/[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 213 by ph10, Wed Aug 15 11:34:14 2007 UTC revision 602 by ph10, Wed May 25 08:29:03 2011 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2007 University of Cambridge             Copyright (c) 1997-2011 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 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      "\\c must be followed by an ASCII character\0"
412      ;
413    
414  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
415  patterns. Note that the tables in chartables are dependent on the locale, and  patterns. Note that the tables in chartables are dependent on the locale, and
# Line 303  For convenience, we use the same bit def Line 427  For convenience, we use the same bit def
427    
428  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
429    
430  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC
431    
432    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
433    UTF-8 mode. */
434    
435  static const unsigned char digitab[] =  static const unsigned char digitab[] =
436    {    {
437    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
# Line 339  static const unsigned char digitab[] = Line 467  static const unsigned char digitab[] =
467    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
468    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
469    
470  #else           /* This is the "abnormal" case, for EBCDIC systems */  #else
471    
472    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
473    
474  static const unsigned char digitab[] =  static const unsigned char digitab[] =
475    {    {
476    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
# Line 420  static BOOL Line 551  static BOOL
551    
552    
553  /*************************************************  /*************************************************
554    *            Find an error text                  *
555    *************************************************/
556    
557    /* The error texts are now all in one long string, to save on relocations. As
558    some of the text is of unknown length, we can't use a table of offsets.
559    Instead, just count through the strings. This is not a performance issue
560    because it happens only when there has been a compilation error.
561    
562    Argument:   the error number
563    Returns:    pointer to the error string
564    */
565    
566    static const char *
567    find_error_text(int n)
568    {
569    const char *s = error_texts;
570    for (; n > 0; n--)
571      {
572      while (*s++ != 0) {};
573      if (*s == 0) return "Error text not found (please report)";
574      }
575    return s;
576    }
577    
578    
579    /*************************************************
580  *            Handle escapes                      *  *            Handle escapes                      *
581  *************************************************/  *************************************************/
582    
# Line 458  ptr--; /* Set Line 615  ptr--; /* Set
615    
616  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
617    
618  /* Non-alphamerics are literals. For digits or letters, do an initial lookup in  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
619  a table. A non-zero result is something that can be returned immediately.  in a table. A non-zero result is something that can be returned immediately.
620  Otherwise further processing may be required. */  Otherwise further processing may be required. */
621    
622  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
623  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */
624  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0) c = i;
625    
626  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
627  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */
628  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
629  #endif  #endif
630    
# Line 483  else Line 640  else
640      /* A number of Perl escapes are not handled by PCRE. We give an explicit      /* A number of Perl escapes are not handled by PCRE. We give an explicit
641      error. */      error. */
642    
643      case 'l':      case CHAR_l:
644      case 'L':      case CHAR_L:
645      case 'N':      case CHAR_u:
646      case 'u':      case CHAR_U:
     case 'U':  
647      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
648      break;      break;
649    
650      /* \g must be followed by a number, either plain or braced. If positive, it      /* \g must be followed by one of a number of specific things:
651      is an absolute backreference. If negative, it is a relative backreference.  
652      This is a Perl 5.10 feature. Perl 5.10 also supports \g{name} as a      (1) A number, either plain or braced. If positive, it is an absolute
653      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
654      unified syntax for back references. As this is synonymous with \k{name}, we      5.10 feature.
655      fudge it up by pretending it really was \k. */  
656        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
657        is part of Perl's movement towards a unified syntax for back references. As
658        this is synonymous with \k{name}, we fudge it up by pretending it really
659        was \k.
660    
661        (3) For Oniguruma compatibility we also support \g followed by a name or a
662        number either in angle brackets or in single quotes. However, these are
663        (possibly recursive) subroutine calls, _not_ backreferences. Just return
664        the -ESC_g code (cf \k). */
665    
666        case CHAR_g:
667        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
668          {
669          c = -ESC_g;
670          break;
671          }
672    
673      case 'g':      /* Handle the Perl-compatible cases */
674      if (ptr[1] == '{')  
675        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
676        {        {
677        const uschar *p;        const uschar *p;
678        for (p = ptr+2; *p != 0 && *p != '}'; p++)        for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
679          if (*p != '-' && (digitab[*p] & ctype_digit) == 0) break;          if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
680        if (*p != 0 && *p != '}')        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
681          {          {
682          c = -ESC_k;          c = -ESC_k;
683          break;          break;
# Line 514  else Line 687  else
687        }        }
688      else braced = FALSE;      else braced = FALSE;
689    
690      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
691        {        {
692        negated = TRUE;        negated = TRUE;
693        ptr++;        ptr++;
# Line 523  else Line 696  else
696    
697      c = 0;      c = 0;
698      while ((digitab[ptr[1]] & ctype_digit) != 0)      while ((digitab[ptr[1]] & ctype_digit) != 0)
699        c = c * 10 + *(++ptr) - '0';        c = c * 10 + *(++ptr) - CHAR_0;
700    
701      if (c < 0)      if (c < 0)   /* Integer overflow */
702        {        {
703        *errorcodeptr = ERR61;        *errorcodeptr = ERR61;
704        break;        break;
705        }        }
706    
707      if (c == 0 || (braced && *(++ptr) != '}'))      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
708        {        {
709        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
710        break;        break;
711        }        }
712    
713        if (c == 0)
714          {
715          *errorcodeptr = ERR58;
716          break;
717          }
718    
719      if (negated)      if (negated)
720        {        {
721        if (c > bracount)        if (c > bracount)
# Line 562  else Line 741  else
741      value is greater than 377, the least significant 8 bits are taken. Inside a      value is greater than 377, the least significant 8 bits are taken. Inside a
742      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
743    
744      case '1': case '2': case '3': case '4': case '5':      case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:
745      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
746    
747      if (!isclass)      if (!isclass)
748        {        {
749        oldptr = ptr;        oldptr = ptr;
750        c -= '0';        c -= CHAR_0;
751        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
752          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - CHAR_0;
753        if (c < 0)        if (c < 0)    /* Integer overflow */
754          {          {
755          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
756          break;          break;
757          }          }
758        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
759          {          {
760          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 588  else Line 767  else
767      generates a binary zero byte and treats the digit as a following literal.      generates a binary zero byte and treats the digit as a following literal.
768      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
769    
770      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
771        {        {
772        ptr--;        ptr--;
773        c = 0;        c = 0;
# Line 601  else Line 780  else
780      to do). Nowadays we allow for larger numbers in UTF-8 mode, but no more      to do). Nowadays we allow for larger numbers in UTF-8 mode, but no more
781      than 3 octal digits. */      than 3 octal digits. */
782    
783      case '0':      case CHAR_0:
784      c -= '0';      c -= CHAR_0;
785      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
786          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
787      if (!utf8 && c > 255) *errorcodeptr = ERR51;      if (!utf8 && c > 255) *errorcodeptr = ERR51;
788      break;      break;
789    
# Line 612  else Line 791  else
791      than 0xff in utf8 mode, but only if the ddd are hex digits. If not, { is      than 0xff in utf8 mode, but only if the ddd are hex digits. If not, { is
792      treated as a data character. */      treated as a data character. */
793    
794      case 'x':      case CHAR_x:
795      if (ptr[1] == '{')      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
796        {        {
797        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
798        int count = 0;        int count = 0;
# Line 622  else Line 801  else
801        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
802          {          {
803          register int cc = *pt++;          register int cc = *pt++;
804          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
805          count++;          count++;
806    
807  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
808          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
809          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
810  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
811          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
812          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
813  #endif  #endif
814          }          }
815    
816        if (*pt == '}')        if (*pt == CHAR_RIGHT_CURLY_BRACKET)
817          {          {
818          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
819          ptr = pt;          ptr = pt;
# Line 650  else Line 829  else
829      c = 0;      c = 0;
830      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
831        {        {
832        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
833        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
834  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
835        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
836        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
837  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
838        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
839        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
840  #endif  #endif
841        }        }
842      break;      break;
843    
844      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
845      This coding is ASCII-specific, but then the whole concept of \cx is      An error is given if the byte following \c is not an ASCII character. This
846        coding is ASCII-specific, but then the whole concept of \cx is
847      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
848    
849      case 'c':      case CHAR_c:
850      c = *(++ptr);      c = *(++ptr);
851      if (c == 0)      if (c == 0)
852        {        {
853        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
854        break;        break;
855        }        }
856    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
857  #ifndef EBCDIC  /* ASCII coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
858      if (c >= 'a' && c <= 'z') c -= 32;        {
859          *errorcodeptr = ERR68;
860          break;
861          }
862        if (c >= CHAR_a && c <= CHAR_z) c -= 32;
863      c ^= 0x40;      c ^= 0x40;
864  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
865      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
866      c ^= 0xC0;      c ^= 0xC0;
867  #endif  #endif
868      break;      break;
869    
870      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
871      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
872      for Perl compatibility, it is a literal. This code looks a bit odd, but      otherwise, for Perl compatibility, it is a literal. This code looks a bit
873      there used to be some cases other than the default, and there may be again      odd, but there used to be some cases other than the default, and there may
874      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
875    
876      default:      default:
877      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 700  else Line 884  else
884      }      }
885    }    }
886    
887    /* Perl supports \N{name} for character names, as well as plain \N for "not
888    newline". PCRE does not support \N{name}. */
889    
890    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET)
891      *errorcodeptr = ERR37;
892    
893    /* If PCRE_UCP is set, we change the values for \d etc. */
894    
895    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
896      c -= (ESC_DU - ESC_D);
897    
898    /* Set the pointer to the final character before returning. */
899    
900  *ptrptr = ptr;  *ptrptr = ptr;
901  return c;  return c;
902  }  }
# Line 740  if (c == 0) goto ERROR_RETURN; Line 937  if (c == 0) goto ERROR_RETURN;
937  /* \P or \p can be followed by a name in {}, optionally preceded by ^ for  /* \P or \p can be followed by a name in {}, optionally preceded by ^ for
938  negation. */  negation. */
939    
940  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
941    {    {
942    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
943      {      {
944      *negptr = TRUE;      *negptr = TRUE;
945      ptr++;      ptr++;
# Line 751  if (c == '{') Line 948  if (c == '{')
948      {      {
949      c = *(++ptr);      c = *(++ptr);
950      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
951      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
952      name[i] = c;      name[i] = c;
953      }      }
954    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
955    name[i] = 0;    name[i] = 0;
956    }    }
957    
# Line 776  top = _pcre_utt_size; Line 973  top = _pcre_utt_size;
973  while (bot < top)  while (bot < top)
974    {    {
975    i = (bot + top) >> 1;    i = (bot + top) >> 1;
976    c = strcmp(name, _pcre_utt[i].name);    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);
977    if (c == 0)    if (c == 0)
978      {      {
979      *dptr = _pcre_utt[i].value;      *dptr = _pcre_utt[i].value;
# Line 819  is_counted_repeat(const uschar *p) Line 1016  is_counted_repeat(const uschar *p)
1016  {  {
1017  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
1018  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
1019  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
1020    
1021  if (*p++ != ',') return FALSE;  if (*p++ != CHAR_COMMA) return FALSE;
1022  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
1023    
1024  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
1025  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
1026    
1027  return (*p == '}');  return (*p == CHAR_RIGHT_CURLY_BRACKET);
1028  }  }
1029    
1030    
# Line 860  int max = -1; Line 1057  int max = -1;
1057  /* Read the minimum value and do a paranoid check: a negative value indicates  /* Read the minimum value and do a paranoid check: a negative value indicates
1058  an integer overflow. */  an integer overflow. */
1059    
1060  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;
1061  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1062    {    {
1063    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 870  if (min < 0 || min > 65535) Line 1067  if (min < 0 || min > 65535)
1067  /* Read the maximum value if there is one, and again do a paranoid on its size.  /* Read the maximum value if there is one, and again do a paranoid on its size.
1068  Also, max must not be less than min. */  Also, max must not be less than min. */
1069    
1070  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1071    {    {
1072    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1073      {      {
1074      max = 0;      max = 0;
1075      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;
1076      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1077        {        {
1078        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 900  return p; Line 1097  return p;
1097    
1098    
1099  /*************************************************  /*************************************************
1100  *       Find forward referenced subpattern       *  *  Subroutine for finding forward reference      *
1101  *************************************************/  *************************************************/
1102    
1103  /* This function scans along a pattern's text looking for capturing  /* This recursive function is called only from find_parens() below. The
1104    top-level call starts at the beginning of the pattern. All other calls must
1105    start at a parenthesis. It scans along a pattern's text looking for capturing
1106  subpatterns, and counting them. If it finds a named pattern that matches the  subpatterns, and counting them. If it finds a named pattern that matches the
1107  name it is given, it returns its number. Alternatively, if the name is NULL, it  name it is given, it returns its number. Alternatively, if the name is NULL, it
1108  returns when it reaches a given numbered subpattern. This is used for forward  returns when it reaches a given numbered subpattern. Recursion is used to keep
1109  references to subpatterns. We know that if (?P< is encountered, the name will  track of subpatterns that reset the capturing group numbers - the (?| feature.
