/[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 323 by ph10, Wed Mar 5 17:23:42 2008 UTC revision 640 by ph10, Mon Jul 25 10:50:28 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) 1997-2008 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 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 142  static const short int escapes[] = { Line 183  static const short int escapes[] = {
183    
184  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is  /* 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  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 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    int   len;    int   len;                 /* Length of verb name */
192    int   op;    int   op;                  /* Op when no arg, or -1 if arg mandatory */
193      int   op_arg;              /* Op when arg present, or -1 if not allowed */
194  } verbitem;  } verbitem;
195    
196  static const char verbnames[] =  static const char verbnames[] =
197    "ACCEPT\0"    "\0"                       /* Empty name is a shorthand for MARK */
198    "COMMIT\0"    STRING_MARK0
199    "F\0"    STRING_ACCEPT0
200    "FAIL\0"    STRING_COMMIT0
201    "PRUNE\0"    STRING_F0
202    "SKIP\0"    STRING_FAIL0
203    "THEN";    STRING_PRUNE0
204      STRING_SKIP0
205  static verbitem verbs[] = {    STRING_THEN;
206    { 6, OP_ACCEPT },  
207    { 6, OP_COMMIT },  static const verbitem verbs[] = {
208    { 1, OP_FAIL },    { 0, -1,        OP_MARK },
209    { 4, OP_FAIL },    { 4, -1,        OP_MARK },
210    { 5, OP_PRUNE },    { 6, OP_ACCEPT, -1 },
211    { 4, OP_SKIP  },    { 6, OP_COMMIT, -1 },
212    { 4, OP_THEN  }    { 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    
221    
222  /* Tables of names of POSIX character classes and their lengths. The names are  /* Tables of names of POSIX character classes and their lengths. The names are
# Line 178  length entry. The first three must be al Line 226  length entry. The first three must be al
226  for handling case independence. */  for handling case independence. */
227    
228  static const char posix_names[] =  static const char posix_names[] =
229    "alpha\0"  "lower\0"  "upper\0"  "alnum\0"  "ascii\0"  "blank\0"    STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
230    "cntrl\0"  "digit\0"  "graph\0"  "print\0"  "punct\0"  "space\0"    STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
231    "word\0"   "xdigit";    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 212  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 224  the number of relocations needed when a Line 320  the number of relocations needed when a
320  it is now one long string. We cannot use a table of offsets, because the  it is now one long string. We cannot use a table of offsets, because the
321  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
322  simply count through to the one we want - this isn't a performance issue  simply count through to the one we want - this isn't a performance issue
323  because these strings are used only when there is a compilation error. */  because these strings are used only when there is a compilation error.
324    
325    Each substring ends with \0 to insert a null character. This includes the final
326    substring, so that the whole string ends with \0\0, which can be detected when
327    counting through. */
328    
329  static const char error_texts[] =  static const char error_texts[] =
330    "no error\0"    "no error\0"
# Line 271  static const char error_texts[] = Line 371  static const char error_texts[] =
371    /* 35 */    /* 35 */
372    "invalid condition (?(0)\0"    "invalid condition (?(0)\0"
373    "\\C not allowed in lookbehind assertion\0"    "\\C not allowed in lookbehind assertion\0"
374    "PCRE does not support \\L, \\l, \\N, \\U, or \\u\0"    "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
375    "number after (?C is > 255\0"    "number after (?C is > 255\0"
376    "closing ) for (?C expected\0"    "closing ) for (?C expected\0"
377    /* 40 */    /* 40 */
# Line 293  static const char error_texts[] = Line 393  static const char error_texts[] =
393    "internal error: previously-checked referenced subpattern not found\0"    "internal error: previously-checked referenced subpattern not found\0"
394    "DEFINE group contains more than one branch\0"    "DEFINE group contains more than one branch\0"
395    /* 55 */    /* 55 */
396    "repeating a DEFINE group is not allowed\0"    "repeating a DEFINE group is not allowed\0"  /** DEAD **/
397    "inconsistent NEWLINE options\0"    "inconsistent NEWLINE options\0"
398    "\\g is not followed by a braced name or an optionally braced non-zero number\0"    "\\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\0"    "a numbered reference must not be zero\0"
400    "(*VERB) with an argument is not supported\0"    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
401    /* 60 */    /* 60 */
402    "(*VERB) not recognized\0"    "(*VERB) not recognized\0"
403    "number is too big\0"    "number is too big\0"
404    "subpattern name expected\0"    "subpattern name expected\0"
405    "digit expected after (?+";    "digit expected after (?+\0"
406      "] is an invalid data character in JavaScript compatibility mode\0"
407      /* 65 */
408      "different names for subpatterns of the same number are not allowed\0"
409      "(*MARK) must have an argument\0"
410      "this version of PCRE is not compiled with PCRE_UCP support\0"
411      "\\c must be followed by an ASCII character\0"
412      "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
413      ;
414    
415  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
416  patterns. Note that the tables in chartables are dependent on the locale, and  patterns. Note that the tables in chartables are dependent on the locale, and
# Line 321  For convenience, we use the same bit def Line 428  For convenience, we use the same bit def
428    
429  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
430    
431  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC
432    
433    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
434    UTF-8 mode. */
435    
436  static const unsigned char digitab[] =  static const unsigned char digitab[] =
437    {    {
438    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
# Line 357  static const unsigned char digitab[] = Line 468  static const unsigned char digitab[] =
468    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
469    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
470    
471  #else           /* This is the "abnormal" case, for EBCDIC systems */  #else
472    
473    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
474    
475  static const unsigned char digitab[] =  static const unsigned char digitab[] =
476    {    {
477    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
# Line 432  static const unsigned char ebcdic_charta Line 546  static const unsigned char ebcdic_charta
546  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
547    
548  static BOOL  static BOOL
549    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,    compile_regex(int, uschar **, const uschar **, int *, BOOL, BOOL, int, int *,
550      int *, int *, branch_chain *, compile_data *, int *);      int *, branch_chain *, compile_data *, int *);
551    
552    
553    
# Line 454  static const char * Line 568  static const char *
568  find_error_text(int n)  find_error_text(int n)
569  {  {
570  const char *s = error_texts;  const char *s = error_texts;
571  for (; n > 0; n--) while (*s++ != 0);  for (; n > 0; n--)
572      {
573      while (*s++ != 0) {};
574      if (*s == 0) return "Error text not found (please report)";
575      }
576  return s;  return s;
577  }  }
578    
579    
580  /*************************************************  /*************************************************
581    *            Check for counted repeat            *
582    *************************************************/
583    
584    /* This function is called when a '{' is encountered in a place where it might
585    start a quantifier. It looks ahead to see if it really is a quantifier or not.
586    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
587    where the ddds are digits.
588    
589    Arguments:
590      p         pointer to the first char after '{'
591    
592    Returns:    TRUE or FALSE
593    */
594    
595    static BOOL
596    is_counted_repeat(const uschar *p)
597    {
598    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
599    while ((digitab[*p] & ctype_digit) != 0) p++;
600    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
601    
602    if (*p++ != CHAR_COMMA) return FALSE;
603    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
604    
605    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
606    while ((digitab[*p] & ctype_digit) != 0) p++;
607    
608    return (*p == CHAR_RIGHT_CURLY_BRACKET);
609    }
610    
611    
612    
613    /*************************************************
614  *            Handle escapes                      *  *            Handle escapes                      *
615  *************************************************/  *************************************************/
616    
# Line 502  if (c == 0) *errorcodeptr = ERR1; Line 653  if (c == 0) *errorcodeptr = ERR1;
653  in a table. A non-zero result is something that can be returned immediately.  in a table. A non-zero result is something that can be returned immediately.
654  Otherwise further processing may be required. */  Otherwise further processing may be required. */
655    
656  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
657  else if (c < '0' || c > 'z') {}                           /* Not alphanumeric */  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */
658  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0) c = i;
659    
660  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
661  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */
# Line 523  else Line 674  else
674      /* A number of Perl escapes are not handled by PCRE. We give an explicit      /* A number of Perl escapes are not handled by PCRE. We give an explicit
675      error. */      error. */
676    
677      case 'l':      case CHAR_l:
678      case 'L':      case CHAR_L:
679      case 'N':      case CHAR_u:
680      case 'u':      case CHAR_U:
     case 'U':  
681      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
682      break;      break;
683    
684      /* \g must be followed by a number, either plain or braced. If positive, it      /* In a character class, \g is just a literal "g". Outside a character
685      is an absolute backreference. If negative, it is a relative backreference.      class, \g must be followed by one of a number of specific things:
686      This is a Perl 5.10 feature. Perl 5.10 also supports \g{name} as a  
687      reference to a named group. This is part of Perl's movement towards a      (1) A number, either plain or braced. If positive, it is an absolute
688      unified syntax for back references. As this is synonymous with \k{name}, we      backreference. If negative, it is a relative backreference. This is a Perl
689      fudge it up by pretending it really was \k. */      5.10 feature.
690    
691        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
692        is part of Perl's movement towards a unified syntax for back references. As
693        this is synonymous with \k{name}, we fudge it up by pretending it really
694        was \k.
695    
696        (3) For Oniguruma compatibility we also support \g followed by a name or a
697        number either in angle brackets or in single quotes. However, these are
698        (possibly recursive) subroutine calls, _not_ backreferences. Just return
699        the -ESC_g code (cf \k). */
700    
701        case CHAR_g:
702        if (isclass) break;
703        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
704          {
705          c = -ESC_g;
706          break;
707          }
708    
709        /* Handle the Perl-compatible cases */
710    
711      case 'g':      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
     if (ptr[1] == '{')  
712        {        {
713        const uschar *p;        const uschar *p;
714        for (p = ptr+2; *p != 0 && *p != '}'; p++)        for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
715          if (*p != '-' && (digitab[*p] & ctype_digit) == 0) break;          if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
716        if (*p != 0 && *p != '}')        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
717          {          {
718          c = -ESC_k;          c = -ESC_k;
719          break;          break;
# Line 554  else Line 723  else
723        }        }
724      else braced = FALSE;      else braced = FALSE;
725    
726      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
727        {        {
728        negated = TRUE;        negated = TRUE;
729        ptr++;        ptr++;
# Line 563  else Line 732  else
732    
733      c = 0;      c = 0;
734      while ((digitab[ptr[1]] & ctype_digit) != 0)      while ((digitab[ptr[1]] & ctype_digit) != 0)
735        c = c * 10 + *(++ptr) - '0';        c = c * 10 + *(++ptr) - CHAR_0;
736    
737      if (c < 0)      if (c < 0)   /* Integer overflow */
738        {        {
739        *errorcodeptr = ERR61;        *errorcodeptr = ERR61;
740        break;        break;
741        }        }
742    
743      if (c == 0 || (braced && *(++ptr) != '}'))      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
744        {        {
745        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
746        break;        break;
747        }        }
748    
749        if (c == 0)
750          {
751          *errorcodeptr = ERR58;
752          break;
753          }
754    
755      if (negated)      if (negated)
756        {        {
757        if (c > bracount)        if (c > bracount)
# Line 602  else Line 777  else
777      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
778      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
779    
780      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:
781      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
782    
783      if (!isclass)      if (!isclass)
784        {        {
785        oldptr = ptr;        oldptr = ptr;
786        c -= '0';        c -= CHAR_0;
787        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
788          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - CHAR_0;
789        if (c < 0)        if (c < 0)    /* Integer overflow */
790          {          {
791          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
792          break;          break;
# Line 628  else Line 803  else
803      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.
804      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
805    
806      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
807        {        {
808        ptr--;        ptr--;
809        c = 0;        c = 0;
# Line 641  else Line 816  else
816      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
817      than 3 octal digits. */      than 3 octal digits. */
818    
819      case '0':      case CHAR_0:
820      c -= '0';      c -= CHAR_0;
821      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
822          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
823      if (!utf8 && c > 255) *errorcodeptr = ERR51;      if (!utf8 && c > 255) *errorcodeptr = ERR51;
824      break;      break;
825    
# Line 652  else Line 827  else
827      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
828      treated as a data character. */      treated as a data character. */
829    
830      case 'x':      case CHAR_x:
831      if (ptr[1] == '{')      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
832        {        {
833        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
834        int count = 0;        int count = 0;
# Line 662  else Line 837  else
837        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
838          {          {
839          register int cc = *pt++;          register int cc = *pt++;
840          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
841          count++;          count++;
842    
843  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
844          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
845          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
846  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
847          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
848          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
849  #endif  #endif
850          }          }
851    
852        if (*pt == '}')        if (*pt == CHAR_RIGHT_CURLY_BRACKET)
853          {          {
854          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
855          ptr = pt;          ptr = pt;
# Line 690  else Line 865  else
865      c = 0;      c = 0;
866      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
867        {        {
868        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
869        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
870  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
871        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
872        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
873  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
874        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
875        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
876  #endif  #endif
877        }        }
878      break;      break;
879    
880      /* 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.
881      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
882        coding is ASCII-specific, but then the whole concept of \cx is
883      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
884    
885      case 'c':      case CHAR_c:
886      c = *(++ptr);      c = *(++ptr);
887      if (c == 0)      if (c == 0)
888        {        {
889        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
890        break;        break;
891        }        }
892    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
893  #ifndef EBCDIC  /* ASCII coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
894      if (c >= 'a' && c <= 'z') c -= 32;        {
895          *errorcodeptr = ERR68;
896          break;
897          }
898        if (c >= CHAR_a && c <= CHAR_z) c -= 32;
899      c ^= 0x40;      c ^= 0x40;
900  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
901      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
902      c ^= 0xC0;      c ^= 0xC0;
903  #endif  #endif
904      break;      break;
# Line 740  else Line 920  else
920      }      }
921    }    }
922    
923    /* Perl supports \N{name} for character names, as well as plain \N for "not
924    newline". PCRE does not support \N{name}. However, it does support
925    quantification such as \N{2,3}. */
926    
927    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
928         !is_counted_repeat(ptr+2))
929      *errorcodeptr = ERR37;
930    
931    /* If PCRE_UCP is set, we change the values for \d etc. */
932    
933    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
934      c -= (ESC_DU - ESC_D);
935    
936    /* Set the pointer to the final character before returning. */
937    
938  *ptrptr = ptr;  *ptrptr = ptr;
939  return c;  return c;
940  }  }
# Line 780  if (c == 0) goto ERROR_RETURN; Line 975  if (c == 0) goto ERROR_RETURN;
975  /* \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
976  negation. */  negation. */
977    
978  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
979    {    {
980    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
981      {      {
982      *negptr = TRUE;      *negptr = TRUE;
983      ptr++;      ptr++;
# Line 791  if (c == '{') Line 986  if (c == '{')
986      {      {
987      c = *(++ptr);      c = *(++ptr);
988      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
989      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
990      name[i] = c;      name[i] = c;
991      }      }
992    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
993    name[i] = 0;    name[i] = 0;
994    }    }
995    
# Line 840  return -1; Line 1035  return -1;
1035    
1036    
1037  /*************************************************  /*************************************************
 *            Check for counted repeat            *  
 *************************************************/  
   
 /* This function is called when a '{' is encountered in a place where it might  
 start a quantifier. It looks ahead to see if it really is a quantifier or not.  
 It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}  
 where the ddds are digits.  
   
