/[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 298 by ph10, Thu Jan 10 17:09:12 2008 UTC revision 746 by ph10, Tue Nov 15 15:07:02 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        /* In a character class, \g is just a literal "g". Outside a character
714        class, \g must be followed by one of a number of specific things:
715    
716      case 'g':      (1) A number, either plain or braced. If positive, it is an absolute
717      if (ptr[1] == '{')      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  for (; *ptr != 0; ptr++)  /* 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    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    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    count++;  Returns:       the number of the found subpattern, or -1 if not found
1423    */
1424    
1425    if (name == NULL && count == lorn) return count;  static int
1426    term = *ptr++;  find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode,
1427    if (term == '<') term = '>';    BOOL utf8)
1428    thisname = ptr;  {
1429    while (*ptr != term) ptr++;  uschar *ptr = (uschar *)cd->start_pattern;
1430    if (name != NULL && lorn == ptr - thisname &&  int count = 0;
1431        strncmp((const char *)name, (const char *)thisname, lorn) == 0)  int rc;
1432      return count;  
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    for (;;)
1439      {
1440      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, &count);
1441      if (rc > 0 || *ptr++ == 0) break;
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 1222  for (;;) Line 1646  for (;;)
1646      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1647    
1648      case OP_EXACT:      case OP_EXACT:
1649        case OP_EXACTI:
1650      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1651      cc += 4;      cc += 4;
1652  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1653      if ((options & PCRE_UTF8) != 0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       {  
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1654  #endif  #endif
1655      break;      break;
1656    
# Line 1252  for (;;) Line 1674  for (;;)
1674      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1675      case OP_WORDCHAR:      case OP_WORDCHAR:
1676      case OP_ANY:      case OP_ANY:
1677        case OP_ALLANY:
1678      branchlength++;      branchlength++;
1679      cc++;      cc++;
1680      break;      break;
# Line 1306  for (;;) Line 1729  for (;;)
1729    
1730    
1731  /*************************************************  /*************************************************
1732  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1733  *************************************************/  *************************************************/
1734    
1735  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1736  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1737    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1738    so that it can be called from pcre_study() when finding the minimum matching
1739    length.
1740    
1741  Arguments:  Arguments:
1742    code        points to start of expression    code        points to start of expression
1743    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1744    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1745    
1746  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
1747  */  */
1748    
1749  static const uschar *  const uschar *
1750  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1751  {  {
1752  for (;;)  for (;;)
1753    {    {
1754    register int c = *code;    register int c = *code;
1755    
1756    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1757    
1758    /* 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 1761  for (;;)
1761    
1762    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1763    
1764      /* Handle recursion */
1765    
1766      else if (c == OP_REVERSE)
1767        {
1768        if (number < 0) return (uschar *)code;
1769        code += _pcre_OP_lengths[c];
1770        }
1771    
1772    /* Handle capturing bracket */    /* Handle capturing bracket */
1773    
1774    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
1775               c == OP_CBRAPOS || c == OP_SCBRAPOS)
1776      {      {
1777      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
1778      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
# Line 1345  for (;;) Line 1781  for (;;)
1781    
1782    /* 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
1783    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
1784    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1785      must add in its length. */
1786    
1787    else    else
1788      {      {
# Line 1369  for (;;) Line 1806  for (;;)
1806        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
1807        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1808        break;        break;
1809    
1810          case OP_MARK:
1811          case OP_PRUNE_ARG:
1812          case OP_SKIP_ARG:
1813          code += code[1];
1814          break;
1815    
1816          case OP_THEN_ARG:
1817          code += code[1];
1818          break;
1819        }        }
1820    
1821      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1383  for (;;) Line 1830  for (;;)
1830      if (utf8) switch(c)      if (utf8) switch(c)
1831        {        {
1832        case OP_CHAR:        case OP_CHAR:
1833        case OP_CHARNC:        case OP_CHARI:
1834        case OP_EXACT:        case OP_EXACT:
1835          case OP_EXACTI:
1836        case OP_UPTO:        case OP_UPTO:
1837          case OP_UPTOI:
1838        case OP_MINUPTO:        case OP_MINUPTO:
1839          case OP_MINUPTOI:
1840        case OP_POSUPTO:        case OP_POSUPTO:
1841          case OP_POSUPTOI:
1842        case OP_STAR:        case OP_STAR:
1843          case OP_STARI:
1844        case OP_MINSTAR:        case OP_MINSTAR:
1845          case OP_MINSTARI:
1846        case OP_POSSTAR:        case OP_POSSTAR:
1847          case OP_POSSTARI:
1848        case OP_PLUS:        case OP_PLUS:
1849          case OP_PLUSI:
1850        case OP_MINPLUS:        case OP_MINPLUS:
1851          case OP_MINPLUSI:
1852        case OP_POSPLUS:        case OP_POSPLUS:
1853          case OP_POSPLUSI:
1854        case OP_QUERY:        case OP_QUERY:
1855          case OP_QUERYI:
1856        case OP_MINQUERY:        case OP_MINQUERY:
1857          case OP_MINQUERYI:
1858        case OP_POSQUERY:        case OP_POSQUERY:
1859          case OP_POSQUERYI:
1860        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1861        break;        break;
1862        }        }
1863    #else
1864        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1865  #endif  #endif
1866      }      }
1867    }    }
# Line 1438  for (;;) Line 1900  for (;;)
1900    
1901    /* 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
1902    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
1903    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1904      must add in its length. */
1905    
1906    else    else
1907      {      {
# Line 1462  for (;;) Line 1925  for (;;)
1925        case OP_TYPEEXACT:        case OP_TYPEEXACT:
1926        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1927        break;        break;
1928    
1929          case OP_MARK:
1930          case OP_PRUNE_ARG:
1931          case OP_SKIP_ARG:
1932          code += code[1];
1933          break;
1934    
1935          case OP_THEN_ARG:
1936          code += code[1];
1937          break;
1938        }        }
1939    
1940      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1476  for (;;) Line 1949  for (;;)
1949      if (utf8) switch(c)      if (utf8) switch(c)
1950        {        {
1951        case OP_CHAR:        case OP_CHAR:
1952        case OP_CHARNC:        case OP_CHARI:
1953        case OP_EXACT:        case OP_EXACT:
1954          case OP_EXACTI:
1955        case OP_UPTO:        case OP_UPTO:
1956          case OP_UPTOI:
1957        case OP_MINUPTO:        case OP_MINUPTO:
1958          case OP_MINUPTOI:
1959        case OP_POSUPTO:        case OP_POSUPTO:
1960          case OP_POSUPTOI:
1961        case OP_STAR:        case OP_STAR:
1962          case OP_STARI:
1963        case OP_MINSTAR:        case OP_MINSTAR:
1964          case OP_MINSTARI:
1965        case OP_POSSTAR:        case OP_POSSTAR:
1966          case OP_POSSTARI:
1967        case OP_PLUS:        case OP_PLUS:
1968          case OP_PLUSI:
1969        case OP_MINPLUS:        case OP_MINPLUS:
1970          case OP_MINPLUSI:
1971        case OP_POSPLUS:        case OP_POSPLUS:
1972          case OP_POSPLUSI:
1973        case OP_QUERY:        case OP_QUERY:
1974          case OP_QUERYI:
1975        case OP_MINQUERY:        case OP_MINQUERY:
1976          case OP_MINQUERYI:
1977        case OP_POSQUERY:        case OP_POSQUERY:
1978          case OP_POSQUERYI:
1979        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1980        break;        break;
1981        }        }
1982    #else
1983        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1984  #endif  #endif
1985      }      }
1986    }    }
# Line 1516  Arguments: Line 2004  Arguments:
2004    code        points to start of search    code        points to start of search
2005    endcode     points to where to stop    endcode     points to where to stop
2006    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
2007      cd          contains pointers to tables etc.
2008    
2009  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2010  */  */
2011    
2012  static BOOL  static BOOL
2013  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
2014      compile_data *cd)
2015  {  {
2016  register int c;  register int c;
2017  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], TRUE);
2018       code < endcode;       code < endcode;
2019       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], TRUE))
2020    {    {
2021    const uschar *ccode;    const uschar *ccode;
2022    
# Line 1542  for (code = first_significant_code(code Line 2032  for (code = first_significant_code(code
2032      continue;      continue;
2033      }      }
2034    
2035      /* For a recursion/subroutine call, if its end has been reached, which
2036      implies a backward reference subroutine call, we can scan it. If it's a
2037      forward reference subroutine call, we can't. To detect forward reference
2038      we have to scan up the list that is kept in the workspace. This function is
2039      called only when doing the real compile, not during the pre-compile that
2040      measures the size of the compiled pattern. */
2041    
2042      if (c == OP_RECURSE)
2043        {
2044        const uschar *scode;
2045        BOOL empty_branch;
2046    
2047        /* Test for forward reference */
2048    
2049        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
2050          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
2051    
2052        /* Not a forward reference, test for completed backward reference */
2053    
2054        empty_branch = FALSE;
2055        scode = cd->start_code + GET(code, 1);
2056        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2057    
2058        /* Completed backwards reference */
2059    
2060        do
2061          {
2062          if (could_be_empty_branch(scode, endcode, utf8, cd))
2063            {
2064            empty_branch = TRUE;
2065            break;
2066            }
2067          scode += GET(scode, 1);
2068          }
2069        while (*scode == OP_ALT);
2070    
2071        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2072        continue;
2073        }
2074    
2075    /* Groups with zero repeats can of course be empty; skip them. */    /* Groups with zero repeats can of course be empty; skip them. */
2076    
2077    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2078          c == OP_BRAPOSZERO)
2079      {      {
2080      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
2081      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
# Line 1552  for (code = first_significant_code(code Line 2083  for (code = first_significant_code(code
2083      continue;      continue;
2084      }      }
2085    
2086      /* A nested group that is already marked as "could be empty" can just be
2087      skipped. */
2088    
2089      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2090          c == OP_SCBRA || c == OP_SCBRAPOS)
2091        {
2092        do code += GET(code, 1); while (*code == OP_ALT);
2093        c = *code;
2094        continue;
2095        }
2096    
2097    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
2098    
2099    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_BRA  || c == OP_BRAPOS ||
2100          c == OP_CBRA || c == OP_CBRAPOS ||
2101          c == OP_ONCE || c == OP_ONCE_NC ||
2102          c == OP_COND)
2103      {      {
2104      BOOL empty_branch;      BOOL empty_branch;
2105      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
2106    
2107      /* Scan a closed bracket */      /* If a conditional group has only one branch, there is a second, implied,
2108        empty branch, so just skip over the conditional, because it could be empty.
