/[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 274 by ph10, Tue Nov 20 10:05:23 2007 UTC revision 745 by ph10, Mon Nov 14 11:41:03 2011 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2007 University of Cambridge             Copyright (c) 1997-2011 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 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 *, int *, branch_chain *, compile_data *, int *);
551    
552    
# 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:
     case 'N':  
     case 'u':  
     case 'U':  
679      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
680      break;      break;
681    
682      /* \g must be followed by a number, either plain or braced. If positive, it      case CHAR_u:
683      is an absolute backreference. If negative, it is a relative backreference.      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
684      This is a Perl 5.10 feature. Perl 5.10 also supports \g{name} as a        {
685      reference to a named group. This is part of Perl's movement towards a        /* In JavaScript, \u must be followed by four hexadecimal numbers.
686      unified syntax for back references. As this is synonymous with \k{name}, we        Otherwise it is a lowercase u letter. */
687      fudge it up by pretending it really was \k. */        if ((digitab[ptr[1]] & ctype_xdigit) != 0 && (digitab[ptr[2]] & ctype_xdigit) != 0
688               && (digitab[ptr[3]] & ctype_xdigit) != 0 && (digitab[ptr[4]] & ctype_xdigit) != 0)
689            {
690            c = 0;
691            for (i = 0; i < 4; ++i)
692              {
693              register int cc = *(++ptr);
694    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
695              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
696              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
697    #else           /* EBCDIC coding */
698              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
699              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
700    #endif
701              }
702            }
703          }
704        else
705          *errorcodeptr = ERR37;
706        break;
707    
708        case CHAR_U:
709        /* In JavaScript, \U is an uppercase U letter. */
710        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
711        break;
712    
713      case 'g':      /* In a character class, \g is just a literal "g". Outside a character
714      if (ptr[1] == '{')      class, \g must be followed by one of a number of specific things:
715    
716        (1) A number, either plain or braced. If positive, it is an absolute
717        backreference. If negative, it is a relative backreference. This is a Perl
718        5.10 feature.
719    
720        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
721        is part of Perl's movement towards a unified syntax for back references. As
722        this is synonymous with \k{name}, we fudge it up by pretending it really
723        was \k.
724    
725        (3) For Oniguruma compatibility we also support \g followed by a name or a
726        number either in angle brackets or in single quotes. However, these are
727        (possibly recursive) subroutine calls, _not_ backreferences. Just return
728        the -ESC_g code (cf \k). */
729    
730        case CHAR_g:
731        if (isclass) break;
732        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
733          {
734          c = -ESC_g;
735          break;
736          }
737    
738        /* Handle the Perl-compatible cases */
739    
740        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
741        {        {
742        const uschar *p;        const uschar *p;
743        for (p = ptr+2; *p != 0 && *p != '}'; p++)        for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
744          if (*p != '-' && (digitab[*p] & ctype_digit) == 0) break;          if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
745        if (*p != 0 && *p != '}')        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
746          {          {
747          c = -ESC_k;          c = -ESC_k;
748          break;          break;
# Line 554  else Line 752  else
752        }        }
753      else braced = FALSE;      else braced = FALSE;
754    
755      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
756        {        {
757        negated = TRUE;        negated = TRUE;
758        ptr++;        ptr++;
# Line 563  else Line 761  else
761    
762      c = 0;      c = 0;
763      while ((digitab[ptr[1]] & ctype_digit) != 0)      while ((digitab[ptr[1]] & ctype_digit) != 0)
764        c = c * 10 + *(++ptr) - '0';        c = c * 10 + *(++ptr) - CHAR_0;
765    
766      if (c < 0)      if (c < 0)   /* Integer overflow */
767        {        {
768        *errorcodeptr = ERR61;        *errorcodeptr = ERR61;
769        break;        break;
770        }        }
771    
772      if (c == 0 || (braced && *(++ptr) != '}'))      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
773        {        {
774        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
775        break;        break;
776        }        }
777    
778        if (c == 0)
779          {
780          *errorcodeptr = ERR58;
781          break;
782          }
783    
784      if (negated)      if (negated)
785        {        {
786        if (c > bracount)        if (c > bracount)
# Line 602  else Line 806  else
806      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
807      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
808    
809      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:
810      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
811    
812      if (!isclass)      if (!isclass)
813        {        {
814        oldptr = ptr;        oldptr = ptr;
815        c -= '0';        c -= CHAR_0;
816        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
817          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - CHAR_0;
818        if (c < 0)        if (c < 0)    /* Integer overflow */
819          {          {
820          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
821          break;          break;
# Line 628  else Line 832  else
832      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.
833      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
834    
835      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
836        {        {
837        ptr--;        ptr--;
838        c = 0;        c = 0;
# Line 641  else Line 845  else
845      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
846      than 3 octal digits. */      than 3 octal digits. */
847    
848      case '0':      case CHAR_0:
849      c -= '0';      c -= CHAR_0;
850      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
851          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
852      if (!utf8 && c > 255) *errorcodeptr = ERR51;      if (!utf8 && c > 255) *errorcodeptr = ERR51;
853      break;      break;
854    
# Line 652  else Line 856  else
856      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
857      treated as a data character. */      treated as a data character. */
858    
859      case 'x':      case CHAR_x:
860      if (ptr[1] == '{')      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
861          {
862          /* In JavaScript, \x must be followed by two hexadecimal numbers.
863          Otherwise it is a lowercase x letter. */
864          if ((digitab[ptr[1]] & ctype_xdigit) != 0 && (digitab[ptr[2]] & ctype_xdigit) != 0)
865            {
866            c = 0;
867            for (i = 0; i < 2; ++i)
868              {
869              register int cc = *(++ptr);
870    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
871              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
872              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
873    #else           /* EBCDIC coding */
874              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
875              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
876    #endif
877              }
878            }
879          break;
880          }
881    
882        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
883        {        {
884        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
885        int count = 0;        int count = 0;
# Line 662  else Line 888  else
888        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
889          {          {
890          register int cc = *pt++;          register int cc = *pt++;
891          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
892          count++;          count++;
893    
894  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
895          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
896          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
897  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
898          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
899          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
900  #endif  #endif
901          }          }
902    
903        if (*pt == '}')        if (*pt == CHAR_RIGHT_CURLY_BRACKET)
904          {          {
905          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
906          ptr = pt;          ptr = pt;
# Line 690  else Line 916  else
916      c = 0;      c = 0;
917      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
918        {        {
919        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
920        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
921  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
922        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
923        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
924  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
925        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
926        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
927  #endif  #endif
928        }        }
929      break;      break;
930    
931      /* 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.
932      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
933        coding is ASCII-specific, but then the whole concept of \cx is
934      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
935    
936      case 'c':      case CHAR_c:
937      c = *(++ptr);      c = *(++ptr);
938      if (c == 0)      if (c == 0)
939        {        {
940        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
941        break;        break;
942        }        }
943    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
944  #ifndef EBCDIC  /* ASCII coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
945      if (c >= 'a' && c <= 'z') c -= 32;        {
946          *errorcodeptr = ERR68;
947          break;
948          }
949        if (c >= CHAR_a && c <= CHAR_z) c -= 32;
950      c ^= 0x40;      c ^= 0x40;
951  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
952      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
953      c ^= 0xC0;      c ^= 0xC0;
954  #endif  #endif
955      break;      break;
# Line 740  else Line 971  else
971      }      }
972    }    }
973    
974    /* Perl supports \N{name} for character names, as well as plain \N for "not
975    newline". PCRE does not support \N{name}. However, it does support
976    quantification such as \N{2,3}. */
977    
978    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
979         !is_counted_repeat(ptr+2))
980      *errorcodeptr = ERR37;
981    
982    /* If PCRE_UCP is set, we change the values for \d etc. */
983    
984    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
985      c -= (ESC_DU - ESC_D);
986    
987    /* Set the pointer to the final character before returning. */
988    
989  *ptrptr = ptr;  *ptrptr = ptr;
990  return c;  return c;
991  }  }
# Line 780  if (c == 0) goto ERROR_RETURN; Line 1026  if (c == 0) goto ERROR_RETURN;
1026  /* \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
1027  negation. */  negation. */
1028    
1029  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
1030    {    {
1031    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1032      {      {
1033      *negptr = TRUE;      *negptr = TRUE;
1034      ptr++;      ptr++;
# Line 791  if (c == '{') Line 1037  if (c == '{')
1037      {      {
1038      c = *(++ptr);      c = *(++ptr);
1039      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
1040      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1041      name[i] = c;      name[i] = c;
1042      }      }
1043    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
1044    name[i] = 0;    name[i] = 0;
1045    }    }
1046    
# Line 840  return -1; Line 1086  return -1;
1086    
1087    
1088  /*************************************************  /*************************************************
 *            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 == '}');  
 }  
   
   
   
 /*************************************************  
1089  *         Read repeat counts                     *  *         Read repeat counts                     *
1090  *************************************************/  *************************************************/
1091    
# Line 900  int max = -1; Line 1113  int max = -1;
1113  /* 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
1114  an integer overflow. */  an integer overflow. */
1115    
1116  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;
1117  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1118    {    {
1119    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 910  if (min < 0 || min > 65535) Line 1123  if (min < 0 || min > 65535)
1123  /* 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.
1124  Also, max must not be less than min. */  Also, max must not be less than min. */
1125    
1126  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1127    {    {
1128    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1129      {      {
1130      max = 0;      max = 0;
1131      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;
1132      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1133        {        {
1134        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 940  return p; Line 1153  return p;
1153    
1154    
1155  /*************************************************  /*************************************************
1156  *       Find forward referenced subpattern       *  *  Subroutine for finding forward reference      *
1157  *************************************************/  *************************************************/
1158    
1159  /* This function scans along a pattern's text looking for capturing  /* This recursive function is called only from find_parens() below. The
1160    top-level call starts at the beginning of the pattern. All other calls must
1161    start at a parenthesis. It scans along a pattern's text looking for capturing
1162  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
1163  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
1164  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
1165  references to subpatterns. We know that if (?P< is encountered, the name will  track of subpatterns that reset the capturing group numbers - the (?| feature.
1166  be terminated by '>' because that is checked in the first pass.  
1167    This function was originally called only from the second pass, in which we know
1168    that if (?< or (?' or (?P< is encountered, the name will be correctly
1169    terminated because that is checked in the first pass. There is now one call to
1170    this function in the first pass, to check for a recursive back reference by
1171    name (so that we can make the whole group atomic). In this case, we need check
1172    only up to the current position in the pattern, and that is still OK because
1173    and previous occurrences will have been checked. To make this work, the test
1174    for "end of pattern" is a check against cd->end_pattern in the main loop,
1175    instead of looking for a binary zero. This means that the special first-pass
1176    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1177    processing items within the loop are OK, because afterwards the main loop will
1178    terminate.)