 Arguments:  
   p         pointer to the first char after '{'  
   
 Returns:    TRUE or FALSE  
 */  
   
 static BOOL  
 is_counted_repeat(const uschar *p)  
 {  
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
 if (*p == '}') return TRUE;  
   
 if (*p++ != ',') return FALSE;  
 if (*p == '}') return TRUE;  
   
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
   
 return (*p == '}');  
 }  
   
   
   
 /*************************************************  
1038  *         Read repeat counts                     *  *         Read repeat counts                     *
1039  *************************************************/  *************************************************/
1040    
# Line 900  int max = -1; Line 1062  int max = -1;
1062  /* 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
1063  an integer overflow. */  an integer overflow. */
1064    
1065  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;
1066  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1067    {    {
1068    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 910  if (min < 0 || min > 65535) Line 1072  if (min < 0 || min > 65535)
1072  /* 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.
1073  Also, max must not be less than min. */  Also, max must not be less than min. */
1074    
1075  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1076    {    {
1077    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1078      {      {
1079      max = 0;      max = 0;
1080      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;
1081      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1082        {        {
1083        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 940  return p; Line 1102  return p;
1102    
1103    
1104  /*************************************************  /*************************************************
1105  *       Find forward referenced subpattern       *  *  Subroutine for finding forward reference      *
1106  *************************************************/  *************************************************/
1107    
1108  /* This function scans along a pattern's text looking for capturing  /* This recursive function is called only from find_parens() below. The
1109    top-level call starts at the beginning of the pattern. All other calls must
1110    start at a parenthesis. It scans along a pattern's text looking for capturing
1111  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
1112  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
1113  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
1114  references to subpatterns. We know that if (?P< is encountered, the name will  track of subpatterns that reset the capturing group numbers - the (?| feature.
1115  be terminated by '>' because that is checked in the first pass.  
1116    This function was originally called only from the second pass, in which we know
1117    that if (?< or (?' or (?P< is encountered, the name will be correctly
1118    terminated because that is checked in the first pass. There is now one call to
1119    this function in the first pass, to check for a recursive back reference by
1120    name (so that we can make the whole group atomic). In this case, we need check
1121    only up to the current position in the pattern, and that is still OK because
1122    and previous occurrences will have been checked. To make this work, the test
1123    for "end of pattern" is a check against cd->end_pattern in the main loop,
1124    instead of looking for a binary zero. This means that the special first-pass
1125    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1126    processing items within the loop are OK, because afterwards the main loop will
1127    terminate.)
1128    
1129  Arguments:  Arguments:
1130    ptr          current position in the pattern    ptrptr       address of the current character pointer (updated)
1131    count        current count of capturing parens so far encountered    cd           compile background data
1132    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1133    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1134    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1135      utf8         TRUE if we are in UTF-8 mode
1136      count        pointer to the current capturing subpattern number (updated)
1137    
1138  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1139  */  */
1140    
1141  static int  static int
1142  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,
1143    BOOL xmode)    BOOL xmode, BOOL utf8, int *count)
1144  {  {
1145  const uschar *thisname;  uschar *ptr = *ptrptr;
1146    int start_count = *count;
1147    int hwm_count = start_count;
1148    BOOL dup_parens = FALSE;
1149    
1150    /* If the first character is a parenthesis, check on the type of group we are
1151    dealing with. The very first call may not start with a parenthesis. */
1152    
1153  for (; *ptr != 0; ptr++)  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1154    {    {
1155    int term;    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1156    
1157      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1158    
1159      /* Handle a normal, unnamed capturing parenthesis. */
1160    
1161      else if (ptr[1] != CHAR_QUESTION_MARK)
1162        {
1163        *count += 1;
1164        if (name == NULL && *count == lorn) return *count;
1165        ptr++;
1166        }
1167    
1168      /* All cases now have (? at the start. Remember when we are in a group
1169      where the parenthesis numbers are duplicated. */
1170    
1171      else if (ptr[2] == CHAR_VERTICAL_LINE)
1172        {
1173        ptr += 3;
1174        dup_parens = TRUE;
1175        }
1176    
1177      /* Handle comments; all characters are allowed until a ket is reached. */
1178    
1179      else if (ptr[2] == CHAR_NUMBER_SIGN)
1180        {
1181        for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1182        goto FAIL_EXIT;
1183        }
1184    
1185      /* Handle a condition. If it is an assertion, just carry on so that it
1186      is processed as normal. If not, skip to the closing parenthesis of the
1187      condition (there can't be any nested parens). */
1188    
1189      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1190        {
1191        ptr += 2;
1192        if (ptr[1] != CHAR_QUESTION_MARK)
1193          {
1194          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1195          if (*ptr != 0) ptr++;
1196          }
1197        }
1198    
1199      /* Start with (? but not a condition. */
1200    
1201      else
1202        {
1203        ptr += 2;
1204        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1205    
1206        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1207    
1208        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1209            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1210          {
1211          int term;
1212          const uschar *thisname;
1213          *count += 1;
1214          if (name == NULL && *count == lorn) return *count;
1215          term = *ptr++;
1216          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1217          thisname = ptr;
1218          while (*ptr != term) ptr++;
1219          if (name != NULL && lorn == ptr - thisname &&
1220              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1221            return *count;
1222          term++;
1223          }
1224        }
1225      }
1226    
1227    /* Past any initial parenthesis handling, scan for parentheses or vertical
1228    bars. Stop if we get to cd->end_pattern. Note that this is important for the
1229    first-pass call when this value is temporarily adjusted to stop at the current
1230    position. So DO NOT change this to a test for binary zero. */
1231    
1232    for (; ptr < cd->end_pattern; ptr++)
1233      {
1234    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1235    
1236    if (*ptr == '\\')    if (*ptr == CHAR_BACKSLASH)
1237      {      {
1238      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1239      if (*ptr == 'Q') for (;;)      if (*ptr == CHAR_Q) for (;;)
1240        {        {
1241        while (*(++ptr) != 0 && *ptr != '\\');        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1242        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1243        if (*(++ptr) == 'E') break;        if (*(++ptr) == CHAR_E) break;
1244        }        }
1245      continue;      continue;
1246      }      }
1247    
1248    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1249      are handled for real. If the first character is '^', skip it. Also, if the
1250      first few characters (either before or after ^) are \Q\E or \E we skip them
1251      too. This makes for compatibility with Perl. Note the use of STR macros to
1252      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1253    
1254    if (*ptr == '[')    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1255      {      {
1256      while (*(++ptr) != ']')      BOOL negate_class = FALSE;
1257        for (;;)
1258          {
1259          if (ptr[1] == CHAR_BACKSLASH)
1260            {
1261            if (ptr[2] == CHAR_E)
1262              ptr+= 2;
1263            else if (strncmp((const char *)ptr+2,
1264                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1265              ptr += 4;
1266            else
1267              break;
1268            }
1269          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1270            {
1271            negate_class = TRUE;
1272            ptr++;
1273            }
1274          else break;
1275          }
1276    
1277        /* If the next character is ']', it is a data character that must be
1278        skipped, except in JavaScript compatibility mode. */
1279    
1280        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1281            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1282          ptr++;
1283    
1284        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1285        {        {
1286        if (*ptr == 0) return -1;        if (*ptr == 0) return -1;
1287        if (*ptr == '\\')        if (*ptr == CHAR_BACKSLASH)
1288          {          {
1289          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) goto FAIL_EXIT;
1290          if (*ptr == 'Q') for (;;)          if (*ptr == CHAR_Q) for (;;)
1291            {            {
1292            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1293            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1294            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1295            }            }
1296          continue;          continue;
1297          }          }
# Line 1008  for (; *ptr != 0; ptr++) Line 1301  for (; *ptr != 0; ptr++)
1301    
1302    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1303    
1304    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1305      {      {
1306      while (*(++ptr) != 0 && *ptr != '\n');      ptr++;
1307      if (*ptr == 0) return -1;      while (*ptr != 0)
1308          {
1309          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1310          ptr++;
1311    #ifdef SUPPORT_UTF8
1312          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
1313    #endif
1314          }
1315        if (*ptr == 0) goto FAIL_EXIT;
1316      continue;      continue;
1317      }      }
1318    
1319    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1320    
1321    if (*ptr != '(') continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != '?' && ptr[1] != '*')  
1322      {      {
1323      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, count);
1324      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1325      continue;      if (*ptr == 0) goto FAIL_EXIT;
1326      }      }
1327    
1328    ptr += 2;    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1329    if (*ptr == 'P') ptr++;                      /* Allow optional P */      {
1330        if (dup_parens && *count < hwm_count) *count = hwm_count;
1331        goto FAIL_EXIT;
1332        }
1333    
1334    /* We have to disambiguate (?<! and (?<= from (?<name> */    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1335        {
1336        if (*count > hwm_count) hwm_count = *count;
1337        *count = start_count;
1338        }
1339      }
1340    
1341    if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  FAIL_EXIT:
1342         *ptr != '\'')  *ptrptr = ptr;
1343      continue;  return -1;
1344    }
1345    
1346    
1347    
1348    
1349    /*************************************************
1350    *       Find forward referenced subpattern       *
1351    *************************************************/
1352    
1353    /* This function scans along a pattern's text looking for capturing
1354    subpatterns, and counting them. If it finds a named pattern that matches the
1355    name it is given, it returns its number. Alternatively, if the name is NULL, it
1356    returns when it reaches a given numbered subpattern. This is used for forward
1357    references to subpatterns. We used to be able to start this scan from the
1358    current compiling point, using the current count value from cd->bracount, and
1359    do it all in a single loop, but the addition of the possibility of duplicate
1360    subpattern numbers means that we have to scan from the very start, in order to
1361    take account of such duplicates, and to use a recursive function to keep track
1362    of the different types of group.
1363    
1364    count++;  Arguments:
1365      cd           compile background data
1366      name         name to seek, or NULL if seeking a numbered subpattern
1367      lorn         name length, or subpattern number if name is NULL
1368      xmode        TRUE if we are in /x mode
1369      utf8         TRUE if we are in UTF-8 mode
1370    
1371    Returns:       the number of the found subpattern, or -1 if not found
1372    */
1373    
1374    if (name == NULL && count == lorn) return count;  static int
1375    term = *ptr++;  find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode,
1376    if (term == '<') term = '>';    BOOL utf8)
1377    thisname = ptr;  {
1378    while (*ptr != term) ptr++;  uschar *ptr = (uschar *)cd->start_pattern;
1379    if (name != NULL && lorn == ptr - thisname &&  int count = 0;
1380        strncmp((const char *)name, (const char *)thisname, lorn) == 0)  int rc;
1381      return count;  
1382    /* If the pattern does not start with an opening parenthesis, the first call
1383    to find_parens_sub() will scan right to the end (if necessary). However, if it
1384    does start with a parenthesis, find_parens_sub() will return when it hits the
1385    matching closing parens. That is why we have to have a loop. */
1386    
1387    for (;;)
1388      {
1389      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, &count);
1390      if (rc > 0 || *ptr++ == 0) break;
1391    }    }
1392    
1393  return -1;  return rc;
1394  }  }
1395    
1396    
1397    
1398    
1399  /*************************************************  /*************************************************
1400  *      Find first significant op code            *  *      Find first significant op code            *
1401  *************************************************/  *************************************************/
1402    
1403  /* This is called by several functions that scan a compiled expression looking  /* This is called by several functions that scan a compiled expression looking
1404  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
1405  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
1406  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
1407  assertions, and also the \b assertion; for others it does not.  does not.
1408    
1409  Arguments:  Arguments:
1410    code         pointer to the start of the group    code         pointer to the start of the group
   options      pointer to external options  
   optbit       the option bit whose changing is significant, or  
                  zero if none are  
1411    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1412    
1413  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1414  */  */
1415    
1416  static const uschar*  static const uschar*
1417  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const uschar *code, BOOL skipassert)
   BOOL skipassert)  
1418  {  {
1419  for (;;)  for (;;)
1420    {    {
1421    switch ((int)*code)    switch ((int)*code)
1422      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1423      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1424      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1425      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
# Line 1100  for (;;) Line 1435  for (;;)
1435    
1436      case OP_CALLOUT:      case OP_CALLOUT:
1437      case OP_CREF:      case OP_CREF:
1438        case OP_NCREF:
1439      case OP_RREF:      case OP_RREF:
1440        case OP_NRREF:
1441      case OP_DEF:      case OP_DEF:
1442      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1443      break;      break;
# Line 1116  for (;;) Line 1453  for (;;)
1453    
1454    
1455  /*************************************************  /*************************************************
1456  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1457  *************************************************/  *************************************************/
1458    
1459  /* 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,
1460  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.
1461  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
1462    temporarily terminated with OP_END when this function is called.
1463    
1464    This function is called when a backward assertion is encountered, so that if it
1465    fails, the error message can point to the correct place in the pattern.
1466    However, we cannot do this when the assertion contains subroutine calls,
1467    because they can be forward references. We solve this by remembering this case
1468    and doing the check at the end; a flag specifies which mode we are running in.
1469    
1470  Arguments:  Arguments:
1471    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1472    options  the compiling options    utf8     TRUE in UTF-8 mode
1473      atend    TRUE if called when the pattern is complete
1474      cd       the "compile data" structure
1475    
1476  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1477                 or -1 if there is no fixed length,
1478               or -2 if \C was encountered               or -2 if \C was encountered
1479                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1480  */  */
1481    
1482  static int  static int
1483  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, BOOL utf8, BOOL atend, compile_data *cd)
1484  {  {
1485  int length = -1;  int length = -1;
1486    
# Line 1145  branch, check the length against that of Line 1493  branch, check the length against that of
1493  for (;;)  for (;;)
1494    {    {
1495    int d;    int d;
1496      uschar *ce, *cs;
1497    register int op = *cc;    register int op = *cc;
1498    switch (op)    switch (op)
1499      {      {
1500        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1501        OP_BRA (normal non-capturing bracket) because the other variants of these
1502        opcodes are all concerned with unlimited repeated groups, which of course
1503        are not of fixed length. They will cause a -1 response from the default
1504        case of this switch. */
1505    
1506      case OP_CBRA:      case OP_CBRA:
1507      case OP_BRA:      case OP_BRA:
1508      case OP_ONCE:      case OP_ONCE:
1509      case OP_COND:      case OP_COND:
1510      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), utf8, atend, cd);
1511      if (d < 0) return d;      if (d < 0) return d;
1512      branchlength += d;      branchlength += d;
1513      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1161  for (;;) Line 1516  for (;;)
1516    
1517      /* Reached end of a branch; if it's a ket it is the end of a nested      /* Reached end of a branch; if it's a ket it is the end of a nested
1518      call. If it's ALT it is an alternation in a nested call. If it is      call. If it's ALT it is an alternation in a nested call. If it is
1519      END it's the end of the outer call. All can be handled by the same code. */      END it's the end of the outer call. All can be handled by the same code.
1520        Note that we must not include the OP_KETRxxx opcodes here, because they
1521        all imply an unlimited repeat. */
1522    
1523      case OP_ALT:      case OP_ALT:
1524      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1525      case OP_END:      case OP_END:
1526      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1527        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
# Line 1175  for (;;) Line 1530  for (;;)
1530      branchlength = 0;      branchlength = 0;
1531      break;      break;
1532    
1533        /* A true recursion implies not fixed length, but a subroutine call may
1534        be OK. If the subroutine is a forward reference, we can't deal with
1535        it until the end of the pattern, so return -3. */
1536    
1537        case OP_RECURSE:
1538        if (!atend) return -3;
1539        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1540        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1541        if (cc > cs && cc < ce) return -1;                /* Recursion */
1542        d = find_fixedlength(cs + 2, utf8, atend, cd);
1543        if (d < 0) return d;
1544        branchlength += d;
1545        cc += 1 + LINK_SIZE;
1546        break;
1547    
1548      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1549    
1550      case OP_ASSERT:      case OP_ASSERT:
# Line 1188  for (;;) Line 1558  for (;;)
1558    
1559      case OP_REVERSE:      case OP_REVERSE:
1560      case OP_CREF:      case OP_CREF:
1561        case OP_NCREF:
1562      case OP_RREF:      case OP_RREF:
1563        case OP_NRREF:
1564      case OP_DEF:      case OP_DEF:
     case OP_OPT:  
1565      case OP_CALLOUT:      case OP_CALLOUT:
1566      case OP_SOD:      case OP_SOD:
1567      case OP_SOM:      case OP_SOM:
1568        case OP_SET_SOM:
1569      case OP_EOD:      case OP_EOD:
1570      case OP_EODN:      case OP_EODN:
1571      case OP_CIRC:      case OP_CIRC:
1572        case OP_CIRCM:
1573      case OP_DOLL:      case OP_DOLL:
1574        case OP_DOLLM:
1575      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1576      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1577      cc += _pcre_OP_lengths[*cc];      cc += _pcre_OP_lengths[*cc];
# Line 1206  for (;;) Line 1580  for (;;)
1580      /* Handle literal characters */      /* Handle literal characters */
1581    
1582      case OP_CHAR:      case OP_CHAR:
1583      case OP_CHARNC:      case OP_CHARI:
1584      case OP_NOT:      case OP_NOT:
1585        case OP_NOTI:
1586      branchlength++;      branchlength++;
1587      cc += 2;      cc += 2;
1588  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1589      if ((options & PCRE_UTF8) != 0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       {  
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1590  #endif  #endif
1591      break;      break;
1592    
# Line 1225  for (;;) Line 1597  for (;;)
1597      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1598      cc += 4;      cc += 4;
1599  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1600      if ((options & PCRE_UTF8) != 0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       {  
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1601  #endif  #endif
1602      break;      break;
1603    
# Line 1252  for (;;) Line 1621  for (;;)
1621      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1622      case OP_WORDCHAR:      case OP_WORDCHAR:
1623      case OP_ANY:      case OP_ANY:
1624        case OP_ALLANY:
1625      branchlength++;      branchlength++;
1626      cc++;      cc++;
1627      break;      break;
# Line 1306  for (;;) Line 1676  for (;;)
1676    
1677    
1678  /*************************************************  /*************************************************
1679  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1680  *************************************************/  *************************************************/
1681    
1682  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1683  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1684    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1685    so that it can be called from pcre_study() when finding the minimum matching
1686    length.
1687    
1688  Arguments:  Arguments:
1689    code        points to start of expression    code        points to start of expression