2109        Otherwise, scan the individual branches of the group. */
2110    
2111      empty_branch = FALSE;      if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
     do  
       {  
       if (!empty_branch && could_be_empty_branch(code, endcode, utf8))  
         empty_branch = TRUE;  
2112        code += GET(code, 1);        code += GET(code, 1);
2113        else
2114          {
2115          empty_branch = FALSE;
2116          do
2117            {
2118            if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
2119              empty_branch = TRUE;
2120            code += GET(code, 1);
2121            }
2122          while (*code == OP_ALT);
2123          if (!empty_branch) return FALSE;   /* All branches are non-empty */
2124        }        }
2125      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
2126      c = *code;      c = *code;
2127      continue;      continue;
2128      }      }
# Line 1630  for (code = first_significant_code(code Line 2183  for (code = first_significant_code(code
2183      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2184      case OP_WORDCHAR:      case OP_WORDCHAR:
2185      case OP_ANY:      case OP_ANY:
2186        case OP_ALLANY:
2187      case OP_ANYBYTE:      case OP_ANYBYTE:
2188      case OP_CHAR:      case OP_CHAR:
2189      case OP_CHARNC:      case OP_CHARI:
2190      case OP_NOT:      case OP_NOT:
2191        case OP_NOTI:
2192      case OP_PLUS:      case OP_PLUS:
2193      case OP_MINPLUS:      case OP_MINPLUS:
2194      case OP_POSPLUS:      case OP_POSPLUS:
# Line 1673  for (code = first_significant_code(code Line 2228  for (code = first_significant_code(code
2228      case OP_KET:      case OP_KET:
2229      case OP_KETRMAX:      case OP_KETRMAX:
2230      case OP_KETRMIN:      case OP_KETRMIN:
2231        case OP_KETRPOS:
2232      case OP_ALT:      case OP_ALT:
2233      return TRUE;      return TRUE;
2234    
# Line 1681  for (code = first_significant_code(code Line 2237  for (code = first_significant_code(code
2237    
2238  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2239      case OP_STAR:      case OP_STAR:
2240        case OP_STARI:
2241      case OP_MINSTAR:      case OP_MINSTAR:
2242        case OP_MINSTARI:
2243      case OP_POSSTAR:      case OP_POSSTAR:
2244        case OP_POSSTARI:
2245      case OP_QUERY:      case OP_QUERY:
2246        case OP_QUERYI:
2247      case OP_MINQUERY:      case OP_MINQUERY:
2248        case OP_MINQUERYI:
2249      case OP_POSQUERY:      case OP_POSQUERY:
2250        case OP_POSQUERYI:
2251        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2252        break;
2253    
2254      case OP_UPTO:      case OP_UPTO:
2255        case OP_UPTOI:
2256      case OP_MINUPTO:      case OP_MINUPTO:
2257        case OP_MINUPTOI:
2258      case OP_POSUPTO:      case OP_POSUPTO:
2259      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_POSUPTOI:
2260        if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2261      break;      break;
2262  #endif  #endif
2263    
2264        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2265        string. */
2266    
2267        case OP_MARK:
2268        case OP_PRUNE_ARG:
2269        case OP_SKIP_ARG:
2270        code += code[1];
2271        break;
2272    
2273        case OP_THEN_ARG:
2274        code += code[1];
2275        break;
2276    
2277        /* None of the remaining opcodes are required to match a character. */
2278    
2279        default:
2280        break;
2281      }      }
2282    }    }
2283    
# Line 1708  return TRUE; Line 2294  return TRUE;
2294  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
2295  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,
2296  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.
2297    This function is called only during the real compile, not during the
2298    pre-compile.
2299    
2300  Arguments:  Arguments:
2301    code        points to start of the recursion    code        points to start of the recursion
2302    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2303    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2304    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2305      cd          pointers to tables etc
2306    
2307  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2308  */  */
2309    
2310  static BOOL  static BOOL
2311  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2312    BOOL utf8)    BOOL utf8, compile_data *cd)
2313  {  {
2314  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2315    {    {
2316    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2317        return FALSE;
2318    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2319    }    }
2320  return TRUE;  return TRUE;
# Line 1756  where Perl recognizes it as the POSIX cl Line 2346  where Perl recognizes it as the POSIX cl
2346  "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,  "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2347  I think.  I think.
2348    
2349    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2350    It seems that the appearance of a nested POSIX class supersedes an apparent
2351    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2352    a digit.
2353    
2354    In Perl, unescaped square brackets may also appear as part of class names. For
2355    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2356    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2357    seem right at all. PCRE does not allow closing square brackets in POSIX class
2358    names.
2359    
2360  Arguments:  Arguments:
2361    ptr      pointer to the initial [    ptr      pointer to the initial [
2362    endptr   where to return the end pointer    endptr   where to return the end pointer
# Line 1770  int terminator; /* Don't combin Line 2371  int terminator; /* Don't combin
2371  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2372  for (++ptr; *ptr != 0; ptr++)  for (++ptr; *ptr != 0; ptr++)
2373    {    {
2374    if (*ptr == '\\' && ptr[1] == ']') ptr++; else    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2375        ptr++;
2376      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2377      else
2378      {      {
2379      if (*ptr == ']') return FALSE;      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
     if (*ptr == terminator && ptr[1] == ']')  
2380        {        {
2381        *endptr = ptr;        *endptr = ptr;
2382        return TRUE;        return TRUE;
2383        }        }
2384        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2385             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2386              ptr[1] == CHAR_EQUALS_SIGN) &&
2387            check_posix_syntax(ptr, endptr))
2388          return FALSE;
2389      }      }
2390    }    }
2391  return FALSE;  return FALSE;
# Line 1824  return -1; Line 2432  return -1;
2432  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2433  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2434  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
2435  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
2436  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
2437  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
2438  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
2439  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2440    OP_END.
2441    
2442  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2443  recursions and subroutine calls. It must also check the list of such references  recursions and subroutine calls. It must also check the list of such references
# Line 1905  auto_callout(uschar *code, const uschar Line 2514  auto_callout(uschar *code, const uschar
2514  {  {
2515  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2516  *code++ = 255;  *code++ = 255;
2517  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2518  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2519  return code + 2*LINK_SIZE;  return code + 2*LINK_SIZE;
2520  }  }
2521    
# Line 1931  Returns: nothing Line 2540  Returns: nothing
2540  static void  static void
2541  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)
2542  {  {
2543  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2544  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2545  }  }
2546    
# Line 1963  get_othercase_range(unsigned int *cptr, Line 2572  get_othercase_range(unsigned int *cptr,
2572  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2573    
2574  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2575    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2576    
2577  if (c > d) return FALSE;  if (c > d) return FALSE;
2578    
# Line 1972  next = othercase + 1; Line 2581  next = othercase + 1;
2581    
2582  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2583    {    {
2584    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2585    next++;    next++;
2586    }    }
2587    
# Line 1981  for (++c; c <= d; c++) Line 2590  for (++c; c <= d; c++)
2590    
2591  return TRUE;  return TRUE;
2592  }  }
2593    
2594    
2595    
2596    /*************************************************
2597    *        Check a character and a property        *
2598    *************************************************/
2599    
2600    /* This function is called by check_auto_possessive() when a property item
2601    is adjacent to a fixed character.