1179    
1180  Arguments:  Arguments:
1181    ptr          current position in the pattern    ptrptr       address of the current character pointer (updated)
1182    count        current count of capturing parens so far encountered    cd           compile background data
1183    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1184    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1185    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1186      utf8         TRUE if we are in UTF-8 mode
1187      count        pointer to the current capturing subpattern number (updated)
1188    
1189  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1190  */  */
1191    
1192  static int  static int
1193  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,
1194    BOOL xmode)    BOOL xmode, BOOL utf8, int *count)
1195  {  {
1196  const uschar *thisname;  uschar *ptr = *ptrptr;
1197    int start_count = *count;
1198    int hwm_count = start_count;
1199    BOOL dup_parens = FALSE;
1200    
1201    /* If the first character is a parenthesis, check on the type of group we are
1202    dealing with. The very first call may not start with a parenthesis. */
1203    
1204  for (; *ptr != 0; ptr++)  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1205    {    {
1206    int term;    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1207    
1208      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1209    
1210      /* Handle a normal, unnamed capturing parenthesis. */
1211    
1212      else if (ptr[1] != CHAR_QUESTION_MARK)
1213        {
1214        *count += 1;
1215        if (name == NULL && *count == lorn) return *count;
1216        ptr++;
1217        }
1218    
1219      /* All cases now have (? at the start. Remember when we are in a group
1220      where the parenthesis numbers are duplicated. */
1221    
1222      else if (ptr[2] == CHAR_VERTICAL_LINE)
1223        {
1224        ptr += 3;
1225        dup_parens = TRUE;
1226        }
1227    
1228      /* Handle comments; all characters are allowed until a ket is reached. */
1229    
1230      else if (ptr[2] == CHAR_NUMBER_SIGN)
1231        {
1232        for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1233        goto FAIL_EXIT;
1234        }
1235    
1236      /* Handle a condition. If it is an assertion, just carry on so that it
1237      is processed as normal. If not, skip to the closing parenthesis of the
1238      condition (there can't be any nested parens). */
1239    
1240      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1241        {
1242        ptr += 2;
1243        if (ptr[1] != CHAR_QUESTION_MARK)
1244          {
1245          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1246          if (*ptr != 0) ptr++;
1247          }
1248        }
1249    
1250      /* Start with (? but not a condition. */
1251    
1252      else
1253        {
1254        ptr += 2;
1255        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1256    
1257        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1258    
1259        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1260            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1261          {
1262          int term;
1263          const uschar *thisname;
1264          *count += 1;
1265          if (name == NULL && *count == lorn) return *count;
1266          term = *ptr++;
1267          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1268          thisname = ptr;
1269          while (*ptr != term) ptr++;
1270          if (name != NULL && lorn == ptr - thisname &&
1271              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1272            return *count;
1273          term++;
1274          }
1275        }
1276      }
1277    
1278    /* Past any initial parenthesis handling, scan for parentheses or vertical
1279    bars. Stop if we get to cd->end_pattern. Note that this is important for the
1280    first-pass call when this value is temporarily adjusted to stop at the current
1281    position. So DO NOT change this to a test for binary zero. */
1282    
1283    for (; ptr < cd->end_pattern; ptr++)
1284      {
1285    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1286    
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      }      }
1298    
1299    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1300      are handled for real. If the first character is '^', skip it. Also, if the
1301      first few characters (either before or after ^) are \Q\E or \E we skip them
1302      too. This makes for compatibility with Perl. Note the use of STR macros to
1303      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1304    
1305    if (*ptr == '[')    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1306      {      {
1307      while (*(++ptr) != ']')      BOOL negate_class = FALSE;
1308        for (;;)
1309          {
1310          if (ptr[1] == CHAR_BACKSLASH)
1311            {
1312            if (ptr[2] == CHAR_E)
1313              ptr+= 2;
1314            else if (strncmp((const char *)ptr+2,
1315                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1316              ptr += 4;
1317            else
1318              break;
1319            }
1320          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1321            {
1322            negate_class = TRUE;
1323            ptr++;
1324            }
1325          else break;
1326          }
1327    
1328        /* If the next character is ']', it is a data character that must be
1329        skipped, except in JavaScript compatibility mode. */
1330    
1331        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1332            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1333          ptr++;
1334    
1335        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1336        {        {
1337        if (*ptr == 0) return -1;        if (*ptr == 0) return -1;
1338        if (*ptr == '\\')        if (*ptr == CHAR_BACKSLASH)
1339          {          {
1340          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) goto FAIL_EXIT;
1341          if (*ptr == 'Q') for (;;)          if (*ptr == CHAR_Q) for (;;)
1342            {            {
1343            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1344            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1345            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1346            }            }
1347          continue;          continue;
1348          }          }
# Line 1008  for (; *ptr != 0; ptr++) Line 1352  for (; *ptr != 0; ptr++)
1352    
1353    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1354    
1355    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1356      {      {
1357      while (*(++ptr) != 0 && *ptr != '\n');      ptr++;
1358      if (*ptr == 0) return -1;      while (*ptr != 0)
1359          {
1360          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1361          ptr++;
1362    #ifdef SUPPORT_UTF8
1363          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
1364    #endif
1365          }
1366        if (*ptr == 0) goto FAIL_EXIT;
1367      continue;      continue;
1368      }      }
1369    
1370    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1371    
1372    if (*ptr != '(') continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != '?' && ptr[1] != '*')  
1373      {      {
1374      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, count);
1375      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1376      continue;      if (*ptr == 0) goto FAIL_EXIT;
1377      }      }
1378    
1379    ptr += 2;    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1380    if (*ptr == 'P') ptr++;                      /* Allow optional P */      {
1381        if (dup_parens && *count < hwm_count) *count = hwm_count;
1382        goto FAIL_EXIT;
1383        }
1384    
1385    /* We have to disambiguate (?<! and (?<= from (?<name> */    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1386        {
1387        if (*count > hwm_count) hwm_count = *count;
1388        *count = start_count;
1389        }
1390      }
1391    
1392    if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  FAIL_EXIT:
1393         *ptr != '\'')  *ptrptr = ptr;
1394      continue;  return -1;
1395    }
1396    
1397    
1398    
1399    
1400    /*************************************************
1401    *       Find forward referenced subpattern       *
1402    *************************************************/
1403    
1404    /* This function scans along a pattern's text looking for capturing
1405    subpatterns, and counting them. If it finds a named pattern that matches the
1406    name it is given, it returns its number. Alternatively, if the name is NULL, it
1407    returns when it reaches a given numbered subpattern. This is used for forward
1408    references to subpatterns. We used to be able to start this scan from the
1409    current compiling point, using the current count value from cd->bracount, and
1410    do it all in a single loop, but the addition of the possibility of duplicate
1411    subpattern numbers means that we have to scan from the very start, in order to
1412    take account of such duplicates, and to use a recursive function to keep track
1413    of the different types of group.
1414    
1415    count++;  Arguments:
1416      cd           compile background data
1417      name         name to seek, or NULL if seeking a numbered subpattern
1418      lorn         name length, or subpattern number if name is NULL
1419      xmode        TRUE if we are in /x mode
1420      utf8         TRUE if we are in UTF-8 mode
1421    
1422    Returns:       the number of the found subpattern, or -1 if not found
1423    */
1424    
1425    static int
1426    find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode,
1427      BOOL utf8)
1428    {
1429    uschar *ptr = (uschar *)cd->start_pattern;
1430    int count = 0;
1431    int rc;
1432    
1433    /* If the pattern does not start with an opening parenthesis, the first call
1434    to find_parens_sub() will scan right to the end (if necessary). However, if it
1435    does start with a parenthesis, find_parens_sub() will return when it hits the
1436    matching closing parens. That is why we have to have a loop. */
1437    
1438    if (name == NULL && count == lorn) return count;  for (;;)
1439    term = *ptr++;    {
1440    if (term == '<') term = '>';    rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, &count);
1441    thisname = ptr;    if (rc > 0 || *ptr++ == 0) break;
   while (*ptr != term) ptr++;  
   if (name != NULL && lorn == ptr - thisname &&  
       strncmp((const char *)name, (const char *)thisname, lorn) == 0)  
     return count;  
1442    }    }
1443    
1444  return -1;  return rc;
1445  }  }
1446    
1447    
1448    
1449    
1450  /*************************************************  /*************************************************
1451  *      Find first significant op code            *  *      Find first significant op code            *
1452  *************************************************/  *************************************************/
1453    
1454  /* This is called by several functions that scan a compiled expression looking  /* This is called by several functions that scan a compiled expression looking
1455  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
1456  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
1457  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
1458  assertions, and also the \b assertion; for others it does not.  does not.
1459    
1460  Arguments:  Arguments:
1461    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  
1462    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1463    
1464  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1465  */  */
1466    
1467  static const uschar*  static const uschar*
1468  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const uschar *code, BOOL skipassert)
   BOOL skipassert)  
1469  {  {
1470  for (;;)  for (;;)
1471    {    {
1472    switch ((int)*code)    switch ((int)*code)
1473      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1474      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1475      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1476      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
# Line 1100  for (;;) Line 1486  for (;;)
1486    
1487      case OP_CALLOUT:      case OP_CALLOUT:
1488      case OP_CREF:      case OP_CREF:
1489        case OP_NCREF:
1490      case OP_RREF:      case OP_RREF:
1491        case OP_NRREF:
1492      case OP_DEF:      case OP_DEF:
1493      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1494      break;      break;
# Line 1116  for (;;) Line 1504  for (;;)
1504    
1505    
1506  /*************************************************  /*************************************************
1507  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1508  *************************************************/  *************************************************/
1509    
1510  /* 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,
1511  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.
1512  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
1513    temporarily terminated with OP_END when this function is called.
1514    
1515    This function is called when a backward assertion is encountered, so that if it
1516    fails, the error message can point to the correct place in the pattern.
1517    However, we cannot do this when the assertion contains subroutine calls,
1518    because they can be forward references. We solve this by remembering this case
1519    and doing the check at the end; a flag specifies which mode we are running in.
1520    
1521  Arguments:  Arguments:
1522    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1523    options  the compiling options    utf8     TRUE in UTF-8 mode
1524      atend    TRUE if called when the pattern is complete
1525      cd       the "compile data" structure
1526    
1527  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1528                 or -1 if there is no fixed length,
1529               or -2 if \C was encountered               or -2 if \C was encountered
1530                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1531  */  */
1532    
1533  static int  static int
1534  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, BOOL utf8, BOOL atend, compile_data *cd)
1535  {  {
1536  int length = -1;  int length = -1;
1537    
# Line 1145  branch, check the length against that of Line 1544  branch, check the length against that of
1544  for (;;)  for (;;)
1545    {    {
1546    int d;    int d;
1547      uschar *ce, *cs;
1548    register int op = *cc;    register int op = *cc;
1549    switch (op)    switch (op)
1550      {      {
1551        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1552        OP_BRA (normal non-capturing bracket) because the other variants of these
1553        opcodes are all concerned with unlimited repeated groups, which of course
1554        are not of fixed length. They will cause a -1 response from the default
1555        case of this switch. */
1556    
1557      case OP_CBRA:      case OP_CBRA:
1558      case OP_BRA:      case OP_BRA:
1559      case OP_ONCE:      case OP_ONCE:
1560        case OP_ONCE_NC:
1561      case OP_COND:      case OP_COND:
1562      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), utf8, atend, cd);
1563      if (d < 0) return d;      if (d < 0) return d;
1564      branchlength += d;      branchlength += d;
1565      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1161  for (;;) Line 1568  for (;;)
1568    
1569      /* 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
1570      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
1571      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.