1110  be terminated by '>' because that is checked in the first pass.  
1111    This function was originally called only from the second pass, in which we know
1112    that if (?< or (?' or (?P< is encountered, the name will be correctly
1113    terminated because that is checked in the first pass. There is now one call to
1114    this function in the first pass, to check for a recursive back reference by
1115    name (so that we can make the whole group atomic). In this case, we need check
1116    only up to the current position in the pattern, and that is still OK because
1117    and previous occurrences will have been checked. To make this work, the test
1118    for "end of pattern" is a check against cd->end_pattern in the main loop,
1119    instead of looking for a binary zero. This means that the special first-pass
1120    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1121    processing items within the loop are OK, because afterwards the main loop will
1122    terminate.)
1123    
1124  Arguments:  Arguments:
1125    ptr          current position in the pattern    ptrptr       address of the current character pointer (updated)
1126    count        current count of capturing parens so far encountered    cd           compile background data
1127    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1128    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1129    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1130      utf8         TRUE if we are in UTF-8 mode
1131      count        pointer to the current capturing subpattern number (updated)
1132    
1133  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1134  */  */
1135    
1136  static int  static int
1137  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,
1138    BOOL xmode)    BOOL xmode, BOOL utf8, int *count)
1139  {  {
1140  const uschar *thisname;  uschar *ptr = *ptrptr;
1141    int start_count = *count;
1142    int hwm_count = start_count;
1143    BOOL dup_parens = FALSE;
1144    
1145    /* If the first character is a parenthesis, check on the type of group we are
1146    dealing with. The very first call may not start with a parenthesis. */
1147    
1148  for (; *ptr != 0; ptr++)  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1149    {    {
1150    int term;    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1151    
1152      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1153    
1154      /* Handle a normal, unnamed capturing parenthesis. */
1155    
1156      else if (ptr[1] != CHAR_QUESTION_MARK)
1157        {
1158        *count += 1;
1159        if (name == NULL && *count == lorn) return *count;
1160        ptr++;
1161        }
1162    
1163      /* All cases now have (? at the start. Remember when we are in a group
1164      where the parenthesis numbers are duplicated. */
1165    
1166      else if (ptr[2] == CHAR_VERTICAL_LINE)
1167        {
1168        ptr += 3;
1169        dup_parens = TRUE;
1170        }
1171    
1172      /* Handle comments; all characters are allowed until a ket is reached. */
1173    
1174      else if (ptr[2] == CHAR_NUMBER_SIGN)
1175        {
1176        for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1177        goto FAIL_EXIT;
1178        }
1179    
1180      /* Handle a condition. If it is an assertion, just carry on so that it
1181      is processed as normal. If not, skip to the closing parenthesis of the
1182      condition (there can't be any nested parens). */
1183    
1184      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1185        {
1186        ptr += 2;
1187        if (ptr[1] != CHAR_QUESTION_MARK)
1188          {
1189          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1190          if (*ptr != 0) ptr++;
1191          }
1192        }
1193    
1194      /* Start with (? but not a condition. */
1195    
1196      else
1197        {
1198        ptr += 2;
1199        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1200    
1201        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1202    
1203        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1204            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1205          {
1206          int term;
1207          const uschar *thisname;
1208          *count += 1;
1209          if (name == NULL && *count == lorn) return *count;
1210          term = *ptr++;
1211          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1212          thisname = ptr;
1213          while (*ptr != term) ptr++;
1214          if (name != NULL && lorn == ptr - thisname &&
1215              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1216            return *count;
1217          term++;
1218          }
1219        }
1220      }
1221    
1222    /* Past any initial parenthesis handling, scan for parentheses or vertical
1223    bars. Stop if we get to cd->end_pattern. Note that this is important for the
1224    first-pass call when this value is temporarily adjusted to stop at the current
1225    position. So DO NOT change this to a test for binary zero. */
1226    
1227    for (; ptr < cd->end_pattern; ptr++)
1228      {
1229    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1230    
1231    if (*ptr == '\\')    if (*ptr == CHAR_BACKSLASH)
1232      {      {
1233      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1234      if (*ptr == 'Q') for (;;)      if (*ptr == CHAR_Q) for (;;)
1235        {        {
1236        while (*(++ptr) != 0 && *ptr != '\\');        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1237        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1238        if (*(++ptr) == 'E') break;        if (*(++ptr) == CHAR_E) break;
1239        }        }
1240      continue;      continue;
1241      }      }
1242    
1243    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1244      are handled for real. If the first character is '^', skip it. Also, if the
1245      first few characters (either before or after ^) are \Q\E or \E we skip them
1246      too. This makes for compatibility with Perl. Note the use of STR macros to
1247      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1248    
1249    if (*ptr == '[')    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1250      {      {
1251      while (*(++ptr) != ']')      BOOL negate_class = FALSE;
1252        for (;;)
1253          {
1254          if (ptr[1] == CHAR_BACKSLASH)
1255            {
1256            if (ptr[2] == CHAR_E)
1257              ptr+= 2;
1258            else if (strncmp((const char *)ptr+2,
1259                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1260              ptr += 4;
1261            else
1262              break;
1263            }
1264          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1265            {
1266            negate_class = TRUE;
1267            ptr++;
1268            }
1269          else break;
1270          }
1271    
1272        /* If the next character is ']', it is a data character that must be
1273        skipped, except in JavaScript compatibility mode. */
1274    
1275        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1276            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1277          ptr++;
1278    
1279        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1280        {        {
1281        if (*ptr == '\\')        if (*ptr == 0) return -1;
1282          if (*ptr == CHAR_BACKSLASH)
1283          {          {
1284          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) goto FAIL_EXIT;
1285          if (*ptr == 'Q') for (;;)          if (*ptr == CHAR_Q) for (;;)
1286            {            {
1287            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1288            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1289            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1290            }            }
1291          continue;          continue;
1292          }          }
# Line 967  for (; *ptr != 0; ptr++) Line 1296  for (; *ptr != 0; ptr++)
1296    
1297    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1298    
1299    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1300      {      {
1301      while (*(++ptr) != 0 && *ptr != '\n');      ptr++;
1302      if (*ptr == 0) return -1;      while (*ptr != 0)
1303          {
1304          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1305          ptr++;
1306    #ifdef SUPPORT_UTF8
1307          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
1308    #endif
1309          }
1310        if (*ptr == 0) goto FAIL_EXIT;
1311      continue;      continue;
1312      }      }
1313    
1314    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1315    
1316    if (*ptr != '(') continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != '?' && ptr[1] != '*')  
1317      {      {
1318      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, count);
1319      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1320      continue;      if (*ptr == 0) goto FAIL_EXIT;
1321      }      }
1322    
1323    ptr += 2;    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1324    if (*ptr == 'P') ptr++;                      /* Allow optional P */      {
1325        if (dup_parens && *count < hwm_count) *count = hwm_count;
1326        goto FAIL_EXIT;
1327        }
1328    
1329    /* We have to disambiguate (?<! and (?<= from (?<name> */    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1330        {
1331        if (*count > hwm_count) hwm_count = *count;
1332        *count = start_count;
1333        }
1334      }
1335    
1336    if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  FAIL_EXIT:
1337         *ptr != '\'')  *ptrptr = ptr;
1338      continue;  return -1;
1339    }
1340    
   count++;  
1341    
1342    if (name == NULL && count == lorn) return count;  
1343    term = *ptr++;  
1344    if (term == '<') term = '>';  /*************************************************
1345    thisname = ptr;  *       Find forward referenced subpattern       *
1346    while (*ptr != term) ptr++;  *************************************************/
1347    if (name != NULL && lorn == ptr - thisname &&  
1348        strncmp((const char *)name, (const char *)thisname, lorn) == 0)  /* This function scans along a pattern's text looking for capturing
1349      return count;  subpatterns, and counting them. If it finds a named pattern that matches the
1350    name it is given, it returns its number. Alternatively, if the name is NULL, it
1351    returns when it reaches a given numbered subpattern. This is used for forward
1352    references to subpatterns. We used to be able to start this scan from the
1353    current compiling point, using the current count value from cd->bracount, and
1354    do it all in a single loop, but the addition of the possibility of duplicate
1355    subpattern numbers means that we have to scan from the very start, in order to
1356    take account of such duplicates, and to use a recursive function to keep track
1357    of the different types of group.
1358    
1359    Arguments:
1360      cd           compile background data
1361      name         name to seek, or NULL if seeking a numbered subpattern
1362      lorn         name length, or subpattern number if name is NULL
1363      xmode        TRUE if we are in /x mode
1364      utf8         TRUE if we are in UTF-8 mode
1365    
1366    Returns:       the number of the found subpattern, or -1 if not found
1367    */
1368    
1369    static int
1370    find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode,
1371      BOOL utf8)
1372    {
1373    uschar *ptr = (uschar *)cd->start_pattern;
1374    int count = 0;
1375    int rc;
1376    
1377    /* If the pattern does not start with an opening parenthesis, the first call
1378    to find_parens_sub() will scan right to the end (if necessary). However, if it
1379    does start with a parenthesis, find_parens_sub() will return when it hits the
1380    matching closing parens. That is why we have to have a loop. */
1381    
1382    for (;;)
1383      {
1384      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, &count);
1385      if (rc > 0 || *ptr++ == 0) break;
1386    }    }
1387    
1388  return -1;  return rc;
1389  }  }
1390    
1391    
1392    
1393    
1394  /*************************************************  /*************************************************
1395  *      Find first significant op code            *  *      Find first significant op code            *
1396  *************************************************/  *************************************************/
1397    
1398  /* This is called by several functions that scan a compiled expression looking  /* This is called by several functions that scan a compiled expression looking
1399  for a fixed first character, or an anchoring op code etc. It skips over things  for a fixed first character, or an anchoring op code etc. It skips over things
1400  that do not influence this. For some calls, a change of option is important.  that do not influence this. For some calls, it makes sense to skip negative
1401  For some calls, it makes sense to skip negative forward and all backward  forward and all backward assertions, and also the \b assertion; for others it
1402  assertions, and also the \b assertion; for others it does not.  does not.
1403    
1404  Arguments:  Arguments:
1405    code         pointer to the start of the group    code         pointer to the start of the group
# Line 1038  for (;;) Line 1419  for (;;)
1419    {    {
1420    switch ((int)*code)    switch ((int)*code)
1421      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1422      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1423      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1424      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
# Line 1059  for (;;) Line 1434  for (;;)
1434    
1435      case OP_CALLOUT:      case OP_CALLOUT:
1436      case OP_CREF:      case OP_CREF:
1437        case OP_NCREF:
1438      case OP_RREF:      case OP_RREF:
1439        case OP_NRREF:
1440      case OP_DEF:      case OP_DEF:
1441      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1442      break;      break;
# Line 1075  for (;;) Line 1452  for (;;)
1452    
1453    
1454  /*************************************************  /*************************************************
1455  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1456  *************************************************/  *************************************************/
1457    
1458  /* 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,
1459  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.
1460  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
1461    temporarily terminated with OP_END when this function is called.
1462    
1463    This function is called when a backward assertion is encountered, so that if it
1464    fails, the error message can point to the correct place in the pattern.
1465    However, we cannot do this when the assertion contains subroutine calls,
1466    because they can be forward references. We solve this by remembering this case
1467    and doing the check at the end; a flag specifies which mode we are running in.