2602    
2603    Arguments:
2604      c            the character
2605      ptype        the property type
2606      pdata        the data for the type
2607      negated      TRUE if it's a negated property (\P or \p{^)
2608    
2609    Returns:       TRUE if auto-possessifying is OK
2610    */
2611    
2612    static BOOL
2613    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2614    {
2615    const ucd_record *prop = GET_UCD(c);
2616    switch(ptype)
2617      {
2618      case PT_LAMP:
2619      return (prop->chartype == ucp_Lu ||
2620              prop->chartype == ucp_Ll ||
2621              prop->chartype == ucp_Lt) == negated;
2622    
2623      case PT_GC:
2624      return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;
2625    
2626      case PT_PC:
2627      return (pdata == prop->chartype) == negated;
2628    
2629      case PT_SC:
2630      return (pdata == prop->script) == negated;
2631    
2632      /* These are specials */
2633    
2634      case PT_ALNUM:
2635      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2636              _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;
2637    
2638      case PT_SPACE:    /* Perl space */
2639      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2640              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2641              == negated;
2642    
2643      case PT_PXSPACE:  /* POSIX space */
2644      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2645              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2646              c == CHAR_FF || c == CHAR_CR)
2647              == negated;
2648    
2649      case PT_WORD:
2650      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2651              _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2652              c == CHAR_UNDERSCORE) == negated;
2653      }
2654    return FALSE;
2655    }
2656  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2657    
2658    
# Line 1994  whether the next thing could possibly ma Line 2666  whether the next thing could possibly ma
2666  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
2667    
2668  Arguments:  Arguments:
2669    op_code       the repeated op code    previous      pointer to the repeated opcode
   this          data for this item, depends on the opcode  
2670    utf8          TRUE in UTF-8 mode    utf8          TRUE in UTF-8 mode
   utf8_char     used for utf8 character bytes, NULL if not relevant  
2671    ptr           next character in pattern    ptr           next character in pattern
2672    options       options bits    options       options bits
2673    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 2006  Returns: TRUE if possessifying is Line 2676  Returns: TRUE if possessifying is
2676  */  */
2677    
2678  static BOOL  static BOOL
2679  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
2680    const uschar *ptr, int options, compile_data *cd)    int options, compile_data *cd)
2681  {  {
2682  int next;  int c, next;
2683    int op_code = *previous++;
2684    
2685  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
2686    
# Line 2018  if ((options & PCRE_EXTENDED) != 0) Line 2689  if ((options & PCRE_EXTENDED) != 0)
2689    for (;;)    for (;;)
2690      {      {
2691      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2692      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2693        {        {
2694        while (*(++ptr) != 0)        ptr++;
2695          while (*ptr != 0)
2696            {
2697          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2698            ptr++;
2699    #ifdef SUPPORT_UTF8
2700            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2701    #endif
2702            }
2703        }        }
2704      else break;      else break;
2705      }      }
# Line 2030  if ((options & PCRE_EXTENDED) != 0) Line 2708  if ((options & PCRE_EXTENDED) != 0)
2708  /* 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
2709  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2710    
2711  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2712    {    {
2713    int temperrorcode = 0;    int temperrorcode = 0;
2714    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
# Line 2055  if ((options & PCRE_EXTENDED) != 0) Line 2733  if ((options & PCRE_EXTENDED) != 0)
2733    for (;;)    for (;;)
2734      {      {
2735      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2736      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2737        {        {
2738        while (*(++ptr) != 0)        ptr++;
2739          while (*ptr != 0)
2740            {
2741          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2742            ptr++;
2743    #ifdef SUPPORT_UTF8
2744            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2745    #endif
2746            }
2747        }        }
2748      else break;      else break;
2749      }      }
# Line 2066  if ((options & PCRE_EXTENDED) != 0) Line 2751  if ((options & PCRE_EXTENDED) != 0)
2751    
2752  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
2753    
2754  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2755    return FALSE;    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2756        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. */  
   
2757    
2758  /* Handle cases when the next item is a character. */  /* Now compare the next item with the previous opcode. First, handle cases when
2759    the next item is a character. */
2760    
2761  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
2762    {    {
2763    case OP_CHAR:    case OP_CHAR:
2764  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2765    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2766    #else
2767      c = *previous;
2768  #endif  #endif
2769    return item != next;    return c != next;
2770    
2771    /* 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
2772    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
2773    high-valued characters. */    high-valued characters. */
2774    
2775    case OP_CHARNC:    case OP_CHARI:
2776  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2777    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2778    #else
2779      c = *previous;
2780  #endif  #endif
2781    if (item == next) return FALSE;    if (c == next) return FALSE;
2782  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2783    if (utf8)    if (utf8)
2784      {      {
2785      unsigned int othercase;      unsigned int othercase;
2786      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2787  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2788      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
2789  #else  #else
2790      othercase = NOTACHAR;      othercase = NOTACHAR;
2791  #endif  #endif
2792      return (unsigned int)item != othercase;      return (unsigned int)c != othercase;
2793      }      }
2794    else    else
2795  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2796    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2797    
2798    /* 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
2799      opcodes are not used for multi-byte characters, because they are coded using
2800      an XCLASS instead. */
2801    
2802    case OP_NOT:    case OP_NOT:
2803    if (next < 0) return FALSE;  /* Not a character */    return (c = *previous) == next;
2804    if (item == next) return TRUE;  
2805    if ((options & PCRE_CASELESS) == 0) return FALSE;    case OP_NOTI:
2806      if ((c = *previous) == next) return TRUE;
2807  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2808    if (utf8)    if (utf8)
2809      {      {
2810      unsigned int othercase;      unsigned int othercase;
2811      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2812  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2813      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
2814  #else  #else
2815      othercase = NOTACHAR;      othercase = NOTACHAR;
2816  #endif  #endif
2817      return (unsigned int)item == othercase;      return (unsigned int)c == othercase;
2818      }      }
2819    else    else
2820  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2821    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2822    
2823      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2824      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
2825    
2826    case OP_DIGIT:    case OP_DIGIT:
2827    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
# Line 2173  if (next >= 0) switch(op_code) Line 2864  if (next >= 0) switch(op_code)
2864      case 0x202f:      case 0x202f:
2865      case 0x205f:      case 0x205f:
2866      case 0x3000:      case 0x3000:
2867      return op_code != OP_HSPACE;      return op_code == OP_NOT_HSPACE;
2868      default:      default:
2869      return op_code == OP_HSPACE;      return op_code != OP_NOT_HSPACE;
2870      }      }
2871    
2872      case OP_ANYNL:
2873    case OP_VSPACE:    case OP_VSPACE:
2874    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
2875    switch(next)    switch(next)
# Line 2189  if (next >= 0) switch(op_code) Line 2881  if (next >= 0) switch(op_code)
2881      case 0x85:      case 0x85:
2882      case 0x2028:      case 0x2028:
2883      case 0x2029:      case 0x2029:
2884      return op_code != OP_VSPACE;      return op_code == OP_NOT_VSPACE;
2885      default:      default:
2886      return op_code == OP_VSPACE;      return op_code != OP_NOT_VSPACE;
2887      }      }
2888    
2889    #ifdef SUPPORT_UCP
2890      case OP_PROP:
2891      return check_char_prop(next, previous[0], previous[1], FALSE);
2892    
2893      case OP_NOTPROP:
2894      return check_char_prop(next, previous[0], previous[1], TRUE);
2895    #endif
2896    
2897    default:    default:
2898    return FALSE;    return FALSE;
2899    }    }
2900    
2901    
2902  /* 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
2903    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
2904    generated only when PCRE_UCP is *not* set, that is, when only ASCII
2905    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
2906    replaced by OP_PROP codes when PCRE_UCP is set. */
2907    
2908  switch(op_code)  switch(op_code)
2909    {    {
2910    case OP_CHAR:    case OP_CHAR:
2911    case OP_CHARNC:    case OP_CHARI:
2912  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2913    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2914    #else
2915      c = *previous;
2916  #endif  #endif
2917    switch(-next)    switch(-next)
2918      {      {
2919      case ESC_d:      case ESC_d:
2920      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
2921    
2922      case ESC_D:      case ESC_D:
2923      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
2924    
2925      case ESC_s:      case ESC_s:
2926      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
2927    
2928      case ESC_S:      case ESC_S:
2929      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
2930    
2931      case ESC_w:      case ESC_w:
2932      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
2933    
2934      case ESC_W:      case ESC_W:
2935      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
2936    
2937      case ESC_h:      case ESC_h:
2938      case ESC_H:      case ESC_H:
2939      switch(item)      switch(c)
2940        {        {
2941        case 0x09:        case 0x09:
2942        case 0x20:        case 0x20:
# Line 2258  switch(op_code) Line 2964  switch(op_code)
2964    
2965      case ESC_v:      case ESC_v:
2966      case ESC_V:      case ESC_V:
2967      switch(item)      switch(c)
2968        {        {
2969        case 0x0a:        case 0x0a:
2970        case 0x0b:        case 0x0b:
# Line 2272  switch(op_code) Line 2978  switch(op_code)
2978        return -next == ESC_v;        return -next == ESC_v;
2979        }        }
2980    
2981        /* When PCRE_UCP is set, these values get generated for \d etc. Find
2982        their substitutions and process them. The result will always be either
2983        -ESC_p or -ESC_P. Then fall through to process those values. */
2984    
2985    #ifdef SUPPORT_UCP
2986        case ESC_du:
2987        case ESC_DU:
2988        case ESC_wu:
2989        case ESC_WU:
2990        case ESC_su:
2991        case ESC_SU:
2992          {
2993          int temperrorcode = 0;
2994          ptr = substitutes[-next - ESC_DU];
2995          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
2996          if (temperrorcode != 0) return FALSE;
2997          ptr++;    /* For compatibility */
2998          }
2999        /* Fall through */
3000    
3001        case ESC_p:
3002        case ESC_P:
3003          {
3004          int ptype, pdata, errorcodeptr;
3005          BOOL negated;
3006    
3007          ptr--;      /* Make ptr point at the p or P */
3008          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
3009          if (ptype < 0) return FALSE;
3010          ptr++;      /* Point past the final curly ket */
3011    
3012          /* If the property item is optional, we have to give up. (When generated
3013          from \d etc by PCRE_UCP, this test will have been applied much earlier,
3014          to the original \d etc. At this point, ptr will point to a zero byte. */
3015    
3016          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3017            strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3018              return FALSE;
3019    
3020          /* Do the property check. */
3021    
3022          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
3023          }
3024    #endif
3025    
3026      default:      default:
3027      return FALSE;      return FALSE;
3028      }      }
3029    
3030      /* In principle, support for Unicode properties should be integrated here as
3031      well. It means re-organizing the above code so as to get hold of the property
3032      values before switching on the op-code. However, I wonder how many patterns
3033      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
3034      these op-codes are never generated.) */
3035    
3036    case OP_DIGIT:    case OP_DIGIT:
3037    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
3038           next == -ESC_h || next == -ESC_v;           next == -ESC_h || next == -ESC_v || next == -ESC_R;
3039    
3040    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
3041    return next == -ESC_d;    return next == -ESC_d;
3042    
3043    case OP_WHITESPACE:    case OP_WHITESPACE:
3044    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
3045    
3046    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
3047    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
3048    
3049    case OP_HSPACE:    case OP_HSPACE:
3050    return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
3051             next == -ESC_w || next == -ESC_v || next == -ESC_R;
3052    
3053    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
3054    return next == -ESC_h;    return next == -ESC_h;
3055    
3056    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
3057      case OP_ANYNL:
3058    case OP_VSPACE:    case OP_VSPACE:
3059    return next == -ESC_V || next == -ESC_d || next == -ESC_w;    return next == -ESC_V || next == -ESC_d || next == -ESC_w;
3060    
3061    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
3062    return next == -ESC_v;    return next == -ESC_v || next == -ESC_R;
3063    
3064    case OP_WORDCHAR:    case OP_WORDCHAR:
3065    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
3066             next == -ESC_v || next == -ESC_R;
3067    
3068    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
3069    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2335  Arguments: Line 3095  Arguments:
3095    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
3096    reqbyteptr     set to the last literal character required, else < 0    reqbyteptr     set to the last literal character required, else < 0
3097    bcptr          points to current branch chain    bcptr          points to current branch chain
3098      cond_depth     conditional nesting depth
3099    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
3100    lengthptr      NULL during the real compile phase    lengthptr      NULL during the real compile phase
