------------------------------------------------------------*/
4212          /* This code is obsolete from release 8.00; the restriction was finally
4213          removed: */
4214    
4215        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4216        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4217    
4218        if (repeat_max != 1) cd->nopartial = TRUE;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4219          /*--------------------------------------------------------------------*/
4220    
4221        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
4222    
# Line 3645  for (;; ptr++) Line 4363  for (;; ptr++)
4363          goto END_REPEAT;          goto END_REPEAT;
4364          }          }
4365    
4366          /*--------------------------------------------------------------------*/
4367          /* This code is obsolete from release 8.00; the restriction was finally
4368          removed: */
4369    
4370        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4371        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4372    
4373        if (repeat_max != 1) cd->nopartial = TRUE;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4374          /*--------------------------------------------------------------------*/
4375    
4376        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4377          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 3673  for (;; ptr++) Line 4396  for (;; ptr++)
4396        {        {
4397        register int i;        register int i;
4398        int ketoffset = 0;        int ketoffset = 0;
4399        int len = code - previous;        int len = (int)(code - previous);
4400        uschar *bralink = NULL;        uschar *bralink = NULL;
4401    
4402        /* Repeating a DEFINE group is pointless */        /* Repeating a DEFINE group is pointless */
# Line 3694  for (;; ptr++) Line 4417  for (;; ptr++)
4417          {          {
4418          register uschar *ket = previous;          register uschar *ket = previous;
4419          do ket += GET(ket, 1); while (*ket != OP_KET);          do ket += GET(ket, 1); while (*ket != OP_KET);
4420          ketoffset = code - ket;          ketoffset = (int)(code - ket);
4421          }          }
4422    
4423        /* 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 3706  for (;; ptr++) Line 4429  for (;; ptr++)
4429    
4430        if (repeat_min == 0)        if (repeat_min == 0)
4431          {          {
4432          /* 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
4433          altogether. */          output altogether, like this:
4434    
4435          if (repeat_max == 0)          ** if (repeat_max == 0)
4436            {          **   {
4437            code = previous;          **   code = previous;
4438            goto END_REPEAT;          **   goto END_REPEAT;
4439            }          **   }
4440    
4441            However, that fails when a group is referenced as a subroutine from
4442            elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it
4443            so that it is skipped on execution. As we don't have a list of which
4444            groups are referenced, we cannot do this selectively.
4445    
4446            If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4447            and do no more at this point. However, we do need to adjust any
4448            OP_RECURSE calls inside the group that refer to the group itself or any
4449            internal or forward referenced group, because the offset is from the
4450            start of the whole regex. Temporarily terminate the pattern while doing
4451            this. */
4452    
4453          /* 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)  
4454            {            {
4455            *code = OP_END;            *code = OP_END;
4456            adjust_recurse(previous, 1, utf8, cd, save_hwm);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
4457            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
4458            code++;            code++;
4459              if (repeat_max == 0)
4460                {
4461                *previous++ = OP_SKIPZERO;
4462                goto END_REPEAT;
4463                }
4464            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4465            }            }
4466    
# Line 3752  for (;; ptr++) Line 4485  for (;; ptr++)
4485            /* We chain together the bracket offset fields that have to be            /* We chain together the bracket offset fields that have to be
4486            filled in later when the ends of the brackets are reached. */            filled in later when the ends of the brackets are reached. */
4487    
4488            offset = (bralink == NULL)? 0 : previous - bralink;            offset = (bralink == NULL)? 0 : (int)(previous - bralink);
4489            bralink = previous;            bralink = previous;
4490            PUTINC(previous, 0, offset);            PUTINC(previous, 0, offset);
4491            }            }
# Line 3773  for (;; ptr++) Line 4506  for (;; ptr++)
4506            {            {
4507            /* 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
4508            just adjust the length as if we had. Do some paranoid checks for            just adjust the length as if we had. Do some paranoid checks for