1468    
1469  Arguments:  Arguments:
1470    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1471    options  the compiling options    options  the compiling options
1472      atend    TRUE if called when the pattern is complete
1473      cd       the "compile data" structure
1474    
1475  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1476                 or -1 if there is no fixed length,
1477               or -2 if \C was encountered               or -2 if \C was encountered
1478                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1479  */  */
1480    
1481  static int  static int
1482  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)
1483  {  {
1484  int length = -1;  int length = -1;
1485    
# Line 1104  branch, check the length against that of Line 1492  branch, check the length against that of
1492  for (;;)  for (;;)
1493    {    {
1494    int d;    int d;
1495      uschar *ce, *cs;
1496    register int op = *cc;    register int op = *cc;
   
1497    switch (op)    switch (op)
1498      {      {
1499      case OP_CBRA:      case OP_CBRA:
1500      case OP_BRA:      case OP_BRA:
1501      case OP_ONCE:      case OP_ONCE:
1502      case OP_COND:      case OP_COND:
1503      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);
1504      if (d < 0) return d;      if (d < 0) return d;
1505      branchlength += d;      branchlength += d;
1506      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1135  for (;;) Line 1523  for (;;)
1523      branchlength = 0;      branchlength = 0;
1524      break;      break;
1525    
1526        /* A true recursion implies not fixed length, but a subroutine call may
1527        be OK. If the subroutine is a forward reference, we can't deal with
1528        it until the end of the pattern, so return -3. */
1529    
1530        case OP_RECURSE:
1531        if (!atend) return -3;
1532        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1533        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1534        if (cc > cs && cc < ce) return -1;                /* Recursion */
1535        d = find_fixedlength(cs + 2, options, atend, cd);
1536        if (d < 0) return d;
1537        branchlength += d;
1538        cc += 1 + LINK_SIZE;
1539        break;
1540    
1541      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1542    
1543      case OP_ASSERT:      case OP_ASSERT:
# Line 1148  for (;;) Line 1551  for (;;)
1551    
1552      case OP_REVERSE:      case OP_REVERSE:
1553      case OP_CREF:      case OP_CREF:
1554        case OP_NCREF:
1555      case OP_RREF:      case OP_RREF:
1556        case OP_NRREF:
1557      case OP_DEF:      case OP_DEF:
     case OP_OPT:  
1558      case OP_CALLOUT:      case OP_CALLOUT:
1559      case OP_SOD:      case OP_SOD:
1560      case OP_SOM:      case OP_SOM:
1561        case OP_SET_SOM:
1562      case OP_EOD:      case OP_EOD:
1563      case OP_EODN:      case OP_EODN:
1564      case OP_CIRC:      case OP_CIRC:
1565        case OP_CIRCM:
1566      case OP_DOLL:      case OP_DOLL:
1567        case OP_DOLLM:
1568      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1569      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1570      cc += _pcre_OP_lengths[*cc];      cc += _pcre_OP_lengths[*cc];
# Line 1166  for (;;) Line 1573  for (;;)
1573      /* Handle literal characters */      /* Handle literal characters */
1574    
1575      case OP_CHAR:      case OP_CHAR:
1576      case OP_CHARNC:      case OP_CHARI:
1577      case OP_NOT:      case OP_NOT:
1578        case OP_NOTI:
1579      branchlength++;      branchlength++;
1580      cc += 2;      cc += 2;
1581  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1582      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1583        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1584  #endif  #endif
1585      break;      break;
1586    
# Line 1185  for (;;) Line 1591  for (;;)
1591      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1592      cc += 4;      cc += 4;
1593  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1594      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1595        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1596  #endif  #endif
1597      break;      break;
1598    
1599      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1600      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1601        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1602      cc += 4;      cc += 4;
1603      break;      break;
1604    
# Line 1211  for (;;) Line 1616  for (;;)
1616      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1617      case OP_WORDCHAR:      case OP_WORDCHAR:
1618      case OP_ANY:      case OP_ANY:
1619        case OP_ALLANY:
1620      branchlength++;      branchlength++;
1621      cc++;      cc++;
1622      break;      break;
# Line 1265  for (;;) Line 1671  for (;;)
1671    
1672    
1673  /*************************************************  /*************************************************
1674  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1675  *************************************************/  *************************************************/
1676    
1677  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1678  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1679    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1680    so that it can be called from pcre_study() when finding the minimum matching
1681    length.
1682    
1683  Arguments:  Arguments:
1684    code        points to start of expression    code        points to start of expression
1685    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1686    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1687    
1688  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
1689  */  */
1690    
1691  static const uschar *  const uschar *
1692  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1693  {  {
1694  for (;;)  for (;;)
1695    {    {
# Line 1293  for (;;) Line 1702  for (;;)
1702    
1703    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1704    
1705      /* Handle recursion */
1706    
1707      else if (c == OP_REVERSE)
1708        {
1709        if (number < 0) return (uschar *)code;
1710        code += _pcre_OP_lengths[c];
1711        }
1712    
1713    /* Handle capturing bracket */    /* Handle capturing bracket */
1714    
1715    else if (c == OP_CBRA)    else if (c == OP_CBRA)
# Line 1302  for (;;) Line 1719  for (;;)
1719      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1720      }      }
1721    
1722    /* In UTF-8 mode, opcodes that are followed by a character may be followed by    /* Otherwise, we can get the item's length from the table, except that for
1723    a multi-byte character. The length in the table is a minimum, so we have to    repeated character types, we have to test for \p and \P, which have an extra
1724    arrange to skip the extra bytes. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1725      must add in its length. */
1726    
1727    else    else
1728      {      {
1729        switch(c)
1730          {
1731          case OP_TYPESTAR:
1732          case OP_TYPEMINSTAR:
1733          case OP_TYPEPLUS:
1734          case OP_TYPEMINPLUS:
1735          case OP_TYPEQUERY:
1736          case OP_TYPEMINQUERY:
1737          case OP_TYPEPOSSTAR:
1738          case OP_TYPEPOSPLUS:
1739          case OP_TYPEPOSQUERY:
1740          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1741          break;
1742    
1743          case OP_TYPEUPTO:
1744          case OP_TYPEMINUPTO:
1745          case OP_TYPEEXACT:
1746          case OP_TYPEPOSUPTO:
1747          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1748          break;
1749    
1750          case OP_MARK:
1751          case OP_PRUNE_ARG:
1752          case OP_SKIP_ARG:
1753          code += code[1];
1754          break;
1755    
1756          case OP_THEN_ARG:
1757          code += code[1+LINK_SIZE];
1758          break;
1759          }
1760    
1761        /* Add in the fixed length from the table */
1762    
1763      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1764    
1765      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1766      a multi-byte character. The length in the table is a minimum, so we have to
1767      arrange to skip the extra bytes. */
1768    
1769  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1770      if (utf8) switch(c)      if (utf8) switch(c)
1771        {        {
1772        case OP_CHAR:        case OP_CHAR:
1773        case OP_CHARNC:        case OP_CHARI:
1774        case OP_EXACT:        case OP_EXACT:
1775          case OP_EXACTI:
1776        case OP_UPTO:        case OP_UPTO:
1777          case OP_UPTOI:
1778        case OP_MINUPTO:        case OP_MINUPTO:
1779          case OP_MINUPTOI:
1780        case OP_POSUPTO:        case OP_POSUPTO:
1781          case OP_POSUPTOI:
1782        case OP_STAR:        case OP_STAR:
1783          case OP_STARI:
1784        case OP_MINSTAR:        case OP_MINSTAR:
1785          case OP_MINSTARI:
1786        case OP_POSSTAR:        case OP_POSSTAR:
1787          case OP_POSSTARI:
1788        case OP_PLUS:        case OP_PLUS:
1789          case OP_PLUSI:
1790        case OP_MINPLUS:        case OP_MINPLUS:
1791          case OP_MINPLUSI:
1792        case OP_POSPLUS:        case OP_POSPLUS:
1793          case OP_POSPLUSI:
1794        case OP_QUERY:        case OP_QUERY:
1795          case OP_QUERYI:
1796        case OP_MINQUERY:        case OP_MINQUERY:
1797          case OP_MINQUERYI:
1798        case OP_POSQUERY:        case OP_POSQUERY:
1799          case OP_POSQUERYI:
1800        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1801        break;        break;
1802        }        }
1803    #else
1804        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1805  #endif  #endif
1806      }      }
1807    }    }
# Line 1366  for (;;) Line 1838  for (;;)
1838    
1839    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1840    
1841    /* Otherwise, we get the item's length from the table. In UTF-8 mode, opcodes    /* Otherwise, we can get the item's length from the table, except that for
1842    that are followed by a character may be followed by a multi-byte character.    repeated character types, we have to test for \p and \P, which have an extra
1843    The length in the table is a minimum, so we have to arrange to skip the extra    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1844    bytes. */    must add in its length. */
1845    
1846    else    else
1847      {      {
1848        switch(c)
1849          {
1850          case OP_TYPESTAR:
1851          case OP_TYPEMINSTAR:
1852          case OP_TYPEPLUS:
1853          case OP_TYPEMINPLUS:
1854          case OP_TYPEQUERY:
1855          case OP_TYPEMINQUERY:
1856          case OP_TYPEPOSSTAR:
1857          case OP_TYPEPOSPLUS:
1858          case OP_TYPEPOSQUERY:
1859          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1860          break;
1861    
1862          case OP_TYPEPOSUPTO:
1863          case OP_TYPEUPTO:
1864          case OP_TYPEMINUPTO:
1865          case OP_TYPEEXACT:
1866          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1867          break;
1868    
1869          case OP_MARK:
1870          case OP_PRUNE_ARG:
1871          case OP_SKIP_ARG:
1872          code += code[1];
1873          break;
1874    
1875          case OP_THEN_ARG:
1876          code += code[1+LINK_SIZE];
1877          break;
1878          }
1879    
1880        /* Add in the fixed length from the table */
1881    
1882      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1883    
1884        /* In UTF-8 mode, opcodes that are followed by a character may be followed
1885        by a multi-byte character. The length in the table is a minimum, so we have
1886        to arrange to skip the extra bytes. */
1887    
1888  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1889      if (utf8) switch(c)      if (utf8) switch(c)
1890        {        {
1891        case OP_CHAR:        case OP_CHAR:
1892        case OP_CHARNC:        case OP_CHARI:
1893        case OP_EXACT:        case OP_EXACT:
1894          case OP_EXACTI:
1895        case OP_UPTO:        case OP_UPTO:
1896          case OP_UPTOI:
1897        case OP_MINUPTO:        case OP_MINUPTO:
1898          case OP_MINUPTOI:
1899        case OP_POSUPTO:        case OP_POSUPTO:
1900          case OP_POSUPTOI:
1901        case OP_STAR:        case OP_STAR:
1902          case OP_STARI:
1903        case OP_MINSTAR:        case OP_MINSTAR:
1904          case OP_MINSTARI:
1905        case OP_POSSTAR:        case OP_POSSTAR:
1906          case OP_POSSTARI:
1907        case OP_PLUS:        case OP_PLUS:
1908          case OP_PLUSI:
1909        case OP_MINPLUS:        case OP_MINPLUS:
1910          case OP_MINPLUSI:
1911        case OP_POSPLUS:        case OP_POSPLUS:
1912          case OP_POSPLUSI:
1913        case OP_QUERY:        case OP_QUERY:
1914          case OP_QUERYI:
1915        case OP_MINQUERY:        case OP_MINQUERY:
1916          case OP_MINQUERYI:
1917        case OP_POSQUERY:        case OP_POSQUERY:
1918          case OP_POSQUERYI:
1919        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1920        break;        break;
1921        }        }
1922    #else
1923        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1924  #endif  #endif
1925      }      }
1926    }    }
# Line 1410  for (;;) Line 1936  for (;;)
1936  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()
1937  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
1938  group that can match nothing. Note that first_significant_code() skips over  group that can match nothing. Note that first_significant_code() skips over
1939  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
1940  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
1941    bracket whose current branch will already have been scanned.
1942    
1943  Arguments:  Arguments:
1944    code        points to start of search    code        points to start of search
1945    endcode     points to where to stop    endcode     points to where to stop
1946    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
1947      cd          contains pointers to tables etc.
1948    
1949  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
1950  */  */
1951    
1952  static BOOL  static BOOL
1953  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
1954      compile_data *cd)
1955  {  {
1956  register int c;  register int c;
1957  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 1433  for (code = first_significant_code(code Line 1962  for (code = first_significant_code(code
1962    
1963    c = *code;    c = *code;
1964    
1965      /* Skip over forward assertions; the other assertions are skipped by
1966      first_significant_code() with a TRUE final argument. */
1967    
1968      if (c == OP_ASSERT)
1969        {
1970        do code += GET(code, 1); while (*code == OP_ALT);
1971        c = *code;
1972        continue;
1973        }
1974    
1975    /* Groups with zero repeats can of course be empty; skip them. */    /* Groups with zero repeats can of course be empty; skip them. */
1976    
1977    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)
1978      {      {
1979      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1980      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
# Line 1443  for (code = first_significant_code(code Line 1982  for (code = first_significant_code(code
1982      continue;      continue;
1983      }      }
1984    
1985      /* For a recursion/subroutine call, if its end has been reached, which
1986      implies a subroutine call, we can scan it. */
1987    
1988      if (c == OP_RECURSE)
1989        {
1990        BOOL empty_branch = FALSE;
1991        const uschar *scode = cd->start_code + GET(code, 1);
1992        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
1993        do
1994          {
1995          if (could_be_empty_branch(scode, endcode, utf8, cd))
1996            {
1997            empty_branch = TRUE;
1998            break;
1999            }
2000          scode += GET(scode, 1);
2001          }
2002        while (*scode == OP_ALT);
2003        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2004        continue;
2005        }
2006    
2007    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
2008    
2009    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 1450  for (code = first_significant_code(code Line 2011  for (code = first_significant_code(code
2011      BOOL empty_branch;      BOOL empty_branch;
2012      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
2013    
2014      /* Scan a closed bracket */      /* If a conditional group has only one branch, there is a second, implied,
2015        empty branch, so just skip over the conditional, because it could be empty.
