/[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 333 by ph10, Thu Apr 10 19:55:57 2008 UTC revision 716 by ph10, Tue Oct 4 16:38:05 2011 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2008 University of Cambridge             Copyright (c) 1997-2011 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 53  supporting internal functions that are n Line 53  supporting internal functions that are n
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56  /* When DEBUG is defined, we need the pcre_printint() function, which is also  /* When PCRE_DEBUG is defined, we need the pcre_printint() function, which is
57  used by pcretest. DEBUG is not defined when building a production library. */  also used by pcretest. PCRE_DEBUG is not defined when building a production
58    library. */
59    
60  #ifdef DEBUG  #ifdef PCRE_DEBUG
61  #include "pcre_printint.src"  #include "pcre_printint.src"
62  #endif  #endif
63    
# Line 91  is 4 there is plenty of room. */ Line 92  is 4 there is plenty of room. */
92    
93  #define COMPILE_WORK_SIZE (4096)  #define COMPILE_WORK_SIZE (4096)
94    
95    /* The overrun tests check for a slightly smaller size so that they detect the
96    overrun before it actually does run off the end of the data block. */
97    
98    #define WORK_SIZE_CHECK (COMPILE_WORK_SIZE - 100)
99    
100    
101  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
102  are simple data values; negative values are for special things like \d and so  are simple data values; negative values are for special things like \d and so
103  on. Zero means further processing is needed (for things like \x), or the escape  on. Zero means further processing is needed (for things like \x), or the escape
104  is invalid. */  is invalid. */
105    
106  #ifndef EBCDIC  /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC
107    
108    /* This is the "normal" table for ASCII systems or for EBCDIC systems running
109    in UTF-8 mode. */
110    
111  static const short int escapes[] = {  static const short int escapes[] = {
112       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,                       0,
113       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,                       0,
114     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */       0,                       0,
115  -ESC_H,      0,      0, -ESC_K,      0,      0,      0,      0,   /* H - O */       0,                       0,
116  -ESC_P, -ESC_Q, -ESC_R, -ESC_S,      0,      0, -ESC_V, -ESC_W,   /* P - W */       0,                       0,
117  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */       CHAR_COLON,              CHAR_SEMICOLON,
118     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
119  -ESC_h,      0,      0, -ESC_k,      0,      0,  ESC_n,      0,   /* h - o */       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
120  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0, -ESC_v, -ESC_w,   /* p - w */       CHAR_COMMERCIAL_AT,      -ESC_A,
121       0,      0, -ESC_z                                            /* x - z */       -ESC_B,                  -ESC_C,
122         -ESC_D,                  -ESC_E,
123         0,                       -ESC_G,
124         -ESC_H,                  0,
125         0,                       -ESC_K,
126         0,                       0,
127         -ESC_N,                  0,
128         -ESC_P,                  -ESC_Q,
129         -ESC_R,                  -ESC_S,
130         0,                       0,
131         -ESC_V,                  -ESC_W,
132         -ESC_X,                  0,
133         -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,
134         CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,
135         CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,
136         CHAR_GRAVE_ACCENT,       7,
137         -ESC_b,                  0,
138         -ESC_d,                  ESC_e,
139         ESC_f,                   0,
140         -ESC_h,                  0,
141         0,                       -ESC_k,
142         0,                       0,
143         ESC_n,                   0,
144         -ESC_p,                  0,
145         ESC_r,                   -ESC_s,
146         ESC_tee,                 0,
147         -ESC_v,                  -ESC_w,
148         0,                       0,
149         -ESC_z
150  };  };
151    
152  #else           /* This is the "abnormal" table for EBCDIC systems */  #else