4509            potential integer overflow. */            potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
4510              integer type when available, otherwise double. */
4511    
4512            if (lengthptr != NULL)            if (lengthptr != NULL)
4513              {              {
4514              int delta = (repeat_min - 1)*length_prevgroup;              int delta = (repeat_min - 1)*length_prevgroup;
4515              if ((double)(repeat_min - 1)*(double)length_prevgroup >              if ((INT64_OR_DOUBLE)(repeat_min - 1)*
4516                                                              (double)INT_MAX ||                    (INT64_OR_DOUBLE)length_prevgroup >
4517                        (INT64_OR_DOUBLE)INT_MAX ||
4518                  OFLOW_MAX - *lengthptr < delta)                  OFLOW_MAX - *lengthptr < delta)
4519                {                {
4520                *errorcodeptr = ERR20;                *errorcodeptr = ERR20;
# Line 3825  for (;; ptr++) Line 4560  for (;; ptr++)
4560          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
4561          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
4562          add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some          add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
4563          paranoid checks to avoid integer overflow. */          paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
4564            a 64-bit integer type when available, otherwise double. */
4565    
4566          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4567            {            {
4568            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4569                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4570            if ((double)repeat_max *            if ((INT64_OR_DOUBLE)repeat_max *
4571                  (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)                  (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4572                    > (double)INT_MAX ||                    > (INT64_OR_DOUBLE)INT_MAX ||
4573                OFLOW_MAX - *lengthptr < delta)                OFLOW_MAX - *lengthptr < delta)
4574              {              {
4575              *errorcodeptr = ERR20;              *errorcodeptr = ERR20;
# Line 3858  for (;; ptr++) Line 4594  for (;; ptr++)
4594              {              {
4595              int offset;              int offset;
4596              *code++ = OP_BRA;              *code++ = OP_BRA;
4597              offset = (bralink == NULL)? 0 : code - bralink;              offset = (bralink == NULL)? 0 : (int)(code - bralink);
4598              bralink = code;              bralink = code;
4599              PUTINC(code, 0, offset);              PUTINC(code, 0, offset);
4600              }              }
# Line 3879  for (;; ptr++) Line 4615  for (;; ptr++)
4615          while (bralink != NULL)          while (bralink != NULL)
4616            {            {
4617            int oldlinkoffset;            int oldlinkoffset;
4618            int offset = code - bralink + 1;            int offset = (int)(code - bralink + 1);
4619            uschar *bra = code - offset;            uschar *bra = code - offset;
4620            oldlinkoffset = GET(bra, 1);            oldlinkoffset = GET(bra, 1);
4621            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;
# Line 3910  for (;; ptr++) Line 4646  for (;; ptr++)
4646            uschar *scode = bracode;            uschar *scode = bracode;
4647            do            do
4648              {              {
4649              if (could_be_empty_branch(scode, ketcode, utf8))              if (could_be_empty_branch(scode, ketcode, utf8, cd))
4650                {                {
4651                *bracode += OP_SBRA - OP_BRA;                *bracode += OP_SBRA - OP_BRA;
4652                break;                break;
# Line 3922  for (;; ptr++) Line 4658  for (;; ptr++)
4658          }          }
4659        }        }
4660    
4661        /* If previous is OP_FAIL, it was generated by an empty class [] in
4662        JavaScript mode. The other ways in which OP_FAIL can be generated, that is
4663        by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
4664        error above. We can just ignore the repeat in JS case. */
4665    
4666        else if (*previous == OP_FAIL) goto END_REPEAT;
4667    
4668      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
4669    
4670      else      else
# Line 3946  for (;; ptr++) Line 4689  for (;; ptr++)
4689      if (possessive_quantifier)      if (possessive_quantifier)
4690        {        {
4691        int len;        int len;
4692        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||  
4693            *tempcode == OP_NOTEXACT)        if (*tempcode == OP_TYPEEXACT)
4694            tempcode += _pcre_OP_lengths[*tempcode] +
4695              ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
4696    
4697          else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
4698            {
4699          tempcode += _pcre_OP_lengths[*tempcode];          tempcode += _pcre_OP_lengths[*tempcode];
4700        len = code - tempcode;  #ifdef SUPPORT_UTF8
4701            if (utf8 && tempcode[-1] >= 0xc0)
4702              tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];
4703    #endif
4704            }
4705    