3101                   points to length accumulator during pre-compile phase                   points to length accumulator during pre-compile phase
# Line 2346  Returns: TRUE on success Line 3107  Returns: TRUE on success
3107  static BOOL  static BOOL
3108  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
3109    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
3110    compile_data *cd, int *lengthptr)    int cond_depth, compile_data *cd, int *lengthptr)
3111  {  {
3112  int repeat_type, op_type;  int repeat_type, op_type;
3113  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
# Line 2355  int greedy_default, greedy_non_default; Line 3116  int greedy_default, greedy_non_default;
3116  int firstbyte, reqbyte;  int firstbyte, reqbyte;
3117  int zeroreqbyte, zerofirstbyte;  int zeroreqbyte, zerofirstbyte;
3118  int req_caseopt, reqvary, tempreqvary;  int req_caseopt, reqvary, tempreqvary;
3119  int options = *optionsptr;  int options = *optionsptr;               /* May change dynamically */
3120  int after_manual_callout = 0;  int after_manual_callout = 0;
3121  int length_prevgroup = 0;  int length_prevgroup = 0;
3122  register int c;  register int c;
# Line 2367  BOOL inescq = FALSE; Line 3128  BOOL inescq = FALSE;
3128  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
3129  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
3130  const uschar *tempptr;  const uschar *tempptr;
3131    const uschar *nestptr = NULL;
3132  uschar *previous = NULL;  uschar *previous = NULL;
3133  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
3134  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
3135  uschar classbits[32];  uschar classbits[32];
3136    
3137    /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3138    must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3139    dynamically as we process the pattern. */
3140    
3141  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3142  BOOL class_utf8;  BOOL class_utf8;
3143  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
3144  uschar *class_utf8data;  uschar *class_utf8data;
3145    uschar *class_utf8data_base;
3146  uschar utf8_char[6];  uschar utf8_char[6];
3147  #else  #else
3148  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
 uschar *utf8_char = NULL;  
3149  #endif  #endif
3150    
3151  #ifdef DEBUG  #ifdef PCRE_DEBUG
3152  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
3153  #endif  #endif
3154    
# Line 2430  for (;; ptr++) Line 3196  for (;; ptr++)
3196    int subfirstbyte;    int subfirstbyte;
3197    int terminator;    int terminator;
3198    int mclength;    int mclength;
3199      int tempbracount;
3200    uschar mcbuffer[8];    uschar mcbuffer[8];
3201    
3202    /* Get next byte in the pattern */    /* Get next byte in the pattern */
3203    
3204    c = *ptr;    c = *ptr;
3205    
3206      /* If we are at the end of a nested substitution, revert to the outer level
3207      string. Nesting only happens one level deep. */
3208    
3209      if (c == 0 && nestptr != NULL)
3210        {
3211        ptr = nestptr;
3212        nestptr = NULL;
3213        c = *ptr;
3214        }
3215    
3216    /* 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
3217    previous cycle of this loop. */    previous cycle of this loop. */
3218    
3219    if (lengthptr != NULL)    if (lengthptr != NULL)
3220      {      {
3221  #ifdef DEBUG  #ifdef PCRE_DEBUG
3222      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3223  #endif  #endif
3224      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
3225        {        {
3226        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3227        goto FAILED;        goto FAILED;
# Line 2466  for (;; ptr++) Line 3243  for (;; ptr++)
3243        goto FAILED;        goto FAILED;
3244        }        }
3245    
3246      *lengthptr += code - last_code;      *lengthptr += (int)(code - last_code);
3247      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));
3248    
3249      /* If "previous" is set and it is not at the start of the work space, move      /* If "previous" is set and it is not at the start of the work space, move
# Line 2493  for (;; ptr++) Line 3270  for (;; ptr++)
3270    /* 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
3271    reference list. */    reference list. */
3272    
3273    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
3274      {      {
3275      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3276      goto FAILED;      goto FAILED;
# Line 2503  for (;; ptr++) Line 3280  for (;; ptr++)
3280    
3281    if (inescq && c != 0)    if (inescq && c != 0)
3282      {      {
3283      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3284        {        {
3285        inescq = FALSE;        inescq = FALSE;
3286        ptr++;        ptr++;
# Line 2529  for (;; ptr++) Line 3306  for (;; ptr++)
3306    /* 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
3307    a quantifier. */    a quantifier. */
3308    
3309    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
3310      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
3311        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
3312    
3313    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
3314         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2540  for (;; ptr++) Line 3318  for (;; ptr++)
3318      previous_callout = NULL;      previous_callout = NULL;
3319      }      }
3320    
3321    /* In extended mode, skip white space and comments */    /* In extended mode, skip white space and comments. */
3322    
3323    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3324      {      {
3325      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
3326      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
3327        {        {
3328        while (*(++ptr) != 0)        ptr++;
3329          while (*ptr != 0)
3330          {          {
3331          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3332            ptr++;
3333    #ifdef SUPPORT_UTF8
3334            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
3335    #endif
3336          }          }
3337        if (*ptr != 0) continue;        if (*ptr != 0) continue;
3338    
# Line 2570  for (;; ptr++) Line 3353  for (;; ptr++)
3353      {      {
3354      /* ===================================================================*/      /* ===================================================================*/
3355      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
3356      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
3357      case ')':      case CHAR_RIGHT_PARENTHESIS:
3358      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
3359      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
3360      *codeptr = code;      *codeptr = code;
# Line 2583  for (;; ptr++) Line 3366  for (;; ptr++)
3366          *errorcodeptr = ERR20;          *errorcodeptr = ERR20;
3367          goto FAILED;          goto FAILED;
3368          }          }
3369        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += (int)(code - last_code);   /* To include callout length */
3370        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
3371        }        }
3372      return TRUE;      return TRUE;
# Line 2593  for (;; ptr++) Line 3376  for (;; ptr++)
3376      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
3377      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3378    
3379      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
3380        previous = NULL;
3381      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3382        {        {
3383        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3384          *code++ = OP_CIRCM;
3385        }        }
3386      previous = NULL;      else *code++ = OP_CIRC;
     *code++ = OP_CIRC;  
3387      break;      break;
3388    
3389      case '$':      case CHAR_DOLLAR_SIGN:
3390      previous = NULL;      previous = NULL;
3391      *code++ = OP_DOLL;      *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
3392      break;      break;
3393    
3394      /* 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
3395      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
3396    
3397      case '.':      case CHAR_DOT:
3398      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3399      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
3400      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3401      previous = code;      previous = code;
3402      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
3403      break;      break;
3404    
3405    
# Line 2630  for (;; ptr++) Line 3414  for (;; ptr++)
3414      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,
3415      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
3416      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.
     */  
3417    
3418      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
3419        default (Perl) mode, it is treated as a data character. */
3420    
3421        case CHAR_RIGHT_SQUARE_BRACKET:
3422        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3423          {
3424          *errorcodeptr = ERR64;
3425          goto FAILED;
3426          }
3427        goto NORMAL_CHAR;
3428    
3429        case CHAR_LEFT_SQUARE_BRACKET:
3430      previous = code;      previous = code;
3431    
3432      /* 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
3433      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. */
3434    
3435      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3436             ptr[1] == CHAR_EQUALS_SIGN) &&
3437          check_posix_syntax(ptr, &tempptr))          check_posix_syntax(ptr, &tempptr))
3438        {        {
3439        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
3440        goto FAILED;        goto FAILED;
3441        }        }
3442    
# Line 2653  for (;; ptr++) Line 3448  for (;; ptr++)
3448      for (;;)      for (;;)
3449        {        {
3450        c = *(++ptr);        c = *(++ptr);
3451        if (c == '\\')        if (c == CHAR_BACKSLASH)
3452          {          {
3453          if (ptr[1] == 'E') ptr++;          if (ptr[1] == CHAR_E)
3454            else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;            ptr++;
3455              else break;          else if (strncmp((const char *)ptr+1,
3456                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
3457              ptr += 3;
3458            else
3459              break;
3460          }          }
3461        else if (!negate_class && c == '^')        else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
3462          negate_class = TRUE;          negate_class = TRUE;
3463        else break;        else break;
3464        }        }
3465    
3466        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
3467        an initial ']' is taken as a data character -- the code below handles
3468        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3469        [^] must match any character, so generate OP_ALLANY. */
3470    
3471        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3472            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3473          {
3474          *code++ = negate_class? OP_ALLANY : OP_FAIL;
3475          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3476          zerofirstbyte = firstbyte;
3477          break;
3478          }
3479    
3480      /* If a class contains a negative special such as \S, we need to flip the      /* If a class contains a negative special such as \S, we need to flip the
3481      negation flag at the end, so that support for characters > 255 works      negation flag at the end, so that support for characters > 255 works
3482      correctly (they are all included in the class). */      correctly (they are all included in the class). */
# Line 2687  for (;; ptr++) Line 3500  for (;; ptr++)
3500  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3501      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
3502      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
3503        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
3504  #endif  #endif
3505    
3506      /* 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 2702  for (;; ptr++) Line 3516  for (;; ptr++)
3516          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3517          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3518          }          }
3519    
3520          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
3521          data and reset the pointer. This is so that very large classes that
3522          contain a zillion UTF-8 characters no longer overwrite the work space
3523          (which is on the stack). */
3524    
3525          if (lengthptr != NULL)
3526            {
3527            *lengthptr += class_utf8data - class_utf8data_base;
3528            class_utf8data = class_utf8data_base;
3529            }
3530    
3531  #endif  #endif
3532    
3533        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
3534    
3535        if (inescq)        if (inescq)
3536          {          {
3537          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3538            {            {
3539            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
3540            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2723  for (;; ptr++) Line 3549  for (;; ptr++)
3549        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3550        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3551    
3552        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3553            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3554            check_posix_syntax(ptr, &tempptr))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3555          {          {
3556          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3557          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3558          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3559          uschar pbits[32];          uschar pbits[32];
3560    
3561          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3562            {            {
3563            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3564            goto FAILED;            goto FAILED;
3565            }            }
3566    
3567          ptr += 2;          ptr += 2;
3568          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3569            {            {
3570            local_negate = TRUE;            local_negate = TRUE;
3571            should_flip_negation = TRUE;  /* Note negative special */            should_flip_negation = TRUE;  /* Note negative special */
3572            ptr++;            ptr++;
3573            }            }
3574    
3575          posix_class = check_posix_name(ptr, tempptr - ptr);          posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3576          if (posix_class < 0)          if (posix_class < 0)
3577            {            {
3578            *errorcodeptr = ERR30;            *errorcodeptr = ERR30;
# Line 2760  for (;; ptr++) Line 3586  for (;; ptr++)
3586          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3587            posix_class = 0;            posix_class = 0;
3588    
3589          /* 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
3590          because we may be adding and subtracting from it, and we don't want to          different escape sequences that use Unicode properties. */
3591          subtract bits that may be in the main map already. At the end we or the  
3592          result into the bit map that is being built. */  #ifdef SUPPORT_UCP
3593            if ((options & PCRE_UCP) != 0)
3594              {
3595              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3596              if (posix_substitutes[pc] != NULL)
3597                {
3598                nestptr = tempptr + 1;
3599                ptr = posix_substitutes[pc] - 1;
3600                continue;
3601                }
3602              }
3603    #endif
3604            /* In the non-UCP case, we build the bit map for the POSIX class in a
3605            chunk of local store because we may be adding and subtracting from it,
3606            and we don't want to subtract bits that may be in the main map already.