2016        Otherwise, scan the individual branches of the group. */
2017    
2018      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;  
2019        code += GET(code, 1);        code += GET(code, 1);
2020        else
2021          {
2022          empty_branch = FALSE;
2023          do
2024            {
2025            if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
2026              empty_branch = TRUE;
2027            code += GET(code, 1);
2028            }
2029          while (*code == OP_ALT);
2030          if (!empty_branch) return FALSE;   /* All branches are non-empty */
2031        }        }
2032      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
2033      c = *code;      c = *code;
2034      continue;      continue;
2035      }      }
# Line 1469  for (code = first_significant_code(code Line 2038  for (code = first_significant_code(code
2038    
2039    switch (c)    switch (c)
2040      {      {
2041      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
2042        cannot be represented just by a bit map. This includes negated single
2043        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
2044        actual length is stored in the compiled code, so we must update "code"
2045        here. */
2046    
2047  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2048      case OP_XCLASS:      case OP_XCLASS:
2049      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
2050      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
2051  #endif  #endif
2052    
# Line 1517  for (code = first_significant_code(code Line 2090  for (code = first_significant_code(code
2090      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2091      case OP_WORDCHAR:      case OP_WORDCHAR:
2092      case OP_ANY:      case OP_ANY:
2093        case OP_ALLANY:
2094      case OP_ANYBYTE:      case OP_ANYBYTE:
2095      case OP_CHAR:      case OP_CHAR:
2096      case OP_CHARNC:      case OP_CHARI:
2097      case OP_NOT:      case OP_NOT:
2098        case OP_NOTI:
2099      case OP_PLUS:      case OP_PLUS:
2100      case OP_MINPLUS:      case OP_MINPLUS:
2101      case OP_POSPLUS:      case OP_POSPLUS:
# Line 1535  for (code = first_significant_code(code Line 2110  for (code = first_significant_code(code
2110      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2111      return FALSE;      return FALSE;
2112    
2113        /* These are going to continue, as they may be empty, but we have to
2114        fudge the length for the \p and \P cases. */
2115    
2116        case OP_TYPESTAR:
2117        case OP_TYPEMINSTAR:
2118        case OP_TYPEPOSSTAR:
2119        case OP_TYPEQUERY:
2120        case OP_TYPEMINQUERY:
2121        case OP_TYPEPOSQUERY:
2122        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2123        break;
2124    
2125        /* Same for these */
2126    
2127        case OP_TYPEUPTO:
2128        case OP_TYPEMINUPTO:
2129        case OP_TYPEPOSUPTO:
2130        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
2131        break;
2132    
2133      /* End of branch */      /* End of branch */
2134    
2135      case OP_KET:      case OP_KET:
# Line 1548  for (code = first_significant_code(code Line 2143  for (code = first_significant_code(code
2143    
2144  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2145      case OP_STAR:      case OP_STAR:
2146        case OP_STARI:
2147      case OP_MINSTAR:      case OP_MINSTAR:
2148        case OP_MINSTARI:
2149      case OP_POSSTAR:      case OP_POSSTAR:
2150        case OP_POSSTARI:
2151      case OP_QUERY:      case OP_QUERY:
2152        case OP_QUERYI:
2153      case OP_MINQUERY:      case OP_MINQUERY:
2154        case OP_MINQUERYI:
2155      case OP_POSQUERY:      case OP_POSQUERY:
2156        case OP_POSQUERYI:
2157        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2158        break;
2159    
2160      case OP_UPTO:      case OP_UPTO:
2161        case OP_UPTOI:
2162      case OP_MINUPTO:      case OP_MINUPTO:
2163        case OP_MINUPTOI:
2164      case OP_POSUPTO:      case OP_POSUPTO:
2165      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_POSUPTOI:
2166        if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2167      break;      break;
2168  #endif  #endif
2169    
2170        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2171        string. */
2172    
2173        case OP_MARK:
2174        case OP_PRUNE_ARG:
2175        case OP_SKIP_ARG:
2176        code += code[1];
2177        break;
2178    
2179        case OP_THEN_ARG:
2180        code += code[1+LINK_SIZE];
2181        break;
2182    
2183        /* None of the remaining opcodes are required to match a character. */
2184    
2185        default:
2186        break;
2187      }      }
2188    }    }
2189    
# Line 1581  Arguments: Line 2206  Arguments:
2206    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2207    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2208    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2209      cd          pointers to tables etc
2210    
2211  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2212  */  */
2213    
2214  static BOOL  static BOOL
2215  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2216    BOOL utf8)    BOOL utf8, compile_data *cd)
2217  {  {
2218  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2219    {    {
2220    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2221        return FALSE;
2222    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2223    }    }
2224  return TRUE;  return TRUE;
# Line 1604  return TRUE; Line 2231  return TRUE;
2231  *************************************************/  *************************************************/
2232    
2233  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2234  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
2235  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2236  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2237    
2238    Originally, this function only recognized a sequence of letters between the
2239    terminators, but it seems that Perl recognizes any sequence of characters,
2240    though of course unknown POSIX names are subsequently rejected. Perl gives an
2241    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2242    didn't consider this to be a POSIX class. Likewise for [:1234:].
2243    
2244    The problem in trying to be exactly like Perl is in the handling of escapes. We
2245    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2246    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2247    below handles the special case of \], but does not try to do any other escape
2248    processing. This makes it different from Perl for cases such as [:l\ower:]
2249    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2250    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2251    I think.
2252    
2253  Argument:  Arguments:
2254    ptr      pointer to the initial [    ptr      pointer to the initial [
2255    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2256    
2257  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2258  */  */
2259    
2260  static BOOL  static BOOL
2261  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
2262  {  {
2263  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2264  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2265  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2266    {    {
2267    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else
2268    return TRUE;      {
2269        if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2270        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2271          {
2272          *endptr = ptr;
2273          return TRUE;
2274          }
2275        }
2276    }    }
2277  return FALSE;  return FALSE;
2278  }  }
# Line 1651  Returns: a value representing the na Line 2297  Returns: a value representing the na
2297  static int  static int
2298  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
2299  {  {
2300    const char *pn = posix_names;
2301  register int yield = 0;  register int yield = 0;
2302  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2303    {    {
2304    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2305      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
2306      pn += posix_name_lengths[yield] + 1;
2307    yield++;    yield++;
2308    }    }
2309  return -1;  return -1;
# Line 1670  return -1; Line 2318  return -1;
2318  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2319  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2320  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
2321  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
2322  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
2323  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
2324  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
2325  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2326    OP_END.
2327    
2328  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2329  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 1697  adjust_recurse(uschar *group, int adjust Line 2346  adjust_recurse(uschar *group, int adjust
2346    uschar *save_hwm)    uschar *save_hwm)
2347  {  {
2348  uschar *ptr = group;  uschar *ptr = group;
2349    
2350  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
2351    {    {
2352    int offset;    int offset;
# Line 1750  auto_callout(uschar *code, const uschar Line 2400  auto_callout(uschar *code, const uschar
2400  {  {
2401  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2402  *code++ = 255;  *code++ = 255;
2403  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2404  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2405  return code + 2*LINK_SIZE;  return code + 2*LINK_SIZE;
2406  }  }
2407    
# Line 1776  Returns: nothing Line 2426  Returns: nothing
2426  static void  static void
2427  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)
2428  {  {
2429  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2430  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2431  }  }
2432    
# Line 1808  get_othercase_range(unsigned int *cptr, Line 2458  get_othercase_range(unsigned int *cptr,
2458  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2459    
2460  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2461    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2462    
2463  if (c > d) return FALSE;  if (c > d) return FALSE;
2464    
# Line 1817  next = othercase + 1; Line 2467  next = othercase + 1;
2467    
2468  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2469    {    {
2470    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2471    next++;    next++;
2472    }    }
2473    
# Line 1826  for (++c; c <= d; c++) Line 2476  for (++c; c <= d; c++)
2476    
2477  return TRUE;  return TRUE;
2478  }  }
2479    
2480    
2481    
2482    /*************************************************
2483    *        Check a character and a property        *
2484    *************************************************/
2485    
2486    /* This function is called by check_auto_possessive() when a property item
2487    is adjacent to a fixed character.
2488    
2489    Arguments:
2490      c            the character
2491      ptype        the property type
2492      pdata        the data for the type
2493      negated      TRUE if it's a negated property (\P or \p{^)
2494    
2495    Returns:       TRUE if auto-possessifying is OK
2496    */
2497    
2498    static BOOL
2499    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2500    {
2501    const ucd_record *prop = GET_UCD(c);
2502    switch(ptype)
2503      {
2504      case PT_LAMP:
2505      return (prop->chartype == ucp_Lu ||
2506              prop->chartype == ucp_Ll ||
2507              prop->chartype == ucp_Lt) == negated;
2508    
2509      case PT_GC:
2510      return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;
2511    
2512      case PT_PC:
2513      return (pdata == prop->chartype) == negated;
2514    
2515      case PT_SC:
2516      return (pdata == prop->script) == negated;
2517    
2518      /* These are specials */
2519    
2520      case PT_ALNUM:
2521      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2522              _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;
2523    
2524      case PT_SPACE:    /* Perl space */
2525      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2526              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2527              == negated;
2528    
2529      case PT_PXSPACE:  /* POSIX space */
2530      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2531              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2532              c == CHAR_FF || c == CHAR_CR)
2533              == negated;
2534    
2535      case PT_WORD:
2536      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2537              _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2538              c == CHAR_UNDERSCORE) == negated;
2539      }
2540    return FALSE;
2541    }
2542  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2543    
2544    
# Line 1839  whether the next thing could possibly ma Line 2552  whether the next thing could possibly ma
2552  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
2553    
2554  Arguments:  Arguments:
2555    op_code       the repeated op code    previous      pointer to the repeated opcode
   this          data for this item, depends on the opcode  
2556    utf8          TRUE in UTF-8 mode    utf8          TRUE in UTF-8 mode
   utf8_char     used for utf8 character bytes, NULL if not relevant  
2557    ptr           next character in pattern    ptr           next character in pattern
2558    options       options bits    options       options bits
2559    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 1851  Returns: TRUE if possessifying is Line 2562  Returns: TRUE if possessifying is
2562  */  */
2563    
2564  static BOOL  static BOOL
2565  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
2566    const uschar *ptr, int options, compile_data *cd)    int options, compile_data *cd)
2567  {  {
2568  int next;  int c, next;
2569    int op_code = *previous++;
2570    
2571  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
2572    
# Line 1863  if ((options & PCRE_EXTENDED) != 0) Line 2575  if ((options & PCRE_EXTENDED) != 0)
2575    for (;;)    for (;;)
2576      {      {
2577      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2578      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2579        {        {
2580        while (*(++ptr) != 0)        ptr++;
2581          while (*ptr != 0)
2582            {
2583          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2584            ptr++;
2585    #ifdef SUPPORT_UTF8
2586            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2587    #endif
2588            }
2589        }        }
2590      else break;      else break;
2591      }      }
# Line 1875  if ((options & PCRE_EXTENDED) != 0) Line 2594  if ((options & PCRE_EXTENDED) != 0)
2594  /* 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
2595  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2596    
2597  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2598    {    {
2599    int temperrorcode = 0;    int temperrorcode = 0;
2600    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
# Line 1900  if ((options & PCRE_EXTENDED) != 0) Line 2619  if ((options & PCRE_EXTENDED) != 0)
2619    for (;;)    for (;;)
2620      {      {
2621      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2622      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2623        {        {
2624        while (*(++ptr) != 0)        ptr++;
2625          while (*ptr != 0)
2626            {
2627          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2628            ptr++;
2629    #ifdef SUPPORT_UTF8
2630            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2631    #endif
2632            }
2633        }        }
2634      else break;      else break;
2635      }      }
# Line 1911  if ((options & PCRE_EXTENDED) != 0) Line 2637  if ((options & PCRE_EXTENDED) != 0)
2637    
2638  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
2639    
2640  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2641    return FALSE;    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2642        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. */  
   
2643    
2644  /* Handle cases when the next item is a character. */  /* Now compare the next item with the previous opcode. First, handle cases when
2645    the next item is a character. */
2646    
2647  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
2648    {    {
2649    case OP_CHAR:    case OP_CHAR:
2650  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2651    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2652    #else
2653      c = *previous;
2654  #endif  #endif
2655    return item != next;    return c != next;
2656    
2657    /* For CHARNC (caseless character) we must check the other case. If we have    /* For CHARI (caseless character) we must check the other case. If we have