4706          len = (int)(code - tempcode);
4707        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4708          {          {
4709          case OP_STAR:  *tempcode = OP_POSSTAR; break;          case OP_STAR:  *tempcode = OP_POSSTAR; break;
# Line 3967  for (;; ptr++) Line 4721  for (;; ptr++)
4721          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4722          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4723    
4724            /* Because we are moving code along, we must ensure that any
4725            pending recursive references are updated. */
4726    
4727          default:          default:
4728            *code = OP_END;
4729            adjust_recurse(tempcode, 1 + LINK_SIZE, utf8, cd, save_hwm);
4730          memmove(tempcode + 1+LINK_SIZE, tempcode, len);          memmove(tempcode + 1+LINK_SIZE, tempcode, len);
4731          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
4732          len += 1 + LINK_SIZE;          len += 1 + LINK_SIZE;
# Line 3994  for (;; ptr++) Line 4753  for (;; ptr++)
4753      lookbehind or option setting or condition or all the other extended      lookbehind or option setting or condition or all the other extended
4754      parenthesis forms.  */      parenthesis forms.  */
4755    
4756      case '(':      case CHAR_LEFT_PARENTHESIS:
4757      newoptions = options;      newoptions = options;
4758      skipbytes = 0;      skipbytes = 0;
4759      bravalue = OP_CBRA;      bravalue = OP_CBRA;
# Line 4003  for (;; ptr++) Line 4762  for (;; ptr++)
4762    
4763      /* First deal with various "verbs" that can be introduced by '*'. */      /* First deal with various "verbs" that can be introduced by '*'. */
4764    
4765      if (*(++ptr) == '*' && (cd->ctypes[ptr[1]] & ctype_letter) != 0)      if (*(++ptr) == CHAR_ASTERISK &&
4766             ((cd->ctypes[ptr[1]] & ctype_letter) != 0 || ptr[1] == ':'))
4767        {        {
4768        int i, namelen;        int i, namelen;
4769        const uschar *name = ++ptr;        int arglen = 0;
4770          const char *vn = verbnames;
4771          const uschar *name = ptr + 1;
4772          const uschar *arg = NULL;
4773        previous = NULL;        previous = NULL;
4774        while ((cd->ctypes[*++ptr] & ctype_letter) != 0);        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
4775        if (*ptr == ':')        namelen = (int)(ptr - name);
4776    
4777          if (*ptr == CHAR_COLON)
4778          {          {
4779          *errorcodeptr = ERR59;   /* Not supported */          arg = ++ptr;
4780          goto FAILED;          while ((cd->ctypes[*ptr] & (ctype_letter|ctype_digit)) != 0
4781              || *ptr == '_') ptr++;
4782            arglen = (int)(ptr - arg);
4783          }          }
4784        if (*ptr != ')')  
4785          if (*ptr != CHAR_RIGHT_PARENTHESIS)
4786          {          {
4787          *errorcodeptr = ERR60;          *errorcodeptr = ERR60;
4788          goto FAILED;          goto FAILED;
4789          }          }
4790        namelen = ptr - name;  
4791          /* Scan the table of verb names */
4792    
4793        for (i = 0; i < verbcount; i++)        for (i = 0; i < verbcount; i++)
4794          {          {
4795          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
4796              strncmp((char *)name, verbs[i].name, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
4797            {            {
4798            *code = verbs[i].op;            /* Check for open captures before ACCEPT */
4799            if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;  
4800            break;            if (verbs[i].op == OP_ACCEPT)
4801                {
4802                open_capitem *oc;
4803                cd->had_accept = TRUE;
4804                for (oc = cd->open_caps; oc != NULL; oc = oc->next)
4805                  {
4806                  *code++ = OP_CLOSE;
4807                  PUT2INC(code, 0, oc->number);
4808                  }
4809                }
4810    
4811              /* Handle the cases with/without an argument */
4812    
4813              if (arglen == 0)
4814                {
4815                if (verbs[i].op < 0)   /* Argument is mandatory */
4816                  {
4817                  *errorcodeptr = ERR66;
4818                  goto FAILED;
4819                  }
4820                *code = verbs[i].op;
4821                if (*code++ == OP_THEN)
4822                  {
4823                  PUT(code, 0, code - bcptr->current_branch - 1);
4824                  code += LINK_SIZE;
4825                  }
4826                }
4827    
4828              else
4829                {
4830                if (verbs[i].op_arg < 0)   /* Argument is forbidden */
4831                  {
4832                  *errorcodeptr = ERR59;
4833                  goto FAILED;
4834                  }
4835                *code = verbs[i].op_arg;
4836                if (*code++ == OP_THEN_ARG)
4837                  {
4838                  PUT(code, 0, code - bcptr->current_branch - 1);
4839                  code += LINK_SIZE;
4840                  }
4841                *code++ = arglen;
4842                memcpy(code, arg, arglen);
4843                code += arglen;
4844                *code++ = 0;
4845                }
4846    
4847              break;  /* Found verb, exit loop */
4848            }            }