1690    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1691    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1692    
1693  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
1694  */  */
1695    
1696  static const uschar *  const uschar *
1697  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1698  {  {
1699  for (;;)  for (;;)
1700    {    {
1701    register int c = *code;    register int c = *code;
1702    
1703    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1704    
1705    /* XCLASS is used for classes that cannot be represented just by a bit    /* XCLASS is used for classes that cannot be represented just by a bit
# Line 1334  for (;;) Line 1708  for (;;)
1708    
1709    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1710    
1711      /* Handle recursion */
1712    
1713      else if (c == OP_REVERSE)
1714        {
1715        if (number < 0) return (uschar *)code;
1716        code += _pcre_OP_lengths[c];
1717        }
1718    
1719    /* Handle capturing bracket */    /* Handle capturing bracket */
1720    
1721    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
1722               c == OP_CBRAPOS || c == OP_SCBRAPOS)
1723      {      {
1724      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
1725      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
# Line 1345  for (;;) Line 1728  for (;;)
1728    
1729    /* Otherwise, we can get the item's length from the table, except that for    /* Otherwise, we can get the item's length from the table, except that for
1730    repeated character types, we have to test for \p and \P, which have an extra    repeated character types, we have to test for \p and \P, which have an extra
1731    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1732      must add in its length. */
1733    
1734    else    else
1735      {      {
# Line 1369  for (;;) Line 1753  for (;;)
1753        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
1754        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1755        break;        break;
1756    
1757          case OP_MARK:
1758          case OP_PRUNE_ARG:
1759          case OP_SKIP_ARG:
1760          code += code[1];
1761          break;
1762    
1763          case OP_THEN_ARG:
1764          code += code[1+LINK_SIZE];
1765          break;
1766        }        }
1767    
1768      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1383  for (;;) Line 1777  for (;;)
1777      if (utf8) switch(c)      if (utf8) switch(c)
1778        {        {
1779        case OP_CHAR:        case OP_CHAR:
1780        case OP_CHARNC:        case OP_CHARI:
1781        case OP_EXACT:        case OP_EXACT:
1782          case OP_EXACTI:
1783        case OP_UPTO:        case OP_UPTO:
1784          case OP_UPTOI:
1785        case OP_MINUPTO:        case OP_MINUPTO:
1786          case OP_MINUPTOI:
1787        case OP_POSUPTO:        case OP_POSUPTO:
1788          case OP_POSUPTOI:
1789        case OP_STAR:        case OP_STAR:
1790          case OP_STARI:
1791        case OP_MINSTAR:        case OP_MINSTAR:
1792          case OP_MINSTARI:
1793        case OP_POSSTAR:        case OP_POSSTAR:
1794          case OP_POSSTARI:
1795        case OP_PLUS:        case OP_PLUS:
1796          case OP_PLUSI:
1797        case OP_MINPLUS:        case OP_MINPLUS:
1798          case OP_MINPLUSI:
1799        case OP_POSPLUS:        case OP_POSPLUS:
1800          case OP_POSPLUSI:
1801        case OP_QUERY:        case OP_QUERY:
1802          case OP_QUERYI:
1803        case OP_MINQUERY:        case OP_MINQUERY:
1804          case OP_MINQUERYI:
1805        case OP_POSQUERY:        case OP_POSQUERY:
1806          case OP_POSQUERYI:
1807        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1808        break;        break;
1809        }        }
1810    #else
1811        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1812  #endif  #endif
1813      }      }
1814    }    }
# Line 1438  for (;;) Line 1847  for (;;)
1847    
1848    /* Otherwise, we can get the item's length from the table, except that for    /* Otherwise, we can get the item's length from the table, except that for
1849    repeated character types, we have to test for \p and \P, which have an extra    repeated character types, we have to test for \p and \P, which have an extra
1850    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1851      must add in its length. */
1852    
1853    else    else
1854      {      {
# Line 1462  for (;;) Line 1872  for (;;)
1872        case OP_TYPEEXACT:        case OP_TYPEEXACT:
1873        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1874        break;        break;
1875    
1876          case OP_MARK:
1877          case OP_PRUNE_ARG:
1878          case OP_SKIP_ARG:
1879          code += code[1];
1880          break;
1881    
1882          case OP_THEN_ARG:
1883          code += code[1+LINK_SIZE];
1884          break;
1885        }        }
1886    
1887      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1476  for (;;) Line 1896  for (;;)
1896      if (utf8) switch(c)      if (utf8) switch(c)
1897        {        {
1898        case OP_CHAR:        case OP_CHAR:
1899        case OP_CHARNC:        case OP_CHARI:
1900        case OP_EXACT:        case OP_EXACT:
1901          case OP_EXACTI:
1902        case OP_UPTO:        case OP_UPTO:
1903          case OP_UPTOI:
1904        case OP_MINUPTO:        case OP_MINUPTO:
1905          case OP_MINUPTOI:
1906        case OP_POSUPTO:        case OP_POSUPTO:
1907          case OP_POSUPTOI:
1908        case OP_STAR:        case OP_STAR:
1909          case OP_STARI:
1910        case OP_MINSTAR:        case OP_MINSTAR:
1911          case OP_MINSTARI:
1912        case OP_POSSTAR:        case OP_POSSTAR:
1913          case OP_POSSTARI:
1914        case OP_PLUS:        case OP_PLUS:
1915          case OP_PLUSI:
1916        case OP_MINPLUS:        case OP_MINPLUS:
1917          case OP_MINPLUSI:
1918        case OP_POSPLUS:        case OP_POSPLUS:
1919          case OP_POSPLUSI:
1920        case OP_QUERY:        case OP_QUERY:
1921          case OP_QUERYI:
1922        case OP_MINQUERY:        case OP_MINQUERY:
1923          case OP_MINQUERYI:
1924        case OP_POSQUERY:        case OP_POSQUERY:
1925          case OP_POSQUERYI:
1926        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1927        break;        break;
1928        }        }
1929    #else
1930        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1931  #endif  #endif
1932      }      }
1933    }    }
# Line 1516  Arguments: Line 1951  Arguments:
1951    code        points to start of search    code        points to start of search
1952    endcode     points to where to stop    endcode     points to where to stop
1953    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
1954      cd          contains pointers to tables etc.
1955    
1956  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
1957  */  */
1958    
1959  static BOOL  static BOOL
1960  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
1961      compile_data *cd)
1962  {  {
1963  register int c;  register int c;
1964  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], TRUE);
1965       code < endcode;       code < endcode;
1966       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], TRUE))
1967    {    {
1968    const uschar *ccode;    const uschar *ccode;
1969    
# Line 1542  for (code = first_significant_code(code Line 1979  for (code = first_significant_code(code
1979      continue;      continue;
1980      }      }
1981    
1982      /* For a recursion/subroutine call, if its end has been reached, which
1983      implies a backward reference subroutine call, we can scan it. If it's a
1984      forward reference subroutine call, we can't. To detect forward reference
1985      we have to scan up the list that is kept in the workspace. This function is
1986      called only when doing the real compile, not during the pre-compile that
1987      measures the size of the compiled pattern. */
1988    
1989      if (c == OP_RECURSE)
1990        {
1991        const uschar *scode;
1992        BOOL empty_branch;
1993    
1994        /* Test for forward reference */
1995    
1996        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
1997          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
1998    
1999        /* Not a forward reference, test for completed backward reference */
2000    
2001        empty_branch = FALSE;
2002        scode = cd->start_code + GET(code, 1);
2003        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2004    
2005        /* Completed backwards reference */
2006    
2007        do
2008          {
2009          if (could_be_empty_branch(scode, endcode, utf8, cd))
2010            {
2011            empty_branch = TRUE;
2012            break;
2013            }
2014          scode += GET(scode, 1);
2015          }
2016        while (*scode == OP_ALT);
2017    
2018        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2019        continue;
2020        }
2021    
2022    /* Groups with zero repeats can of course be empty; skip them. */    /* Groups with zero repeats can of course be empty; skip them. */
2023    
2024    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2025          c == OP_BRAPOSZERO)
2026      {      {
2027      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
2028      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
# Line 1552  for (code = first_significant_code(code Line 2030  for (code = first_significant_code(code
2030      continue;      continue;
2031      }      }
2032    
2033      /* A nested group that is already marked as "could be empty" can just be
2034      skipped. */
2035    
2036      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2037          c == OP_SCBRA || c == OP_SCBRAPOS)
2038        {
2039        do code += GET(code, 1); while (*code == OP_ALT);
2040        c = *code;
2041        continue;
2042        }
2043    
2044    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
2045    
2046    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_BRA  || c == OP_BRAPOS ||
2047          c == OP_CBRA || c == OP_CBRAPOS ||
2048          c == OP_ONCE || c == OP_COND)
2049      {      {
2050      BOOL empty_branch;      BOOL empty_branch;
2051      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
2052    
2053      /* Scan a closed bracket */      /* If a conditional group has only one branch, there is a second, implied,
2054        empty branch, so just skip over the conditional, because it could be empty.
2055        Otherwise, scan the individual branches of the group. */
2056    
2057      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;  
2058        code += GET(code, 1);        code += GET(code, 1);
2059        else
2060          {
2061          empty_branch = FALSE;
2062          do
2063            {
2064            if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
2065              empty_branch = TRUE;
2066            code += GET(code, 1);
2067            }
2068          while (*code == OP_ALT);
2069          if (!empty_branch) return FALSE;   /* All branches are non-empty */
2070        }        }
2071      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
2072      c = *code;      c = *code;
2073      continue;      continue;
2074      }      }
# Line 1630  for (code = first_significant_code(code Line 2129  for (code = first_significant_code(code
2129      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2130      case OP_WORDCHAR:      case OP_WORDCHAR:
2131      case OP_ANY:      case OP_ANY:
2132        case OP_ALLANY:
2133      case OP_ANYBYTE:      case OP_ANYBYTE:
2134      case OP_CHAR:      case OP_CHAR:
2135      case OP_CHARNC:      case OP_CHARI:
2136      case OP_NOT:      case OP_NOT:
2137        case OP_NOTI:
2138      case OP_PLUS:      case OP_PLUS:
2139      case OP_MINPLUS:      case OP_MINPLUS:
2140      case OP_POSPLUS:      case OP_POSPLUS:
# Line 1673  for (code = first_significant_code(code Line 2174  for (code = first_significant_code(code
2174      case OP_KET:      case OP_KET:
2175      case OP_KETRMAX:      case OP_KETRMAX:
2176      case OP_KETRMIN:      case OP_KETRMIN:
2177        case OP_KETRPOS:
2178      case OP_ALT:      case OP_ALT:
2179      return TRUE;      return TRUE;
2180    
# Line 1681  for (code = first_significant_code(code Line 2183  for (code = first_significant_code(code
2183    
2184  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2185      case OP_STAR:      case OP_STAR:
2186        case OP_STARI:
2187      case OP_MINSTAR:      case OP_MINSTAR:
2188        case OP_MINSTARI:
2189      case OP_POSSTAR:      case OP_POSSTAR:
2190        case OP_POSSTARI:
2191      case OP_QUERY:      case OP_QUERY:
2192        case OP_QUERYI:
2193      case OP_MINQUERY:      case OP_MINQUERY:
2194        case OP_MINQUERYI:
2195      case OP_POSQUERY:      case OP_POSQUERY:
2196        case OP_POSQUERYI:
2197        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2198        break;
2199    
2200      case OP_UPTO:      case OP_UPTO:
2201        case OP_UPTOI:
2202      case OP_MINUPTO:      case OP_MINUPTO:
2203        case OP_MINUPTOI:
2204      case OP_POSUPTO:      case OP_POSUPTO:
2205      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_POSUPTOI:
2206        if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2207      break;      break;
2208  #endif  #endif
2209    
2210        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2211        string. */
2212    
2213        case OP_MARK:
2214        case OP_PRUNE_ARG:
2215        case OP_SKIP_ARG:
2216        code += code[1];
2217        break;
2218    
2219        case OP_THEN_ARG:
2220        code += code[1+LINK_SIZE];
2221        break;
2222    
2223        /* None of the remaining opcodes are required to match a character. */
2224    
2225        default:
2226        break;
2227      }      }
2228    }    }
2229    
# Line 1708  return TRUE; Line 2240  return TRUE;
2240  the current branch of the current pattern to see if it could match the empty  the current branch of the current pattern to see if it could match the empty
2241  string. If it could, we must look outwards for branches at other levels,  string. If it could, we must look outwards for branches at other levels,
2242  stopping when we pass beyond the bracket which is the subject of the recursion.  stopping when we pass beyond the bracket which is the subject of the recursion.
2243    This function is called only during the real compile, not during the
2244    pre-compile.
2245    
2246  Arguments:  Arguments:
2247    code        points to start of the recursion    code        points to start of the recursion
2248    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2249    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2250    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2251      cd          pointers to tables etc
2252    
2253  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2254  */  */
2255    
2256  static BOOL  static BOOL
2257  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2258    BOOL utf8)    BOOL utf8, compile_data *cd)
2259  {  {
2260  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2261    {    {
2262    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2263        return FALSE;
2264    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2265    }    }
2266  return TRUE;  return TRUE;
# Line 1756  where Perl recognizes it as the POSIX cl Line 2292  where Perl recognizes it as the POSIX cl
2292  "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,  "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2293  I think.  I think.
2294    
2295    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2296    It seems that the appearance of a nested POSIX class supersedes an apparent
2297    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2298    a digit. Also, unescaped square brackets may also appear as part of class
2299    names. For example, [:a[:abc]b:] gives unknown class "[:abc]b:]"in Perl.
2300    
2301  Arguments:  Arguments:
2302    ptr      pointer to the initial [    ptr      pointer to the initial [
2303    endptr   where to return the end pointer    endptr   where to return the end pointer
# Line 1770  int terminator; /* Don't combin Line 2312  int terminator; /* Don't combin
2312  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2313  for (++ptr; *ptr != 0; ptr++)  for (++ptr; *ptr != 0; ptr++)
2314    {    {
2315    if (*ptr == '\\' && ptr[1] == ']') ptr++; else    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2316        ptr++;
2317      else
2318      {      {
2319      if (*ptr == ']') return FALSE;      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
     if (*ptr == terminator && ptr[1] == ']')  
2320        {        {
2321        *endptr = ptr;        *endptr = ptr;
2322        return TRUE;        return TRUE;
2323        }        }
2324        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2325             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2326              ptr[1] == CHAR_EQUALS_SIGN) &&
2327            check_posix_syntax(ptr, endptr))
2328          return FALSE;
2329      }      }
2330    }    }
2331  return FALSE;  return FALSE;
# Line 1824  return -1; Line 2372  return -1;
2372  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2373  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2374  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
2375  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
2376  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
2377  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
2378  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
2379  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2380    OP_END.
2381    
2382  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2383  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 1905  auto_callout(uschar *code, const uschar Line 2454  auto_callout(uschar *code, const uschar
2454  {  {
2455  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2456  *code++ = 255;  *code++ = 255;
2457  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2458  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2459  return code + 2*LINK_SIZE;  return code + 2*LINK_SIZE;
2460  }  }
2461    
# Line 1931  Returns: nothing Line 2480  Returns: nothing
2480  static void  static void
2481  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)
2482  {  {
2483  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2484  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2485  }  }
2486    
# Line 1963  get_othercase_range(unsigned int *cptr, Line 2512  get_othercase_range(unsigned int *cptr,
2512  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2513    
2514  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2515    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2516    
2517  if (c > d) return FALSE;  if (c > d) return FALSE;
2518    
# Line 1972  next = othercase + 1; Line 2521  next = othercase + 1;
2521    
2522  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2523    {    {
2524    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2525    next++;    next++;
2526    }    }
2527    
# Line 1981  for (++c; c <= d; c++) Line 2530  for (++c; c <= d; c++)
2530    
2531  return TRUE;  return TRUE;
2532  }  }
2533    
2534    
2535    
2536    /*************************************************
2537    *        Check a character and a property        *
2538    *************************************************/
2539    
2540    /* This function is called by check_auto_possessive() when a property item
2541    is adjacent to a fixed character.
2542    
2543    Arguments:
2544      c            the character
2545      ptype        the property type
2546      pdata        the data for the type
2547      negated      TRUE if it's a negated property (\P or \p{^)
2548    
2549    Returns:       TRUE if auto-possessifying is OK
2550    */
2551    
2552    static BOOL
2553    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2554    {
2555    const ucd_record *prop = GET_UCD(c);
2556    switch(ptype)
2557      {
2558      case PT_LAMP:
2559      return (prop->chartype == ucp_Lu ||
2560              prop->chartype == ucp_Ll ||
2561              prop->chartype == ucp_Lt) == negated;
2562    
2563      case PT_GC:
2564      return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;
2565    
2566      case PT_PC:
2567      return (pdata == prop->chartype) == negated;
2568    
2569      case PT_SC:
2570      return (pdata == prop->script) == negated;
2571    
2572      /* These are specials */
2573    
2574      case PT_ALNUM:
2575      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2576              _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;
2577    
2578      case PT_SPACE:    /* Perl space */
2579      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2580              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2581              == negated;
2582    
2583      case PT_PXSPACE:  /* POSIX space */
2584      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2585              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2586              c == CHAR_FF || c == CHAR_CR)
2587              == negated;
2588    
2589      case PT_WORD:
2590      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2591              _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2592              c == CHAR_UNDERSCORE) == negated;
2593      }
2594    return FALSE;
2595    }
2596  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2597    
2598    
# Line 1994  whether the next thing could possibly ma Line 2606  whether the next thing could possibly ma
2606  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
2607    
2608  Arguments:  Arguments:
2609    op_code       the repeated op code    previous      pointer to the repeated opcode
   this          data for this item, depends on the opcode  
2610    utf8          TRUE in UTF-8 mode    utf8          TRUE in UTF-8 mode
   utf8_char     used for utf8 character bytes, NULL if not relevant  
2611    ptr           next character in pattern    ptr           next character in pattern
2612    options       options bits    options       options bits
2613    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 2006  Returns: TRUE if possessifying is Line 2616  Returns: TRUE if possessifying is
2616  */  */
2617    
2618  static BOOL  static BOOL
2619  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
2620    const uschar *ptr, int options, compile_data *cd)    int options, compile_data *cd)
2621  {  {
2622  int next;  int c, next;
2623    int op_code = *previous++;
2624    
2625  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
2626    
# Line 2018  if ((options & PCRE_EXTENDED) != 0) Line 2629  if ((options & PCRE_EXTENDED) != 0)
2629    for (;;)    for (;;)
2630      {      {
2631      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2632      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2633        {        {
2634        while (*(++ptr) != 0)        ptr++;
2635          while (*ptr != 0)
2636            {
2637          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2638            ptr++;
2639    #ifdef SUPPORT_UTF8
2640            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2641    #endif
2642            }
2643        }        }
2644      else break;      else break;
2645      }      }
# Line 2030  if ((options & PCRE_EXTENDED) != 0) Line 2648  if ((options & PCRE_EXTENDED) != 0)
2648  /* 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
2649  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2650    
2651  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2652    {    {
2653    int temperrorcode = 0;    int temperrorcode = 0;
2654    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
# Line 2055  if ((options & PCRE_EXTENDED) != 0) Line 2673  if ((options & PCRE_EXTENDED) != 0)
2673    for (;;)    for (;;)
2674      {      {
2675      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2676      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2677        {        {
2678        while (*(++ptr) != 0)        ptr++;
2679          while (*ptr != 0)
2680            {
2681          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2682            ptr++;
2683    #ifdef SUPPORT_UTF8
2684            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2685    #endif
2686            }
2687        }        }
2688      else break;      else break;
2689      }      }
# Line 2066  if ((options & PCRE_EXTENDED) != 0) Line 2691  if ((options & PCRE_EXTENDED) != 0)
2691    
2692  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
2693    
2694  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2695    return FALSE;    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2696        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. */  
   