1572        Note that we must not include the OP_KETRxxx opcodes here, because they
1573        all imply an unlimited repeat. */
1574    
1575      case OP_ALT:      case OP_ALT:
1576      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1577      case OP_END:      case OP_END:
1578      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1579        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
# Line 1175  for (;;) Line 1582  for (;;)
1582      branchlength = 0;      branchlength = 0;
1583      break;      break;
1584    
1585        /* A true recursion implies not fixed length, but a subroutine call may
1586        be OK. If the subroutine is a forward reference, we can't deal with
1587        it until the end of the pattern, so return -3. */
1588    
1589        case OP_RECURSE:
1590        if (!atend) return -3;
1591        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1592        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1593        if (cc > cs && cc < ce) return -1;                /* Recursion */
1594        d = find_fixedlength(cs + 2, utf8, atend, cd);
1595        if (d < 0) return d;
1596        branchlength += d;
1597        cc += 1 + LINK_SIZE;
1598        break;
1599    
1600      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1601    
1602      case OP_ASSERT:      case OP_ASSERT:
# Line 1188  for (;;) Line 1610  for (;;)
1610    
1611      case OP_REVERSE:      case OP_REVERSE:
1612      case OP_CREF:      case OP_CREF:
1613        case OP_NCREF:
1614      case OP_RREF:      case OP_RREF:
1615        case OP_NRREF:
1616      case OP_DEF:      case OP_DEF:
     case OP_OPT:  
1617      case OP_CALLOUT:      case OP_CALLOUT:
1618      case OP_SOD:      case OP_SOD:
1619      case OP_SOM:      case OP_SOM:
1620        case OP_SET_SOM:
1621      case OP_EOD:      case OP_EOD:
1622      case OP_EODN:      case OP_EODN:
1623      case OP_CIRC:      case OP_CIRC:
1624        case OP_CIRCM:
1625      case OP_DOLL:      case OP_DOLL:
1626        case OP_DOLLM:
1627      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1628      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1629      cc += _pcre_OP_lengths[*cc];      cc += _pcre_OP_lengths[*cc];
# Line 1206  for (;;) Line 1632  for (;;)
1632      /* Handle literal characters */      /* Handle literal characters */
1633    
1634      case OP_CHAR:      case OP_CHAR:
1635      case OP_CHARNC:      case OP_CHARI:
1636      case OP_NOT:      case OP_NOT:
1637        case OP_NOTI:
1638      branchlength++;      branchlength++;
1639      cc += 2;      cc += 2;
1640  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1641      if ((options & PCRE_UTF8) != 0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       {  
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1642  #endif  #endif
1643      break;      break;
1644    
# Line 1225  for (;;) Line 1649  for (;;)
1649      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1650      cc += 4;      cc += 4;
1651  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1652      if ((options & PCRE_UTF8) != 0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       {  
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1653  #endif  #endif
1654      break;      break;
1655    
# Line 1252  for (;;) Line 1673  for (;;)
1673      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1674      case OP_WORDCHAR:      case OP_WORDCHAR:
1675      case OP_ANY:      case OP_ANY:
1676        case OP_ALLANY:
1677      branchlength++;      branchlength++;
1678      cc++;      cc++;
1679      break;      break;
# Line 1306  for (;;) Line 1728  for (;;)
1728    
1729    
1730  /*************************************************  /*************************************************
1731  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1732  *************************************************/  *************************************************/
1733    
1734  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1735  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1736    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1737    so that it can be called from pcre_study() when finding the minimum matching
1738    length.
1739    
1740  Arguments:  Arguments:
1741    code        points to start of expression    code        points to start of expression
1742    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1743    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1744    
1745  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
1746  */  */
1747    
1748  static const uschar *  const uschar *
1749  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1750  {  {
1751  for (;;)  for (;;)
1752    {    {
1753    register int c = *code;    register int c = *code;
1754    
1755    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1756    
1757    /* 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 1760  for (;;)
1760    
1761    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1762    
1763      /* Handle recursion */
1764    
1765      else if (c == OP_REVERSE)
1766        {
1767        if (number < 0) return (uschar *)code;
1768        code += _pcre_OP_lengths[c];
1769        }
1770    
1771    /* Handle capturing bracket */    /* Handle capturing bracket */
1772    
1773    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
1774               c == OP_CBRAPOS || c == OP_SCBRAPOS)
1775      {      {
1776      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
1777      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
# Line 1345  for (;;) Line 1780  for (;;)
1780    
1781    /* 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
1782    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
1783    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1784      must add in its length. */
1785    
1786    else    else
1787      {      {
# Line 1369  for (;;) Line 1805  for (;;)
1805        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
1806        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1807        break;        break;
1808    
1809          case OP_MARK:
1810          case OP_PRUNE_ARG:
1811          case OP_SKIP_ARG:
1812          code += code[1];
1813          break;
1814    
1815          case OP_THEN_ARG:
1816          code += code[1];
1817          break;
1818        }        }
1819    
1820      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1383  for (;;) Line 1829  for (;;)
1829      if (utf8) switch(c)      if (utf8) switch(c)
1830        {        {
1831        case OP_CHAR:        case OP_CHAR:
1832        case OP_CHARNC:        case OP_CHARI:
1833        case OP_EXACT:        case OP_EXACT:
1834          case OP_EXACTI:
1835        case OP_UPTO:        case OP_UPTO:
1836          case OP_UPTOI:
1837        case OP_MINUPTO:        case OP_MINUPTO:
1838          case OP_MINUPTOI:
1839        case OP_POSUPTO:        case OP_POSUPTO:
1840          case OP_POSUPTOI:
1841        case OP_STAR:        case OP_STAR:
1842          case OP_STARI:
1843        case OP_MINSTAR:        case OP_MINSTAR:
1844          case OP_MINSTARI:
1845        case OP_POSSTAR:        case OP_POSSTAR:
1846          case OP_POSSTARI:
1847        case OP_PLUS:        case OP_PLUS:
1848          case OP_PLUSI:
1849        case OP_MINPLUS:        case OP_MINPLUS:
1850          case OP_MINPLUSI:
1851        case OP_POSPLUS:        case OP_POSPLUS:
1852          case OP_POSPLUSI:
1853        case OP_QUERY:        case OP_QUERY:
1854          case OP_QUERYI:
1855        case OP_MINQUERY:        case OP_MINQUERY:
1856          case OP_MINQUERYI:
1857        case OP_POSQUERY:        case OP_POSQUERY:
1858          case OP_POSQUERYI:
1859        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1860        break;        break;
1861        }        }
1862    #else
1863        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1864  #endif  #endif
1865      }      }
1866    }    }
# Line 1438  for (;;) Line 1899  for (;;)
1899    
1900    /* 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
1901    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
1902    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1903      must add in its length. */
1904    
1905    else    else
1906      {      {
# Line 1462  for (;;) Line 1924  for (;;)
1924        case OP_TYPEEXACT:        case OP_TYPEEXACT:
1925        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1926        break;        break;
1927    
1928          case OP_MARK:
1929          case OP_PRUNE_ARG:
1930          case OP_SKIP_ARG:
1931          code += code[1];
1932          break;
1933    
1934          case OP_THEN_ARG:
1935          code += code[1];
1936          break;
1937        }        }
1938    
1939      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1476  for (;;) Line 1948  for (;;)
1948      if (utf8) switch(c)      if (utf8) switch(c)
1949        {        {
1950        case OP_CHAR:        case OP_CHAR:
1951        case OP_CHARNC:        case OP_CHARI:
1952        case OP_EXACT:        case OP_EXACT:
1953          case OP_EXACTI:
1954        case OP_UPTO:        case OP_UPTO:
1955          case OP_UPTOI:
1956        case OP_MINUPTO:        case OP_MINUPTO:
1957          case OP_MINUPTOI:
1958        case OP_POSUPTO:        case OP_POSUPTO:
1959          case OP_POSUPTOI:
1960        case OP_STAR:        case OP_STAR:
1961          case OP_STARI:
1962        case OP_MINSTAR:        case OP_MINSTAR:
1963          case OP_MINSTARI:
1964        case OP_POSSTAR:        case OP_POSSTAR:
1965          case OP_POSSTARI:
1966        case OP_PLUS:        case OP_PLUS:
1967          case OP_PLUSI:
1968        case OP_MINPLUS:        case OP_MINPLUS:
1969          case OP_MINPLUSI:
1970        case OP_POSPLUS:        case OP_POSPLUS:
1971          case OP_POSPLUSI:
1972        case OP_QUERY:        case OP_QUERY:
1973          case OP_QUERYI:
1974        case OP_MINQUERY:        case OP_MINQUERY:
1975          case OP_MINQUERYI:
1976        case OP_POSQUERY:        case OP_POSQUERY:
1977          case OP_POSQUERYI:
1978        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1979        break;        break;
1980        }        }
1981    #else
1982        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1983  #endif  #endif
1984      }      }
1985    }    }
# Line 1508  for (;;) Line 1995  for (;;)
1995  can match the empty string or not. It is called from could_be_empty()  can match the empty string or not. It is called from could_be_empty()
1996  below and from compile_branch() when checking for an unlimited repeat of a  below and from compile_branch() when checking for an unlimited repeat of a
1997  group that can match nothing. Note that first_significant_code() skips over  group that can match nothing. Note that first_significant_code() skips over
1998  assertions. If we hit an unclosed bracket, we return "empty" - this means we've  backward and negative forward assertions when its final argument is TRUE. If we
1999  struck an inner bracket whose current branch will already have been scanned.  hit an unclosed bracket, we return "empty" - this means we've struck an inner
2000    bracket whose current branch will already have been scanned.
2001    
2002  Arguments:  Arguments:
2003    code        points to start of search    code        points to start of search
2004    endcode     points to where to stop    endcode     points to where to stop
2005    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
2006      cd          contains pointers to tables etc.
2007    
2008  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2009  */  */
2010    
2011  static BOOL  static BOOL
2012  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
2013      compile_data *cd)
2014  {  {
2015  register int c;  register int c;
2016  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], TRUE);
2017       code < endcode;       code < endcode;
2018       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], TRUE))
2019    {    {
2020    const uschar *ccode;    const uschar *ccode;
2021    
2022    c = *code;    c = *code;
2023    
2024    /* Groups with zero repeats can of course be empty; skip them. */    /* Skip over forward assertions; the other assertions are skipped by
2025      first_significant_code() with a TRUE final argument. */
2026    
2027    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_ASSERT)
2028      {      {
     code += _pcre_OP_lengths[c];  
2029      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
2030      c = *code;      c = *code;
2031      continue;      continue;
2032      }      }
2033    
2034    /* For other groups, scan the branches. */    /* For a recursion/subroutine call, if its end has been reached, which
2035      implies a backward reference subroutine call, we can scan it. If it's a
2036      forward reference subroutine call, we can't. To detect forward reference
2037      we have to scan up the list that is kept in the workspace. This function is
2038      called only when doing the real compile, not during the pre-compile that
2039      measures the size of the compiled pattern. */
2040    
2041    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_RECURSE)
2042      {      {
2043        const uschar *scode;
2044      BOOL empty_branch;      BOOL empty_branch;
     if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */  
2045    
2046      /* Scan a closed bracket */      /* Test for forward reference */
2047    
2048        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
2049          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
2050    
2051        /* Not a forward reference, test for completed backward reference */
2052    
2053      empty_branch = FALSE;      empty_branch = FALSE;
2054        scode = cd->start_code + GET(code, 1);
2055        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2056    
2057        /* Completed backwards reference */
2058    
2059      do      do
2060        {        {
2061        if (!empty_branch && could_be_empty_branch(code, endcode, utf8))        if (could_be_empty_branch(scode, endcode, utf8, cd))
2062            {
2063          empty_branch = TRUE;          empty_branch = TRUE;
2064            break;
2065            }
2066          scode += GET(scode, 1);
2067          }
2068        while (*scode == OP_ALT);
2069    
2070        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2071        continue;
2072        }
2073    
2074      /* Groups with zero repeats can of course be empty; skip them. */
2075    
2076      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2077          c == OP_BRAPOSZERO)
2078        {
2079        code += _pcre_OP_lengths[c];
2080        do code += GET(code, 1); while (*code == OP_ALT);
2081        c = *code;
2082        continue;
2083        }
2084    
2085      /* A nested group that is already marked as "could be empty" can just be
2086      skipped. */
2087    
2088      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2089          c == OP_SCBRA || c == OP_SCBRAPOS)
2090        {
2091        do code += GET(code, 1); while (*code == OP_ALT);
2092        c = *code;
2093        continue;
2094        }
2095    
2096      /* For other groups, scan the branches. */
2097    
2098      if (c == OP_BRA  || c == OP_BRAPOS ||
2099          c == OP_CBRA || c == OP_CBRAPOS ||
2100          c == OP_ONCE || c == OP_ONCE_NC ||
2101          c == OP_COND)
2102        {
2103        BOOL empty_branch;
2104        if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
2105    
2106        /* If a conditional group has only one branch, there is a second, implied,
2107        empty branch, so just skip over the conditional, because it could be empty.