4849    
4850            vn += verbs[i].len + 1;
4851          }          }
4852        if (i < verbcount) continue;  
4853        *errorcodeptr = ERR60;        if (i < verbcount) continue;    /* Successfully handled a verb */
4854          *errorcodeptr = ERR60;          /* Verb not recognized */
4855        goto FAILED;        goto FAILED;
4856        }        }
4857    
4858      /* Deal with the extended parentheses; all are introduced by '?', and the      /* Deal with the extended parentheses; all are introduced by '?', and the
4859      appearance of any of them means that this is not a capturing group. */      appearance of any of them means that this is not a capturing group. */
4860    
4861      else if (*ptr == '?')      else if (*ptr == CHAR_QUESTION_MARK)
4862        {        {
4863        int i, set, unset, namelen;        int i, set, unset, namelen;
4864        int *optset;        int *optset;
# Line 4047  for (;; ptr++) Line 4867  for (;; ptr++)
4867    
4868        switch (*(++ptr))        switch (*(++ptr))
4869          {          {
4870          case '#':                 /* Comment; skip to ket */          case CHAR_NUMBER_SIGN:                 /* Comment; skip to ket */
4871          ptr++;          ptr++;
4872          while (*ptr != 0 && *ptr != ')') ptr++;          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
4873          if (*ptr == 0)          if (*ptr == 0)
4874            {            {
4875            *errorcodeptr = ERR18;            *errorcodeptr = ERR18;
# Line 4059  for (;; ptr++) Line 4879  for (;; ptr++)
4879    
4880    
4881          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4882          case '|':                 /* Reset capture count for each branch */          case CHAR_VERTICAL_LINE:  /* Reset capture count for each branch */
4883          reset_bracount = TRUE;          reset_bracount = TRUE;
4884          /* Fall through */          /* Fall through */
4885    
4886          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4887          case ':':                 /* Non-capturing bracket */          case CHAR_COLON:          /* Non-capturing bracket */
4888          bravalue = OP_BRA;          bravalue = OP_BRA;
4889          ptr++;          ptr++;
4890          break;          break;
4891    
4892    
4893          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4894          case '(':          case CHAR_LEFT_PARENTHESIS:
4895          bravalue = OP_COND;       /* Conditional group */          bravalue = OP_COND;       /* Conditional group */
4896    
4897          /* 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 4091  for (;; ptr++) Line 4911  for (;; ptr++)
4911          the switch. This will take control down to where bracketed groups,          the switch. This will take control down to where bracketed groups,
4912          including assertions, are processed. */          including assertions, are processed. */
4913    
4914          if (ptr[1] == '?' && (ptr[2] == '=' || ptr[2] == '!' || ptr[2] == '<'))          if (ptr[1] == CHAR_QUESTION_MARK && (ptr[2] == CHAR_EQUALS_SIGN ||
4915                ptr[2] == CHAR_EXCLAMATION_MARK || ptr[2] == CHAR_LESS_THAN_SIGN))
4916            break;            break;
4917    
4918          /* 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 4103  for (;; ptr++) Line 4924  for (;; ptr++)
4924    
4925          /* Check for a test for recursion in a named group. */          /* Check for a test for recursion in a named group. */
4926    
4927          if (ptr[1] == 'R' && ptr[2] == '&')          if (ptr[1] == CHAR_R && ptr[2] == CHAR_AMPERSAND)
4928            {            {
4929            terminator = -1;            terminator = -1;
4930            ptr += 2;            ptr += 2;
# Line 4113  for (;; ptr++) Line 4934  for (;; ptr++)
4934          /* Check for a test for a named group's having been set, using the Perl          /* Check for a test for a named group's having been set, using the Perl
4935          syntax (?(<name>) or (?('name') */          syntax (?(<name>) or (?('name') */
4936    
4937          else if (ptr[1] == '<')          else if (ptr[1] == CHAR_LESS_THAN_SIGN)
4938            {            {
4939            terminator = '>';            terminator = CHAR_GREATER_THAN_SIGN;
4940            ptr++;            ptr++;
4941            }            }
4942          else if (ptr[1] == '\'')          else if (ptr[1] == CHAR_APOSTROPHE)
4943            {            {
4944            terminator = '\'';            terminator = CHAR_APOSTROPHE;
4945            ptr++;            ptr++;
4946            }            }
4947          else          else
4948            {            {
4949            terminator = 0;            terminator = 0;
4950            if (ptr[1] == '-' || ptr[1] == '+') refsign = *(++ptr);            if (ptr[1] == CHAR_MINUS || ptr[1] == CHAR_PLUS) refsign = *(++ptr);
4951            }            }
4952    
4953          /* We now expect to read a name; any thing else is an error */          /* We now expect to read a name; any thing else is an error */
# Line 4146  for (;; ptr++) Line 4967  for (;; ptr++)
4967            {            {
4968            if (recno >= 0)            if (recno >= 0)
4969              recno = ((digitab[*ptr] & ctype_digit) != 0)?              recno = ((digitab[*ptr] & ctype_digit) != 0)?