3607            At the end we or the result into the bit map that is being built. */
3608    
3609          posix_class *= 3;          posix_class *= 3;
3610    
# Line 2807  for (;; ptr++) Line 3648  for (;; ptr++)
3648    
3649        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3650        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
3651        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
3652        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
3653        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
3654        character in them, so set class_charcount bigger than one. */        are either treated as literal characters (by default), or are faulted if
3655          PCRE_EXTRA is set. */
3656    
3657        if (c == '\\')        if (c == CHAR_BACKSLASH)
3658          {          {
3659          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3660          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3661    
3662          if (-c == ESC_b) c = '\b';       /* \b is backspace in a class */          if (-c == ESC_b) c = CHAR_BS;    /* \b is backspace in a class */
         else if (-c == ESC_X) c = 'X';   /* \X is literal X in a class */  
         else if (-c == ESC_R) c = 'R';   /* \R is literal R in a class */  
3663          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3664            {            {
3665            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3666              {              {
3667              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3668              }              }
# Line 2836  for (;; ptr++) Line 3676  for (;; ptr++)
3676            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
3677            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3678    
3679            /* Save time by not doing this in the pre-compile phase. */            switch (-c)
   
           if (lengthptr == NULL) switch (-c)  
3680              {              {
3681    #ifdef SUPPORT_UCP
3682                case ESC_du:     /* These are the values given for \d etc */
3683                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3684                case ESC_wu:     /* escape sequence with an appropriate \p */
3685                case ESC_WU:     /* or \P to test Unicode properties instead */
3686                case ESC_su:     /* of the default ASCII testing. */
3687                case ESC_SU:
3688                nestptr = ptr;
3689                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3690                class_charcount -= 2;                /* Undo! */
3691                continue;
3692    #endif
3693              case ESC_d:              case ESC_d:
3694              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3695              continue;              continue;
# Line 2858  for (;; ptr++) Line 3708  for (;; ptr++)
3708              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3709              continue;              continue;
3710    
3711                /* Perl 5.004 onwards omits VT from \s, but we must preserve it
3712                if it was previously set by something earlier in the character
3713                class. */
3714    
3715              case ESC_s:              case ESC_s:
3716              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];              classbits[0] |= cbits[cbit_space];
3717              classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= cbits[cbit_space+1] & ~0x08;
3718                for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3719              continue;              continue;
3720    
3721              case ESC_S:              case ESC_S:
# Line 2869  for (;; ptr++) Line 3724  for (;; ptr++)
3724              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3725              continue;              continue;
3726    
3727              default:    /* Not recognized; fall through */              case ESC_h:
             break;      /* Need "default" setting to stop compiler warning. */  
             }  
   
           /* In the pre-compile phase, just do the recognition. */  
   
           else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||  
                    c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;  
   
           /* We need to deal with \H, \h, \V, and \v in both phases because  
           they use extra memory. */  
   
           if (-c == ESC_h)  
             {  
3728              SETBIT(classbits, 0x09); /* VT */              SETBIT(classbits, 0x09); /* VT */
3729              SETBIT(classbits, 0x20); /* SPACE */              SETBIT(classbits, 0x20); /* SPACE */
3730              SETBIT(classbits, 0xa0); /* NSBP */              SETBIT(classbits, 0xa0); /* NSBP */
# Line 2906  for (;; ptr++) Line 3748  for (;; ptr++)
3748                }                }
3749  #endif  #endif
3750              continue;              continue;
             }  
3751    
3752            if (-c == ESC_H)              case ESC_H:
             {  
3753              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3754                {                {
3755                int x = 0xff;                int x = 0xff;
# Line 2951  for (;; ptr++) Line 3791  for (;; ptr++)
3791                }                }
3792  #endif  #endif
3793              continue;              continue;
             }  
3794    
3795            if (-c == ESC_v)              case ESC_v:
             {  
3796              SETBIT(classbits, 0x0a); /* LF */              SETBIT(classbits, 0x0a); /* LF */
3797              SETBIT(classbits, 0x0b); /* VT */              SETBIT(classbits, 0x0b); /* VT */
3798              SETBIT(classbits, 0x0c); /* FF */              SETBIT(classbits, 0x0c); /* FF */
# Line 2970  for (;; ptr++) Line 3808  for (;; ptr++)
3808                }                }
3809  #endif  #endif
3810              continue;              continue;
             }  
3811    
3812            if (-c == ESC_V)              case ESC_V:
             {  
3813              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3814                {                {
3815                int x = 0xff;                int x = 0xff;
# Line 3003  for (;; ptr++) Line 3839  for (;; ptr++)
3839                }                }
3840  #endif  #endif
3841              continue;              continue;
             }  
   
           /* We need to deal with \P and \p in both phases. */  
3842    
3843  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3844            if (-c == ESC_p || -c == ESC_P)              case ESC_p:
3845              {              case ESC_P:
3846              BOOL negated;                {
3847              int pdata;                BOOL negated;
3848              int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                int pdata;
3849              if (ptype < 0) goto FAILED;                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3850              class_utf8 = TRUE;                if (ptype < 0) goto FAILED;
3851              *class_utf8data++ = ((-c == ESC_p) != negated)?                class_utf8 = TRUE;
3852                XCL_PROP : XCL_NOTPROP;                *class_utf8data++ = ((-c == ESC_p) != negated)?
3853              *class_utf8data++ = ptype;                  XCL_PROP : XCL_NOTPROP;
3854              *class_utf8data++ = pdata;                *class_utf8data++ = ptype;
3855              class_charcount -= 2;   /* Not a < 256 character */                *class_utf8data++ = pdata;
3856              continue;                class_charcount -= 2;   /* Not a < 256 character */
3857              }                continue;
3858                  }
3859  #endif  #endif
3860            /* Unrecognized escapes are faulted if PCRE is running in its              /* Unrecognized escapes are faulted if PCRE is running in its
3861            strict mode. By default, for compatibility with Perl, they are              strict mode. By default, for compatibility with Perl, they are
3862            treated as literals. */              treated as literals. */
3863    
3864            if ((options & PCRE_EXTRA) != 0)              default:
3865              {              if ((options & PCRE_EXTRA) != 0)
3866              *errorcodeptr = ERR7;                {
3867              goto FAILED;                *errorcodeptr = ERR7;
3868                  goto FAILED;
3869                  }
3870                class_charcount -= 2;  /* Undo the default count from above */
3871                c = *ptr;              /* Get the final character and fall through */
3872                break;
3873              }              }
   
           class_charcount -= 2;  /* Undo the default count from above */  
           c = *ptr;              /* Get the final character and fall through */  
3874            }            }
3875    
3876          /* 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 3048  for (;; ptr++) Line 3884  for (;; ptr++)
3884        entirely. The code for handling \Q and \E is messy. */        entirely. The code for handling \Q and \E is messy. */
3885    
3886        CHECK_RANGE:        CHECK_RANGE:
3887        while (ptr[1] == '\\' && ptr[2] == 'E')        while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3888          {          {
3889          inescq = FALSE;          inescq = FALSE;
3890          ptr += 2;          ptr += 2;
# Line 3058  for (;; ptr++) Line 3894  for (;; ptr++)
3894    
3895        /* Remember \r or \n */        /* Remember \r or \n */
3896    
3897        if (c == '\r' || c == '\n') cd->external_flags |= PCRE_HASCRORLF;        if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3898    
3899        /* Check for range */        /* Check for range */
3900    
3901        if (!inescq && ptr[1] == '-')        if (!inescq && ptr[1] == CHAR_MINUS)
3902          {          {
3903          int d;          int d;
3904          ptr += 2;          ptr += 2;
3905          while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3906    
3907          /* 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
3908          mode. */          mode. */
3909    
3910          while (*ptr == '\\' && ptr[1] == 'Q')          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3911            {            {
3912            ptr += 2;            ptr += 2;
3913            if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }            if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3914                { ptr += 2; continue; }
3915            inescq = TRUE;            inescq = TRUE;
3916            break;            break;
3917            }            }
3918    
3919          if (*ptr == 0 || (!inescq && *ptr == ']'))          if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3920            {            {
3921            ptr = oldptr;            ptr = oldptr;
3922            goto LONE_SINGLE_CHARACTER;            goto LONE_SINGLE_CHARACTER;
# Line 3098  for (;; ptr++) Line 3935  for (;; ptr++)
3935          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
3936          in such circumstances. */          in such circumstances. */
3937    
3938          if (!inescq && d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3939            {            {
3940            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3941            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3942    
3943            /* \b is backspace; \X is literal X; \R is literal R; any other            /* \b is backspace; any other special means the '-' was literal */
           special means the '-' was literal */  
3944    
3945            if (d < 0)            if (d < 0)
3946              {              {
3947              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  
3948                {                {
3949                ptr = oldptr;                ptr = oldptr;
3950                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 3131  for (;; ptr++) Line 3965  for (;; ptr++)
3965    
3966          /* Remember \r or \n */          /* Remember \r or \n */
3967    
3968          if (d == '\r' || d == '\n') cd->external_flags |= PCRE_HASCRORLF;          if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3969    
3970          /* 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
3971          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 3251  for (;; ptr++) Line 4085  for (;; ptr++)
4085          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
4086            {            {
4087            unsigned int othercase;            unsigned int othercase;
4088            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)            if ((othercase = UCD_OTHERCASE(c)) != c)
4089              {              {
4090              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
4091              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 3276  for (;; ptr++) Line 4110  for (;; ptr++)
4110          }          }
4111        }        }
4112    
4113      /* 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.