153    
154    /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
155    
156  static const short int escapes[] = {  static const short int escapes[] = {
157  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
158  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 130  static const short int escapes[] = { Line 171  static const short int escapes[] = {
171  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
172  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,
173  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
174  /*  D0 */   '}',     0, -ESC_K,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,-ESC_N,      0, -ESC_P,
175  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
176  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
177  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
# Line 142  static const short int escapes[] = { Line 183  static const short int escapes[] = {
183    
184  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
185  searched linearly. Put all the names into a single string, in order to reduce  searched linearly. Put all the names into a single string, in order to reduce
186  the number of relocations when a shared library is dynamically linked. */  the number of relocations when a shared library is dynamically linked. The
187    string is built from string macros so that it works in UTF-8 mode on EBCDIC
188    platforms. */
189    
190  typedef struct verbitem {  typedef struct verbitem {
191    int   len;    int   len;                 /* Length of verb name */
192    int   op;    int   op;                  /* Op when no arg, or -1 if arg mandatory */
193      int   op_arg;              /* Op when arg present, or -1 if not allowed */
194  } verbitem;  } verbitem;
195    
196  static const char verbnames[] =  static const char verbnames[] =
197    "ACCEPT\0"    "\0"                       /* Empty name is a shorthand for MARK */
198    "COMMIT\0"    STRING_MARK0
199    "F\0"    STRING_ACCEPT0
200    "FAIL\0"    STRING_COMMIT0
201    "PRUNE\0"    STRING_F0
202    "SKIP\0"    STRING_FAIL0
203    "THEN";    STRING_PRUNE0
204      STRING_SKIP0
205      STRING_THEN;
206    
207  static const verbitem verbs[] = {  static const verbitem verbs[] = {
208    { 6, OP_ACCEPT },    { 0, -1,        OP_MARK },
209    { 6, OP_COMMIT },    { 4, -1,        OP_MARK },
210    { 1, OP_FAIL },    { 6, OP_ACCEPT, -1 },
211    { 4, OP_FAIL },    { 6, OP_COMMIT, -1 },
212    { 5, OP_PRUNE },    { 1, OP_FAIL,   -1 },
213    { 4, OP_SKIP  },    { 4, OP_FAIL,   -1 },
214    { 4, OP_THEN  }    { 5, OP_PRUNE,  OP_PRUNE_ARG },
215      { 4, OP_SKIP,   OP_SKIP_ARG  },
216      { 4, OP_THEN,   OP_THEN_ARG  }
217  };  };
218    
219  static const int verbcount = sizeof(verbs)/sizeof(verbitem);  static const int verbcount = sizeof(verbs)/sizeof(verbitem);
# 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, angle-bracketed, or quoted name/number or by a plain number\0"    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
399    "a numbered reference must not be zero\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:
679      case 'N':      case CHAR_u:
680      case 'u':      case CHAR_U:
     case 'U':  
681      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
682      break;      break;
683    
684      /* \g must be followed by one of a number of specific things:      /* In a character class, \g is just a literal "g". Outside a character
685        class, \g must be followed by one of a number of specific things:
686    
687      (1) A number, either plain or braced. If positive, it is an absolute      (1) A number, either plain or braced. If positive, it is an absolute
688      backreference. If negative, it is a relative backreference. This is a Perl      backreference. If negative, it is a relative backreference. This is a Perl
689      5.10 feature.      5.10 feature.
690    
691      (2) Perl 5.10 also supports \g{name} as a reference to a named group. This      (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
692      is part of Perl's movement towards a unified syntax for back references. As      is part of Perl's movement towards a unified syntax for back references. As
693      this is synonymous with \k{name}, we fudge it up by pretending it really      this is synonymous with \k{name}, we fudge it up by pretending it really
694      was \k.      was \k.