4970                recno * 10 + *ptr - '0' : -1;                recno * 10 + *ptr - CHAR_0 : -1;
4971            ptr++;            ptr++;
4972            }            }
4973          namelen = ptr - name;          namelen = (int)(ptr - name);
4974    
4975          if ((terminator > 0 && *ptr++ != terminator) || *ptr++ != ')')          if ((terminator > 0 && *ptr++ != terminator) ||
4976                *ptr++ != CHAR_RIGHT_PARENTHESIS)
4977            {            {
4978            ptr--;      /* Error offset */            ptr--;      /* Error offset */
4979            *errorcodeptr = ERR26;            *errorcodeptr = ERR26;
# Line 4173  for (;; ptr++) Line 4995  for (;; ptr++)
4995              *errorcodeptr = ERR58;              *errorcodeptr = ERR58;
4996              goto FAILED;              goto FAILED;
4997              }              }
4998            if (refsign == '-')            recno = (refsign == CHAR_MINUS)?
4999                cd->bracount - recno + 1 : recno +cd->bracount;
5000              if (recno <= 0 || recno > cd->final_bracount)
5001              {              {
5002              recno = cd->bracount - recno + 1;              *errorcodeptr = ERR15;
5003              if (recno <= 0)              goto FAILED;
               {  
               *errorcodeptr = ERR15;  
               goto FAILED;  
               }  
5004              }              }
           else recno += cd->bracount;  
5005            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
5006            break;            break;
5007            }            }
5008    
5009          /* Otherwise (did not start with "+" or "-"), start by looking for the          /* Otherwise (did not start with "+" or "-"), start by looking for the
5010          name. */          name. If we find a name, add one to the opcode to change OP_CREF or
5011            OP_RREF into OP_NCREF or OP_NRREF. These behave exactly the same,
5012            except they record that the reference was originally to a name. The
5013            information is used to check duplicate names. */
5014    
5015          slot = cd->name_table;          slot = cd->name_table;
5016          for (i = 0; i < cd->names_found; i++)          for (i = 0; i < cd->names_found; i++)
# Line 4203  for (;; ptr++) Line 5025  for (;; ptr++)
5025            {            {
5026            recno = GET2(slot, 0);            recno = GET2(slot, 0);
5027            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
5028              code[1+LINK_SIZE]++;
5029            }            }
5030    
5031          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
5032    
5033          else if ((i = find_parens(ptr, cd->bracount, name, namelen,          else if ((i = find_parens(cd, name, namelen,
5034                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0)) > 0)
5035            {            {
5036            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
5037              code[1+LINK_SIZE]++;
5038            }            }
5039    
5040          /* If terminator == 0 it means that the name followed directly after          /* If terminator == 0 it means that the name followed directly after
# Line 4228  for (;; ptr++) Line 5052  for (;; ptr++)
5052          /* Check for (?(R) for recursion. Allow digits after R to specify a          /* Check for (?(R) for recursion. Allow digits after R to specify a
5053          specific group number. */          specific group number. */
5054    
5055          else if (*name == 'R')          else if (*name == CHAR_R)
5056            {            {
5057            recno = 0;            recno = 0;
5058            for (i = 1; i < namelen; i++)            for (i = 1; i < namelen; i++)
# Line 4238  for (;; ptr++) Line 5062  for (;; ptr++)
5062                *errorcodeptr = ERR15;                *errorcodeptr = ERR15;
5063                goto FAILED;                goto FAILED;
5064                }                }
5065              recno = recno * 10 + name[i] - '0';              recno = recno * 10 + name[i] - CHAR_0;
5066              }              }
5067            if (recno == 0) recno = RREF_ANY;            if (recno == 0) recno = RREF_ANY;
5068            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */
# Line 4248  for (;; ptr++) Line 5072  for (;; ptr++)
5072          /* Similarly, check for the (?(DEFINE) "condition", which is always          /* Similarly, check for the (?(DEFINE) "condition", which is always
5073          false. */          false. */
5074    
5075          else if (namelen == 6 && strncmp((char *)name, "DEFINE", 6) == 0)          else if (namelen == 6 && strncmp((char *)name, STRING_DEFINE, 6) == 0)
5076            {            {
5077            code[1+LINK_SIZE] = OP_DEF;            code[1+LINK_SIZE] = OP_DEF;
5078            skipbytes = 1;            skipbytes = 1;
5079            }            }
5080    
5081          /* Check for the "name" actually being a subpattern number. */          /* Check for the "name" actually being a subpattern number. We are
5082            in the second pass here, so final_bracount is set. */
5083    
5084          else if (recno > 0)          else if (recno > 0 && recno <= cd->final_bracount)
5085            {            {
5086            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
5087            }            }
# Line 4272  for (;; ptr++) Line 5097  for (;; ptr++)
5097    
5098    