4114        If we are at the end of an internal nested string, revert to the outer
4115        string. */
4116    
4117        while (((c = *(++ptr)) != 0 ||
4118               (nestptr != NULL &&
4119                 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != 0)) &&
4120               (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
4121    
4122      while ((c = *(++ptr)) != 0 && (c != ']' || inescq));      /* Check for missing terminating ']' */
4123    
4124      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)
4125        {        {
4126        *errorcodeptr = ERR6;        *errorcodeptr = ERR6;
4127        goto FAILED;        goto FAILED;
4128        }        }
4129    
   
 /* 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  
   
   
4130      /* 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
4131      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
4132      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 3312  we set the flag only if there is a liter Line 4134  we set the flag only if there is a liter
4134    
4135      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
4136      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
4137      operate on single-bytes only. This is an historical hangover. Maybe one day      operate on single-bytes characters only. This is an historical hangover.
4138      we can tidy these opcodes to handle multi-byte characters.      Maybe one day we can tidy these opcodes to handle multi-byte characters.
4139    
4140      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
4141      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.
4142      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
4143      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
4144      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
4145      reqbyte, save the previous value for reinstating. */      of reqbyte, save the previous value for reinstating. */
4146    
4147  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4148      if (class_charcount == 1 && !class_utf8 &&      if (class_charcount == 1 && !class_utf8 &&
# Line 3331  we set the flag only if there is a liter Line 4153  we set the flag only if there is a liter
4153        {        {
4154        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
4155    
4156        /* The OP_NOT opcode works on one-byte characters only. */        /* The OP_NOT[I] opcodes work on one-byte characters only. */
4157    
4158        if (negate_class)        if (negate_class)
4159          {          {
4160          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
4161          zerofirstbyte = firstbyte;          zerofirstbyte = firstbyte;
4162          *code++ = OP_NOT;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
4163          *code++ = class_lastchar;          *code++ = class_lastchar;
4164          break;          break;
4165          }          }
# Line 3368  we set the flag only if there is a liter Line 4190  we set the flag only if there is a liter
4190    
4191      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
4192      extended class, with its own opcode, unless there was a negated special      extended class, with its own opcode, unless there was a negated special
4193      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
4194      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
4195      (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
4196      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
4197        actual compiled code. */
4198    
4199  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4200      if (class_utf8 && !should_flip_negation)      if (class_utf8 && (!should_flip_negation || (options & PCRE_UCP) != 0))
4201        {        {
4202        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
4203        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 3400  we set the flag only if there is a liter Line 4223  we set the flag only if there is a liter
4223        }        }
4224  #endif  #endif
4225    
4226      /* 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
4227      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
4228      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
4229      map into the code vector, negating it if necessary. */      (non-UCP) in the class. Then copy the 32-byte map into the code vector,
4230        negating it if necessary. */
4231    
4232      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
4233      if (negate_class)      if (negate_class)
# Line 3423  we set the flag only if there is a liter Line 4247  we set the flag only if there is a liter
4247      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
4248      has been tested above. */      has been tested above. */
4249    
4250      case '{':      case CHAR_LEFT_CURLY_BRACKET:
4251      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
4252      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
4253      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
4254      goto REPEAT;      goto REPEAT;
4255    
4256      case '*':      case CHAR_ASTERISK:
4257      repeat_min = 0;      repeat_min = 0;
4258      repeat_max = -1;      repeat_max = -1;
4259      goto REPEAT;      goto REPEAT;
4260    
4261      case '+':      case CHAR_PLUS:
4262      repeat_min = 1;      repeat_min = 1;
4263      repeat_max = -1;      repeat_max = -1;
4264      goto REPEAT;      goto REPEAT;
4265    
4266      case '?':      case CHAR_QUESTION_MARK:
4267      repeat_min = 0;      repeat_min = 0;
4268      repeat_max = 1;      repeat_max = 1;
4269    
# Line 3463  we set the flag only if there is a liter Line 4287  we set the flag only if there is a liter
4287      op_type = 0;                    /* Default single-char op codes */      op_type = 0;                    /* Default single-char op codes */
4288      possessive_quantifier = FALSE;  /* Default not possessive quantifier */      possessive_quantifier = FALSE;  /* Default not possessive quantifier */
4289    
4290      /* 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
4291      for an inserted OP_ONCE for the additional '+' extension. */      insert something before it. */
4292    
4293      tempcode = previous;      tempcode = previous;
4294    
# Line 3474  we set the flag only if there is a liter Line 4298  we set the flag only if there is a liter
4298      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
4299      repeat type to the non-default. */      repeat type to the non-default. */
4300    
4301      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
4302        {        {
4303        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
4304        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
4305        ptr++;        ptr++;
4306        }        }
4307      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
4308        {        {
4309        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
4310        ptr++;        ptr++;
4311        }        }
4312      else repeat_type = greedy_default;      else repeat_type = greedy_default;
4313    
4314        /* If previous was a recursion call, wrap it in atomic brackets so that
4315        previous becomes the atomic group. All recursions were so wrapped in the
4316        past, but it no longer happens for non-repeated recursions. In fact, the
4317        repeated ones could be re-implemented independently so as not to need this,
4318        but for the moment we rely on the code for repeating groups. */
4319    
4320        if (*previous == OP_RECURSE)
4321          {
4322          memmove(previous + 1 + LINK_SIZE, previous, 1 + LINK_SIZE);
4323          *previous = OP_ONCE;
4324          PUT(previous, 1, 2 + 2*LINK_SIZE);
4325          previous[2 + 2*LINK_SIZE] = OP_KET;
4326          PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
4327          code += 2 + 2 * LINK_SIZE;
4328          length_prevgroup = 3 + 3*LINK_SIZE;
4329    
4330          /* When actually compiling, we need to check whether this was a forward
4331          reference, and if so, adjust the offset. */
4332    
4333          if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
4334            {
4335            int offset = GET(cd->hwm, -LINK_SIZE);
4336            if (offset == previous + 1 - cd->start_code)
4337              PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
4338            }
4339          }
4340    
4341        /* Now handle repetition for the different types of item. */
4342    
4343      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
4344      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
4345      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
4346      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
4347      instead.  */      instead.  */
4348    
4349      if (*previous == OP_CHAR || *previous == OP_CHARNC)      if (*previous == OP_CHAR || *previous == OP_CHARI)
4350        {        {
4351          op_type = (*previous == OP_CHAR)? 0 : OP_STARI - OP_STAR;
4352    
4353        /* 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
4354        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
4355        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 3527  we set the flag only if there is a liter Line 4382  we set the flag only if there is a liter
4382    
4383        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4384            repeat_max < 0 &&            repeat_max < 0 &&
4385            check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
             options, cd))  
4386          {          {
4387          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4388          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3540  we set the flag only if there is a liter Line 4394  we set the flag only if there is a liter
4394      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
4395      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-
4396      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
4397      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
4398      currently used only for single-byte chars. */      are currently used only for single-byte chars. */
4399    
4400      else if (*previous == OP_NOT)      else if (*previous == OP_NOT || *previous == OP_NOTI)
4401        {        {
4402        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = ((*previous == OP_NOT)? OP_NOTSTAR : OP_NOTSTARI) - OP_STAR;
4403        c = previous[1];        c = previous[1];
4404        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4405            repeat_max < 0 &&            repeat_max < 0 &&
4406            check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4407          {          {
4408          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4409          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3573  we set the flag only if there is a liter Line 4427  we set the flag only if there is a liter
4427    
4428        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4429            repeat_max < 0 &&            repeat_max < 0 &&
4430            check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4431          {          {
4432          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4433          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3595  we set the flag only if there is a liter Line 4449  we set the flag only if there is a liter
4449    
4450        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
4451    
4452          /*--------------------------------------------------------------------*/
4453          /* This code is obsolete from release 8.00; the restriction was finally
4454          removed: */
4455    
4456        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4457        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4458    
4459        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4460          /*--------------------------------------------------------------------*/
4461    
4462        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
4463    
# Line 3737  we set the flag only if there is a liter Line 4596  we set the flag only if there is a liter
4596  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4597               *previous == OP_XCLASS ||               *previous == OP_XCLASS ||
4598  #endif  #endif
4599               *previous == OP_REF)               *previous == OP_REF ||
4600                 *previous == OP_REFI)
4601        {        {
4602        if (repeat_max == 0)        if (repeat_max == 0)
4603          {          {
# Line 3745  we set the flag only if there is a liter Line 4605  we set the flag only if there is a liter
4605          goto END_REPEAT;          goto END_REPEAT;
4606          }          }
4607    
4608          /*--------------------------------------------------------------------*/
4609          /* This code is obsolete from release 8.00; the restriction was finally
4610          removed: */
4611    
4612        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4613        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4614    
4615        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4616          /*--------------------------------------------------------------------*/
4617    
4618        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4619          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 3766  we set the flag only if there is a liter Line 4631  we set the flag only if there is a liter
4631        }        }
4632    
4633      /* 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
4634      cases. */      cases. Note that at this point we can encounter only the "basic" bracket
4635        opcodes such as BRA and CBRA, as this is the place where they get converted
4636        into the more special varieties such as BRAPOS and SBRA. A test for >=
4637        OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
4638        ASSERTBACK_NOT, ONCE, BRA, CBRA, and COND. Originally, PCRE did not allow
4639        repetition of assertions, but now it does, for Perl compatibility. */
4640    
4641      else if (*previous == OP_BRA  || *previous == OP_CBRA ||      else if (*previous >= OP_ASSERT && *previous <= OP_COND)
              *previous == OP_ONCE || *previous == OP_COND)  
4642        {        {
4643        register int i;        register int i;
4644        int ketoffset = 0;        int len = (int)(code - previous);
       int len = code - previous;  
4645        uschar *bralink = NULL;        uschar *bralink = NULL;
4646          uschar *brazeroptr = NULL;
4647    
4648        /* Repeating a DEFINE group is pointless */        /* Repeating a DEFINE group is pointless, but Perl allows the syntax, so
4649          we just ignore the repeat. */
4650    
4651        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
4652          {          goto END_REPEAT;
4653          *errorcodeptr = ERR55;  
4654          goto FAILED;        /* There is no sense in actually repeating assertions. The only potential
4655          }        use of repetition is in cases when the assertion is optional. Therefore,
4656          if the minimum is greater than zero, just ignore the repeat. If the
4657          maximum is not not zero or one, set it to 1. */
4658    
4659        /* If the maximum repeat count is unlimited, find the end of the bracket        if (*previous < OP_ONCE)    /* Assertion */
4660        by scanning through from the start, and compute the offset back to it          {
4661        from the current code pointer. There may be an OP_OPT setting following          if (repeat_min > 0) goto END_REPEAT;
4662        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;  
4663          }          }
4664    
4665        /* 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 3806  we set the flag only if there is a liter Line 4671  we set the flag only if there is a liter
4671    
4672        if (repeat_min == 0)        if (repeat_min == 0)
4673          {          {
4674          /* 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
4675          altogether. */          output altogether, like this:
   
         if (repeat_max == 0)  
           {  
           code = previous;  
           goto END_REPEAT;  
           }  
4676    
4677          /* If the maximum is 1 or unlimited, we just have to stick in the          ** if (repeat_max == 0)
4678          BRAZERO and do no more at this point. However, we do need to adjust          **   {
4679          any OP_RECURSE calls inside the group that refer to the group itself or          **   code = previous;
4680          any internal or forward referenced group, because the offset is from          **   goto END_REPEAT;
4681          the start of the whole regex. Temporarily terminate the pattern while          **   }
4682          doing this. */  
4683            However, that fails when a group or a subgroup within it is referenced
4684            as a subroutine from elsewhere in the pattern, so now we stick in
4685            OP_SKIPZERO in front of it so that it is skipped on execution. As we
4686            don't have a list of which groups are referenced, we cannot do this
4687            selectively.