2697    
2698  /* Handle cases when the next item is a character. */  /* Now compare the next item with the previous opcode. First, handle cases when
2699    the next item is a character. */
2700    
2701  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
2702    {    {
2703    case OP_CHAR:    case OP_CHAR:
2704  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2705    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2706    #else
2707      c = *previous;
2708  #endif  #endif
2709    return item != next;    return c != next;
2710    
2711    /* 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
2712    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
2713    high-valued characters. */    high-valued characters. */
2714    
2715    case OP_CHARNC:    case OP_CHARI:
2716  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2717    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2718    #else
2719      c = *previous;
2720  #endif  #endif
2721    if (item == next) return FALSE;    if (c == next) return FALSE;
2722  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2723    if (utf8)    if (utf8)
2724      {      {
2725      unsigned int othercase;      unsigned int othercase;
2726      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2727  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2728      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
2729  #else  #else
2730      othercase = NOTACHAR;      othercase = NOTACHAR;
2731  #endif  #endif
2732      return (unsigned int)item != othercase;      return (unsigned int)c != othercase;
2733      }      }
2734    else    else
2735  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2736    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2737    
2738    /* 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
2739      opcodes are not used for multi-byte characters, because they are coded using
2740      an XCLASS instead. */
2741    
2742    case OP_NOT:    case OP_NOT:
2743    if (item == next) return TRUE;    return (c = *previous) == next;
2744    if ((options & PCRE_CASELESS) == 0) return FALSE;  
2745      case OP_NOTI:
2746      if ((c = *previous) == next) return TRUE;
2747  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2748    if (utf8)    if (utf8)
2749      {      {
2750      unsigned int othercase;      unsigned int othercase;
2751      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2752  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2753      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
2754  #else  #else
2755      othercase = NOTACHAR;      othercase = NOTACHAR;
2756  #endif  #endif
2757      return (unsigned int)item == othercase;      return (unsigned int)c == othercase;
2758      }      }
2759    else    else
2760  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2761    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2762    
2763      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2764      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
2765    
2766    case OP_DIGIT:    case OP_DIGIT:
2767    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
# Line 2172  if (next >= 0) switch(op_code) Line 2804  if (next >= 0) switch(op_code)
2804      case 0x202f:      case 0x202f:
2805      case 0x205f:      case 0x205f:
2806      case 0x3000:      case 0x3000:
2807      return op_code != OP_HSPACE;      return op_code == OP_NOT_HSPACE;
2808      default:      default:
2809      return op_code == OP_HSPACE;      return op_code != OP_NOT_HSPACE;
2810      }      }
2811    
2812      case OP_ANYNL:
2813    case OP_VSPACE:    case OP_VSPACE:
2814    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
2815    switch(next)    switch(next)
# Line 2188  if (next >= 0) switch(op_code) Line 2821  if (next >= 0) switch(op_code)
2821      case 0x85:      case 0x85:
2822      case 0x2028:      case 0x2028:
2823      case 0x2029:      case 0x2029:
2824      return op_code != OP_VSPACE;      return op_code == OP_NOT_VSPACE;
2825      default:      default:
2826      return op_code == OP_VSPACE;      return op_code != OP_NOT_VSPACE;
2827      }      }
2828    
2829    #ifdef SUPPORT_UCP
2830      case OP_PROP:
2831      return check_char_prop(next, previous[0], previous[1], FALSE);
2832    
2833      case OP_NOTPROP:
2834      return check_char_prop(next, previous[0], previous[1], TRUE);
2835    #endif
2836    
2837    default:    default:
2838    return FALSE;    return FALSE;
2839    }    }
2840    
2841    
2842  /* 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
2843    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
2844    generated only when PCRE_UCP is *not* set, that is, when only ASCII
2845    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
2846    replaced by OP_PROP codes when PCRE_UCP is set. */
2847    
2848  switch(op_code)  switch(op_code)
2849    {    {
2850    case OP_CHAR:    case OP_CHAR:
2851    case OP_CHARNC:    case OP_CHARI:
2852  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2853    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2854    #else
2855      c = *previous;
2856  #endif  #endif
2857    switch(-next)    switch(-next)
2858      {      {
2859      case ESC_d:      case ESC_d:
2860      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
2861    
2862      case ESC_D:      case ESC_D:
2863      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
2864    
2865      case ESC_s:      case ESC_s:
2866      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
2867    
2868      case ESC_S:      case ESC_S:
2869      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
2870    
2871      case ESC_w:      case ESC_w:
2872      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
2873    
2874      case ESC_W:      case ESC_W:
2875      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
2876    
2877      case ESC_h:      case ESC_h:
2878      case ESC_H:      case ESC_H:
2879      switch(item)      switch(c)
2880        {        {
2881        case 0x09:        case 0x09:
2882        case 0x20:        case 0x20:
# Line 2257  switch(op_code) Line 2904  switch(op_code)
2904    
2905      case ESC_v:      case ESC_v:
2906      case ESC_V:      case ESC_V:
2907      switch(item)      switch(c)
2908        {        {
2909        case 0x0a:        case 0x0a:
2910        case 0x0b:        case 0x0b:
# Line 2271  switch(op_code) Line 2918  switch(op_code)
2918        return -next == ESC_v;        return -next == ESC_v;
2919        }        }
2920    
2921        /* When PCRE_UCP is set, these values get generated for \d etc. Find
2922        their substitutions and process them. The result will always be either
2923        -ESC_p or -ESC_P. Then fall through to process those values. */
2924    
2925    #ifdef SUPPORT_UCP
2926        case ESC_du:
2927        case ESC_DU:
2928        case ESC_wu:
2929        case ESC_WU:
2930        case ESC_su:
2931        case ESC_SU:
2932          {
2933          int temperrorcode = 0;
2934          ptr = substitutes[-next - ESC_DU];
2935          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
2936          if (temperrorcode != 0) return FALSE;
2937          ptr++;    /* For compatibility */
2938          }
2939        /* Fall through */
2940    
2941        case ESC_p:
2942        case ESC_P:
2943          {
2944          int ptype, pdata, errorcodeptr;
2945          BOOL negated;
2946    
2947          ptr--;      /* Make ptr point at the p or P */
2948          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
2949          if (ptype < 0) return FALSE;
2950          ptr++;      /* Point past the final curly ket */
2951    
2952          /* If the property item is optional, we have to give up. (When generated
2953          from \d etc by PCRE_UCP, this test will have been applied much earlier,
2954          to the original \d etc. At this point, ptr will point to a zero byte. */
2955    
2956          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2957            strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2958              return FALSE;
2959    
2960          /* Do the property check. */
2961    
2962          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
2963          }
2964    #endif
2965    
2966      default:      default:
2967      return FALSE;      return FALSE;
2968      }      }
2969    
2970      /* In principle, support for Unicode properties should be integrated here as
2971      well. It means re-organizing the above code so as to get hold of the property
2972      values before switching on the op-code. However, I wonder how many patterns
2973      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
2974      these op-codes are never generated.) */
2975    
2976    case OP_DIGIT:    case OP_DIGIT:
2977    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2978           next == -ESC_h || next == -ESC_v;           next == -ESC_h || next == -ESC_v || next == -ESC_R;
2979    
2980    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
2981    return next == -ESC_d;    return next == -ESC_d;
2982    
2983    case OP_WHITESPACE:    case OP_WHITESPACE:
2984    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
2985    
2986    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
2987    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2988    
2989    case OP_HSPACE:    case OP_HSPACE:
2990    return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
2991             next == -ESC_w || next == -ESC_v || next == -ESC_R;
2992    
2993    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
2994    return next == -ESC_h;    return next == -ESC_h;
2995    
2996    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
2997      case OP_ANYNL:
2998    case OP_VSPACE:    case OP_VSPACE:
2999    return next == -ESC_V || next == -ESC_d || next == -ESC_w;    return next == -ESC_V || next == -ESC_d || next == -ESC_w;
3000    
3001    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
3002    return next == -ESC_v;    return next == -ESC_v || next == -ESC_R;
3003    
3004    case OP_WORDCHAR:    case OP_WORDCHAR:
3005    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
3006             next == -ESC_v || next == -ESC_R;
3007    
3008    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
3009    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2354  int greedy_default, greedy_non_default; Line 3055  int greedy_default, greedy_non_default;
3055  int firstbyte, reqbyte;  int firstbyte, reqbyte;
3056  int zeroreqbyte, zerofirstbyte;  int zeroreqbyte, zerofirstbyte;
3057  int req_caseopt, reqvary, tempreqvary;  int req_caseopt, reqvary, tempreqvary;
3058  int options = *optionsptr;  int options = *optionsptr;               /* May change dynamically */
3059  int after_manual_callout = 0;  int after_manual_callout = 0;
3060  int length_prevgroup = 0;  int length_prevgroup = 0;
3061  register int c;  register int c;
# Line 2366  BOOL inescq = FALSE; Line 3067  BOOL inescq = FALSE;
3067  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
3068  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
3069  const uschar *tempptr;  const uschar *tempptr;
3070    const uschar *nestptr = NULL;
3071  uschar *previous = NULL;  uschar *previous = NULL;
3072  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
3073  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
3074  uschar classbits[32];  uschar classbits[32];
3075    
3076    /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3077    must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3078    dynamically as we process the pattern. */
3079    
3080  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3081  BOOL class_utf8;  BOOL class_utf8;
3082  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
# Line 2382  BOOL utf8 = FALSE; Line 3088  BOOL utf8 = FALSE;
3088  uschar *utf8_char = NULL;  uschar *utf8_char = NULL;
3089  #endif  #endif
3090    
3091  #ifdef DEBUG  #ifdef PCRE_DEBUG
3092  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
3093  #endif  #endif
3094    
# Line 2436  for (;; ptr++) Line 3142  for (;; ptr++)
3142    
3143    c = *ptr;    c = *ptr;
3144    
3145      /* If we are at the end of a nested substitution, revert to the outer level
3146      string. Nesting only happens one level deep. */
3147    
3148      if (c == 0 && nestptr != NULL)
3149        {
3150        ptr = nestptr;
3151        nestptr = NULL;
3152        c = *ptr;
3153        }
3154    
3155    /* 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
3156    previous cycle of this loop. */    previous cycle of this loop. */
3157    
3158    if (lengthptr != NULL)    if (lengthptr != NULL)
3159      {      {
3160  #ifdef DEBUG  #ifdef PCRE_DEBUG
3161      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3162  #endif  #endif
3163      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
3164        {        {
3165        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3166        goto FAILED;        goto FAILED;
# Line 2466  for (;; ptr++) Line 3182  for (;; ptr++)
3182        goto FAILED;        goto FAILED;
3183        }        }
3184    
3185      *lengthptr += code - last_code;      *lengthptr += (int)(code - last_code);
3186      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));
3187    
3188      /* 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 2493  for (;; ptr++) Line 3209  for (;; ptr++)
3209    /* 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
3210    reference list. */    reference list. */
3211    
3212    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
3213      {      {
3214      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3215      goto FAILED;      goto FAILED;
# Line 2503  for (;; ptr++) Line 3219  for (;; ptr++)
3219    
3220    if (inescq && c != 0)    if (inescq && c != 0)
3221      {      {
3222      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3223        {        {
3224        inescq = FALSE;        inescq = FALSE;
3225        ptr++;        ptr++;
# Line 2529  for (;; ptr++) Line 3245  for (;; ptr++)
3245    /* 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
3246    a quantifier. */    a quantifier. */
3247    
3248    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
3249      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
3250        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
3251    
3252    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
3253         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2540  for (;; ptr++) Line 3257  for (;; ptr++)
3257      previous_callout = NULL;      previous_callout = NULL;
3258      }      }
3259    
3260    /* In extended mode, skip white space and comments */    /* In extended mode, skip white space and comments. */
3261    
3262    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3263      {      {
3264      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
3265      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
3266        {        {
3267        while (*(++ptr) != 0)        ptr++;
3268          while (*ptr != 0)
3269          {          {
3270          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3271            ptr++;
3272    #ifdef SUPPORT_UTF8
3273            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
3274    #endif
3275          }          }
3276        if (*ptr != 0) continue;        if (*ptr != 0) continue;
3277    
# Line 2570  for (;; ptr++) Line 3292  for (;; ptr++)
3292      {      {
3293      /* ===================================================================*/      /* ===================================================================*/
3294      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
3295      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
3296      case ')':      case CHAR_RIGHT_PARENTHESIS:
3297      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
3298      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
3299      *codeptr = code;      *codeptr = code;
# Line 2583  for (;; ptr++) Line 3305  for (;; ptr++)
3305          *errorcodeptr = ERR20;          *errorcodeptr = ERR20;
3306          goto FAILED;          goto FAILED;
3307          }          }
3308        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += (int)(code - last_code);   /* To include callout length */
3309        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
3310        }        }
3311      return TRUE;      return TRUE;
# Line 2593  for (;; ptr++) Line 3315  for (;; ptr++)
3315      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
3316      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3317    
3318      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
3319        previous = NULL;
3320      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3321        {        {
3322        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3323          *code++ = OP_CIRCM;
3324        }        }
3325      previous = NULL;      else *code++ = OP_CIRC;
     *code++ = OP_CIRC;  
3326      break;      break;
3327    
3328      case '$':      case CHAR_DOLLAR_SIGN:
3329      previous = NULL;      previous = NULL;
3330      *code++ = OP_DOLL;      *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
3331      break;      break;
3332    
3333      /* 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
3334      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
3335    
3336      case '.':      case CHAR_DOT:
3337      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3338      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
3339      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3340      previous = code;      previous = code;
3341      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
3342      break;      break;
3343    
3344    
# Line 2630  for (;; ptr++) Line 3353  for (;; ptr++)
3353      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,
3354      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
3355      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.
     */  
3356    
3357      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
3358        default (Perl) mode, it is treated as a data character. */
3359    
3360        case CHAR_RIGHT_SQUARE_BRACKET:
3361        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3362          {
3363          *errorcodeptr = ERR64;
3364          goto FAILED;
3365          }
3366        goto NORMAL_CHAR;
3367    
3368        case CHAR_LEFT_SQUARE_BRACKET:
3369      previous = code;      previous = code;
3370    
3371      /* 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
3372      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. */
3373    
3374      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3375             ptr[1] == CHAR_EQUALS_SIGN) &&
3376          check_posix_syntax(ptr, &tempptr))          check_posix_syntax(ptr, &tempptr))
3377        {        {
3378        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
3379        goto FAILED;        goto FAILED;
3380        }        }
3381    
# Line 2653  for (;; ptr++) Line 3387  for (;; ptr++)
3387      for (;;)      for (;;)
3388        {        {
3389        c = *(++ptr);        c = *(++ptr);
3390        if (c == '\\')        if (c == CHAR_BACKSLASH)
3391          {          {
3392          if (ptr[1] == 'E') ptr++;          if (ptr[1] == CHAR_E)
3393            else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;            ptr++;
3394              else break;          else if (strncmp((const char *)ptr+1,
3395                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
3396              ptr += 3;
3397            else
3398              break;
3399          }          }
3400        else if (!negate_class && c == '^')        else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
3401          negate_class = TRUE;          negate_class = TRUE;
3402        else break;        else break;
3403        }        }
3404    
3405        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
3406        an initial ']' is taken as a data character -- the code below handles
3407        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3408        [^] must match any character, so generate OP_ALLANY. */
3409    
3410        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3411            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3412          {
3413          *code++ = negate_class? OP_ALLANY : OP_FAIL;
3414          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3415          zerofirstbyte = firstbyte;
3416          break;
3417          }
3418    
3419      /* If a class contains a negative special such as \S, we need to flip the      /* If a class contains a negative special such as \S, we need to flip the
3420      negation flag at the end, so that support for characters > 255 works      negation flag at the end, so that support for characters > 255 works
3421      correctly (they are all included in the class). */      correctly (they are all included in the class). */
# Line 2721  for (;; ptr++) Line 3473  for (;; ptr++)
3473    
3474        if (inescq)        if (inescq)
3475          {          {
3476          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3477            {            {
3478            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
3479            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2736  for (;; ptr++) Line 3488  for (;; ptr++)
3488        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3489        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3490    
3491        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3492            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3493            check_posix_syntax(ptr, &tempptr))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3494          {          {
3495          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3496          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3497          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3498          uschar pbits[32];          uschar pbits[32];
3499    
3500          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3501            {            {
3502            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3503            goto FAILED;            goto FAILED;
3504            }            }
3505    
3506          ptr += 2;          ptr += 2;
3507          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3508            {            {
3509            local_negate = TRUE;            local_negate = TRUE;
3510            should_flip_negation = TRUE;  /* Note negative special */            should_flip_negation = TRUE;  /* Note negative special */
3511            ptr++;            ptr++;
3512            }            }
3513    
3514          posix_class = check_posix_name(ptr, tempptr - ptr);          posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3515          if (posix_class < 0)          if (posix_class < 0)
3516            {            {
3517            *errorcodeptr = ERR30;            *errorcodeptr = ERR30;
# Line 2773  for (;; ptr++) Line 3525  for (;; ptr++)
3525          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3526            posix_class = 0;            posix_class = 0;
3527    
3528          /* 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
3529          because we may be adding and subtracting from it, and we don't want to          different escape sequences that use Unicode properties. */
3530          subtract bits that may be in the main map already. At the end we or the  
3531          result into the bit map that is being built. */  #ifdef SUPPORT_UCP
3532            if ((options & PCRE_UCP) != 0)
3533              {
3534              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3535              if (posix_substitutes[pc] != NULL)
3536                {
3537                nestptr = tempptr + 1;
3538                ptr = posix_substitutes[pc] - 1;
3539                continue;
3540                }
3541              }
3542    #endif
3543            /* In the non-UCP case, we build the bit map for the POSIX class in a
3544            chunk of local store because we may be adding and subtracting from it,
3545            and we don't want to subtract bits that may be in the main map already.
3546            At the end we or the result into the bit map that is being built. */
3547    
3548          posix_class *= 3;          posix_class *= 3;
3549    
# Line 2820  for (;; ptr++) Line 3587  for (;; ptr++)
3587    
3588        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3589        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
3590        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
3591        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
3592        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
3593        character in them, so set class_charcount bigger than one. */        are either treated as literal characters (by default), or are faulted if
3594          PCRE_EXTRA is set. */
3595    
3596        if (c == '\\')        if (c == CHAR_BACKSLASH)
3597          {          {
3598          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3599          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3600    
3601          if (-c == ESC_b) c = '\b';       /* \b is backspace 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 */  
3602          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3603            {            {
3604            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3605              {              {
3606              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3607              }              }
# Line 2849  for (;; ptr++) Line 3615  for (;; ptr++)
3615            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
3616            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3617    
3618            /* Save time by not doing this in the pre-compile phase. */            switch (-c)
   