2658    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
2659    high-valued characters. */    high-valued characters. */
2660    
2661    case OP_CHARNC:    case OP_CHARI:
2662  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2663    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2664    #else
2665      c = *previous;
2666  #endif  #endif
2667    if (item == next) return FALSE;    if (c == next) return FALSE;
2668  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2669    if (utf8)    if (utf8)
2670      {      {
2671      unsigned int othercase;      unsigned int othercase;
2672      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2673  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2674      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
2675  #else  #else
2676      othercase = NOTACHAR;      othercase = NOTACHAR;
2677  #endif  #endif
2678      return (unsigned int)item != othercase;      return (unsigned int)c != othercase;
2679      }      }
2680    else    else
2681  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2682    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2683    
2684    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT and OP_NOTI, the data is always a single-byte character. These
2685      opcodes are not used for multi-byte characters, because they are coded using
2686      an XCLASS instead. */
2687    
2688    case OP_NOT:    case OP_NOT:
2689    if (next < 0) return FALSE;  /* Not a character */    return (c = *previous) == next;
2690    if (item == next) return TRUE;  
2691    if ((options & PCRE_CASELESS) == 0) return FALSE;    case OP_NOTI:
2692      if ((c = *previous) == next) return TRUE;
2693  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2694    if (utf8)    if (utf8)
2695      {      {
2696      unsigned int othercase;      unsigned int othercase;
2697      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2698  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2699      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
2700  #else  #else
2701      othercase = NOTACHAR;      othercase = NOTACHAR;
2702  #endif  #endif
2703      return (unsigned int)item == othercase;      return (unsigned int)c == othercase;
2704      }      }
2705    else    else
2706  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2707    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2708    
2709      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2710      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
2711    
2712    case OP_DIGIT:    case OP_DIGIT:
2713    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
# Line 2018  if (next >= 0) switch(op_code) Line 2750  if (next >= 0) switch(op_code)
2750      case 0x202f:      case 0x202f:
2751      case 0x205f:      case 0x205f:
2752      case 0x3000:      case 0x3000:
2753      return op_code != OP_HSPACE;      return op_code == OP_NOT_HSPACE;
2754      default:      default:
2755      return op_code == OP_HSPACE;      return op_code != OP_NOT_HSPACE;
2756      }      }
2757    
2758      case OP_ANYNL:
2759    case OP_VSPACE:    case OP_VSPACE:
2760    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
2761    switch(next)    switch(next)
# Line 2034  if (next >= 0) switch(op_code) Line 2767  if (next >= 0) switch(op_code)
2767      case 0x85:      case 0x85:
2768      case 0x2028:      case 0x2028:
2769      case 0x2029:      case 0x2029:
2770      return op_code != OP_VSPACE;      return op_code == OP_NOT_VSPACE;
2771      default:      default:
2772      return op_code == OP_VSPACE;      return op_code != OP_NOT_VSPACE;
2773      }      }
2774    
2775    #ifdef SUPPORT_UCP
2776      case OP_PROP:
2777      return check_char_prop(next, previous[0], previous[1], FALSE);
2778    
2779      case OP_NOTPROP:
2780      return check_char_prop(next, previous[0], previous[1], TRUE);
2781    #endif
2782    
2783    default:    default:
2784    return FALSE;    return FALSE;
2785    }    }
2786    
2787    
2788  /* 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
2789    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
2790    generated only when PCRE_UCP is *not* set, that is, when only ASCII
2791    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
2792    replaced by OP_PROP codes when PCRE_UCP is set. */
2793    
2794  switch(op_code)  switch(op_code)
2795    {    {
2796    case OP_CHAR:    case OP_CHAR:
2797    case OP_CHARNC:    case OP_CHARI:
2798  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2799    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2800    #else
2801      c = *previous;
2802  #endif  #endif
2803    switch(-next)    switch(-next)
2804      {      {
2805      case ESC_d:      case ESC_d:
2806      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
2807    
2808      case ESC_D:      case ESC_D:
2809      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
2810    
2811      case ESC_s:      case ESC_s:
2812      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
2813    
2814      case ESC_S:      case ESC_S:
2815      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
2816    
2817      case ESC_w:      case ESC_w:
2818      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
2819    
2820      case ESC_W:      case ESC_W:
2821      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
2822    
2823      case ESC_h:      case ESC_h:
2824      case ESC_H:      case ESC_H:
2825      switch(item)      switch(c)
2826        {        {
2827        case 0x09:        case 0x09:
2828        case 0x20:        case 0x20:
# Line 2103  switch(op_code) Line 2850  switch(op_code)
2850    
2851      case ESC_v:      case ESC_v:
2852      case ESC_V:      case ESC_V:
2853      switch(item)      switch(c)
2854        {        {
2855        case 0x0a:        case 0x0a:
2856        case 0x0b:        case 0x0b:
# Line 2117  switch(op_code) Line 2864  switch(op_code)
2864        return -next == ESC_v;        return -next == ESC_v;
2865        }        }
2866    
2867        /* When PCRE_UCP is set, these values get generated for \d etc. Find
2868        their substitutions and process them. The result will always be either
2869        -ESC_p or -ESC_P. Then fall through to process those values. */
2870    
2871    #ifdef SUPPORT_UCP
2872        case ESC_du:
2873        case ESC_DU:
2874        case ESC_wu:
2875        case ESC_WU:
2876        case ESC_su:
2877        case ESC_SU:
2878          {
2879          int temperrorcode = 0;
2880          ptr = substitutes[-next - ESC_DU];
2881          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
2882          if (temperrorcode != 0) return FALSE;
2883          ptr++;    /* For compatibility */
2884          }
2885        /* Fall through */
2886    
2887        case ESC_p:
2888        case ESC_P:
2889          {
2890          int ptype, pdata, errorcodeptr;
2891          BOOL negated;
2892    
2893          ptr--;      /* Make ptr point at the p or P */
2894          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
2895          if (ptype < 0) return FALSE;
2896          ptr++;      /* Point past the final curly ket */
2897    
2898          /* If the property item is optional, we have to give up. (When generated
2899          from \d etc by PCRE_UCP, this test will have been applied much earlier,
2900          to the original \d etc. At this point, ptr will point to a zero byte. */
2901    
2902          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2903            strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2904              return FALSE;
2905    
2906          /* Do the property check. */
2907    
2908          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
2909          }
2910    #endif
2911    
2912      default:      default:
2913      return FALSE;      return FALSE;
2914      }      }
2915    
2916      /* In principle, support for Unicode properties should be integrated here as
2917      well. It means re-organizing the above code so as to get hold of the property
2918      values before switching on the op-code. However, I wonder how many patterns
2919      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
2920      these op-codes are never generated.) */
2921    
2922    case OP_DIGIT:    case OP_DIGIT:
2923    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2924           next == -ESC_h || next == -ESC_v;           next == -ESC_h || next == -ESC_v || next == -ESC_R;
2925    
2926    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
2927    return next == -ESC_d;    return next == -ESC_d;
2928    
2929    case OP_WHITESPACE:    case OP_WHITESPACE:
2930    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
2931    
2932    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
2933    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2934    
2935    case OP_HSPACE:    case OP_HSPACE:
2936    return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
2937             next == -ESC_w || next == -ESC_v || next == -ESC_R;
2938    
2939    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
2940    return next == -ESC_h;    return next == -ESC_h;
2941    
2942    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
2943      case OP_ANYNL:
2944    case OP_VSPACE:    case OP_VSPACE:
2945    return next == -ESC_V || next == -ESC_d || next == -ESC_w;    return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2946    
2947    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
2948    return next == -ESC_v;    return next == -ESC_v || next == -ESC_R;
2949    
2950    case OP_WORDCHAR:    case OP_WORDCHAR:
2951    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
2952             next == -ESC_v || next == -ESC_R;
2953    
2954    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
2955    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2212  BOOL inescq = FALSE; Line 3013  BOOL inescq = FALSE;
3013  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
3014  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
3015  const uschar *tempptr;  const uschar *tempptr;
3016    const uschar *nestptr = NULL;
3017  uschar *previous = NULL;  uschar *previous = NULL;
3018  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
3019  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
# Line 2221  uschar classbits[32]; Line 3023  uschar classbits[32];
3023  BOOL class_utf8;  BOOL class_utf8;
3024  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
3025  uschar *class_utf8data;  uschar *class_utf8data;
3026    uschar *class_utf8data_base;
3027  uschar utf8_char[6];  uschar utf8_char[6];
3028  #else  #else
3029  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
3030  uschar *utf8_char = NULL;  uschar *utf8_char = NULL;
3031  #endif  #endif
3032    
3033  #ifdef DEBUG  #ifdef PCRE_DEBUG
3034  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
3035  #endif  #endif
3036    
# Line 2260  req_caseopt = ((options & PCRE_CASELESS) Line 3063  req_caseopt = ((options & PCRE_CASELESS)
3063  for (;; ptr++)  for (;; ptr++)
3064    {    {
3065    BOOL negate_class;    BOOL negate_class;
3066      BOOL should_flip_negation;
3067    BOOL possessive_quantifier;    BOOL possessive_quantifier;
3068    BOOL is_quantifier;    BOOL is_quantifier;
3069    BOOL is_recurse;    BOOL is_recurse;
# Line 2280  for (;; ptr++) Line 3084  for (;; ptr++)
3084    
3085    c = *ptr;    c = *ptr;
3086    
3087      /* If we are at the end of a nested substitution, revert to the outer level
3088      string. Nesting only happens one level deep. */
3089    
3090      if (c == 0 && nestptr != NULL)
3091        {
3092        ptr = nestptr;
3093        nestptr = NULL;
3094        c = *ptr;
3095        }
3096    
3097    /* 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
3098    previous cycle of this loop. */    previous cycle of this loop. */
3099    
3100    if (lengthptr != NULL)    if (lengthptr != NULL)
3101      {      {
3102  #ifdef DEBUG  #ifdef PCRE_DEBUG
3103      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3104  #endif  #endif
3105      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
3106        {        {
3107        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3108        goto FAILED;        goto FAILED;
# Line 2310  for (;; ptr++) Line 3124  for (;; ptr++)
3124        goto FAILED;        goto FAILED;
3125        }        }
3126    
3127      *lengthptr += code - last_code;      *lengthptr += (int)(code - last_code);
3128      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));
3129    
3130      /* 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 2337  for (;; ptr++) Line 3151  for (;; ptr++)
3151    /* 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
3152    reference list. */    reference list. */
3153    
3154    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
3155      {      {
3156      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3157      goto FAILED;      goto FAILED;
# Line 2347  for (;; ptr++) Line 3161  for (;; ptr++)
3161    
3162    if (inescq && c != 0)    if (inescq && c != 0)
3163      {      {
3164      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3165        {        {
3166        inescq = FALSE;        inescq = FALSE;
3167        ptr++;        ptr++;
# Line 2373  for (;; ptr++) Line 3187  for (;; ptr++)
3187    /* 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
3188    a quantifier. */    a quantifier. */
3189    
3190    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
3191      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
3192        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
3193    
3194    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
3195         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2389  for (;; ptr++) Line 3204  for (;; ptr++)
3204    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3205      {      {
3206      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
3207      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
3208        {        {
3209        while (*(++ptr) != 0)        ptr++;
3210          while (*ptr != 0)
3211          {          {
3212          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3213            ptr++;
3214    #ifdef SUPPORT_UTF8
3215            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
3216    #endif
3217          }          }
3218        if (*ptr != 0) continue;        if (*ptr != 0) continue;
3219    
# Line 2414  for (;; ptr++) Line 3234  for (;; ptr++)
3234      {      {
3235      /* ===================================================================*/      /* ===================================================================*/
3236      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
3237      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
3238      case ')':      case CHAR_RIGHT_PARENTHESIS:
3239      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
3240      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
3241      *codeptr = code;      *codeptr = code;
# Line 2427  for (;; ptr++) Line 3247  for (;; ptr++)
3247          *errorcodeptr = ERR20;          *errorcodeptr = ERR20;
3248          goto FAILED;          goto FAILED;
3249          }          }
3250        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += (int)(code - last_code);   /* To include callout length */
3251        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
3252        }        }
3253      return TRUE;      return TRUE;
# Line 2437  for (;; ptr++) Line 3257  for (;; ptr++)
3257      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
3258      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3259    
3260      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
3261        previous = NULL;
3262      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3263        {        {
3264        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3265          *code++ = OP_CIRCM;
3266        }        }
3267      previous = NULL;      else *code++ = OP_CIRC;
     *code++ = OP_CIRC;  
3268      break;      break;
3269    
3270      case '$':      case CHAR_DOLLAR_SIGN:
3271      previous = NULL;      previous = NULL;
3272      *code++ = OP_DOLL;      *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
3273      break;      break;
3274    
3275      /* 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
3276      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
3277    
3278      case '.':      case CHAR_DOT:
3279      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3280      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
3281      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3282      previous = code;      previous = code;
3283      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
3284      break;      break;
3285    
3286    
# Line 2474  for (;; ptr++) Line 3295  for (;; ptr++)
3295      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,
3296      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
3297      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.