4688    
4689            If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4690            and do no more at this point. However, we do need to adjust any
4691            OP_RECURSE calls inside the group that refer to the group itself or any
4692            internal or forward referenced group, because the offset is from the
4693            start of the whole regex. Temporarily terminate the pattern while doing
4694            this. */
4695    
4696          if (repeat_max <= 1)          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
4697            {            {
4698            *code = OP_END;            *code = OP_END;
4699            adjust_recurse(previous, 1, utf8, cd, save_hwm);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
4700            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
4701            code++;            code++;
4702              if (repeat_max == 0)
4703                {
4704                *previous++ = OP_SKIPZERO;
4705                goto END_REPEAT;
4706                }
4707              brazeroptr = previous;    /* Save for possessive optimizing */
4708            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4709            }            }
4710    
# Line 3852  we set the flag only if there is a liter Line 4729  we set the flag only if there is a liter
4729            /* We chain together the bracket offset fields that have to be            /* We chain together the bracket offset fields that have to be
4730            filled in later when the ends of the brackets are reached. */            filled in later when the ends of the brackets are reached. */
4731    
4732            offset = (bralink == NULL)? 0 : previous - bralink;            offset = (bralink == NULL)? 0 : (int)(previous - bralink);
4733            bralink = previous;            bralink = previous;
4734            PUTINC(previous, 0, offset);            PUTINC(previous, 0, offset);
4735            }            }
# Line 3873  we set the flag only if there is a liter Line 4750  we set the flag only if there is a liter
4750            {            {
4751            /* 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
4752            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
4753            potential integer overflow. */            potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
4754              integer type when available, otherwise double. */
4755    
4756            if (lengthptr != NULL)            if (lengthptr != NULL)
4757              {              {
4758              int delta = (repeat_min - 1)*length_prevgroup;              int delta = (repeat_min - 1)*length_prevgroup;
4759              if ((double)(repeat_min - 1)*(double)length_prevgroup >              if ((INT64_OR_DOUBLE)(repeat_min - 1)*
4760                                                              (double)INT_MAX ||                    (INT64_OR_DOUBLE)length_prevgroup >
4761                        (INT64_OR_DOUBLE)INT_MAX ||
4762                  OFLOW_MAX - *lengthptr < delta)                  OFLOW_MAX - *lengthptr < delta)
4763                {                {
4764                *errorcodeptr = ERR20;                *errorcodeptr = ERR20;
# Line 3925  we set the flag only if there is a liter Line 4804  we set the flag only if there is a liter
4804          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
4805          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
4806          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
4807          paranoid checks to avoid integer overflow. */          paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
4808            a 64-bit integer type when available, otherwise double. */
4809    
4810          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4811            {            {
4812            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4813                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4814            if ((double)repeat_max *            if ((INT64_OR_DOUBLE)repeat_max *
4815                  (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)                  (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4816                    > (double)INT_MAX ||                    > (INT64_OR_DOUBLE)INT_MAX ||
4817                OFLOW_MAX - *lengthptr < delta)                OFLOW_MAX - *lengthptr < delta)
4818              {              {
4819              *errorcodeptr = ERR20;              *errorcodeptr = ERR20;
# Line 3958  we set the flag only if there is a liter Line 4838  we set the flag only if there is a liter
4838              {              {
4839              int offset;              int offset;
4840              *code++ = OP_BRA;              *code++ = OP_BRA;
4841              offset = (bralink == NULL)? 0 : code - bralink;              offset = (bralink == NULL)? 0 : (int)(code - bralink);
4842              bralink = code;              bralink = code;
4843              PUTINC(code, 0, offset);              PUTINC(code, 0, offset);
4844              }              }
# Line 3979  we set the flag only if there is a liter Line 4859  we set the flag only if there is a liter
4859          while (bralink != NULL)          while (bralink != NULL)
4860            {            {
4861            int oldlinkoffset;            int oldlinkoffset;
4862            int offset = code - bralink + 1;            int offset = (int)(code - bralink + 1);
4863            uschar *bra = code - offset;            uschar *bra = code - offset;
4864            oldlinkoffset = GET(bra, 1);            oldlinkoffset = GET(bra, 1);
4865            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;
# Line 3989  we set the flag only if there is a liter Line 4869  we set the flag only if there is a liter
4869            }            }
4870          }          }
4871    
4872        /* If the maximum is unlimited, set a repeater in the final copy. We        /* If the maximum is unlimited, set a repeater in the final copy. For
4873        can't just offset backwards from the current code point, because we        ONCE brackets, that's all we need to do. However, possessively repeated
4874        don't know if there's been an options resetting after the ket. The        ONCE brackets can be converted into non-capturing brackets, as the
4875        correct offset was computed above.        behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to
4876          deal with possessive ONCEs specially.
4877    
4878          Otherwise, if the quantifier was possessive, we convert the BRA code to
4879          the POS form, and the KET code to KETRPOS. (It turns out to be convenient
4880          at runtime to detect this kind of subpattern at both the start and at the
4881          end.) The use of special opcodes makes it possible to reduce greatly the
4882          stack usage in pcre_exec(). If the group is preceded by OP_BRAZERO,
4883          convert this to OP_BRAPOSZERO. Then cancel the possessive flag so that
4884          the default action below, of wrapping everything inside atomic brackets,
4885          does not happen.
4886    
4887        Then, when we are doing the actual compile phase, check to see whether        Then, when we are doing the actual compile phase, check to see whether
4888        this group is a non-atomic one that could match an empty string. If so,        this group is one that could match an empty string. If so, convert the
4889        convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so        initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so that runtime
4890        that runtime checking can be done. [This check is also applied to        checking can be done. [This check is also applied to ONCE groups at
4891        atomic groups at runtime, but in a different way.] */        runtime, but in a different way.] */
4892    
4893        else        else
4894          {          {
4895          uschar *ketcode = code - ketoffset;          uschar *ketcode = code - 1 - LINK_SIZE;
4896          uschar *bracode = ketcode - GET(ketcode, 1);          uschar *bracode = ketcode - GET(ketcode, 1);
4897          *ketcode = OP_KETRMAX + repeat_type;  
4898          if (lengthptr == NULL && *bracode != OP_ONCE)          if ((*bracode == OP_ONCE || *bracode == OP_ONCE_NC) &&
4899                possessive_quantifier) *bracode = OP_BRA;
4900    
4901            if (*bracode == OP_ONCE || *bracode == OP_ONCE_NC)
4902              *ketcode = OP_KETRMAX + repeat_type;
4903            else
4904            {            {
4905            uschar *scode = bracode;            if (possessive_quantifier)
4906            do              {
4907                *bracode += 1;                   /* Switch to xxxPOS opcodes */
4908                *ketcode = OP_KETRPOS;
4909                if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
4910                possessive_quantifier = FALSE;
4911                }
4912              else *ketcode = OP_KETRMAX + repeat_type;
4913    
4914              if (lengthptr == NULL)
4915              {              {
4916              if (could_be_empty_branch(scode, ketcode, utf8))              uschar *scode = bracode;
4917                do
4918                {                {
4919                *bracode += OP_SBRA - OP_BRA;                if (could_be_empty_branch(scode, ketcode, utf8, cd))
4920                break;                  {
4921                    *bracode += OP_SBRA - OP_BRA;
4922                    break;
4923                    }
4924                  scode += GET(scode, 1);
4925                }                }
4926              scode += GET(scode, 1);              while (*scode == OP_ALT);
4927              }              }
           while (*scode == OP_ALT);  
4928            }            }
4929          }          }
4930        }        }
4931    
4932        /* If previous is OP_FAIL, it was generated by an empty class [] in
4933        JavaScript mode. The other ways in which OP_FAIL can be generated, that is
4934        by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
4935        error above. We can just ignore the repeat in JS case. */
4936    
4937        else if (*previous == OP_FAIL) goto END_REPEAT;
4938    
4939      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
4940    
4941      else      else
# Line 4031  we set the flag only if there is a liter Line 4945  we set the flag only if there is a liter
4945        }        }
4946    
4947      /* If the character following a repeat is '+', or if certain optimization      /* If the character following a repeat is '+', or if certain optimization
4948      tests above succeeded, possessive_quantifier is TRUE. For some of the      tests above succeeded, possessive_quantifier is TRUE. For some opcodes,
4949      simpler opcodes, there is an special alternative opcode for this. For      there are special alternative opcodes for this case. For anything else, we
4950      anything else, we wrap the entire repeated item inside OP_ONCE brackets.      wrap the entire repeated item inside OP_ONCE brackets. Logically, the '+'
4951      The '+' notation is just syntactic sugar, taken from Sun's Java package,      notation is just syntactic sugar, taken from Sun's Java package, but the
4952      but the special opcodes can optimize it a bit. The repeated item starts at      special opcodes can optimize it.