2108        Otherwise, scan the individual branches of the group. */
2109    
2110        if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
2111        code += GET(code, 1);        code += GET(code, 1);
2112        else
2113          {
2114          empty_branch = FALSE;
2115          do
2116            {
2117            if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
2118              empty_branch = TRUE;
2119            code += GET(code, 1);
2120            }
2121          while (*code == OP_ALT);
2122          if (!empty_branch) return FALSE;   /* All branches are non-empty */
2123        }        }
2124      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
2125      c = *code;      c = *code;
2126      continue;      continue;
2127      }      }
# Line 1619  for (code = first_significant_code(code Line 2182  for (code = first_significant_code(code
2182      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2183      case OP_WORDCHAR:      case OP_WORDCHAR:
2184      case OP_ANY:      case OP_ANY:
2185        case OP_ALLANY:
2186      case OP_ANYBYTE:      case OP_ANYBYTE:
2187      case OP_CHAR:      case OP_CHAR:
2188      case OP_CHARNC:      case OP_CHARI:
2189      case OP_NOT:      case OP_NOT:
2190        case OP_NOTI:
2191      case OP_PLUS:      case OP_PLUS:
2192      case OP_MINPLUS:      case OP_MINPLUS:
2193      case OP_POSPLUS:      case OP_POSPLUS:
# Line 1662  for (code = first_significant_code(code Line 2227  for (code = first_significant_code(code
2227      case OP_KET:      case OP_KET:
2228      case OP_KETRMAX:      case OP_KETRMAX:
2229      case OP_KETRMIN:      case OP_KETRMIN:
2230        case OP_KETRPOS:
2231      case OP_ALT:      case OP_ALT:
2232      return TRUE;      return TRUE;
2233    
# Line 1670  for (code = first_significant_code(code Line 2236  for (code = first_significant_code(code
2236    
2237  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2238      case OP_STAR:      case OP_STAR:
2239        case OP_STARI:
2240      case OP_MINSTAR:      case OP_MINSTAR:
2241        case OP_MINSTARI:
2242      case OP_POSSTAR:      case OP_POSSTAR:
2243        case OP_POSSTARI:
2244      case OP_QUERY:      case OP_QUERY:
2245        case OP_QUERYI:
2246      case OP_MINQUERY:      case OP_MINQUERY:
2247        case OP_MINQUERYI:
2248      case OP_POSQUERY:      case OP_POSQUERY:
2249        case OP_POSQUERYI:
2250        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2251        break;
2252    
2253      case OP_UPTO:      case OP_UPTO:
2254        case OP_UPTOI:
2255      case OP_MINUPTO:      case OP_MINUPTO:
2256        case OP_MINUPTOI:
2257      case OP_POSUPTO:      case OP_POSUPTO:
2258      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_POSUPTOI:
2259        if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2260      break;      break;
2261  #endif  #endif
2262    
2263        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2264        string. */
2265    
2266        case OP_MARK:
2267        case OP_PRUNE_ARG:
2268        case OP_SKIP_ARG:
2269        code += code[1];
2270        break;
2271    
2272        case OP_THEN_ARG:
2273        code += code[1];
2274        break;
2275    
2276        /* None of the remaining opcodes are required to match a character. */
2277    
2278        default:
2279        break;
2280      }      }
2281    }    }
2282    
# Line 1697  return TRUE; Line 2293  return TRUE;
2293  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
2294  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,
2295  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.
2296    This function is called only during the real compile, not during the
2297    pre-compile.
2298    
2299  Arguments:  Arguments:
2300    code        points to start of the recursion    code        points to start of the recursion
2301    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2302    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2303    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2304      cd          pointers to tables etc
2305    
2306  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2307  */  */
2308    
2309  static BOOL  static BOOL
2310  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2311    BOOL utf8)    BOOL utf8, compile_data *cd)
2312  {  {
2313  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2314    {    {
2315    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2316        return FALSE;
2317    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2318    }    }
2319  return TRUE;  return TRUE;
# Line 1726  return TRUE; Line 2326  return TRUE;
2326  *************************************************/  *************************************************/
2327    
2328  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2329  encountered in a character class. It checks whether this is followed by an  encountered in a character class. It checks whether this is followed by a
2330  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2331  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2332    
2333    Originally, this function only recognized a sequence of letters between the
2334    terminators, but it seems that Perl recognizes any sequence of characters,
2335    though of course unknown POSIX names are subsequently rejected. Perl gives an
2336    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2337    didn't consider this to be a POSIX class. Likewise for [:1234:].
2338    
2339    The problem in trying to be exactly like Perl is in the handling of escapes. We
2340    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2341    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2342    below handles the special case of \], but does not try to do any other escape
2343    processing. This makes it different from Perl for cases such as [:l\ower:]
2344    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2345    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2346    I think.
2347    
2348    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2349    It seems that the appearance of a nested POSIX class supersedes an apparent
2350    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2351    a digit.
2352    
2353    In Perl, unescaped square brackets may also appear as part of class names. For
2354    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2355    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2356    seem right at all. PCRE does not allow closing square brackets in POSIX class
2357    names.
2358    
2359  Argument:  Arguments:
2360    ptr      pointer to the initial [    ptr      pointer to the initial [
2361    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2362    
2363  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2364  */  */
2365    
2366  static BOOL  static BOOL
2367  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
2368  {  {
2369  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2370  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2371  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2372    {    {
2373    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2374    return TRUE;      ptr++;
2375      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2376      else
2377        {
2378        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2379          {
2380          *endptr = ptr;
2381          return TRUE;
2382          }
2383        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2384             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2385              ptr[1] == CHAR_EQUALS_SIGN) &&
2386            check_posix_syntax(ptr, endptr))
2387          return FALSE;
2388        }
2389    }    }
2390  return FALSE;  return FALSE;
2391  }  }
# Line 1794  return -1; Line 2431  return -1;
2431  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2432  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2433  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
2434  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
2435  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
2436  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
2437  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
2438  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2439    OP_END.
2440    
2441  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2442  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 1875  auto_callout(uschar *code, const uschar Line 2513  auto_callout(uschar *code, const uschar
2513  {  {
2514  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2515  *code++ = 255;  *code++ = 255;
2516  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2517  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2518  return code + 2*LINK_SIZE;  return code + 2*LINK_SIZE;
2519  }  }
2520    
# Line 1901  Returns:             nothing Line 2539  Returns:             nothing
2539  static void  static void
2540  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)
2541  {  {
2542  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2543  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2544  }  }
2545    
# Line 1933  get_othercase_range(unsigned int *cptr, Line 2571  get_othercase_range(unsigned int *cptr,
2571  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2572    
2573  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2574    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2575    
2576  if (c > d) return FALSE;  if (c > d) return FALSE;
2577    
# Line 1942  next = othercase + 1; Line 2580  next = othercase + 1;
2580    
2581  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2582    {    {
2583    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2584    next++;    next++;
2585    }    }
2586    
# Line 1951  for (++c; c <= d; c++) Line 2589  for (++c; c <= d; c++)
2589    
2590  return TRUE;  return TRUE;
2591  }  }
2592    
2593    
2594    
2595    /*************************************************
2596    *        Check a character and a property        *
2597    *************************************************/
2598    
2599    /* This function is called by check_auto_possessive() when a property item
2600    is adjacent to a fixed character.
2601    
2602    Arguments:
2603      c            the character
2604      ptype        the property type
2605      pdata        the data for the type
2606      negated      TRUE if it's a negated property (\P or \p{^)
2607    
2608    Returns:       TRUE if auto-possessifying is OK
2609    */
2610    
2611    static BOOL
2612    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2613    {
2614    const ucd_record *prop = GET_UCD(c);
2615    switch(ptype)
2616      {
2617      case PT_LAMP:
2618      return (prop->chartype == ucp_Lu ||
2619              prop->chartype == ucp_Ll ||
2620              prop->chartype == ucp_Lt) == negated;
2621    
2622      case PT_GC:
2623      return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;
2624    
2625      case PT_PC:
2626      return (pdata == prop->chartype) == negated;
2627    
2628      case PT_SC:
2629      return (pdata == prop->script) == negated;
2630    
2631      /* These are specials */
2632    
2633      case PT_ALNUM:
2634      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2635              _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;
2636    
2637      case PT_SPACE:    /* Perl space */
2638      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2639              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2640              == negated;
2641    
2642      case PT_PXSPACE:  /* POSIX space */
2643      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2644              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2645              c == CHAR_FF || c == CHAR_CR)
2646              == negated;
2647    
2648      case PT_WORD:
2649      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2650              _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2651              c == CHAR_UNDERSCORE) == negated;
2652      }
2653    return FALSE;
2654    }
2655  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2656    
2657    
# Line 1964  whether the next thing could possibly ma Line 2665  whether the next thing could possibly ma
2665  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
2666    
2667  Arguments:  Arguments:
2668    op_code       the repeated op code    previous      pointer to the repeated opcode
   this          data for this item, depends on the opcode  
2669    utf8          TRUE in UTF-8 mode    utf8          TRUE in UTF-8 mode
   utf8_char     used for utf8 character bytes, NULL if not relevant  
2670    ptr           next character in pattern    ptr           next character in pattern
2671    options       options bits    options       options bits
2672    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 1976  Returns:        TRUE if possessifying is Line 2675  Returns:        TRUE if possessifying is
2675  */  */
2676    
2677  static BOOL  static BOOL
2678  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
2679    const uschar *ptr, int options, compile_data *cd)    int options, compile_data *cd)
2680  {  {
2681  int next;  int c, next;
2682    int op_code = *previous++;
2683    
2684  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
2685    
# Line 1988  if ((options & PCRE_EXTENDED) != 0) Line 2688  if ((options & PCRE_EXTENDED) != 0)
2688    for (;;)    for (;;)
2689      {      {
2690      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2691      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2692        {        {
2693        while (*(++ptr) != 0)        ptr++;
2694          while (*ptr != 0)
2695            {
2696          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2697            ptr++;
2698    #ifdef SUPPORT_UTF8
2699            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2700    #endif
2701            }
2702        }        }
2703      else break;      else break;
2704      }      }
# Line 2000  if ((options & PCRE_EXTENDED) != 0) Line 2707  if ((options & PCRE_EXTENDED) != 0)
2707  /* 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
2708  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2709    
2710  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2711    {    {
2712    int temperrorcode = 0;    int temperrorcode = 0;
2713    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