     */  
3298    
3299      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
3300        default (Perl) mode, it is treated as a data character. */
3301    
3302        case CHAR_RIGHT_SQUARE_BRACKET:
3303        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3304          {
3305          *errorcodeptr = ERR64;
3306          goto FAILED;
3307          }
3308        goto NORMAL_CHAR;
3309    
3310        case CHAR_LEFT_SQUARE_BRACKET:
3311      previous = code;      previous = code;
3312    
3313      /* 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
3314      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. */
3315    
3316      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3317          check_posix_syntax(ptr, &tempptr, cd))           ptr[1] == CHAR_EQUALS_SIGN) &&
3318            check_posix_syntax(ptr, &tempptr))
3319        {        {
3320        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
3321        goto FAILED;        goto FAILED;
3322        }        }
3323    
# Line 2497  for (;; ptr++) Line 3329  for (;; ptr++)
3329      for (;;)      for (;;)
3330        {        {
3331        c = *(++ptr);        c = *(++ptr);
3332        if (c == '\\')        if (c == CHAR_BACKSLASH)
3333          {          {
3334          if (ptr[1] == 'E') ptr++;          if (ptr[1] == CHAR_E)
3335            else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;            ptr++;
3336              else break;          else if (strncmp((const char *)ptr+1,
3337                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
3338              ptr += 3;
3339            else
3340              break;
3341          }          }
3342        else if (!negate_class && c == '^')        else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
3343          negate_class = TRUE;          negate_class = TRUE;
3344        else break;        else break;
3345        }        }
3346    
3347        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
3348        an initial ']' is taken as a data character -- the code below handles
3349        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3350        [^] must match any character, so generate OP_ALLANY. */
3351    
3352        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3353            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3354          {
3355          *code++ = negate_class? OP_ALLANY : OP_FAIL;
3356          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3357          zerofirstbyte = firstbyte;
3358          break;
3359          }
3360    
3361        /* If a class contains a negative special such as \S, we need to flip the
3362        negation flag at the end, so that support for characters > 255 works
3363        correctly (they are all included in the class). */
3364    
3365        should_flip_negation = FALSE;
3366    
3367      /* 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
3368      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
3369      valued UTF-8 characters, we don't yet do any optimization. */      valued UTF-8 characters, we don't yet do any optimization. */
# Line 2525  for (;; ptr++) Line 3381  for (;; ptr++)
3381  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3382      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
3383      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
3384        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
3385  #endif  #endif
3386    
3387      /* 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 2540  for (;; ptr++) Line 3397  for (;; ptr++)
3397          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3398          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3399          }          }
3400    
3401          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
3402          data and reset the pointer. This is so that very large classes that
3403          contain a zillion UTF-8 characters no longer overwrite the work space
3404          (which is on the stack). */
3405    
3406          if (lengthptr != NULL)
3407            {
3408            *lengthptr += class_utf8data - class_utf8data_base;
3409            class_utf8data = class_utf8data_base;
3410            }
3411    
3412  #endif  #endif
3413    
3414        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
3415    
3416        if (inescq)        if (inescq)
3417          {          {
3418          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3419            {            {
3420            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
3421            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2561  for (;; ptr++) Line 3430  for (;; ptr++)
3430        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3431        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3432    
3433        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3434            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3435            check_posix_syntax(ptr, &tempptr, cd))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3436          {          {
3437          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3438          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3439          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3440          uschar pbits[32];          uschar pbits[32];
3441    
3442          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3443            {            {
3444            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3445            goto FAILED;            goto FAILED;
3446            }            }
3447    
3448          ptr += 2;          ptr += 2;
3449          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3450            {            {
3451            local_negate = TRUE;            local_negate = TRUE;
3452              should_flip_negation = TRUE;  /* Note negative special */
3453            ptr++;            ptr++;
3454            }            }
3455    
3456          posix_class = check_posix_name(ptr, tempptr - ptr);          posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3457          if (posix_class < 0)          if (posix_class < 0)
3458            {            {
3459            *errorcodeptr = ERR30;            *errorcodeptr = ERR30;
# Line 2597  for (;; ptr++) Line 3467  for (;; ptr++)
3467          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3468            posix_class = 0;            posix_class = 0;
3469    
3470          /* 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
3471          because we may be adding and subtracting from it, and we don't want to          different escape sequences that use Unicode properties. */
3472          subtract bits that may be in the main map already. At the end we or the  
3473          result into the bit map that is being built. */  #ifdef SUPPORT_UCP
3474            if ((options & PCRE_UCP) != 0)
3475              {
3476              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3477              if (posix_substitutes[pc] != NULL)
3478                {
3479                nestptr = tempptr + 1;
3480                ptr = posix_substitutes[pc] - 1;
3481                continue;
3482                }
3483              }
3484    #endif
3485            /* In the non-UCP case, we build the bit map for the POSIX class in a
3486            chunk of local store because we may be adding and subtracting from it,
3487            and we don't want to subtract bits that may be in the main map already.
3488            At the end we or the result into the bit map that is being built. */
3489    
3490          posix_class *= 3;          posix_class *= 3;
3491    
# Line 2644  for (;; ptr++) Line 3529  for (;; ptr++)
3529    
3530        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3531        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
3532        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
3533        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
3534        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
3535        character in them, so set class_charcount bigger than one. */        are either treated as literal characters (by default), or are faulted if
3536          PCRE_EXTRA is set. */
3537    
3538        if (c == '\\')        if (c == CHAR_BACKSLASH)
3539          {          {
3540          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3541          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3542    
3543          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 */  
3544          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3545            {            {
3546            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3547              {              {
3548              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3549              }              }
3550            else inescq = TRUE;            else inescq = TRUE;
3551            continue;            continue;
3552            }            }
3553            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
3554    
3555          if (c < 0)          if (c < 0)
3556            {            {
3557            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
3558            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3559    
3560            /* Save time by not doing this in the pre-compile phase. */            switch (-c)
   
           if (lengthptr == NULL) switch (-c)  
3561              {              {
3562    #ifdef SUPPORT_UCP
3563                case ESC_du:     /* These are the values given for \d etc */
3564                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3565                case ESC_wu:     /* escape sequence with an appropriate \p */
3566                case ESC_WU:     /* or \P to test Unicode properties instead */
3567                case ESC_su:     /* of the default ASCII testing. */
3568                case ESC_SU:
3569                nestptr = ptr;
3570                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3571                class_charcount -= 2;                /* Undo! */
3572                continue;
3573    #endif
3574              case ESC_d:              case ESC_d:
3575              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3576              continue;              continue;
3577    
3578              case ESC_D:              case ESC_D:
3579                should_flip_negation = TRUE;
3580              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
3581              continue;              continue;
3582    
# Line 2689  for (;; ptr++) Line 3585  for (;; ptr++)
3585              continue;              continue;
3586    
3587              case ESC_W:              case ESC_W:
3588                should_flip_negation = TRUE;
3589              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3590              continue;              continue;
3591    
3592                /* Perl 5.004 onwards omits VT from \s, but we must preserve it
3593                if it was previously set by something earlier in the character
3594                class. */
3595    
3596              case ESC_s:              case ESC_s:
3597              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];              classbits[0] |= cbits[cbit_space];
3598              classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= cbits[cbit_space+1] & ~0x08;
3599                for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3600              continue;              continue;
3601    
3602              case ESC_S:              case ESC_S:
3603                should_flip_negation = TRUE;
3604              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
3605              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3606              continue;              continue;
3607    
3608              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)  
             {  
3609              SETBIT(classbits, 0x09); /* VT */              SETBIT(classbits, 0x09); /* VT */
3610              SETBIT(classbits, 0x20); /* SPACE */              SETBIT(classbits, 0x20); /* SPACE */
3611              SETBIT(classbits, 0xa0); /* NSBP */              SETBIT(classbits, 0xa0); /* NSBP */
# Line 2742  for (;; ptr++) Line 3629  for (;; ptr++)
3629                }                }
3630  #endif  #endif
3631              continue;              continue;
             }  
3632    
3633            if (-c == ESC_H)              case ESC_H:
             {  
3634              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3635                {                {
3636                int x = 0xff;                int x = 0xff;
# Line 2787  for (;; ptr++) Line 3672  for (;; ptr++)
3672                }                }
3673  #endif  #endif
3674              continue;              continue;
             }  
3675    
3676            if (-c == ESC_v)              case ESC_v:
             {  
3677              SETBIT(classbits, 0x0a); /* LF */              SETBIT(classbits, 0x0a); /* LF */
3678              SETBIT(classbits, 0x0b); /* VT */              SETBIT(classbits, 0x0b); /* VT */
3679              SETBIT(classbits, 0x0c); /* FF */              SETBIT(classbits, 0x0c); /* FF */
# Line 2806  for (;; ptr++) Line 3689  for (;; ptr++)
3689                }                }
3690  #endif  #endif
3691              continue;              continue;
             }  
3692    
3693            if (-c == ESC_V)              case ESC_V:
             {  
3694              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3695                {                {
3696                int x = 0xff;                int x = 0xff;
# Line 2839  for (;; ptr++) Line 3720  for (;; ptr++)
3720                }                }
3721  #endif  #endif
3722              continue;              continue;
             }  
   
           /* We need to deal with \P and \p in both phases. */  
3723    
3724  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3725            if (-c == ESC_p || -c == ESC_P)              case ESC_p:
3726              {              case ESC_P:
3727              BOOL negated;                {
3728              int pdata;                BOOL negated;
3729              int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                int pdata;
3730              if (ptype < 0) goto FAILED;                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3731              class_utf8 = TRUE;                if (ptype < 0) goto FAILED;
3732              *class_utf8data++ = ((-c == ESC_p) != negated)?                class_utf8 = TRUE;