5099          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5100          case '=':                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
5101          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
5102          ptr++;          ptr++;
5103          break;          break;
5104    
5105    
5106          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5107          case '!':                 /* Negative lookahead */          case CHAR_EXCLAMATION_MARK:            /* Negative lookahead */
5108          ptr++;          ptr++;
5109          if (*ptr == ')')          /* Optimize (?!) */          if (*ptr == CHAR_RIGHT_PARENTHESIS)    /* Optimize (?!) */
5110            {            {
5111            *code++ = OP_FAIL;            *code++ = OP_FAIL;
5112            previous = NULL;            previous = NULL;
# Line 4292  for (;; ptr++) Line 5117  for (;; ptr++)
5117    
5118    
5119          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5120          case '<':                 /* Lookbehind or named define */          case CHAR_LESS_THAN_SIGN:              /* Lookbehind or named define */
5121          switch (ptr[1])          switch (ptr[1])
5122            {            {
5123            case '=':               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
5124            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
5125            ptr += 2;            ptr += 2;
5126            break;            break;
5127    
5128            case '!':               /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
5129            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
5130            ptr += 2;            ptr += 2;
5131            break;            break;
# Line 4315  for (;; ptr++) Line 5140  for (;; ptr++)
5140    
5141    
5142          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5143          case '>':                 /* One-time brackets */          case CHAR_GREATER_THAN_SIGN:           /* One-time brackets */
5144          bravalue = OP_ONCE;          bravalue = OP_ONCE;
5145          ptr++;          ptr++;
5146          break;          break;
5147    
5148    
5149          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5150          case 'C':                 /* Callout - may be followed by digits; */          case CHAR_C:                 /* Callout - may be followed by digits; */
5151          previous_callout = code;  /* Save for later completion */          previous_callout = code;  /* Save for later completion */
5152          after_manual_callout = 1; /* Skip one item before completing */          after_manual_callout = 1; /* Skip one item before completing */
5153          *code++ = OP_CALLOUT;          *code++ = OP_CALLOUT;
5154            {            {
5155            int n = 0;            int n = 0;
5156            while ((digitab[*(++ptr)] & ctype_digit) != 0)            while ((digitab[*(++ptr)] & ctype_digit) != 0)
5157              n = n * 10 + *ptr - '0';              n = n * 10 + *ptr - CHAR_0;
5158            if (*ptr != ')')            if (*ptr != CHAR_RIGHT_PARENTHESIS)
5159              {              {
5160              *errorcodeptr = ERR39;              *errorcodeptr = ERR39;
5161              goto FAILED;              goto FAILED;
# Line 4341  for (;; ptr++) Line 5166  for (;; ptr++)
5166              goto FAILED;              goto FAILED;
5167              }              }
5168            *code++ = n;            *code++ = n;
5169            PUT(code, 0, ptr - cd->start_pattern + 1);  /* Pattern offset */            PUT(code, 0, (int)(ptr - cd->start_pattern + 1)); /* Pattern offset */
5170            PUT(code, LINK_SIZE, 0);                    /* Default length */            PUT(code, LINK_SIZE, 0);                          /* Default length */
5171            code += 2 * LINK_SIZE;            code += 2 * LINK_SIZE;
5172            }            }
5173          previous = NULL;          previous = NULL;
# Line 4350  for (;; ptr++) Line 5175  for (;; ptr++)
5175    
5176    
5177          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5178          case 'P':                 /* Python-style named subpattern handling */          case CHAR_P:              /* Python-style named subpattern handling */
5179          if (*(++ptr) == '=' || *ptr == '>')  /* Reference or recursion */          if (*(++ptr) == CHAR_EQUALS_SIGN ||
5180                *ptr == CHAR_GREATER_THAN_SIGN)  /* Reference or recursion */
5181            {            {
5182            is_recurse = *ptr == '>';            is_recurse = *ptr == CHAR_GREATER_THAN_SIGN;
5183            terminator = ')';            terminator = CHAR_RIGHT_PARENTHESIS;
5184            goto NAMED_REF_OR_RECURSE;            goto NAMED_REF_OR_RECURSE;
5185            }            }
5186          else if (*ptr != '<')    /* Test for Python-style definition */          else if (*ptr != CHAR_LESS_THAN_SIGN)  /* Test for Python-style defn */
5187            {            {
5188            *errorcodeptr = ERR41;            *errorcodeptr = ERR41;
5189            goto FAILED;            goto FAILED;
# Line 4367  for (;; ptr++) Line 5193  for (;; ptr++)
5193    
5194          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5195          DEFINE_NAME:    /* Come here from (?< handling */          DEFINE_NAME:    /* Come here from (?< handling */
5196          case '\'':          case CHAR_APOSTROPHE:
5197            {            {
5198            terminator = (*ptr == '<')? '>' : '\'';            terminator = (*ptr == CHAR_LESS_THAN_SIGN)?