# Line 2025  if ((options & PCRE_EXTENDED) != 0) Line 2732  if ((options & PCRE_EXTENDED) != 0)
2732    for (;;)    for (;;)
2733      {      {
2734      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2735      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2736        {        {
2737        while (*(++ptr) != 0)        ptr++;
2738          while (*ptr != 0)
2739            {
2740          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2741            ptr++;
2742    #ifdef SUPPORT_UTF8
2743            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2744    #endif
2745            }
2746        }        }
2747      else break;      else break;
2748      }      }
# Line 2036  if ((options & PCRE_EXTENDED) != 0) Line 2750  if ((options & PCRE_EXTENDED) != 0)
2750    
2751  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
2752    
2753  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2754    return FALSE;    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2755        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. */  
   
2756    
2757  /* Handle cases when the next item is a character. */  /* Now compare the next item with the previous opcode. First, handle cases when
2758    the next item is a character. */
2759    
2760  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
2761    {    {
2762    case OP_CHAR:    case OP_CHAR:
2763  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2764    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2765    #else
2766      c = *previous;
2767  #endif  #endif
2768    return item != next;    return c != next;
2769    
2770    /* 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
2771    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
2772    high-valued characters. */    high-valued characters. */
2773    
2774    case OP_CHARNC:    case OP_CHARI:
2775  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2776    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2777    #else
2778      c = *previous;
2779  #endif  #endif
2780    if (item == next) return FALSE;    if (c == next) return FALSE;
2781  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2782    if (utf8)    if (utf8)
2783      {      {
2784      unsigned int othercase;      unsigned int othercase;
2785      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2786  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2787      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
2788  #else  #else
2789      othercase = NOTACHAR;      othercase = NOTACHAR;
2790  #endif  #endif
2791      return (unsigned int)item != othercase;      return (unsigned int)c != othercase;
2792      }      }
2793    else    else
2794  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2795    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2796    
2797    /* 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
2798      opcodes are not used for multi-byte characters, because they are coded using
2799      an XCLASS instead. */
2800    
2801    case OP_NOT:    case OP_NOT:
2802    if (next < 0) return FALSE;  /* Not a character */    return (c = *previous) == next;
2803    if (item == next) return TRUE;  
2804    if ((options & PCRE_CASELESS) == 0) return FALSE;    case OP_NOTI:
2805      if ((c = *previous) == next) return TRUE;
2806  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2807    if (utf8)    if (utf8)
2808      {      {
2809      unsigned int othercase;      unsigned int othercase;
2810      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2811  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2812      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
2813  #else  #else
2814      othercase = NOTACHAR;      othercase = NOTACHAR;
2815  #endif  #endif
2816      return (unsigned int)item == othercase;      return (unsigned int)c == othercase;
2817      }      }
2818    else    else
2819  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2820    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2821    
2822      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2823      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
2824    
2825    case OP_DIGIT:    case OP_DIGIT:
2826    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
# Line 2143  if (next >= 0) switch(op_code) Line 2863  if (next >= 0) switch(op_code)
2863      case 0x202f:      case 0x202f:
2864      case 0x205f:      case 0x205f:
2865      case 0x3000:      case 0x3000:
2866      return op_code != OP_HSPACE;      return op_code == OP_NOT_HSPACE;
2867      default:      default:
2868      return op_code == OP_HSPACE;      return op_code != OP_NOT_HSPACE;
2869      }      }
2870    
2871      case OP_ANYNL:
2872    case OP_VSPACE:    case OP_VSPACE:
2873    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
2874    switch(next)    switch(next)
# Line 2159  if (next >= 0) switch(op_code) Line 2880  if (next >= 0) switch(op_code)
2880      case 0x85:      case 0x85:
2881      case 0x2028:      case 0x2028:
2882      case 0x2029:      case 0x2029:
2883      return op_code != OP_VSPACE;      return op_code == OP_NOT_VSPACE;
2884      default:      default:
2885      return op_code == OP_VSPACE;      return op_code != OP_NOT_VSPACE;
2886      }      }
2887    
2888    #ifdef SUPPORT_UCP
2889      case OP_PROP:
2890      return check_char_prop(next, previous[0], previous[1], FALSE);
2891    
2892      case OP_NOTPROP:
2893      return check_char_prop(next, previous[0], previous[1], TRUE);
2894    #endif
2895    
2896    default:    default:
2897    return FALSE;    return FALSE;
2898    }    }
2899    
2900    
2901  /* 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
2902    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
2903    generated only when PCRE_UCP is *not* set, that is, when only ASCII
2904    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
2905    replaced by OP_PROP codes when PCRE_UCP is set. */
2906    
2907  switch(op_code)  switch(op_code)
2908    {    {
2909    case OP_CHAR:    case OP_CHAR:
2910    case OP_CHARNC:    case OP_CHARI:
2911  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2912    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2913    #else
2914      c = *previous;
2915  #endif  #endif
2916    switch(-next)    switch(-next)
2917      {      {
2918      case ESC_d:      case ESC_d:
2919      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
2920    
2921      case ESC_D:      case ESC_D:
2922      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
2923    
2924      case ESC_s:      case ESC_s:
2925      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
2926    
2927      case ESC_S:      case ESC_S:
2928      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
2929    
2930      case ESC_w:      case ESC_w:
2931      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
2932    
2933      case ESC_W:      case ESC_W:
2934      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
2935    
2936      case ESC_h:      case ESC_h:
2937      case ESC_H:      case ESC_H:
2938      switch(item)      switch(c)
2939        {        {
2940        case 0x09:        case 0x09:
2941        case 0x20:        case 0x20:
# Line 2228  switch(op_code) Line 2963  switch(op_code)
2963    
2964      case ESC_v:      case ESC_v:
2965      case ESC_V:      case ESC_V:
2966      switch(item)      switch(c)
2967        {        {
2968        case 0x0a:        case 0x0a:
2969        case 0x0b:        case 0x0b:
# Line 2242  switch(op_code) Line 2977  switch(op_code)
2977        return -next == ESC_v;        return -next == ESC_v;
2978        }        }
2979    
2980        /* When PCRE_UCP is set, these values get generated for \d etc. Find
2981        their substitutions and process them. The result will always be either
2982        -ESC_p or -ESC_P. Then fall through to process those values. */
2983    
2984    #ifdef SUPPORT_UCP
2985        case ESC_du:
2986        case ESC_DU:
2987        case ESC_wu:
2988        case ESC_WU:
2989        case ESC_su:
2990        case ESC_SU:
2991          {
2992          int temperrorcode = 0;
2993          ptr = substitutes[-next - ESC_DU];
2994          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
2995          if (temperrorcode != 0) return FALSE;
2996          ptr++;    /* For compatibility */
2997          }
2998        /* Fall through */
2999    
3000        case ESC_p:
3001        case ESC_P:
3002          {
3003          int ptype, pdata, errorcodeptr;
3004          BOOL negated;
3005    
3006          ptr--;      /* Make ptr point at the p or P */
3007          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
3008          if (ptype < 0) return FALSE;
3009          ptr++;      /* Point past the final curly ket */
3010    
3011          /* If the property item is optional, we have to give up. (When generated
3012          from \d etc by PCRE_UCP, this test will have been applied much earlier,
3013          to the original \d etc. At this point, ptr will point to a zero byte. */
3014    
3015          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3016            strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3017              return FALSE;
3018    
3019          /* Do the property check. */
3020    
3021          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
3022          }
3023    #endif
3024    
3025      default:      default:
3026      return FALSE;      return FALSE;
3027      }      }
3028    
3029      /* In principle, support for Unicode properties should be integrated here as
3030      well. It means re-organizing the above code so as to get hold of the property
3031      values before switching on the op-code. However, I wonder how many patterns
3032      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
3033      these op-codes are never generated.) */
3034    
3035    case OP_DIGIT:    case OP_DIGIT:
3036    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
3037           next == -ESC_h || next == -ESC_v;           next == -ESC_h || next == -ESC_v || next == -ESC_R;
3038    
3039    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
3040    return next == -ESC_d;    return next == -ESC_d;
3041    
3042    case OP_WHITESPACE:    case OP_WHITESPACE:
3043    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
3044    
3045    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
3046    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
3047    
3048    case OP_HSPACE:    case OP_HSPACE:
3049    return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
3050             next == -ESC_w || next == -ESC_v || next == -ESC_R;
3051    
3052    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
3053    return next == -ESC_h;    return next == -ESC_h;
3054    
3055    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
3056      case OP_ANYNL:
3057    case OP_VSPACE:    case OP_VSPACE:
3058    return next == -ESC_V || next == -ESC_d || next == -ESC_w;    return next == -ESC_V || next == -ESC_d || next == -ESC_w;
3059    
3060    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
3061    return next == -ESC_v;    return next == -ESC_v || next == -ESC_R;
3062    
3063    case OP_WORDCHAR:    case OP_WORDCHAR:
3064    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
3065             next == -ESC_v || next == -ESC_R;
3066    
3067    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
3068    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2305  Arguments: Line 3094  Arguments:
3094    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
3095    reqbyteptr     set to the last literal character required, else < 0    reqbyteptr     set to the last literal character required, else < 0
3096    bcptr          points to current branch chain    bcptr          points to current branch chain
3097      cond_depth     conditional nesting depth
3098    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
3099    lengthptr      NULL during the real compile phase    lengthptr      NULL during the real compile phase
3100                   points to length accumulator during pre-compile phase                   points to length accumulator during pre-compile phase
# Line 2316  Returns:         TRUE on success Line 3106  Returns:         TRUE on success
3106  static BOOL  static BOOL
3107  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
3108    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
3109    compile_data *cd, int *lengthptr)    int cond_depth, compile_data *cd, int *lengthptr)
3110  {  {
3111  int repeat_type, op_type;  int repeat_type, op_type;
3112  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
# Line 2325  int greedy_default, greedy_non_default; Line 3115  int greedy_default, greedy_non_default;
3115  int firstbyte, reqbyte;  int firstbyte, reqbyte;
3116  int zeroreqbyte, zerofirstbyte;  int zeroreqbyte, zerofirstbyte;
3117  int req_caseopt, reqvary, tempreqvary;  int req_caseopt, reqvary, tempreqvary;
3118  int options = *optionsptr;  int options = *optionsptr;               /* May change dynamically */
3119  int after_manual_callout = 0;  int after_manual_callout = 0;
3120  int length_prevgroup = 0;  int length_prevgroup = 0;
3121  register int c;  register int c;
# Line 2337  BOOL inescq = FALSE; Line 3127  BOOL inescq = FALSE;
3127  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
3128  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
3129  const uschar *tempptr;  const uschar *tempptr;
3130    const uschar *nestptr = NULL;
3131  uschar *previous = NULL;  uschar *previous = NULL;
3132  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
3133  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
3134  uschar classbits[32];  uschar classbits[32];
3135    
3136    /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3137    must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3138    dynamically as we process the pattern. */
3139    
3140  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3141  BOOL class_utf8;  BOOL class_utf8;
3142  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
3143  uschar *class_utf8data;  uschar *class_utf8data;
3144    uschar *class_utf8data_base;
3145  uschar utf8_char[6];  uschar utf8_char[6];
3146  #else  #else
3147  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
 uschar *utf8_char = NULL;  
3148  #endif  #endif
3149    
3150  #ifdef DEBUG  #ifdef PCRE_DEBUG