           if (lengthptr == NULL) switch (-c)  
3619              {              {
3620    #ifdef SUPPORT_UCP
3621                case ESC_du:     /* These are the values given for \d etc */
3622                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3623                case ESC_wu:     /* escape sequence with an appropriate \p */
3624                case ESC_WU:     /* or \P to test Unicode properties instead */
3625                case ESC_su:     /* of the default ASCII testing. */
3626                case ESC_SU:
3627                nestptr = ptr;
3628                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3629                class_charcount -= 2;                /* Undo! */
3630                continue;
3631    #endif
3632              case ESC_d:              case ESC_d:
3633              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3634              continue;              continue;
# Line 2871  for (;; ptr++) Line 3647  for (;; ptr++)
3647              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3648              continue;              continue;
3649    
3650                /* Perl 5.004 onwards omits VT from \s, but we must preserve it
3651                if it was previously set by something earlier in the character
3652                class. */
3653    
3654              case ESC_s:              case ESC_s:
3655              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];              classbits[0] |= cbits[cbit_space];
3656              classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= cbits[cbit_space+1] & ~0x08;
3657                for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3658              continue;              continue;
3659    
3660              case ESC_S:              case ESC_S:
# Line 2882  for (;; ptr++) Line 3663  for (;; ptr++)
3663              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3664              continue;              continue;
3665    
3666              default:    /* Not recognized; fall through */              case ESC_h:
             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)  
             {  
3667              SETBIT(classbits, 0x09); /* VT */              SETBIT(classbits, 0x09); /* VT */
3668              SETBIT(classbits, 0x20); /* SPACE */              SETBIT(classbits, 0x20); /* SPACE */
3669              SETBIT(classbits, 0xa0); /* NSBP */              SETBIT(classbits, 0xa0); /* NSBP */
# Line 2919  for (;; ptr++) Line 3687  for (;; ptr++)
3687                }                }
3688  #endif  #endif
3689              continue;              continue;
             }  
3690    
3691            if (-c == ESC_H)              case ESC_H:
             {  
3692              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3693                {                {
3694                int x = 0xff;                int x = 0xff;
# Line 2964  for (;; ptr++) Line 3730  for (;; ptr++)
3730                }                }
3731  #endif  #endif
3732              continue;              continue;
             }  
3733    
3734            if (-c == ESC_v)              case ESC_v:
             {  
3735              SETBIT(classbits, 0x0a); /* LF */              SETBIT(classbits, 0x0a); /* LF */
3736              SETBIT(classbits, 0x0b); /* VT */              SETBIT(classbits, 0x0b); /* VT */
3737              SETBIT(classbits, 0x0c); /* FF */              SETBIT(classbits, 0x0c); /* FF */
# Line 2983  for (;; ptr++) Line 3747  for (;; ptr++)
3747                }                }
3748  #endif  #endif
3749              continue;              continue;
             }  
3750    
3751            if (-c == ESC_V)              case ESC_V:
             {  
3752              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3753                {                {
3754                int x = 0xff;                int x = 0xff;
# Line 3016  for (;; ptr++) Line 3778  for (;; ptr++)
3778                }                }
3779  #endif  #endif
3780              continue;              continue;
             }  
   