3733                XCL_PROP : XCL_NOTPROP;                *class_utf8data++ = ((-c == ESC_p) != negated)?
3734              *class_utf8data++ = ptype;                  XCL_PROP : XCL_NOTPROP;
3735              *class_utf8data++ = pdata;                *class_utf8data++ = ptype;
3736              class_charcount -= 2;   /* Not a < 256 character */                *class_utf8data++ = pdata;
3737              continue;                class_charcount -= 2;   /* Not a < 256 character */
3738              }                continue;
3739                  }
3740  #endif  #endif
3741            /* Unrecognized escapes are faulted if PCRE is running in its              /* Unrecognized escapes are faulted if PCRE is running in its
3742            strict mode. By default, for compatibility with Perl, they are              strict mode. By default, for compatibility with Perl, they are
3743            treated as literals. */              treated as literals. */
3744    
3745            if ((options & PCRE_EXTRA) != 0)              default:
3746              {              if ((options & PCRE_EXTRA) != 0)
3747              *errorcodeptr = ERR7;                {
3748              goto FAILED;                *errorcodeptr = ERR7;
3749                  goto FAILED;
3750                  }
3751                class_charcount -= 2;  /* Undo the default count from above */
3752                c = *ptr;              /* Get the final character and fall through */
3753                break;
3754              }              }
   
           class_charcount -= 2;  /* Undo the default count from above */  
           c = *ptr;              /* Get the final character and fall through */  
3755            }            }
3756    
3757          /* 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 2884  for (;; ptr++) Line 3765  for (;; ptr++)
3765        entirely. The code for handling \Q and \E is messy. */        entirely. The code for handling \Q and \E is messy. */
3766    
3767        CHECK_RANGE:        CHECK_RANGE:
3768        while (ptr[1] == '\\' && ptr[2] == 'E')        while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3769          {          {
3770          inescq = FALSE;          inescq = FALSE;
3771          ptr += 2;          ptr += 2;
# Line 2892  for (;; ptr++) Line 3773  for (;; ptr++)
3773    
3774        oldptr = ptr;        oldptr = ptr;
3775    
3776        if (!inescq && ptr[1] == '-')        /* Remember \r or \n */
3777    
3778          if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3779    
3780          /* Check for range */
3781    
3782          if (!inescq && ptr[1] == CHAR_MINUS)
3783          {          {
3784          int d;          int d;
3785          ptr += 2;          ptr += 2;
3786          while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3787    
3788          /* 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
3789          mode. */          mode. */
3790    
3791          while (*ptr == '\\' && ptr[1] == 'Q')          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3792            {            {
3793            ptr += 2;            ptr += 2;
3794            if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }            if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3795                { ptr += 2; continue; }
3796            inescq = TRUE;            inescq = TRUE;
3797            break;            break;
3798            }            }
3799    
3800          if (*ptr == 0 || (!inescq && *ptr == ']'))          if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3801            {            {
3802            ptr = oldptr;            ptr = oldptr;
3803            goto LONE_SINGLE_CHARACTER;            goto LONE_SINGLE_CHARACTER;
# Line 2928  for (;; ptr++) Line 3816  for (;; ptr++)
3816          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
3817          in such circumstances. */          in such circumstances. */
3818    
3819          if (!inescq && d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3820            {            {
3821            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3822            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3823    
3824            /* \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 */  
3825    
3826            if (d < 0)            if (d < 0)
3827              {              {
3828              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  
3829                {                {
3830                ptr = oldptr;                ptr = oldptr;
3831                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 2959  for (;; ptr++) Line 3844  for (;; ptr++)
3844    
3845          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3846    
3847            /* Remember \r or \n */
3848    
3849            if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3850    
3851          /* 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
3852          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
3853          matching for characters > 127 is available only if UCP support is          matching for characters > 127 is available only if UCP support is
# Line 3077  for (;; ptr++) Line 3966  for (;; ptr++)
3966          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3967            {            {
3968            unsigned int othercase;            unsigned int othercase;
3969            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)            if ((othercase = UCD_OTHERCASE(c)) != c)
3970              {              {
3971              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3972              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 3102  for (;; ptr++) Line 3991  for (;; ptr++)
3991          }          }
3992        }        }
3993    
3994      /* 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.
3995        If we are at the end of an internal nested string, revert to the outer
3996        string. */
3997    
3998        while (((c = *(++ptr)) != 0 ||
3999               (nestptr != NULL &&
4000                 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != 0)) &&
4001               (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
4002    
4003      while ((c = *(++ptr)) != 0 && (c != ']' || inescq));      /* Check for missing terminating ']' */
4004    
4005      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)
4006        {        {
4007        *errorcodeptr = ERR6;        *errorcodeptr = ERR6;
4008        goto FAILED;        goto FAILED;
4009        }        }
4010    
4011      /* 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
4012      less than 256. In non-UTF-8 mode we can always optimize. In UTF-8 mode, we      less than 256. As long as there were no characters >= 128 and there was no
4013      can optimize the negative case only if there were no characters >= 128      use of \p or \P, in other words, no use of any XCLASS features, we can
4014      because OP_NOT and the related opcodes like OP_NOTSTAR operate on      optimize.
4015      single-bytes only. This is an historical hangover. Maybe one day we can  
4016      tidy these opcodes to handle multi-byte characters.      In UTF-8 mode, we can optimize the negative case only if there were no
4017        characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
4018        operate on single-bytes characters only. This is an historical hangover.
4019        Maybe one day we can tidy these opcodes to handle multi-byte characters.
4020    
4021      The optimization throws away the bit map. We turn the item into a      The optimization throws away the bit map. We turn the item into a
4022      1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note      1-character OP_CHAR[I] if it's positive, or OP_NOT[I] if it's negative.
4023      that OP_NOT does not support multibyte characters. In the positive case, it      Note that OP_NOT[I] does not support multibyte characters. In the positive
4024      can cause firstbyte to be set. Otherwise, there can be no first char if      case, it can cause firstbyte to be set. Otherwise, there can be no first
4025      this item is first, whatever repeat count may follow. In the case of      char if this item is first, whatever repeat count may follow. In the case
4026      reqbyte, save the previous value for reinstating. */      of reqbyte, save the previous value for reinstating. */
4027    
4028  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4029      if (class_charcount == 1 &&      if (class_charcount == 1 && !class_utf8 &&
4030            (!utf8 ||        (!utf8 || !negate_class || class_lastchar < 128))
           (!class_utf8 && (!negate_class || class_lastchar < 128))))  
   
4031  #else  #else
4032      if (class_charcount == 1)      if (class_charcount == 1)
4033  #endif  #endif
4034        {        {
4035        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
4036    
4037        /* The OP_NOT opcode works on one-byte characters only. */        /* The OP_NOT[I] opcodes work on one-byte characters only. */
4038    
4039        if (negate_class)        if (negate_class)
4040          {          {
4041          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
4042          zerofirstbyte = firstbyte;          zerofirstbyte = firstbyte;
4043          *code++ = OP_NOT;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
4044          *code++ = class_lastchar;          *code++ = class_lastchar;
4045          break;          break;
4046          }          }
# Line 3173  for (;; ptr++) Line 4070  for (;; ptr++)
4070      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
4071    
4072      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
4073      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode, unless there was a negated special
4074      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
4075        characters > 255 are in the class, so any that were explicitly given as
4076        well can be ignored. If (when there are explicit characters > 255 that must
4077        be listed) there are no characters < 256, we can omit the bitmap in the
4078        actual compiled code. */
4079    
4080  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4081      if (class_utf8)      if (class_utf8 && (!should_flip_negation || (options & PCRE_UCP) != 0))
4082        {        {
4083        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
4084        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 3203  for (;; ptr++) Line 4104  for (;; ptr++)
4104        }        }
4105  #endif  #endif
4106    
4107      /* 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
4108      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
4109      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
4110      setting must remain unchanged after any kind of repeat. */      (non-UCP) in the class. Then copy the 32-byte map into the code vector,
4111        negating it if necessary. */
4112    
4113        *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
4114      if (negate_class)      if (negate_class)
4115        {        {
       *code++ = OP_NCLASS;  
4116        if (lengthptr == NULL)    /* Save time in the pre-compile phase */        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
4117          for (c = 0; c < 32; c++) code[c] = ~classbits[c];          for (c = 0; c < 32; c++) code[c] = ~classbits[c];
4118        }        }
4119      else      else
4120        {        {
       *code++ = OP_CLASS;  
4121        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
4122        }        }
4123      code += 32;      code += 32;
# Line 3227  for (;; ptr++) Line 4128  for (;; ptr++)
4128      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
4129      has been tested above. */      has been tested above. */
4130    
4131      case '{':      case CHAR_LEFT_CURLY_BRACKET:
4132      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
4133      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
4134      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
4135      goto REPEAT;      goto REPEAT;
4136    
4137      case '*':      case CHAR_ASTERISK:
4138      repeat_min = 0;      repeat_min = 0;
4139      repeat_max = -1;      repeat_max = -1;
4140      goto REPEAT;      goto REPEAT;
4141    
4142      case '+':      case CHAR_PLUS:
4143      repeat_min = 1;      repeat_min = 1;
4144      repeat_max = -1;      repeat_max = -1;
4145      goto REPEAT;      goto REPEAT;
4146    
4147      case '?':      case CHAR_QUESTION_MARK:
4148      repeat_min = 0;      repeat_min = 0;
4149      repeat_max = 1;      repeat_max = 1;
4150    
# Line 3278  for (;; ptr++) Line 4179  for (;; ptr++)
4179      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
4180      repeat type to the non-default. */      repeat type to the non-default. */
4181    
4182      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
4183        {        {
4184        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
4185        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
4186        ptr++;        ptr++;
4187        }        }
4188      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
4189        {        {
4190        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
4191        ptr++;        ptr++;
# Line 3297  for (;; ptr++) Line 4198  for (;; ptr++)
4198      the first thing in a branch because the x will have gone into firstbyte      the first thing in a branch because the x will have gone into firstbyte
4199      instead.  */      instead.  */
4200    
4201      if (*previous == OP_CHAR || *previous == OP_CHARNC)      if (*previous == OP_CHAR || *previous == OP_CHARI)
4202        {        {
4203          op_type = (*previous == OP_CHAR)? 0 : OP_STARI - OP_STAR;
4204    
4205        /* Deal with UTF-8 characters that take up more than one byte. It's        /* Deal with UTF-8 characters that take up more than one byte. It's
4206        easier to write this out separately than try to macrify it. Use c to        easier to write this out separately than try to macrify it. Use c to
4207        hold the length of the character in bytes, plus 0x80 to flag that it's a        hold the length of the character in bytes, plus 0x80 to flag that it's a
# Line 3331  for (;; ptr++) Line 4234  for (;; ptr++)
4234    
4235        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4236            repeat_max < 0 &&            repeat_max < 0 &&
4237            check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
             options, cd))  
4238          {          {
4239          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4240          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3344  for (;; ptr++) Line 4246  for (;; ptr++)
4246      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
4247      one of the special opcodes, replacing it. The code is shared with single-      one of the special opcodes, replacing it. The code is shared with single-
4248      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
4249      repeat_type. We can also test for auto-possessification. OP_NOT is      repeat_type. We can also test for auto-possessification. OP_NOT and OP_NOTI
4250      currently used only for single-byte chars. */      are currently used only for single-byte chars. */
4251    
4252      else if (*previous == OP_NOT)      else if (*previous == OP_NOT || *previous == OP_NOTI)
4253        {        {
4254        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = ((*previous == OP_NOT)? OP_NOTSTAR : OP_NOTSTARI) - OP_STAR;
4255        c = previous[1];        c = previous[1];
4256        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4257            repeat_max < 0 &&            repeat_max < 0 &&
4258            check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4259          {          {
4260          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4261          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3377  for (;; ptr++) Line 4279  for (;; ptr++)
4279    
4280        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4281            repeat_max < 0 &&            repeat_max < 0 &&
4282            check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4283          {          {
4284          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4285          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3399  for (;; ptr++) Line 4301  for (;; ptr++)
4301    
4302        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
4303    
4304          /*--------------------------------------------------------------------*/
4305          /* This code is obsolete from release 8.00; the restriction was finally
4306          removed: */
4307    
4308        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4309        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4310    
4311        if (repeat_max != 1) cd->nopartial = TRUE;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4312          /*--------------------------------------------------------------------*/
4313    
4314        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
4315    
# Line 3541  for (;; ptr++) Line 4448  for (;; ptr++)
4448  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4449               *previous == OP_XCLASS ||               *previous == OP_XCLASS ||
4450  #endif  #endif
4451               *previous == OP_REF)               *previous == OP_REF ||
4452                 *previous == OP_REFI)
4453        {        {
4454        if (repeat_max == 0)        if (repeat_max == 0)
4455          {          {
# Line 3549  for (;; ptr++) Line 4457  for (;; ptr++)
4457          goto END_REPEAT;          goto END_REPEAT;
4458          }          }
4459    
4460          /*--------------------------------------------------------------------*/
4461          /* This code is obsolete from release 8.00; the restriction was finally
4462          removed: */
4463    
4464        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4465        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4466    
4467        if (repeat_max != 1) cd->nopartial = TRUE;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4468          /*--------------------------------------------------------------------*/
4469    
4470        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4471          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 3577  for (;; ptr++) Line 4490  for (;; ptr++)
4490        {        {
4491        register int i;        register int i;
4492        int ketoffset = 0;        int ketoffset = 0;
4493        int len = code - previous;        int len = (int)(code - previous);
4494        uschar *bralink = NULL;        uschar *bralink = NULL;
4495    
4496        /* Repeating a DEFINE group is pointless */        /* Repeating a DEFINE group is pointless */
# Line 3590  for (;; ptr++) Line 4503  for (;; ptr++)
4503    
4504        /* If the maximum repeat count is unlimited, find the end of the bracket        /* If the maximum repeat count is unlimited, find the end of the bracket
4505        by scanning through from the start, and compute the offset back to it        by scanning through from the start, and compute the offset back to it
4506        from the current code pointer. There may be an OP_OPT setting following        from the current code pointer. */
       the final KET, so we can't find the end just by going back from the code  
       pointer. */  
4507    
4508        if (repeat_max == -1)        if (repeat_max == -1)
4509          {          {
4510          register uschar *ket = previous;          register uschar *ket = previous;
4511          do ket += GET(ket, 1); while (*ket != OP_KET);          do ket += GET(ket, 1); while (*ket != OP_KET);
4512          ketoffset = code - ket;          ketoffset = (int)(code - ket);
4513          }          }
4514    
4515        /* 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 3610  for (;; ptr++) Line 4521  for (;; ptr++)
4521    
4522        if (repeat_min == 0)        if (repeat_min == 0)
4523          {          {
4524          /* 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
4525          altogether. */          output altogether, like this:
   
         if (repeat_max == 0)  
           {  
           code = previous;  
           goto END_REPEAT;  
           }  
4526    
4527          /* If the maximum is 1 or unlimited, we just have to stick in the          ** if (repeat_max == 0)
4528          BRAZERO and do no more at this point. However, we do need to adjust          **   {
4529          any OP_RECURSE calls inside the group that refer to the group itself or          **   code = previous;
4530          any internal or forward referenced group, because the offset is from          **   goto END_REPEAT;
4531          the start of the whole regex. Temporarily terminate the pattern while          **   }
4532          doing this. */  
4533            However, that fails when a group is referenced as a subroutine from
4534            elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it
4535            so that it is skipped on execution. As we don't have a list of which
4536            groups are referenced, we cannot do this selectively.