3151  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
3152  #endif  #endif
3153    
# Line 2385  req_caseopt = ((options & PCRE_CASELESS) Line 3180  req_caseopt = ((options & PCRE_CASELESS)
3180  for (;; ptr++)  for (;; ptr++)
3181    {    {
3182    BOOL negate_class;    BOOL negate_class;
3183    BOOL should_flip_negation;    BOOL should_flip_negation;
3184    BOOL possessive_quantifier;    BOOL possessive_quantifier;
3185    BOOL is_quantifier;    BOOL is_quantifier;
3186    BOOL is_recurse;    BOOL is_recurse;
# Line 2400  for (;; ptr++) Line 3195  for (;; ptr++)
3195    int subfirstbyte;    int subfirstbyte;
3196    int terminator;    int terminator;
3197    int mclength;    int mclength;
3198      int tempbracount;
3199    uschar mcbuffer[8];    uschar mcbuffer[8];
3200    
3201    /* Get next byte in the pattern */    /* Get next byte in the pattern */
3202    
3203    c = *ptr;    c = *ptr;
3204    
3205      /* If we are at the end of a nested substitution, revert to the outer level
3206      string. Nesting only happens one level deep. */
3207    
3208      if (c == 0 && nestptr != NULL)
3209        {
3210        ptr = nestptr;
3211        nestptr = NULL;
3212        c = *ptr;
3213        }
3214    
3215    /* 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
3216    previous cycle of this loop. */    previous cycle of this loop. */
3217    
3218    if (lengthptr != NULL)    if (lengthptr != NULL)
3219      {      {
3220  #ifdef DEBUG  #ifdef PCRE_DEBUG
3221      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3222  #endif  #endif
3223      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
3224        {        {
3225        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3226        goto FAILED;        goto FAILED;
# Line 2436  for (;; ptr++) Line 3242  for (;; ptr++)
3242        goto FAILED;        goto FAILED;
3243        }        }
3244    
3245      *lengthptr += code - last_code;      *lengthptr += (int)(code - last_code);
3246      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));
3247    
3248      /* 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 2463  for (;; ptr++) Line 3269  for (;; ptr++)
3269    /* 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
3270    reference list. */    reference list. */
3271    
3272    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
3273      {      {
3274      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3275      goto FAILED;      goto FAILED;
# Line 2473  for (;; ptr++) Line 3279  for (;; ptr++)
3279    
3280    if (inescq && c != 0)    if (inescq && c != 0)
3281      {      {
3282      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3283        {        {
3284        inescq = FALSE;        inescq = FALSE;
3285        ptr++;        ptr++;
# Line 2499  for (;; ptr++) Line 3305  for (;; ptr++)
3305    /* 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
3306    a quantifier. */    a quantifier. */
3307    
3308    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
3309      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
3310        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
3311    
3312    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
3313         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2510  for (;; ptr++) Line 3317  for (;; ptr++)
3317      previous_callout = NULL;      previous_callout = NULL;
3318      }      }
3319    
3320    /* In extended mode, skip white space and comments */    /* In extended mode, skip white space and comments. */
3321    
3322    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3323      {      {
3324      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
3325      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
3326        {        {
3327        while (*(++ptr) != 0)        ptr++;
3328          while (*ptr != 0)
3329          {          {
3330          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3331            ptr++;
3332    #ifdef SUPPORT_UTF8
3333            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
3334    #endif
3335          }          }
3336        if (*ptr != 0) continue;        if (*ptr != 0) continue;
3337    
# Line 2540  for (;; ptr++) Line 3352  for (;; ptr++)
3352      {      {
3353      /* ===================================================================*/      /* ===================================================================*/
3354      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
3355      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
3356      case ')':      case CHAR_RIGHT_PARENTHESIS:
3357      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
3358      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
3359      *codeptr = code;      *codeptr = code;
# Line 2553  for (;; ptr++) Line 3365  for (;; ptr++)
3365          *errorcodeptr = ERR20;          *errorcodeptr = ERR20;
3366          goto FAILED;          goto FAILED;
3367          }          }
3368        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += (int)(code - last_code);   /* To include callout length */
3369        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
3370        }        }
3371      return TRUE;      return TRUE;
# Line 2563  for (;; ptr++) Line 3375  for (;; ptr++)
3375      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
3376      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3377    
3378      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
3379        previous = NULL;
3380      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3381        {        {
3382        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3383          *code++ = OP_CIRCM;
3384        }        }
3385      previous = NULL;      else *code++ = OP_CIRC;
     *code++ = OP_CIRC;  
3386      break;      break;
3387    
3388      case '$':      case CHAR_DOLLAR_SIGN:
3389      previous = NULL;      previous = NULL;
3390      *code++ = OP_DOLL;      *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
3391      break;      break;
3392    
3393      /* 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
3394      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
3395    
3396      case '.':      case CHAR_DOT:
3397      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3398      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
3399      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3400      previous = code;      previous = code;
3401      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
3402      break;      break;
3403    
3404    
# Line 2600  for (;; ptr++) Line 3413  for (;; ptr++)
3413      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,
3414      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
3415      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.
     */  
3416    
3417      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
3418        default (Perl) mode, it is treated as a data character. */
3419    
3420        case CHAR_RIGHT_SQUARE_BRACKET:
3421        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3422          {
3423          *errorcodeptr = ERR64;
3424          goto FAILED;
3425          }
3426        goto NORMAL_CHAR;
3427    
3428        case CHAR_LEFT_SQUARE_BRACKET:
3429      previous = code;      previous = code;
3430    
3431      /* 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
3432      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. */
3433    
3434      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3435          check_posix_syntax(ptr, &tempptr, cd))           ptr[1] == CHAR_EQUALS_SIGN) &&
3436            check_posix_syntax(ptr, &tempptr))
3437        {        {
3438        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
3439        goto FAILED;        goto FAILED;
3440        }        }
3441    
# Line 2623  for (;; ptr++) Line 3447  for (;; ptr++)
3447      for (;;)      for (;;)
3448        {        {
3449        c = *(++ptr);        c = *(++ptr);
3450        if (c == '\\')        if (c == CHAR_BACKSLASH)
3451          {          {
3452          if (ptr[1] == 'E') ptr++;          if (ptr[1] == CHAR_E)
3453            else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;            ptr++;
3454              else break;          else if (strncmp((const char *)ptr+1,
3455                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
3456              ptr += 3;
3457            else
3458              break;
3459          }          }
3460        else if (!negate_class && c == '^')        else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
3461          negate_class = TRUE;          negate_class = TRUE;
3462        else break;        else break;
3463        }        }
3464    
3465      /* If a class contains a negative special such as \S, we need to flip the      /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
3466      negation flag at the end, so that support for characters > 255 works      an initial ']' is taken as a data character -- the code below handles
3467        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3468        [^] must match any character, so generate OP_ALLANY. */
3469    
3470        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3471            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3472          {
3473          *code++ = negate_class? OP_ALLANY : OP_FAIL;
3474          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3475          zerofirstbyte = firstbyte;
3476          break;
3477          }
3478    
3479        /* If a class contains a negative special such as \S, we need to flip the
3480        negation flag at the end, so that support for characters > 255 works
3481      correctly (they are all included in the class). */      correctly (they are all included in the class). */
3482    
3483      should_flip_negation = FALSE;      should_flip_negation = FALSE;
# Line 2657  for (;; ptr++) Line 3499  for (;; ptr++)
3499  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3500      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
3501      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
3502        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
3503  #endif  #endif
3504    
3505      /* Process characters until ] is reached. By writing this as a "do" it      /* Process characters until ] is reached. By writing this as a "do" it
# Line 2672  for (;; ptr++) Line 3515  for (;; ptr++)
3515          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3516          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3517          }          }
3518    
3519          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
3520          data and reset the pointer. This is so that very large classes that
3521          contain a zillion UTF-8 characters no longer overwrite the work space
3522          (which is on the stack). */
3523    
3524          if (lengthptr != NULL)
3525            {
3526            *lengthptr += class_utf8data - class_utf8data_base;
3527            class_utf8data = class_utf8data_base;
3528            }
3529    
3530  #endif  #endif
3531    
3532        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
3533    
3534        if (inescq)        if (inescq)
3535          {          {
3536          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3537            {            {
3538            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
3539            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2693  for (;; ptr++) Line 3548  for (;; ptr++)
3548        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3549        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3550    
3551        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3552            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3553            check_posix_syntax(ptr, &tempptr, cd))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3554          {          {
3555          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3556          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3557          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3558          uschar pbits[32];          uschar pbits[32];
3559    
3560          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3561            {            {
3562            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3563            goto FAILED;            goto FAILED;
3564            }            }
3565    
3566          ptr += 2;          ptr += 2;
3567          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3568            {            {
3569            local_negate = TRUE;            local_negate = TRUE;
3570            should_flip_negation = TRUE;  /* Note negative special */            should_flip_negation = TRUE;  /* Note negative special */
3571            ptr++;            ptr++;
3572            }            }
3573    
3574          posix_class = check_posix_name(ptr, tempptr - ptr);          posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3575          if (posix_class < 0)          if (posix_class < 0)
3576            {            {
3577            *errorcodeptr = ERR30;            *errorcodeptr = ERR30;
# Line 2730  for (;; ptr++) Line 3585  for (;; ptr++)
3585          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3586            posix_class = 0;            posix_class = 0;
3587    
3588          /* 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
3589          because we may be adding and subtracting from it, and we don't want to          different escape sequences that use Unicode properties. */
3590          subtract bits that may be in the main map already. At the end we or the  
3591          result into the bit map that is being built. */  #ifdef SUPPORT_UCP
3592            if ((options & PCRE_UCP) != 0)
3593              {
3594              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3595              if (posix_substitutes[pc] != NULL)
3596                {
3597                nestptr = tempptr + 1;
3598                ptr = posix_substitutes[pc] - 1;
3599                continue;
3600                }
3601              }
3602    #endif
3603            /* In the non-UCP case, we build the bit map for the POSIX class in a
3604            chunk of local store because we may be adding and subtracting from it,
3605            and we don't want to subtract bits that may be in the main map already.