           /* We need to deal with \P and \p in both phases. */  
3781    
3782  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3783            if (-c == ESC_p || -c == ESC_P)              case ESC_p:
3784              {              case ESC_P:
3785              BOOL negated;                {
3786              int pdata;                BOOL negated;
3787              int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                int pdata;
3788              if (ptype < 0) goto FAILED;                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3789              class_utf8 = TRUE;                if (ptype < 0) goto FAILED;
3790              *class_utf8data++ = ((-c == ESC_p) != negated)?                class_utf8 = TRUE;
3791                XCL_PROP : XCL_NOTPROP;                *class_utf8data++ = ((-c == ESC_p) != negated)?
3792              *class_utf8data++ = ptype;                  XCL_PROP : XCL_NOTPROP;
3793              *class_utf8data++ = pdata;                *class_utf8data++ = ptype;
3794              class_charcount -= 2;   /* Not a < 256 character */                *class_utf8data++ = pdata;
3795              continue;                class_charcount -= 2;   /* Not a < 256 character */
3796              }                continue;
3797                  }
3798  #endif  #endif
3799            /* Unrecognized escapes are faulted if PCRE is running in its              /* Unrecognized escapes are faulted if PCRE is running in its
3800            strict mode. By default, for compatibility with Perl, they are              strict mode. By default, for compatibility with Perl, they are
3801            treated as literals. */              treated as literals. */
3802    
3803            if ((options & PCRE_EXTRA) != 0)              default:
3804              {              if ((options & PCRE_EXTRA) != 0)
3805              *errorcodeptr = ERR7;                {
3806              goto FAILED;                *errorcodeptr = ERR7;
3807                  goto FAILED;
3808                  }
3809                class_charcount -= 2;  /* Undo the default count from above */
3810                c = *ptr;              /* Get the final character and fall through */
3811                break;
3812              }              }
   
           class_charcount -= 2;  /* Undo the default count from above */  
           c = *ptr;              /* Get the final character and fall through */  
3813            }            }
3814    
3815          /* 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 3061  for (;; ptr++) Line 3823  for (;; ptr++)
3823        entirely. The code for handling \Q and \E is messy. */        entirely. The code for handling \Q and \E is messy. */
3824    
3825        CHECK_RANGE:        CHECK_RANGE:
3826        while (ptr[1] == '\\' && ptr[2] == 'E')        while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3827          {          {
3828          inescq = FALSE;          inescq = FALSE;
3829          ptr += 2;          ptr += 2;
# Line 3071  for (;; ptr++) Line 3833  for (;; ptr++)
3833    
3834        /* Remember \r or \n */        /* Remember \r or \n */
3835    
3836        if (c == '\r' || c == '\n') cd->external_flags |= PCRE_HASCRORLF;        if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3837    
3838        /* Check for range */        /* Check for range */
3839    
3840        if (!inescq && ptr[1] == '-')        if (!inescq && ptr[1] == CHAR_MINUS)
3841          {          {
3842          int d;          int d;
3843          ptr += 2;          ptr += 2;
3844          while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3845    
3846          /* 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
3847          mode. */          mode. */
3848    
3849          while (*ptr == '\\' && ptr[1] == 'Q')          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3850            {            {
3851            ptr += 2;            ptr += 2;
3852            if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }            if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3853                { ptr += 2; continue; }
3854            inescq = TRUE;            inescq = TRUE;
3855            break;            break;
3856            }            }
3857    
3858          if (*ptr == 0 || (!inescq && *ptr == ']'))          if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3859            {            {
3860            ptr = oldptr;            ptr = oldptr;
3861            goto LONE_SINGLE_CHARACTER;            goto LONE_SINGLE_CHARACTER;
# Line 3111  for (;; ptr++) Line 3874  for (;; ptr++)
3874          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
3875          in such circumstances. */          in such circumstances. */
3876    
3877          if (!inescq && d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3878            {            {
3879            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3880            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3881    
3882            /* \b is backspace; \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 */  
3883    
3884            if (d < 0)            if (d < 0)
3885              {              {
3886              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  
3887                {                {
3888                ptr = oldptr;                ptr = oldptr;
3889                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 3144  for (;; ptr++) Line 3904  for (;; ptr++)
3904    
3905          /* Remember \r or \n */          /* Remember \r or \n */
3906    
3907          if (d == '\r' || d == '\n') cd->external_flags |= PCRE_HASCRORLF;          if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3908    
3909          /* 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
3910          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
# Line 3264  for (;; ptr++) Line 4024  for (;; ptr++)
4024          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
4025            {            {
4026            unsigned int othercase;            unsigned int othercase;
4027            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)            if ((othercase = UCD_OTHERCASE(c)) != c)
4028              {              {
4029              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
4030              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 3289  for (;; ptr++) Line 4049  for (;; ptr++)
4049          }          }
4050        }        }
4051    
4052      /* 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.
4053        If we are at the end of an internal nested string, revert to the outer
4054        string. */
4055    
4056        while (((c = *(++ptr)) != 0 ||
4057               (nestptr != NULL &&
4058                 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != 0)) &&
4059               (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
4060    
4061      while ((c = *(++ptr)) != 0 && (c != ']' || inescq));      /* Check for missing terminating ']' */
4062    
4063      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)
4064        {        {
4065        *errorcodeptr = ERR6;        *errorcodeptr = ERR6;
4066        goto FAILED;        goto FAILED;
4067        }        }
4068    
   
 /* This code has been disabled because it would mean that \s counts as  
 an explicit \r or \n reference, and that's not really what is wanted. Now  
 we set the flag only if there is a literal "\r" or "\n" in the class. */  
   
 #if 0  
     /* Remember whether \r or \n are in this class */  
   
     if (negate_class)  
       {  
       if ((classbits[1] & 0x24) != 0x24) cd->external_flags |= PCRE_HASCRORLF;  
       }  
     else  
       {  
       if ((classbits[1] & 0x24) != 0) cd->external_flags |= PCRE_HASCRORLF;  
       }  
 #endif  
   