4537    
4538            If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4539            and do no more at this point. However, we do need to adjust any
4540            OP_RECURSE calls inside the group that refer to the group itself or any
4541            internal or forward referenced group, because the offset is from the
4542            start of the whole regex. Temporarily terminate the pattern while doing
4543            this. */
4544    
4545          if (repeat_max <= 1)          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
4546            {            {
4547            *code = OP_END;            *code = OP_END;
4548            adjust_recurse(previous, 1, utf8, cd, save_hwm);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
4549            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
4550            code++;            code++;
4551              if (repeat_max == 0)
4552                {
4553                *previous++ = OP_SKIPZERO;
4554                goto END_REPEAT;
4555                }
4556            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4557            }            }
4558    
# Line 3656  for (;; ptr++) Line 4577  for (;; ptr++)
4577            /* We chain together the bracket offset fields that have to be            /* We chain together the bracket offset fields that have to be
4578            filled in later when the ends of the brackets are reached. */            filled in later when the ends of the brackets are reached. */
4579    
4580            offset = (bralink == NULL)? 0 : previous - bralink;            offset = (bralink == NULL)? 0 : (int)(previous - bralink);
4581            bralink = previous;            bralink = previous;
4582            PUTINC(previous, 0, offset);            PUTINC(previous, 0, offset);
4583            }            }
# Line 3677  for (;; ptr++) Line 4598  for (;; ptr++)
4598            {            {
4599            /* 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
4600            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
4601            potential integer overflow. */            potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
4602              integer type when available, otherwise double. */
4603    
4604            if (lengthptr != NULL)            if (lengthptr != NULL)
4605              {              {
4606              int delta = (repeat_min - 1)*length_prevgroup;              int delta = (repeat_min - 1)*length_prevgroup;
4607              if ((double)(repeat_min - 1)*(double)length_prevgroup >              if ((INT64_OR_DOUBLE)(repeat_min - 1)*
4608                                                              (double)INT_MAX ||                    (INT64_OR_DOUBLE)length_prevgroup >
4609                        (INT64_OR_DOUBLE)INT_MAX ||
4610                  OFLOW_MAX - *lengthptr < delta)                  OFLOW_MAX - *lengthptr < delta)
4611                {                {
4612                *errorcodeptr = ERR20;                *errorcodeptr = ERR20;
# Line 3729  for (;; ptr++) Line 4652  for (;; ptr++)
4652          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
4653          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
4654          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
4655          paranoid checks to avoid integer overflow. */          paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
4656            a 64-bit integer type when available, otherwise double. */
4657    
4658          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4659            {            {
4660            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4661                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4662            if ((double)repeat_max *            if ((INT64_OR_DOUBLE)repeat_max *
4663                  (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)                  (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4664                    > (double)INT_MAX ||                    > (INT64_OR_DOUBLE)INT_MAX ||
4665                OFLOW_MAX - *lengthptr < delta)                OFLOW_MAX - *lengthptr < delta)
4666              {              {
4667              *errorcodeptr = ERR20;              *errorcodeptr = ERR20;
# Line 3762  for (;; ptr++) Line 4686  for (;; ptr++)
4686              {              {
4687              int offset;              int offset;
4688              *code++ = OP_BRA;              *code++ = OP_BRA;
4689              offset = (bralink == NULL)? 0 : code - bralink;              offset = (bralink == NULL)? 0 : (int)(code - bralink);
4690              bralink = code;              bralink = code;
4691              PUTINC(code, 0, offset);              PUTINC(code, 0, offset);
4692              }              }
# Line 3783  for (;; ptr++) Line 4707  for (;; ptr++)
4707          while (bralink != NULL)          while (bralink != NULL)
4708            {            {
4709            int oldlinkoffset;            int oldlinkoffset;
4710            int offset = code - bralink + 1;            int offset = (int)(code - bralink + 1);
4711            uschar *bra = code - offset;            uschar *bra = code - offset;
4712            oldlinkoffset = GET(bra, 1);            oldlinkoffset = GET(bra, 1);
4713            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;
# Line 3814  for (;; ptr++) Line 4738  for (;; ptr++)
4738            uschar *scode = bracode;            uschar *scode = bracode;
4739            do            do
4740              {              {
4741              if (could_be_empty_branch(scode, ketcode, utf8))              if (could_be_empty_branch(scode, ketcode, utf8, cd))
4742                {                {
4743                *bracode += OP_SBRA - OP_BRA;                *bracode += OP_SBRA - OP_BRA;
4744                break;                break;
# Line 3826  for (;; ptr++) Line 4750  for (;; ptr++)
4750          }          }
4751        }        }
4752    
4753        /* If previous is OP_FAIL, it was generated by an empty class [] in
4754        JavaScript mode. The other ways in which OP_FAIL can be generated, that is
4755        by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
4756        error above. We can just ignore the repeat in JS case. */
4757    
4758        else if (*previous == OP_FAIL) goto END_REPEAT;
4759    
4760      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
4761    
4762      else      else
# Line 3850  for (;; ptr++) Line 4781  for (;; ptr++)
4781      if (possessive_quantifier)      if (possessive_quantifier)
4782        {        {
4783        int len;        int len;
4784        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||  
4785            *tempcode == OP_NOTEXACT)        if (*tempcode == OP_TYPEEXACT)
4786            tempcode += _pcre_OP_lengths[*tempcode] +
4787              ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
4788    
4789          else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
4790            {
4791          tempcode += _pcre_OP_lengths[*tempcode];          tempcode += _pcre_OP_lengths[*tempcode];
4792        len = code - tempcode;  #ifdef SUPPORT_UTF8
4793            if (utf8 && tempcode[-1] >= 0xc0)
4794              tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];
4795    #endif
4796            }
4797    
4798          len = (int)(code - tempcode);
4799        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4800          {          {
4801          case OP_STAR:  *tempcode = OP_POSSTAR; break;          case OP_STAR:  *tempcode = OP_POSSTAR; break;
# Line 3861  for (;; ptr++) Line 4803  for (;; ptr++)
4803          case OP_QUERY: *tempcode = OP_POSQUERY; break;          case OP_QUERY: *tempcode = OP_POSQUERY; break;
4804          case OP_UPTO:  *tempcode = OP_POSUPTO; break;          case OP_UPTO:  *tempcode = OP_POSUPTO; break;
4805    
4806          case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;          case OP_STARI:  *tempcode = OP_POSSTARI; break;
4807          case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;          case OP_PLUSI:  *tempcode = OP_POSPLUSI; break;
4808          case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;          case OP_QUERYI: *tempcode = OP_POSQUERYI; break;
4809          case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;          case OP_UPTOI:  *tempcode = OP_POSUPTOI; break;
4810    
4811          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;
4812          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;
4813          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4814          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4815    
4816            case OP_NOTSTARI:  *tempcode = OP_NOTPOSSTARI; break;
4817            case OP_NOTPLUSI:  *tempcode = OP_NOTPOSPLUSI; break;
4818            case OP_NOTQUERYI: *tempcode = OP_NOTPOSQUERYI; break;
4819            case OP_NOTUPTOI:  *tempcode = OP_NOTPOSUPTOI; break;
4820    
4821            case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;
4822            case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;
4823            case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
4824            case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;
4825    
4826            /* Because we are moving code along, we must ensure that any
4827            pending recursive references are updated. */
4828    
4829          default:          default:
4830            *code = OP_END;
4831            adjust_recurse(tempcode, 1 + LINK_SIZE, utf8, cd, save_hwm);
4832          memmove(tempcode + 1+LINK_SIZE, tempcode, len);          memmove(tempcode + 1+LINK_SIZE, tempcode, len);
4833          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
4834          len += 1 + LINK_SIZE;          len += 1 + LINK_SIZE;
# Line 3898  for (;; ptr++) Line 4855  for (;; ptr++)
4855      lookbehind or option setting or condition or all the other extended      lookbehind or option setting or condition or all the other extended
4856      parenthesis forms.  */      parenthesis forms.  */
4857    
4858      case '(':      case CHAR_LEFT_PARENTHESIS:
4859      newoptions = options;      newoptions = options;
4860      skipbytes = 0;      skipbytes = 0;
4861      bravalue = OP_CBRA;      bravalue = OP_CBRA;
# Line 3907  for (;; ptr++) Line 4864  for (;; ptr++)
4864    
4865      /* First deal with various "verbs" that can be introduced by '*'. */      /* First deal with various "verbs" that can be introduced by '*'. */
4866    
4867      if (*(++ptr) == '*' && (cd->ctypes[ptr[1]] & ctype_letter) != 0)      if (*(++ptr) == CHAR_ASTERISK &&
4868             ((cd->ctypes[ptr[1]] & ctype_letter) != 0 || ptr[1] == ':'))
4869        {        {
4870        int i, namelen;        int i, namelen;
4871        const uschar *name = ++ptr;        int arglen = 0;
4872          const char *vn = verbnames;
4873          const uschar *name = ptr + 1;
4874          const uschar *arg = NULL;
4875        previous = NULL;        previous = NULL;
4876        while ((cd->ctypes[*++ptr] & ctype_letter) != 0);        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
4877        if (*ptr == ':')        namelen = (int)(ptr - name);
4878    
4879          if (*ptr == CHAR_COLON)
4880          {          {
4881          *errorcodeptr = ERR59;   /* Not supported */          arg = ++ptr;
4882          goto FAILED;          while ((cd->ctypes[*ptr] & (ctype_letter|ctype_digit)) != 0
4883              || *ptr == '_') ptr++;
4884            arglen = (int)(ptr - arg);
4885          }          }
4886        if (*ptr != ')')  
4887          if (*ptr != CHAR_RIGHT_PARENTHESIS)
4888          {          {
4889          *errorcodeptr = ERR60;          *errorcodeptr = ERR60;
4890          goto FAILED;          goto FAILED;
4891          }          }
4892        namelen = ptr - name;  
4893          /* Scan the table of verb names */
4894    
4895        for (i = 0; i < verbcount; i++)        for (i = 0; i < verbcount; i++)
4896          {          {
4897          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
4898              strncmp((char *)name, verbs[i].name, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
4899            {            {
4900            *code = verbs[i].op;            /* Check for open captures before ACCEPT */
4901            if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;  
4902            break;            if (verbs[i].op == OP_ACCEPT)
4903                {
4904                open_capitem *oc;
4905                cd->had_accept = TRUE;
4906                for (oc = cd->open_caps; oc != NULL; oc = oc->next)
4907                  {
4908                  *code++ = OP_CLOSE;
4909                  PUT2INC(code, 0, oc->number);
4910                  }
4911                }
4912    
4913              /* Handle the cases with/without an argument */
4914    
4915              if (arglen == 0)
4916                {
4917                if (verbs[i].op < 0)   /* Argument is mandatory */
4918                  {
4919                  *errorcodeptr = ERR66;
4920                  goto FAILED;
4921                  }
4922                *code = verbs[i].op;
4923                if (*code++ == OP_THEN)
4924                  {
4925                  PUT(code, 0, code - bcptr->current_branch - 1);
4926                  code += LINK_SIZE;
4927                  }
4928                }
4929    
4930              else
4931                {
4932                if (verbs[i].op_arg < 0)   /* Argument is forbidden */
4933                  {
4934                  *errorcodeptr = ERR59;
4935                  goto FAILED;
4936                  }
4937                *code = verbs[i].op_arg;
4938                if (*code++ == OP_THEN_ARG)
4939                  {
4940                  PUT(code, 0, code - bcptr->current_branch - 1);
4941                  code += LINK_SIZE;
4942                  }
4943                *code++ = arglen;
4944                memcpy(code, arg, arglen);
4945                code += arglen;
4946                *code++ = 0;
4947                }
4948    
4949              break;  /* Found verb, exit loop */
4950            }            }
4951    
4952            vn += verbs[i].len + 1;
4953          }          }