3606            At the end we or the result into the bit map that is being built. */
3607    
3608          posix_class *= 3;          posix_class *= 3;
3609    
# Line 2777  for (;; ptr++) Line 3647  for (;; ptr++)
3647    
3648        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3649        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
3650        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
3651        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
3652        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
3653        character in them, so set class_charcount bigger than one. */        are either treated as literal characters (by default), or are faulted if
3654          PCRE_EXTRA is set. */
3655    
3656        if (c == '\\')        if (c == CHAR_BACKSLASH)
3657          {          {
3658          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3659          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3660    
3661          if (-c == ESC_b) c = '\b';       /* \b is backslash in a class */          if (-c == ESC_b) c = CHAR_BS;    /* \b is backspace in a class */
         else if (-c == ESC_X) c = 'X';   /* \X is literal X in a class */  
         else if (-c == ESC_R) c = 'R';   /* \R is literal R in a class */  
3662          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3663            {            {
3664            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3665              {              {
3666              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3667              }              }
# Line 2806  for (;; ptr++) Line 3675  for (;; ptr++)
3675            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
3676            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3677    
3678            /* Save time by not doing this in the pre-compile phase. */            switch (-c)
   
           if (lengthptr == NULL) switch (-c)  
3679              {              {
3680    #ifdef SUPPORT_UCP
3681                case ESC_du:     /* These are the values given for \d etc */
3682                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3683                case ESC_wu:     /* escape sequence with an appropriate \p */
3684                case ESC_WU:     /* or \P to test Unicode properties instead */
3685                case ESC_su:     /* of the default ASCII testing. */
3686                case ESC_SU:
3687                nestptr = ptr;
3688                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3689                class_charcount -= 2;                /* Undo! */
3690                continue;
3691    #endif
3692              case ESC_d:              case ESC_d:
3693              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3694              continue;              continue;
3695    
3696              case ESC_D:              case ESC_D:
3697              should_flip_negation = TRUE;              should_flip_negation = TRUE;
3698              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
3699              continue;              continue;
3700    
# Line 2824  for (;; ptr++) Line 3703  for (;; ptr++)
3703              continue;              continue;
3704    
3705              case ESC_W:              case ESC_W:
3706              should_flip_negation = TRUE;              should_flip_negation = TRUE;
3707              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3708              continue;              continue;
3709    
3710                /* Perl 5.004 onwards omits VT from \s, but we must preserve it
3711                if it was previously set by something earlier in the character
3712                class. */
3713    
3714              case ESC_s:              case ESC_s:
3715              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];              classbits[0] |= cbits[cbit_space];
3716              classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= cbits[cbit_space+1] & ~0x08;
3717                for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3718              continue;              continue;
3719    
3720              case ESC_S:              case ESC_S:
3721              should_flip_negation = TRUE;              should_flip_negation = TRUE;
3722              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
3723              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3724              continue;              continue;
3725    
3726              case ESC_E: /* Perl ignores an orphan \E */              case ESC_h:
             continue;  
   
             default:    /* Not recognized; fall through */  
             break;      /* Need "default" setting to stop compiler warning. */  
             }  
   
           /* In the pre-compile phase, just do the recognition. */  
   
           else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||  
                    c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;  
   
           /* We need to deal with \H, \h, \V, and \v in both phases because  
           they use extra memory. */  
   
           if (-c == ESC_h)  
             {  
3727              SETBIT(classbits, 0x09); /* VT */              SETBIT(classbits, 0x09); /* VT */
3728              SETBIT(classbits, 0x20); /* SPACE */              SETBIT(classbits, 0x20); /* SPACE */
3729              SETBIT(classbits, 0xa0); /* NSBP */              SETBIT(classbits, 0xa0); /* NSBP */
# Line 2879  for (;; ptr++) Line 3747  for (;; ptr++)
3747                }                }
3748  #endif  #endif
3749              continue;              continue;
             }  
3750    
3751            if (-c == ESC_H)              case ESC_H:
             {  
3752              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3753                {                {
3754                int x = 0xff;                int x = 0xff;
# Line 2924  for (;; ptr++) Line 3790  for (;; ptr++)
3790                }                }
3791  #endif  #endif
3792              continue;              continue;
             }  
3793    
3794            if (-c == ESC_v)              case ESC_v:
             {  
3795              SETBIT(classbits, 0x0a); /* LF */              SETBIT(classbits, 0x0a); /* LF */
3796              SETBIT(classbits, 0x0b); /* VT */              SETBIT(classbits, 0x0b); /* VT */
3797              SETBIT(classbits, 0x0c); /* FF */              SETBIT(classbits, 0x0c); /* FF */
# Line 2943  for (;; ptr++) Line 3807  for (;; ptr++)
3807                }                }
3808  #endif  #endif
3809              continue;              continue;
             }  
3810    
3811            if (-c == ESC_V)              case ESC_V:
             {  
3812              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3813                {                {
3814                int x = 0xff;                int x = 0xff;
# Line 2976  for (;; ptr++) Line 3838  for (;; ptr++)
3838                }                }
3839  #endif  #endif
3840              continue;              continue;
             }  
   
           /* We need to deal with \P and \p in both phases. */  
3841    
3842  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3843            if (-c == ESC_p || -c == ESC_P)              case ESC_p:
3844              {              case ESC_P:
3845              BOOL negated;                {
3846              int pdata;                BOOL negated;
3847              int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                int pdata;
3848              if (ptype < 0) goto FAILED;                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3849              class_utf8 = TRUE;                if (ptype < 0) goto FAILED;
3850              *class_utf8data++ = ((-c == ESC_p) != negated)?                class_utf8 = TRUE;
3851                XCL_PROP : XCL_NOTPROP;                *class_utf8data++ = ((-c == ESC_p) != negated)?
3852              *class_utf8data++ = ptype;                  XCL_PROP : XCL_NOTPROP;
3853              *class_utf8data++ = pdata;                *class_utf8data++ = ptype;
3854              class_charcount -= 2;   /* Not a < 256 character */                *class_utf8data++ = pdata;
3855              continue;                class_charcount -= 2;   /* Not a < 256 character */
3856              }                continue;
3857                  }
3858  #endif  #endif
3859            /* Unrecognized escapes are faulted if PCRE is running in its              /* Unrecognized escapes are faulted if PCRE is running in its
3860            strict mode. By default, for compatibility with Perl, they are              strict mode. By default, for compatibility with Perl, they are
3861            treated as literals. */              treated as literals. */
3862    
3863            if ((options & PCRE_EXTRA) != 0)              default:
3864              {              if ((options & PCRE_EXTRA) != 0)
3865              *errorcodeptr = ERR7;                {
3866              goto FAILED;                *errorcodeptr = ERR7;
3867                  goto FAILED;
3868                  }
3869                class_charcount -= 2;  /* Undo the default count from above */
3870                c = *ptr;              /* Get the final character and fall through */
3871                break;
3872              }              }
   
           class_charcount -= 2;  /* Undo the default count from above */  
           c = *ptr;              /* Get the final character and fall through */  
3873            }            }
3874    
3875          /* 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 3021  for (;; ptr++) Line 3883  for (;; ptr++)
3883        entirely. The code for handling \Q and \E is messy. */        entirely. The code for handling \Q and \E is messy. */
3884    
3885        CHECK_RANGE:        CHECK_RANGE:
3886        while (ptr[1] == '\\' && ptr[2] == 'E')        while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3887          {          {
3888          inescq = FALSE;          inescq = FALSE;
3889          ptr += 2;          ptr += 2;
# Line 3031  for (;; ptr++) Line 3893  for (;; ptr++)
3893    
3894        /* Remember \r or \n */        /* Remember \r or \n */
3895    
3896        if (c == '\r' || c == '\n') cd->external_flags |= PCRE_HASCRORLF;        if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3897    
3898        /* Check for range */        /* Check for range */
3899    
3900        if (!inescq && ptr[1] == '-')        if (!inescq && ptr[1] == CHAR_MINUS)
3901          {          {
3902          int d;          int d;
3903          ptr += 2;          ptr += 2;
3904          while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3905    
3906          /* 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
3907          mode. */          mode. */
3908    
3909          while (*ptr == '\\' && ptr[1] == 'Q')          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3910            {            {
3911            ptr += 2;            ptr += 2;
3912            if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }            if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3913                { ptr += 2; continue; }
3914            inescq = TRUE;            inescq = TRUE;
3915            break;            break;
3916            }            }
3917    
3918          if (*ptr == 0 || (!inescq && *ptr == ']'))          if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3919            {            {
3920            ptr = oldptr;            ptr = oldptr;
3921            goto LONE_SINGLE_CHARACTER;            goto LONE_SINGLE_CHARACTER;
# Line 3071  for (;; ptr++) Line 3934  for (;; ptr++)
3934          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
3935          in such circumstances. */          in such circumstances. */
3936    
3937          if (!inescq && d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3938            {            {
3939            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3940            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3941    
3942            /* \b is backslash; \X is literal X; \R is literal R; any other            /* \b is backspace; any other special means the '-' was literal */
           special means the '-' was literal */  
3943    
3944            if (d < 0)            if (d < 0)
3945              {              {
3946              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  
3947                {                {
3948                ptr = oldptr;                ptr = oldptr;
3949                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 3104  for (;; ptr++) Line 3964  for (;; ptr++)
3964    
3965          /* Remember \r or \n */          /* Remember \r or \n */
3966    
3967          if (d == '\r' || d == '\n') cd->external_flags |= PCRE_HASCRORLF;          if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3968    
3969          /* 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
3970          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 3224  for (;; ptr++) Line 4084  for (;; ptr++)
4084          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
4085            {            {
4086            unsigned int othercase;            unsigned int othercase;
4087            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)            if ((othercase = UCD_OTHERCASE(c)) != c)
4088              {              {
4089              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
4090              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 3249  for (;; ptr++) Line 4109  for (;; ptr++)
4109          }          }
4110        }        }
4111    
4112      /* 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.
4113        If we are at the end of an internal nested string, revert to the outer
4114        string. */
4115    
4116        while (((c = *(++ptr)) != 0 ||
4117               (nestptr != NULL &&
4118                 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != 0)) &&
4119               (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
4120    
4121      while ((c = *(++ptr)) != 0 && (c != ']' || inescq));      /* Check for missing terminating ']' */
4122    
4123      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)
4124        {        {
4125        *errorcodeptr = ERR6;        *errorcodeptr = ERR6;
4126        goto FAILED;        goto FAILED;
4127        }        }
4128    
   
 /* 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  
   
   
4129      /* 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
4130      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
4131      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 3285  we set the flag only if there is a liter Line 4133  we set the flag only if there is a liter
4133    
4134      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
4135      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
4136      operate on single-bytes only. This is an historical hangover. Maybe one day      operate on single-bytes characters only. This is an historical hangover.
4137      we can tidy these opcodes to handle multi-byte characters.      Maybe one day we can tidy these opcodes to handle multi-byte characters.
4138    
4139      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
4140      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.