   
4069      /* 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
4070      less than 256. As long as there were no characters >= 128 and there was no      less than 256. As long as there were no characters >= 128 and there was no
4071      use of \p or \P, in other words, no use of any XCLASS features, we can      use of \p or \P, in other words, no use of any XCLASS features, we can
# Line 3325  we set the flag only if there is a liter Line 4073  we set the flag only if there is a liter
4073    
4074      In UTF-8 mode, we can optimize the negative case only if there were no      In UTF-8 mode, we can optimize the negative case only if there were no
4075      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
4076      operate on single-bytes only. This is an historical hangover. Maybe one day      operate on single-bytes characters only. This is an historical hangover.
4077      we can tidy these opcodes to handle multi-byte characters.      Maybe one day we can tidy these opcodes to handle multi-byte characters.
4078    
4079      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
4080      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.
4081      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
4082      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
4083      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
4084      reqbyte, save the previous value for reinstating. */      of reqbyte, save the previous value for reinstating. */
4085    
4086  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4087      if (class_charcount == 1 && !class_utf8 &&      if (class_charcount == 1 && !class_utf8 &&
# Line 3344  we set the flag only if there is a liter Line 4092  we set the flag only if there is a liter
4092        {        {
4093        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
4094    
4095        /* The OP_NOT opcode works on one-byte characters only. */        /* The OP_NOT[I] opcodes work on one-byte characters only. */
4096    
4097        if (negate_class)        if (negate_class)
4098          {          {
4099          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
4100          zerofirstbyte = firstbyte;          zerofirstbyte = firstbyte;
4101          *code++ = OP_NOT;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
4102          *code++ = class_lastchar;          *code++ = class_lastchar;
4103          break;          break;
4104          }          }
# Line 3381  we set the flag only if there is a liter Line 4129  we set the flag only if there is a liter
4129    
4130      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
4131      extended class, with its own opcode, unless there was a negated special      extended class, with its own opcode, unless there was a negated special
4132      such as \S in the class, because in that case all characters > 255 are in      such as \S in the class, and PCRE_UCP is not set, because in that case all
4133      the class, so any that were explicitly given as well can be ignored. If      characters > 255 are in the class, so any that were explicitly given as
4134      (when there are explicit characters > 255 that must be listed) there are no      well can be ignored. If (when there are explicit characters > 255 that must
4135      characters < 256, we can omit the bitmap in the actual compiled code. */      be listed) there are no characters < 256, we can omit the bitmap in the
4136        actual compiled code. */
4137    
4138  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4139      if (class_utf8 && !should_flip_negation)      if (class_utf8 && (!should_flip_negation || (options & PCRE_UCP) != 0))
4140        {        {
4141        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
4142        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 3413  we set the flag only if there is a liter Line 4162  we set the flag only if there is a liter
4162        }        }
4163  #endif  #endif
4164    
4165      /* If there are no characters > 255, set the opcode to OP_CLASS or      /* If there are no characters > 255, or they are all to be included or
4166      OP_NCLASS, depending on whether the whole class was negated and whether      excluded, set the opcode to OP_CLASS or OP_NCLASS, depending on whether the
4167      there were negative specials such as \S in the class. Then copy the 32-byte      whole class was negated and whether there were negative specials such as \S
4168      map into the code vector, negating it if necessary. */      (non-UCP) in the class. Then copy the 32-byte map into the code vector,
4169        negating it if necessary. */
4170    
4171      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
4172      if (negate_class)      if (negate_class)
# Line 3436  we set the flag only if there is a liter Line 4186  we set the flag only if there is a liter
4186      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
4187      has been tested above. */      has been tested above. */
4188    
4189      case '{':      case CHAR_LEFT_CURLY_BRACKET:
4190      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
4191      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
4192      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
4193      goto REPEAT;      goto REPEAT;
4194    
4195      case '*':      case CHAR_ASTERISK:
4196      repeat_min = 0;      repeat_min = 0;
4197      repeat_max = -1;      repeat_max = -1;
4198      goto REPEAT;      goto REPEAT;
4199    
4200      case '+':      case CHAR_PLUS:
4201      repeat_min = 1;      repeat_min = 1;
4202      repeat_max = -1;      repeat_max = -1;
4203      goto REPEAT;      goto REPEAT;
4204    
4205      case '?':      case CHAR_QUESTION_MARK:
4206      repeat_min = 0;      repeat_min = 0;
4207      repeat_max = 1;      repeat_max = 1;
4208    
# Line 3476  we set the flag only if there is a liter Line 4226  we set the flag only if there is a liter
4226      op_type = 0;                    /* Default single-char op codes */      op_type = 0;                    /* Default single-char op codes */
4227      possessive_quantifier = FALSE;  /* Default not possessive quantifier */      possessive_quantifier = FALSE;  /* Default not possessive quantifier */
4228    
4229      /* Save start of previous item, in case we have to move it up to make space      /* Save start of previous item, in case we have to move it up in order to
4230      for an inserted OP_ONCE for the additional '+' extension. */      insert something before it. */
4231    
4232      tempcode = previous;      tempcode = previous;
4233    
# Line 3487  we set the flag only if there is a liter Line 4237  we set the flag only if there is a liter
4237      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
4238      repeat type to the non-default. */      repeat type to the non-default. */
4239    
4240      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
4241        {        {
4242        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
4243        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
4244        ptr++;        ptr++;
4245        }        }
4246      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
4247        {        {
4248        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
4249        ptr++;        ptr++;
4250        }        }
4251      else repeat_type = greedy_default;      else repeat_type = greedy_default;
4252    
4253        /* If previous was a recursion call, wrap it in atomic brackets so that
4254        previous becomes the atomic group. All recursions were so wrapped in the
4255        past, but it no longer happens for non-repeated recursions. In fact, the
4256        repeated ones could be re-implemented independently so as not to need this,
4257        but for the moment we rely on the code for repeating groups. */
4258    
4259        if (*previous == OP_RECURSE)
4260          {
4261          memmove(previous + 1 + LINK_SIZE, previous, 1 + LINK_SIZE);
4262          *previous = OP_ONCE;
4263          PUT(previous, 1, 2 + 2*LINK_SIZE);
4264          previous[2 + 2*LINK_SIZE] = OP_KET;
4265          PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
4266          code += 2 + 2 * LINK_SIZE;
4267          length_prevgroup = 3 + 3*LINK_SIZE;
4268    
4269          /* When actually compiling, we need to check whether this was a forward
4270          reference, and if so, adjust the offset. */
4271    
4272          if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
4273            {
4274            int offset = GET(cd->hwm, -LINK_SIZE);
4275            if (offset == previous + 1 - cd->start_code)
4276              PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
4277            }
4278          }
4279    
4280        /* Now handle repetition for the different types of item. */
4281    
4282      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
4283      repeat item instead. If a char item has a minumum of more than one, ensure      repeat item instead. If a char item has a minumum of more than one, ensure
# Line 3506  we set the flag only if there is a liter Line 4285  we set the flag only if there is a liter
4285      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
4286      instead.  */      instead.  */
4287    
4288      if (*previous == OP_CHAR || *previous == OP_CHARNC)      if (*previous == OP_CHAR || *previous == OP_CHARI)
4289        {        {
4290          op_type = (*previous == OP_CHAR)? 0 : OP_STARI - OP_STAR;
4291    
4292        /* 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
4293        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
4294        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 3540  we set the flag only if there is a liter Line 4321  we set the flag only if there is a liter
4321    
4322        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4323            repeat_max < 0 &&            repeat_max < 0 &&
4324            check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
             options, cd))  
4325          {          {
4326          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4327          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3553  we set the flag only if there is a liter Line 4333  we set the flag only if there is a liter
4333      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
4334      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-
4335      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
4336      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
4337      currently used only for single-byte chars. */      are currently used only for single-byte chars. */
4338    
4339      else if (*previous == OP_NOT)      else if (*previous == OP_NOT || *previous == OP_NOTI)
4340        {        {
4341        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = ((*previous == OP_NOT)? OP_NOTSTAR : OP_NOTSTARI) - OP_STAR;
4342        c = previous[1];        c = previous[1];
4343        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4344            repeat_max < 0 &&            repeat_max < 0 &&
4345            check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4346          {          {
4347          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4348          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3586  we set the flag only if there is a liter Line 4366  we set the flag only if there is a liter
4366    
4367        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4368            repeat_max < 0 &&            repeat_max < 0 &&
4369            check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4370          {          {
4371          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4372          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3608  we set the flag only if there is a liter Line 4388  we set the flag only if there is a liter
4388    
4389        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
4390    
4391          /*--------------------------------------------------------------------*/
4392          /* This code is obsolete from release 8.00; the restriction was finally
4393          removed: */
4394    
4395        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4396        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4397    
4398        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4399          /*--------------------------------------------------------------------*/
4400    
4401        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
4402    
# Line 3750  we set the flag only if there is a liter Line 4535  we set the flag only if there is a liter
4535  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4536               *previous == OP_XCLASS ||               *previous == OP_XCLASS ||
4537  #endif  #endif
4538               *previous == OP_REF)               *previous == OP_REF ||
4539                 *previous == OP_REFI)
4540        {        {
4541        if (repeat_max == 0)        if (repeat_max == 0)
4542          {          {
# Line 3758  we set the flag only if there is a liter Line 4544  we set the flag only if there is a liter
4544          goto END_REPEAT;          goto END_REPEAT;
4545          }          }
4546    
4547          /*--------------------------------------------------------------------*/
4548          /* This code is obsolete from release 8.00; the restriction was finally
4549          removed: */
4550    
4551        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4552        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4553    
4554        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4555          /*--------------------------------------------------------------------*/
4556    
4557        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4558          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 3779  we set the flag only if there is a liter Line 4570  we set the flag only if there is a liter
4570        }        }
4571    
4572      /* If previous was a bracket group, we may have to replicate it in certain      /* If previous was a bracket group, we may have to replicate it in certain
4573      cases. */      cases. Note that at this point we can encounter only the "basic" bracket
4574        opcodes such as BRA and CBRA, as this is the place where they get converted
4575        into the more special varieties such as BRAPOS and SBRA. A test for >=
4576        OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
4577        ASSERTBACK_NOT, ONCE, BRA, CBRA, and COND. Originally, PCRE did not allow
4578        repetition of assertions, but now it does, for Perl compatibility. */
4579    
4580      else if (*previous == OP_BRA  || *previous == OP_CBRA ||      else if (*previous >= OP_ASSERT && *previous <= OP_COND)
              *previous == OP_ONCE || *previous == OP_COND)  
4581        {        {
4582        register int i;        register int i;
4583        int ketoffset = 0;        int len = (int)(code - previous);
       int len = code - previous;  
4584        uschar *bralink = NULL;        uschar *bralink = NULL;
4585          uschar *brazeroptr = NULL;
4586        /* Repeating a DEFINE group is pointless */  
4587          /* Repeating a DEFINE group is pointless, but Perl allows the syntax, so
4588          we just ignore the repeat. */
4589    
4590        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
4591          {          goto END_REPEAT;
         *errorcodeptr = ERR55;  
         goto FAILED;  
         }  
4592    
4593        /* If the maximum repeat count is unlimited, find the end of the bracket        /* There is no sense in actually repeating assertions. The only potential
4594        by scanning through from the start, and compute the offset back to it        use of repetition is in cases when the assertion is optional. Therefore,
4595        from the current code pointer. There may be an OP_OPT setting following        if the minimum is greater than zero, just ignore the repeat. If the
4596        the final KET, so we can't find the end just by going back from the code        maximum is not not zero or one, set it to 1. */
4597        pointer. */  
4598          if (*previous < OP_ONCE)    /* Assertion */
4599        if (repeat_max == -1)          {
4600          {          if (repeat_min > 0) goto END_REPEAT;
4601          register uschar *ket = previous;          if (repeat_max < 0 || repeat_max > 1) repeat_max = 1;
4602          do ket += GET(ket, 1); while (*ket != OP_KET);          }
         ketoffset = code - ket;  
         }  
4603    
4604        /* 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
4605        OP_BRAZERO in front of it, and because the group appears once in the        OP_BRAZERO in front of it, and because the group appears once in the
# Line 3819  we set the flag only if there is a liter Line 4610  we set the flag only if there is a liter
4610    
4611        if (repeat_min == 0)        if (repeat_min == 0)
4612          {          {
4613          /* 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
4614          altogether. */          output altogether, like this:
4615    
4616          if (repeat_max == 0)          ** if (repeat_max == 0)
4617            {          **   {
4618            code = previous;          **   code = previous;
4619            goto END_REPEAT;          **   goto END_REPEAT;
4620            }          **   }
4621    
4622            However, that fails when a group or a subgroup within it is referenced
4623            as a subroutine from elsewhere in the pattern, so now we stick in
4624            OP_SKIPZERO in front of it so that it is skipped on execution. As we
4625            don't have a list of which groups are referenced, we cannot do this
4626            selectively.
4627    
4628            If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4629            and do no more at this point. However, we do need to adjust any
4630            OP_RECURSE calls inside the group that refer to the group itself or any
4631            internal or forward referenced group, because the offset is from the
4632            start of the whole regex. Temporarily terminate the pattern while doing
4633            this. */
4634    
4635          /* If the maximum is 1 or unlimited, we just have to stick in the          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
         BRAZERO and do no more at this point. However, we do need to adjust  
         any OP_RECURSE calls inside the group that refer to the group itself or  
         any internal or forward referenced group, because the offset is from  
         the start of the whole regex. Temporarily terminate the pattern while  
         doing this. */  
   