5199                CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
5200            name = ++ptr;            name = ++ptr;
5201    
5202            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
5203            namelen = ptr - name;            namelen = (int)(ptr - name);
5204    
5205            /* In the pre-compile phase, just do a syntax check. */            /* In the pre-compile phase, just do a syntax check. */
5206    
# Line 4400  for (;; ptr++) Line 5227  for (;; ptr++)
5227                }                }
5228              }              }
5229    
5230            /* In the real compile, create the entry in the table */            /* In the real compile, create the entry in the table, maintaining
5231              alphabetical order. Duplicate names for different numbers are
5232              permitted only if PCRE_DUPNAMES is set. Duplicate names for the same
5233              number are always OK. (An existing number can be re-used if (?|
5234              appears in the pattern.) In either event, a duplicate name results in
5235              a duplicate entry in the table, even if the number is the same. This
5236              is because the number of names, and hence the table size, is computed
5237              in the pre-compile, and it affects various numbers and pointers which
5238              would all have to be modified, and the compiled code moved down, if
5239              duplicates with the same number were omitted from the table. This
5240              doesn't seem worth the hassle. However, *different* names for the
5241              same number are not permitted. */
5242    
5243            else            else
5244              {              {
5245                BOOL dupname = FALSE;
5246              slot = cd->name_table;              slot = cd->name_table;
5247    
5248              for (i = 0; i < cd->names_found; i++)              for (i = 0; i < cd->names_found; i++)
5249                {                {
5250                int crc = memcmp(name, slot+2, namelen);                int crc = memcmp(name, slot+2, namelen);
# Line 4412  for (;; ptr++) Line 5252  for (;; ptr++)
5252                  {                  {
5253                  if (slot[2+namelen] == 0)                  if (slot[2+namelen] == 0)
5254                    {                    {
5255                    if ((options & PCRE_DUPNAMES) == 0)                    if (GET2(slot, 0) != cd->bracount + 1 &&
5256                          (options & PCRE_DUPNAMES) == 0)
5257                      {                      {
5258                      *errorcodeptr = ERR43;                      *errorcodeptr = ERR43;
5259                      goto FAILED;                      goto FAILED;
5260                      }                      }
5261                      else dupname = TRUE;
5262                    }                    }
5263                  else crc = -1;      /* Current name is substring */                  else crc = -1;      /* Current name is a substring */
5264                  }                  }
5265    
5266                  /* Make space in the table and break the loop for an earlier
5267                  name. For a duplicate or later name, carry on. We do this for
5268                  duplicates so that in the simple case (when ?(| is not used) they
5269                  are in order of their numbers. */
5270    
5271                if (crc < 0)                if (crc < 0)
5272                  {                  {
5273                  memmove(slot + cd->name_entry_size, slot,                  memmove(slot + cd->name_entry_size, slot,
5274                    (cd->names_found - i) * cd->name_entry_size);                    (cd->names_found - i) * cd->name_entry_size);
5275                  break;                  break;
5276                  }                  }
5277    
5278                  /* Continue the loop for a later or duplicate name */
5279    
5280                slot += cd->name_entry_size;                slot += cd->name_entry_size;
5281                }                }
5282    
5283                /* For non-duplicate names, check for a duplicate number before
5284                adding the new name. */
5285    
5286                if (!dupname)
5287                  {
5288                  uschar *cslot = cd->name_table;
5289                  for (i = 0; i < cd->names_found; i++)
5290                    {
5291                    if (cslot != slot)
5292                      {
5293                      if (GET2(cslot, 0) == cd->bracount + 1)
5294                        {
5295                        *errorcodeptr = ERR65;
5296                        goto FAILED;
5297                        }
5298                      }
5299                    else i--;
5300                    cslot += cd->name_entry_size;
5301                    }
5302                  }
5303    
5304              PUT2(slot, 0, cd->bracount + 1);              PUT2(slot, 0, cd->bracount + 1);
5305              memcpy(slot + 2, name, namelen);              memcpy(slot + 2, name, namelen);
5306              slot[2+namelen] = 0;              slot[2+namelen] = 0;
5307              }              }
5308            }            }
5309    
5310          /* In both cases, count the number of names we've encountered. */          /* In both pre-compile and compile, count the number of names we've