4141      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
4142      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
4143      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
4144      reqbyte, save the previous value for reinstating. */      of reqbyte, save the previous value for reinstating. */
4145    
4146  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4147      if (class_charcount == 1 && !class_utf8 &&      if (class_charcount == 1 && !class_utf8 &&
# Line 3304  we set the flag only if there is a liter Line 4152  we set the flag only if there is a liter
4152        {        {
4153        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
4154    
4155        /* The OP_NOT opcode works on one-byte characters only. */        /* The OP_NOT[I] opcodes work on one-byte characters only. */
4156    
4157        if (negate_class)        if (negate_class)
4158          {          {
4159          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
4160          zerofirstbyte = firstbyte;          zerofirstbyte = firstbyte;
4161          *code++ = OP_NOT;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
4162          *code++ = class_lastchar;          *code++ = class_lastchar;
4163          break;          break;
4164          }          }
# Line 3340  we set the flag only if there is a liter Line 4188  we set the flag only if there is a liter
4188      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
4189    
4190      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
4191      extended class, with its own opcode, unless there was a negated special      extended class, with its own opcode, unless there was a negated special
4192      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
4193      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
4194      (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
4195      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
4196        actual compiled code. */
4197    
4198  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4199      if (class_utf8 && !should_flip_negation)      if (class_utf8 && (!should_flip_negation || (options & PCRE_UCP) != 0))
4200        {        {
4201        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
4202        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 3373  we set the flag only if there is a liter Line 4222  we set the flag only if there is a liter
4222        }        }
4223  #endif  #endif
4224    
4225      /* 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
4226      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
4227      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
4228      map into the code vector, negating it if necessary. */      (non-UCP) in the class. Then copy the 32-byte map into the code vector,
4229        negating it if necessary. */
4230    
4231      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
4232      if (negate_class)      if (negate_class)
4233        {        {
# Line 3396  we set the flag only if there is a liter Line 4246  we set the flag only if there is a liter
4246      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
4247      has been tested above. */      has been tested above. */
4248    
4249      case '{':      case CHAR_LEFT_CURLY_BRACKET:
4250      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
4251      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
4252      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
4253      goto REPEAT;      goto REPEAT;
4254    
4255      case '*':      case CHAR_ASTERISK:
4256      repeat_min = 0;      repeat_min = 0;
4257      repeat_max = -1;      repeat_max = -1;
4258      goto REPEAT;      goto REPEAT;
4259    
4260      case '+':      case CHAR_PLUS:
4261      repeat_min = 1;      repeat_min = 1;
4262      repeat_max = -1;      repeat_max = -1;
4263      goto REPEAT;      goto REPEAT;
4264    
4265      case '?':      case CHAR_QUESTION_MARK:
4266      repeat_min = 0;      repeat_min = 0;
4267      repeat_max = 1;      repeat_max = 1;
4268    
# Line 3436  we set the flag only if there is a liter Line 4286  we set the flag only if there is a liter
4286      op_type = 0;                    /* Default single-char op codes */      op_type = 0;                    /* Default single-char op codes */
4287      possessive_quantifier = FALSE;  /* Default not possessive quantifier */      possessive_quantifier = FALSE;  /* Default not possessive quantifier */
4288    
4289      /* 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
4290      for an inserted OP_ONCE for the additional '+' extension. */      insert something before it. */
4291    
4292      tempcode = previous;      tempcode = previous;
4293    
# Line 3447  we set the flag only if there is a liter Line 4297  we set the flag only if there is a liter
4297      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
4298      repeat type to the non-default. */      repeat type to the non-default. */
4299    
4300      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
4301        {        {
4302        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
4303        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
4304        ptr++;        ptr++;
4305        }        }
4306      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
4307        {        {
4308        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
4309        ptr++;        ptr++;
4310        }        }
4311      else repeat_type = greedy_default;      else repeat_type = greedy_default;
4312    
4313        /* If previous was a recursion call, wrap it in atomic brackets so that
4314        previous becomes the atomic group. All recursions were so wrapped in the
4315        past, but it no longer happens for non-repeated recursions. In fact, the
4316        repeated ones could be re-implemented independently so as not to need this,
4317        but for the moment we rely on the code for repeating groups. */
4318    
4319        if (*previous == OP_RECURSE)
4320          {
4321          memmove(previous + 1 + LINK_SIZE, previous, 1 + LINK_SIZE);
4322          *previous = OP_ONCE;
4323          PUT(previous, 1, 2 + 2*LINK_SIZE);
4324          previous[2 + 2*LINK_SIZE] = OP_KET;
4325          PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
4326          code += 2 + 2 * LINK_SIZE;
4327          length_prevgroup = 3 + 3*LINK_SIZE;
4328    
4329          /* When actually compiling, we need to check whether this was a forward
4330          reference, and if so, adjust the offset. */
4331    
4332          if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
4333            {
4334            int offset = GET(cd->hwm, -LINK_SIZE);
4335            if (offset == previous + 1 - cd->start_code)
4336              PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
4337            }
4338          }
4339    
4340        /* Now handle repetition for the different types of item. */
4341    
4342      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
4343      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
4344      that it is set in reqbyte - it might not be if a sequence such as x{3} is      that it is set in reqbyte - it might not be if a sequence such as x{3} is
4345      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
4346      instead.  */      instead.  */
4347    
4348      if (*previous == OP_CHAR || *previous == OP_CHARNC)      if (*previous == OP_CHAR || *previous == OP_CHARI)
4349        {        {
4350          op_type = (*previous == OP_CHAR)? 0 : OP_STARI - OP_STAR;
4351    
4352        /* 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
4353        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
4354        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 3500  we set the flag only if there is a liter Line 4381  we set the flag only if there is a liter
4381    
4382        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4383            repeat_max < 0 &&            repeat_max < 0 &&
4384            check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
             options, cd))  
4385          {          {
4386          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4387          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3513  we set the flag only if there is a liter Line 4393  we set the flag only if there is a liter
4393      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
4394      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-
4395      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
4396      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
4397      currently used only for single-byte chars. */      are currently used only for single-byte chars. */
4398    
4399      else if (*previous == OP_NOT)      else if (*previous == OP_NOT || *previous == OP_NOTI)
4400        {        {
4401        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = ((*previous == OP_NOT)? OP_NOTSTAR : OP_NOTSTARI) - OP_STAR;
4402        c = previous[1];        c = previous[1];
4403        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4404            repeat_max < 0 &&            repeat_max < 0 &&
4405            check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4406          {          {
4407          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4408          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3546  we set the flag only if there is a liter Line 4426  we set the flag only if there is a liter
4426    
4427        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4428            repeat_max < 0 &&            repeat_max < 0 &&
4429            check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4430          {          {
4431          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4432          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3568  we set the flag only if there is a liter Line 4448  we set the flag only if there is a liter
4448    
4449        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
4450    
4451          /*--------------------------------------------------------------------*/
4452          /* This code is obsolete from release 8.00; the restriction was finally
4453          removed: */
4454    
4455        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4456        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4457    
4458        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4459          /*--------------------------------------------------------------------*/
4460    
4461        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
4462    
# Line 3710  we set the flag only if there is a liter Line 4595  we set the flag only if there is a liter
4595  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4596               *previous == OP_XCLASS ||               *previous == OP_XCLASS ||
4597  #endif  #endif
4598               *previous == OP_REF)               *previous == OP_REF ||
4599                 *previous == OP_REFI)
4600        {        {
4601        if (repeat_max == 0)        if (repeat_max == 0)
4602          {          {
# Line 3718  we set the flag only if there is a liter Line 4604  we set the flag only if there is a liter
4604          goto END_REPEAT;          goto END_REPEAT;
4605          }          }
4606    
4607          /*--------------------------------------------------------------------*/
4608          /* This code is obsolete from release 8.00; the restriction was finally
4609          removed: */
4610    
4611        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4612        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4613    
4614        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4615          /*--------------------------------------------------------------------*/
4616    
4617        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4618          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 3739  we set the flag only if there is a liter Line 4630  we set the flag only if there is a liter
4630        }        }
4631    
4632      /* 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
4633      cases. */      cases. Note that at this point we can encounter only the "basic" bracket
4634        opcodes such as BRA and CBRA, as this is the place where they get converted
4635        into the more special varieties such as BRAPOS and SBRA. A test for >=
4636        OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
4637        ASSERTBACK_NOT, ONCE, BRA, CBRA, and COND. Originally, PCRE did not allow
4638        repetition of assertions, but now it does, for Perl compatibility. */
4639    
4640      else if (*previous == OP_BRA  || *previous == OP_CBRA ||      else if (*previous >= OP_ASSERT && *previous <= OP_COND)
              *previous == OP_ONCE || *previous == OP_COND)  
4641        {        {
4642        register int i;        register int i;
4643        int ketoffset = 0;        int len = (int)(code - previous);
       int len = code - previous;  
4644        uschar *bralink = NULL;        uschar *bralink = NULL;
4645          uschar *brazeroptr = NULL;
4646    
4647        /* Repeating a DEFINE group is pointless */        /* Repeating a DEFINE group is pointless, but Perl allows the syntax, so
4648          we just ignore the repeat. */
4649    
4650        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
4651          {          goto END_REPEAT;
4652          *errorcodeptr = ERR55;  
4653          goto FAILED;        /* There is no sense in actually repeating assertions. The only potential
4654          }        use of repetition is in cases when the assertion is optional. Therefore,
4655          if the minimum is greater than zero, just ignore the repeat. If the
4656          maximum is not not zero or one, set it to 1. */
4657    
4658        /* If the maximum repeat count is unlimited, find the end of the bracket        if (*previous < OP_ONCE)    /* Assertion */
4659        by scanning through from the start, and compute the offset back to it          {
4660        from the current code pointer. There may be an OP_OPT setting following          if (repeat_min > 0) goto END_REPEAT;
4661        the final KET, so we can't find the end just by going back from the code          if (repeat_max < 0 || repeat_max > 1) repeat_max = 1;
       pointer. */  
   
       if (repeat_max == -1)  
         {  
         register uschar *ket = previous;  
         do ket += GET(ket, 1); while (*ket != OP_KET);  
         ketoffset = code - ket;  
4662          }          }
4663    
4664        /* The case of a zero minimum is special because of the need to stick        /* The case of a zero minimum is special because of the need to stick
# Line 3779  we set the flag only if there is a liter Line 4670  we set the flag only if there is a liter
4670    
4671        if (repeat_min == 0)        if (repeat_min == 0)
4672          {          {
4673          /* 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
4674          altogether. */          output altogether, like this:
4675    
4676          if (repeat_max == 0)          ** if (repeat_max == 0)
4677            {          **   {
4678            code = previous;          **   code = previous;
4679            goto END_REPEAT;          **   goto END_REPEAT;
4680            }          **   }
4681    
4682            However, that fails when a group or a subgroup within it is referenced
4683            as a subroutine from elsewhere in the pattern, so now we stick in
4684            OP_SKIPZERO in front of it so that it is skipped on execution. As we
4685            don't have a list of which groups are referenced, we cannot do this
4686            selectively.
4687    
4688            If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4689            and do no more at this point. However, we do need to adjust any
4690            OP_RECURSE calls inside the group that refer to the group itself or any
4691            internal or forward referenced group, because the offset is from the
4692            start of the whole regex. Temporarily terminate the pattern while doing
4693            this. */
4694    
4695          /* 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)  
4696            {            {
4697            *code = OP_END;            *code = OP_END;
4698            adjust_recurse(previous, 1, utf8, cd, save_hwm);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
4699            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
4700            code++;            code++;
4701              if (repeat_max == 0)
4702                {
4703                *previous++ = OP_SKIPZERO;
4704                goto END_REPEAT;
4705                }
4706              brazeroptr = previous;    /* Save for possessive optimizing */
4707            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4708            }            }
4709    
# Line 3825  we set the flag only if there is a liter Line 4728  we set the flag only if there is a liter
4728            /* We chain together the bracket offset fields that have to be            /* We chain together the bracket offset fields that have to be
4729            filled in later when the ends of the brackets are reached. */            filled in later when the ends of the brackets are reached. */
4730    
4731            offset = (bralink == NULL)? 0 : previous - bralink;            offset = (bralink == NULL)? 0 : (int)(previous - bralink);
4732            bralink = previous;            bralink = previous;
4733            PUTINC(previous, 0, offset);            PUTINC(previous, 0, offset);
4734            }            }
# Line 3846  we set the flag only if there is a liter Line 4749  we set the flag only if there is a liter
4749            {            {
4750            /* 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
4751            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
4752            potential integer overflow. */            potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
4753              integer type when available, otherwise double. */
4754    
4755            if (lengthptr != NULL)            if (lengthptr != NULL)
4756              {              {
4757              int delta = (repeat_min - 1)*length_prevgroup;              int delta = (repeat_min - 1)*length_prevgroup;
4758              if ((double)(repeat_min - 1)*(double)length_prevgroup >              if ((INT64_OR_DOUBLE)(repeat_min - 1)*
4759                                                              (double)INT_MAX ||                    (INT64_OR_DOUBLE)length_prevgroup >
4760                        (INT64_OR_DOUBLE)INT_MAX ||
4761                  OFLOW_MAX - *lengthptr < delta)                  OFLOW_MAX - *lengthptr < delta)
4762                {                {
4763                *errorcodeptr = ERR20;                *errorcodeptr = ERR20;
# Line 3898  we set the flag only if there is a liter Line 4803  we set the flag only if there is a liter
4803          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
4804          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
4805          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
4806          paranoid checks to avoid integer overflow. */          paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
4807            a 64-bit integer type when available, otherwise double. */
4808    
4809          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4810            {            {
4811            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4812                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4813            if ((double)repeat_max *            if ((INT64_OR_DOUBLE)repeat_max *
4814                  (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)                  (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4815                    > (double)INT_MAX ||                    > (INT64_OR_DOUBLE)INT_MAX ||
4816                OFLOW_MAX - *lengthptr < delta)                OFLOW_MAX - *lengthptr < delta)
4817              {              {
4818              *errorcodeptr = ERR20;              *errorcodeptr = ERR20;
# Line 3931  we set the flag only if there is a liter Line 4837  we set the flag only if there is a liter
4837              {              {
4838              int offset;              int offset;
4839              *code++ = OP_BRA;              *code++ = OP_BRA;
4840              offset = (bralink == NULL)? 0 : code - bralink;              offset = (bralink == NULL)? 0 : (int)(code - bralink);
4841              bralink = code;              bralink = code;
4842              PUTINC(code, 0, offset);              PUTINC(code, 0, offset);
4843              }              }
# Line 3952  we set the flag onl