         if (repeat_max <= 1)  
4636            {            {
4637            *code = OP_END;            *code = OP_END;
4638            adjust_recurse(previous, 1, utf8, cd, save_hwm);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
4639            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
4640            code++;            code++;
4641              if (repeat_max == 0)
4642                {
4643                *previous++ = OP_SKIPZERO;
4644                goto END_REPEAT;
4645                }
4646              brazeroptr = previous;    /* Save for possessive optimizing */
4647            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4648            }            }
4649    
# Line 3865  we set the flag only if there is a liter Line 4668  we set the flag only if there is a liter
4668            /* We chain together the bracket offset fields that have to be            /* We chain together the bracket offset fields that have to be
4669            filled in later when the ends of the brackets are reached. */            filled in later when the ends of the brackets are reached. */
4670    
4671            offset = (bralink == NULL)? 0 : previous - bralink;            offset = (bralink == NULL)? 0 : (int)(previous - bralink);
4672            bralink = previous;            bralink = previous;
4673            PUTINC(previous, 0, offset);            PUTINC(previous, 0, offset);
4674            }            }
# Line 3886  we set the flag only if there is a liter Line 4689  we set the flag only if there is a liter
4689            {            {
4690            /* 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
4691            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
4692            potential integer overflow. */            potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
4693              integer type when available, otherwise double. */
4694    
4695            if (lengthptr != NULL)            if (lengthptr != NULL)
4696              {              {
4697              int delta = (repeat_min - 1)*length_prevgroup;              int delta = (repeat_min - 1)*length_prevgroup;
4698              if ((double)(repeat_min - 1)*(double)length_prevgroup >              if ((INT64_OR_DOUBLE)(repeat_min - 1)*
4699                                                              (double)INT_MAX ||                    (INT64_OR_DOUBLE)length_prevgroup >
4700                        (INT64_OR_DOUBLE)INT_MAX ||
4701                  OFLOW_MAX - *lengthptr < delta)                  OFLOW_MAX - *lengthptr < delta)
4702                {                {
4703                *errorcodeptr = ERR20;                *errorcodeptr = ERR20;
# Line 3938  we set the flag only if there is a liter Line 4743  we set the flag only if there is a liter
4743          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
4744          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
4745          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
4746          paranoid checks to avoid integer overflow. */          paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
4747            a 64-bit integer type when available, otherwise double. */
4748    
4749          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4750            {            {
4751            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4752                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4753            if ((double)repeat_max *            if ((INT64_OR_DOUBLE)repeat_max *
4754                  (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)                  (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4755                    > (double)INT_MAX ||                    > (INT64_OR_DOUBLE)INT_MAX ||
4756                OFLOW_MAX - *lengthptr < delta)                OFLOW_MAX - *lengthptr < delta)
4757              {              {
4758              *errorcodeptr = ERR20;              *errorcodeptr = ERR20;
# Line 3971  we set the flag only if there is a liter Line 4777  we set the flag only if there is a liter
4777              {              {
4778              int offset;              int offset;
4779              *code++ = OP_BRA;              *code++ = OP_BRA;
4780              offset = (bralink == NULL)? 0 : code - bralink;              offset = (bralink == NULL)? 0 : (int)(code - bralink);
4781              bralink = code;              bralink = code;
4782              PUTINC(code, 0, offset);              PUTINC(code, 0, offset);
4783              }              }
# Line 3992  we set the flag only if there is a liter Line 4798  we set the flag only if there is a liter
4798          while (bralink != NULL)          while (bralink != NULL)
4799            {            {
4800            int oldlinkoffset;            int oldlinkoffset;
4801            int offset = code - bralink + 1;            int offset = (int)(code - bralink + 1);
4802            uschar *bra = code - offset;            uschar *bra = code - offset;
4803            oldlinkoffset = GET(bra, 1);            oldlinkoffset = GET(bra, 1);
4804            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;
# Line 4002  we set the flag only if there is a liter Line 4808  we set the flag only if there is a liter
4808            }            }
4809          }          }
4810    
4811        /* If the maximum is unlimited, set a repeater in the final copy. We        /* If the maximum is unlimited, set a repeater in the final copy. For
4812        can't just offset backwards from the current code point, because we        ONCE brackets, that's all we need to do. However, possessively repeated
4813        don't know if there's been an options resetting after the ket. The        ONCE brackets can be converted into non-capturing brackets, as the
4814        correct offset was computed above.        behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to
4815          deal with possessive ONCEs specially.
4816    
4817          Otherwise, if the quantifier was possessive, we convert the BRA code to
4818          the POS form, and the KET code to KETRPOS. (It turns out to be convenient
4819          at runtime to detect this kind of subpattern at both the start and at the
4820          end.) The use of special opcodes makes it possible to reduce greatly the
4821          stack usage in pcre_exec(). If the group is preceded by OP_BRAZERO,
4822          convert this to OP_BRAPOSZERO. Then cancel the possessive flag so that
4823          the default action below, of wrapping everything inside atomic brackets,
4824          does not happen.
4825    
4826        Then, when we are doing the actual compile phase, check to see whether        Then, when we are doing the actual compile phase, check to see whether
4827        this group is a non-atomic one that could match an empty string. If so,        this group is one that could match an empty string. If so, convert the
4828        convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so        initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so that runtime
4829        that runtime checking can be done. [This check is also applied to        checking can be done. [This check is also applied to ONCE groups at
4830        atomic groups at runtime, but in a different way.] */        runtime, but in a different way.] */
4831    
4832        else        else
4833          {          {
4834          uschar *ketcode = code - ketoffset;          uschar *ketcode = code - 1 - LINK_SIZE;
4835          uschar *bracode = ketcode - GET(ketcode, 1);          uschar *bracode = ketcode - GET(ketcode, 1);
4836          *ketcode = OP_KETRMAX + repeat_type;  
4837          if (lengthptr == NULL && *bracode != OP_ONCE)          if (*bracode == OP_ONCE && possessive_quantifier) *bracode = OP_BRA;
4838            if (*bracode == OP_ONCE)
4839              *ketcode = OP_KETRMAX + repeat_type;
4840            else
4841            {            {
4842            uschar *scode = bracode;            if (possessive_quantifier)
4843            do              {
4844                *bracode += 1;                   /* Switch to xxxPOS opcodes */
4845                *ketcode = OP_KETRPOS;
4846                if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
4847                possessive_quantifier = FALSE;
4848                }
4849              else *ketcode = OP_KETRMAX + repeat_type;
4850    
4851              if (lengthptr == NULL)
4852              {              {
4853              if (could_be_empty_branch(scode, ketcode, utf8))              uschar *scode = bracode;
4854                do
4855                {                {
4856                *bracode += OP_SBRA - OP_BRA;                if (could_be_empty_branch(scode, ketcode, utf8, cd))
4857                break;                  {
4858                    *bracode += OP_SBRA - OP_BRA;
4859                    break;
4860                    }
4861                  scode += GET(scode, 1);
4862                }                }
4863              scode += GET(scode, 1);              while (*scode == OP_ALT);
4864              }              }
           while (*scode == OP_ALT);  
4865            }            }
4866          }          }
4867        }        }
4868    
4869        /* If previous is OP_FAIL, it was generated by an empty class [] in
4870        JavaScript mode. The other ways in which OP_FAIL can be generated, that is
4871        by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
4872        error above. We can just ignore the repeat in JS case. */
4873    
4874        else if (*previous == OP_FAIL) goto END_REPEAT;
4875    
4876      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
4877    
4878      else      else
# Line 4044  we set the flag only if there is a liter Line 4882  we set the flag only if there is a liter
4882        }        }
4883    
4884      /* If the character following a repeat is '+', or if certain optimization      /* If the character following a repeat is '+', or if certain optimization
4885      tests above succeeded, possessive_quantifier is TRUE. For some of the      tests above succeeded, possessive_quantifier is TRUE. For some opcodes,
4886      simpler opcodes, there is an special alternative opcode for this. For      there are special alternative opcodes for this case. For anything else, we
4887      anything else, we wrap the entire repeated item inside OP_ONCE brackets.      wrap the entire repeated item inside OP_ONCE brackets. Logically, the '+'
4888      The '+' notation is just syntactic sugar, taken from Sun's Java package,      notation is just syntactic sugar, taken from Sun's Java package, but the
4889      but the special opcodes can optimize it a bit. The repeated item starts at      special opcodes can optimize it.
4890      tempcode, not at previous, which might be the first part of a string whose  
4891      (former) last char we repeated.      Possessively repeated subpatterns have already been handled in the code
4892        just above, so possessive_quantifier is always FALSE for them at this
4893        stage.
4894    
4895        Note that the repeated item starts at tempcode, not at previous, which
4896        might be the first part of a string whose (former) last char we repeated.
4897    
4898      Possessifying an 'exact' quantifier has no effect, so we can ignore it. But      Possessifying an 'exact' quantifier has no effect, so we can ignore it. But
4899      an 'upto' may follow. We skip over an 'exact' item, and then test the      an 'upto' may follow. We skip over an 'exact' item, and then test the
# Line 4059  we set the flag only if there is a liter Line 4902  we set the flag only if there is a liter
4902      if (possessive_quantifier)      if (possessive_quantifier)
4903        {        {
4904        int len;        int len;
4905        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||  
4906            *tempcode == OP_NOTEXACT)        if (*tempcode == OP_TYPEEXACT)
4907          tempcode += _pcre_OP_lengths[*tempcode] +          tempcode += _pcre_OP_lengths[*tempcode] +
4908            ((*tempcode == OP_TYPEEXACT &&            ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
4909               (tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP))? 2:0);  
4910        len = code - tempcode;        else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
4911            {
4912            tempcode += _pcre_OP_lengths[*tempcode];
4913    #ifdef SUPPORT_UTF8
4914            if (utf8 && tempcode[-1] >= 0xc0)
4915              tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];
4916    #endif
4917            }
4918    
4919          len = (int)(code - tempcode);
4920        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4921          {          {
4922          case OP_STAR:  *tempcode = OP_POSSTAR; break;          case OP_STAR:  *tempcode = OP_POSSTAR; break;
# Line 4072  we set the flag only if there is a liter Line 4924  we set the flag only if there is a liter
4924          case OP_QUERY: *tempcode = OP_POSQUERY; break;          case OP_QUERY: *tempcode = OP_POSQUERY; break;
4925          case OP_UPTO:  *tempcode = OP_POSUPTO; break;          case OP_UPTO:  *tempcode = OP_POSUPTO; break;
4926    
4927          case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;          case OP_STARI:  *tempcode = OP_POSSTARI; break;
4928          case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;          case OP_PLUSI:  *tempcode = OP_POSPLUSI; break;
4929          case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;          case OP_QUERYI: *tempcode = OP_POSQUERYI; break;
4930          case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;          case OP_UPTOI:  *tempcode = OP_POSUPTOI; break;
4931    
4932          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;
4933          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;
4934          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4935          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4936    
4937            case OP_NOTSTARI:  *tempcode = OP_NOTPOSSTARI; break;
4938            case OP_NOTPLUSI:  *tempcode = OP_NOTPOSPLUSI; break;
4939            case OP_NOTQUERYI: *tempcode = OP_NOTPOSQUERYI; break;
4940            case OP_NOTUPTOI:  *tempcode = OP_NOTPOSUPTOI; break;
4941    
4942            case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;
4943            case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;
4944            case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
4945            case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;
4946    
4947            /* Because we are moving code along, we must ensure that any
4948            pending recursive references are updated. */
4949    
4950          default:          default:
4951            *code = OP_END;
4952            adjust_recurse(tempcode, 1 + LINK_SIZE, utf8, cd, save_hwm);
4953          memmove(tempcode + 1+LINK_SIZE, tempcode, len);          memmove(tempcode + 1+LINK_SIZE, tempcode, len);
4954          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
4955          len += 1 + LINK_SIZE;          len += 1 + LINK_SIZE;
# Line 4109  we set the flag only if there is a liter Line 4976  we set the flag only if there is a liter
4976      lookbehind or option setting or condition or all the other extended      lookbehind or option setting or condition or all the other extended
4977      parenthesis forms.  */      parenthesis forms.  */
4978    
4979      case '(':      case CHAR_LEFT_PARENTHESIS:
4980      newoptions = options;      newoptions = options;
4981      skipbytes = 0;      skipbytes = 0;
4982      bravalue = OP_CBRA;      bravalue = OP_CBRA;
# Line 4118  we set the flag only if there is a liter Line 4985  we set the flag only if there is a liter
4985    
4986      /* First deal with various "verbs" that can be introduced by '*'. */      /* First deal with various "verbs" that can be introduced by '*'. */
4987    
4988      if (*(++ptr) == '*' && (cd->ctypes[ptr[1]] & ctype_letter) != 0)      if (*(++ptr) == CHAR_ASTERISK &&
4989             ((cd->ctypes[ptr[1]] & ctype_letter) != 0 || ptr[1] == ':'))
4990        {        {
4991        int i, namelen;        int i, namelen;
4992          int arglen = 0;
4993        const char *vn = verbnames;        const char *vn = verbnames;
4994        const uschar *name = ++ptr;        const uschar *name = ptr + 1;
4995          const uschar *arg = NULL;
4996        previous = NULL;        previous = NULL;
4997        while ((cd->ctypes[*++ptr] & ctype_letter) != 0);        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
4998        if (*ptr == ':')        namelen = (int)(ptr - name);
4999    
5000          if (*ptr == CHAR_COLON)
5001          {          {
5002          *errorcodeptr = ERR59;   /* Not supported */          arg = ++ptr;
5003          goto FAILED;          while ((cd->ctypes[*ptr] & (ctype_letter|ctype_digit)) != 0
5004              || *ptr == '_') ptr++;
5005            arglen = (int)(ptr - arg);
5006          }          }
5007        if (*ptr != ')')  
5008          if (*ptr != CHAR_RIGHT_PARENTHESIS)
5009          {          {
5010          *errorcodeptr = ERR60;          *errorcodeptr = ERR60;
5011          goto FAILED;          goto FAILED;
5012          }          }
5013        namelen = ptr - name;  
5014          /* Scan the table of verb names */
5015    
5016        for (i = 0; i < verbcount; i++)        for (i = 0; i < verbcount; i++)
5017          {          {
5018          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
5019              strncmp((char *)name, vn, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
5020            {            {
5021            *code = verbs[i].op;            /* Check for open captures before ACCEPT and convert it to
5022            if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;            ASSERT_ACCEPT if in an assertion. */
5023            break;  
5024              if (verbs[i].op == OP_ACCEPT)
5025                {
5026                open_capitem *oc;
5027                if (arglen != 0)
5028                  {
5029                  *errorcodeptr = ERR59;
5030                  goto FAILED;
5031                  }
5032                cd->had_accept = TRUE;
5033                for (oc = cd->open_caps; oc != NULL; oc = oc->next)
5034                  {
5035                  *code++ = OP_CLOSE;
5036                  PUT2INC(code, 0, oc->number);
5037                  }
5038                *code++ = (cd->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT;
5039                }
5040    
5041              /* Handle other cases with/without an argument */
5042    
5043              else if (arglen == 0)
5044                {
5045                if (verbs[i].op < 0)   /* Argument is mandatory */
5046                  {
5047                  *errorcodeptr = ERR66;
5048                  goto FAILED;
5049                  }
5050                *code = verbs[i].op;
5051                if (*code++ == OP_THEN)
5052                  {
5053                  PUT(code, 0, code - bcptr->current_branch - 1);
5054                  code += LINK_SIZE;
5055                  }
5056                }
5057    
5058              else
5059                {
5060                if (verbs[i].op_arg < 0)   /* Argument is forbidden */
5061                  {
5062                  *errorcodeptr = ERR59;
5063                  goto FAILED;
5064                  }
5065                *code = verbs[i].op_arg;
5066                if (*code++ == OP_THEN_ARG)
5067                  {
5068                  PUT(code, 0, code - bcptr->current_branch - 1);
5069                  code += LINK_SIZE;
5070                  }
5071                *code++ = arglen;
5072                memcpy(code, arg, arglen);
5073                code += arglen;
5074                *code++ = 0;
5075                }
5076    
5077              break;  /* Found verb, exit loop */
5078            }            }
5079    
5080          vn += verbs[i].len + 1;          vn += verbs[i].len + 1;
5081          }          }
5082        if (i < verbcount) continue;  
5083        *errorcodeptr = ERR60;        if (i < verbcount) continue;    /* Successfully handled a verb */
5084          *errorcodeptr = ERR60;          /* Verb not recognized */
5085        goto FAILED;        goto FAILED;
5086        }        }
5087    
5088      /* Deal with the extended parentheses; all are introduced by '?', and the      /* Deal with the extended parentheses; all are introduced by '?', and the
5089      appearance of any of them means that this is not a capturing group. */      appearance of any of them means that this is not a capturing group. */
5090    
5091      else if (*ptr == '?')      else if (*ptr == CHAR_QUESTION_MARK)
5092        {        {
5093        int i, set, unset, namelen;        int i, set, unset, namelen;
5094        int *optset;        int *optset;
# Line 4164  we set the flag only if there is a liter Line 5097  we set the flag only if there is a liter
5097    
5098        switch (*(++ptr))        switch (*(++ptr))
5099          {          {
5100          case '#':                 /* Comment; skip to ket */          case CHAR_NUMBER_SIGN:                 /* Comment; skip to ket */
5101          ptr++;          ptr++;
5102          while (*ptr != 0 && *ptr != ')') ptr++;          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
5103          if (*ptr == 0)          if (*ptr == 0)
5104            {            {
5105            *errorcodeptr = ERR18;            *errorcodeptr = ERR18;
# Line 4176  we set the flag only if there is a liter Line 5109  we set the flag only if there is a liter
5109    
5110    
5111          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5112          case '|':                 /* Reset capture count for each branch */          case CHAR_VERTICAL_LINE:  /* Reset capture count for each branch */
5113          reset_bracount = TRUE;          reset_bracount = TRUE;
5114          /* Fall through */          /* Fall through */
5115    
5116