5311            encountered. */
5312    
         ptr++;                    /* Move past > or ' */  
5313          cd->names_found++;          cd->names_found++;
5314            ptr++;                    /* Move past > or ' */
5315          goto NUMBERED_GROUP;          goto NUMBERED_GROUP;
5316    
5317    
5318          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5319          case '&':                 /* Perl recursion/subroutine syntax */          case CHAR_AMPERSAND:            /* Perl recursion/subroutine syntax */
5320          terminator = ')';          terminator = CHAR_RIGHT_PARENTHESIS;
5321          is_recurse = TRUE;          is_recurse = TRUE;
5322          /* Fall through */          /* Fall through */
5323    
5324          /* We come here from the Python syntax above that handles both          /* We come here from the Python syntax above that handles both
5325          references (?P=name) and recursion (?P>name), as well as falling          references (?P=name) and recursion (?P>name), as well as falling
5326          through from the Perl recursion syntax (?&name). */          through from the Perl recursion syntax (?&name). We also come here from
5327            the Perl \k<name> or \k'name' back reference syntax and the \k{name}
5328            .NET syntax, and the Oniguruma \g<...> and \g'...' subroutine syntax. */
5329    
5330          NAMED_REF_OR_RECURSE:          NAMED_REF_OR_RECURSE:
5331          name = ++ptr;          name = ++ptr;
5332          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
5333          namelen = ptr - name;          namelen = (int)(ptr - name);
5334    
5335          /* In the pre-compile phase, do a syntax check and set a dummy          /* In the pre-compile phase, do a syntax check and set a dummy
5336          reference number. */          reference number. */
5337    
5338          if (lengthptr != NULL)          if (lengthptr != NULL)
5339            {            {
5340              if (namelen == 0)
5341                {
5342                *errorcodeptr = ERR62;
5343                goto FAILED;
5344                }
5345            if (*ptr != terminator)            if (*ptr != terminator)
5346              {              {
5347              *errorcodeptr = ERR42;              *errorcodeptr = ERR42;
# Line 4475  for (;; ptr++) Line 5355  for (;; ptr++)
5355            recno = 0;            recno = 0;
5356            }            }
5357    
5358          /* In the real compile, seek the name in the table */          /* In the real compile, seek the name in the table. We check the name
5359            first, and then check that we have reached the end of the name in the
5360            table. That way, if the name that is longer than any in the table,
5361            the comparison will fail without reading beyond the table entry. */
5362    
5363          else          else
5364            {            {
5365            slot = cd->name_table;            slot = cd->name_table;
5366            for (i = 0; i < cd->names_found; i++)            for (i = 0; i < cd->names_found; i++)
5367              {              {
5368              if (strncmp((char *)name, (char *)slot+2, namelen) == 0) break;              if (strncmp((char *)name, (char *)slot+2, namelen) == 0 &&
5369                    slot[2+namelen] == 0)
5370                  break;
5371              slot += cd->name_entry_size;              slot += cd->name_entry_size;
5372              }              }
5373    
# Line 4491  for (;; ptr++) Line 5376  for (;; ptr++)
5376              recno = GET2(slot, 0);              recno = GET2(slot, 0);
5377              }              }
5378            else if ((recno =                /* Forward back reference */            else if ((recno =                /* Forward back reference */
5379                      find_parens(ptr, cd->bracount, name, namelen,                      find_parens(cd, name, namelen,
5380                        (options & PCRE_EXTENDED) != 0)) <= 0)                        (options & PCRE_EXTENDED) != 0)) <= 0)
5381              {              {
5382              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
# Line 4507  for (;; ptr++) Line 5392  for (;; ptr++)
5392    
5393    
5394          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5395          case 'R':                 /* Recursion */          case CHAR_R:              /* Recursion */
5396          ptr++;                    /* Same as (?0)      */          ptr++;                    /* Same as (?0)      */
5397          /* Fall through */          /* Fall through */
5398    
5399    
5400          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5401          case '-': case '+':          case CHAR_MINUS: case CHAR_PLUS:  /* Recursion or subroutine */
5402          case '0': case '1': case '2': case '3': case '4':   /* Recursion or */          case CHAR_0: case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4:
5403          case '5': case '6': case '7': case '8': case '9':   /* subroutine */          case CHAR_5: case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
5404            {            {
5405            const uschar *called;            const uschar *called;
5406              terminator = CHAR_RIGHT_PARENTHESIS;
5407    
5408             &n