4953      tempcode, not at previous, which might be the first part of a string whose  
4954      (former) last char we repeated.      Possessively repeated subpatterns have already been handled in the code
4955        just above, so possessive_quantifier is always FALSE for them at this
4956        stage.
4957    
4958        Note that the repeated item starts at tempcode, not at previous, which
4959        might be the first part of a string whose (former) last char we repeated.
4960    
4961      Possessifying an 'exact' quantifier has no effect, so we can ignore it. But      Possessifying an 'exact' quantifier has no effect, so we can ignore it. But
4962      an 'upto' may follow. We skip over an 'exact' item, and then test the      an 'upto' may follow. We skip over an 'exact' item, and then test the
# Line 4046  we set the flag only if there is a liter Line 4965  we set the flag only if there is a liter
4965      if (possessive_quantifier)      if (possessive_quantifier)
4966        {        {
4967        int len;        int len;
4968        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||  
4969            *tempcode == OP_NOTEXACT)        if (*tempcode == OP_TYPEEXACT)
4970          tempcode += _pcre_OP_lengths[*tempcode] +          tempcode += _pcre_OP_lengths[*tempcode] +
4971            ((*tempcode == OP_TYPEEXACT &&            ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
4972               (tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP))? 2:0);  
4973        len = code - tempcode;        else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
4974            {
4975            tempcode += _pcre_OP_lengths[*tempcode];
4976    #ifdef SUPPORT_UTF8
4977            if (utf8 && tempcode[-1] >= 0xc0)
4978              tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];
4979    #endif
4980            }
4981    
4982          len = (int)(code - tempcode);
4983        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4984          {          {
4985          case OP_STAR:  *tempcode = OP_POSSTAR; break;          case OP_STAR:  *tempcode = OP_POSSTAR; break;
# Line 4059  we set the flag only if there is a liter Line 4987  we set the flag only if there is a liter
4987          case OP_QUERY: *tempcode = OP_POSQUERY; break;          case OP_QUERY: *tempcode = OP_POSQUERY; break;
4988          case OP_UPTO:  *tempcode = OP_POSUPTO; break;          case OP_UPTO:  *tempcode = OP_POSUPTO; break;
4989    
4990          case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;          case OP_STARI:  *tempcode = OP_POSSTARI; break;
4991          case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;          case OP_PLUSI:  *tempcode = OP_POSPLUSI; break;
4992          case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;          case OP_QUERYI: *tempcode = OP_POSQUERYI; break;
4993          case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;          case OP_UPTOI:  *tempcode = OP_POSUPTOI; break;
4994    
4995          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;
4996          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;
4997          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4998          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4999    
5000            case OP_NOTSTARI:  *tempcode = OP_NOTPOSSTARI; break;
5001            case OP_NOTPLUSI:  *tempcode = OP_NOTPOSPLUSI; break;
5002            case OP_NOTQUERYI: *tempcode = OP_NOTPOSQUERYI; break;
5003            case OP_NOTUPTOI:  *tempcode = OP_NOTPOSUPTOI; break;
5004    
5005            case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;
5006            case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;
5007            case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
5008            case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;
5009    
5010            /* Because we are moving code along, we must ensure that any
5011            pending recursive references are updated. */
5012    
5013          default:          default:
5014            *code = OP_END;
5015            adjust_recurse(tempcode, 1 + LINK_SIZE, utf8, cd, save_hwm);
5016          memmove(tempcode + 1+LINK_SIZE, tempcode, len);          memmove(tempcode + 1+LINK_SIZE, tempcode, len);
5017          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
5018          len += 1 + LINK_SIZE;          len += 1 + LINK_SIZE;
# Line 4096  we set the flag only if there is a liter Line 5039  we set the flag only if there is a liter
5039      lookbehind or option setting or condition or all the other extended      lookbehind or option setting or condition or all the other extended
5040      parenthesis forms.  */      parenthesis forms.  */
5041    
5042      case '(':      case CHAR_LEFT_PARENTHESIS:
5043      newoptions = options;      newoptions = options;
5044      skipbytes = 0;      skipbytes = 0;
5045      bravalue = OP_CBRA;      bravalue = OP_CBRA;
# Line 4105  we set the flag only if there is a liter Line 5048  we set the flag only if there is a liter
5048    
5049      /* First deal with various "verbs" that can be introduced by '*'. */      /* First deal with various "verbs" that can be introduced by '*'. */
5050    
5051      if (*(++ptr) == '*' && (cd->ctypes[ptr[1]] & ctype_letter) != 0)      if (*(++ptr) == CHAR_ASTERISK &&
5052             ((cd->ctypes[ptr[1]] & ctype_letter) != 0 || ptr[1] == ':'))
5053        {        {
5054        int i, namelen;        int i, namelen;
5055          int arglen = 0;
5056        const char *vn = verbnames;        const char *vn = verbnames;
5057        const uschar *name = ++ptr;        const uschar *name = ptr + 1;
5058          const uschar *arg = NULL;
5059        previous = NULL;        previous = NULL;
5060        while ((cd->ctypes[*++ptr] & ctype_letter) != 0);        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
5061        if (*ptr == ':')        namelen = (int)(ptr - name);
5062    
5063          /* It appears that Perl allows any characters whatsoever, other than
5064          a closing parenthesis, to appear in arguments, so we no longer insist on
5065          letters, digits, and underscores. */
5066    
5067          if (*ptr == CHAR_COLON)
5068          {          {
5069          *errorcodeptr = ERR59;   /* Not supported */          arg = ++ptr;
5070          goto FAILED;          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
5071            arglen = (int)(ptr - arg);
5072          }          }
5073        if (*ptr != ')')  
5074          if (*ptr != CHAR_RIGHT_PARENTHESIS)
5075          {          {
5076          *errorcodeptr = ERR60;          *errorcodeptr = ERR60;
5077          goto FAILED;          goto FAILED;
5078          }          }
5079        namelen = ptr - name;  
5080          /* Scan the table of verb names */
5081    
5082        for (i = 0; i < verbcount; i++)        for (i = 0; i < verbcount; i++)
5083          {          {
5084          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
5085              strncmp((char *)name, vn, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
5086            {            {
5087            *code = verbs[i].op;            /* Check for open captures before ACCEPT and convert it to
5088            if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;            ASSERT_ACCEPT if in an assertion. */
5089            break;  
5090              if (verbs[i].op == OP_ACCEPT)
5091                {
5092                open_capitem *oc;
5093                if (arglen != 0)
5094                  {
5095                  *errorcodeptr = ERR59;
5096                  goto FAILED;
5097                  }
5098                cd->had_accept = TRUE;
5099                for (oc = cd->open_caps; oc != NULL; oc = oc->next)
5100                  {
5101                  *code++ = OP_CLOSE;
5102                  PUT2INC(code, 0, oc->number);
5103                  }
5104                *code++ = (cd->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT;
5105    
5106                /* Do not set firstbyte after *ACCEPT */
5107                if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
5108                }
5109    
5110              /* Handle other cases with/without an argument */
5111    
5112              else if (arglen == 0)
5113                {
5114                if (verbs[i].op < 0)   /* Argument is mandatory */
5115                  {
5116                  *errorcodeptr = ERR66;
5117                  goto FAILED;
5118                  }
5119                *code = verbs[i].op;
5120                if (*code++ == OP_THEN) cd->external_flags |= PCRE_HASTHEN;
5121                }
5122    
5123              else
5124                {
5125                if (verbs[i].op_arg < 0)   /* Argument is forbidden */
5126                  {
5127                  *errorcodeptr = ERR59;
5128                  goto FAILED;
5129                  }
5130                *code = verbs[i].op_arg;
5131                if (*code++ == OP_THEN_ARG) cd->external_flags |= PCRE_HASTHEN;
5132                *code++ = arglen;
5133                memcpy(code, arg, arglen);
5134                code += arglen;
5135                *code++ = 0;
5136                }
5137    
5138              break;  /* Found verb, exit loop */
5139            }            }
5140    
5141          vn += verbs[i].len + 1;          vn += verbs[i].len + 1;
5142          }          }
5143        if (i < verbcount) continue;  
5144        *errorcodeptr = ERR60;        if (i < verbcount) continue;    /* Successfully handled a verb */
5145          *errorcodeptr = ERR60;          /* Verb not recognized */
5146        goto FAILED;        goto FAILED;
5147        }        }
5148    
5149      /* Deal with the extended parentheses; all are introduced by '?', and the      /* Deal with the extended parentheses; all are introduced by '?', and the
5150      appearance of any of them means that this is not a capturing group. */      appearance of any of them means that this is not a capturing group. */
5151    
5152      else if (*ptr == '?')      else if (*ptr == CHAR_QUESTION_MARK)
5153        {        {
5154        int i, set, unset, namelen;        int i, set, unset, namelen;
5155        int *optset;        int *optset;
# Line 4151  we set the flag only if there is a liter Line 5158  we set the flag only if there is a liter
5158    
5159        switch (*(++ptr))        switch (*(++ptr))
5160          {          {
5161          case '#':                 /* Comment; skip to ket */          case CHAR_NUMBER_SIGN:                 /* Comment; skip to ket */
5162          ptr++;          ptr++;
5163          while (*ptr != 0 && *ptr != ')') ptr++;          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
5164          if (*ptr == 0)          if (*ptr == 0)
5165            {            {
5166            *errorcodeptr = ERR18;            *errorcodeptr = ERR18;
# Line 4163  we set the flag only if there is a liter Line 5170  we set the flag only if there is a liter
5170    
5171    
5172          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5173          case '|':                 /* Reset capture count for each branch */          case CHAR_VERTICAL_LINE:  /* Reset capture count for each branch */
5174          reset_bracount = TRUE;          reset_bracount = TRUE;
5175          /* Fall through */          /* Fall through */
5176    
5177          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5178          case ':':                 /* Non-capturing bracket */          case CHAR_COLON:          /* Non-capturing bracket */
5179          bravalue = OP_BRA;          bravalue = OP_BRA;
5180          ptr++;          ptr++;
5181          break;          break;
5182    
5183    
5184          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5185          case '(':          case CHAR_LEFT_PARENTHESIS:
5186          bravalue = OP_COND;       /* Conditional group */          bravalue = OP_COND;       /* Conditional group */
5187    
5188          /* A condition can be an assertion, a number (referring to a numbered          /* A condition can be an assertion, a number (referring to a numbered
# Line 4195