695    
696      (3) For Oniguruma compatibility we also support \g followed by a name or a      (3) For Oniguruma compatibility we also support \g followed by a name or a
697      number either in angle brackets or in single quotes. However, these are      number either in angle brackets or in single quotes. However, these are
698      (possibly recursive) subroutine calls, _not_ backreferences. Just return      (possibly recursive) subroutine calls, _not_ backreferences. Just return
699      the -ESC_g code (cf \k). */      the -ESC_g code (cf \k). */
700    
701      case 'g':      case CHAR_g:
702      if (ptr[1] == '<' || ptr[1] == '\'')      if (isclass) break;
703        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
704        {        {
705        c = -ESC_g;        c = -ESC_g;
706        break;        break;
707        }        }
708    
709      /* Handle the Perl-compatible cases */      /* Handle the Perl-compatible cases */
710    
711      if (ptr[1] == '{')      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
712        {        {
713        const uschar *p;        const uschar *p;
714        for (p = ptr+2; *p != 0 && *p != '}'; p++)        for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
715          if (*p != '-' && (digitab[*p] & ctype_digit) == 0) break;          if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
716        if (*p != 0 && *p != '}')        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
717          {          {
718          c = -ESC_k;          c = -ESC_k;
719          break;          break;
# Line 571  else Line 723  else
723        }        }
724      else braced = FALSE;      else braced = FALSE;
725    
726      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
727        {        {
728        negated = TRUE;        negated = TRUE;
729        ptr++;        ptr++;
# Line 580  else Line 732  else
732    
733      c = 0;      c = 0;
734      while ((digitab[ptr[1]] & ctype_digit) != 0)      while ((digitab[ptr[1]] & ctype_digit) != 0)
735        c = c * 10 + *(++ptr) - '0';        c = c * 10 + *(++ptr) - CHAR_0;
736    
737      if (c < 0)   /* Integer overflow */      if (c < 0)   /* Integer overflow */
738        {        {
739        *errorcodeptr = ERR61;        *errorcodeptr = ERR61;
740        break;        break;
741        }        }
742    
743      if (braced && *(++ptr) != '}')      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
744        {        {
745        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
746        break;        break;
747        }        }
748    
749      if (c == 0)      if (c == 0)
750        {        {
751        *errorcodeptr = ERR58;        *errorcodeptr = ERR58;
752        break;        break;
753        }        }
754    
755      if (negated)      if (negated)
756        {        {
# Line 625  else Line 777  else
777      value is greater than 377, the least significant 8 bits are taken. Inside a      value is greater than 377, the least significant 8 bits are taken. Inside a
778      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
779    
780      case '1': case '2': case '3': case '4': case '5':      case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:
781      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
782    
783      if (!isclass)      if (!isclass)
784        {        {
785        oldptr = ptr;        oldptr = ptr;
786        c -= '0';        c -= CHAR_0;
787        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
788          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - CHAR_0;
789        if (c < 0)    /* Integer overflow */        if (c < 0)    /* Integer overflow */
790          {          {
791          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
# Line 651  else Line 803  else
803      generates a binary zero byte and treats the digit as a following literal.      generates a binary zero byte and treats the digit as a following literal.
804      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
805    
806      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
807        {        {
808        ptr--;        ptr--;
809        c = 0;        c = 0;
# Line 664  else Line 816  else
816      to do). Nowadays we allow for larger numbers in UTF-8 mode, but no more      to do). Nowadays we allow for larger numbers in UTF-8 mode, but no more
817      than 3 octal digits. */      than 3 octal digits. */
818    
819      case '0':      case CHAR_0:
820      c -= '0';      c -= CHAR_0;
821      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
822          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
823      if (!utf8 && c > 255) *errorcodeptr = ERR51;      if (!utf8 && c > 255) *errorcodeptr = ERR51;
824      break;      break;
825    
# Line 675  else Line 827  else
827      than 0xff in utf8 mode, but only if the ddd are hex digits. If not, { is      than 0xff in utf8 mode, but only if the ddd are hex digits. If not, { is
828      treated as a data character. */      treated as a data character. */
829    
830      case 'x':      case CHAR_x:
831      if (ptr[1] == '{')      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
832        {        {
833        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
834        int count = 0;        int count = 0;
# Line 685  else Line 837  else
837        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
838          {          {
839          register int cc = *pt++;          register int cc = *pt++;
840          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
841          count++;          count++;
842    
843  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
844          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
845          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
846  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
847          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
848          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
849  #endif  #endif
850          }          }
851    
852        if (*pt == '}')        if (*pt == CHAR_RIGHT_CURLY_BRACKET)
853          {          {
854          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
855          ptr = pt;          ptr = pt;
# Line 713  else Line 865  else
865      c = 0;      c = 0;
866      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
867        {        {
868        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
869        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
870  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
871        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
872        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
873  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
874        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
875        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
876  #endif  #endif
877        }        }
878      break;      break;
879    
880      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
881      This coding is ASCII-specific, but then the whole concept of \cx is      An error is given if the byte following \c is not an ASCII character. This
882        coding is ASCII-specific, but then the whole concept of \cx is
883      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
884    
885      case 'c':      case CHAR_c:
886      c = *(++ptr);      c = *(++ptr);
887      if (c == 0)      if (c == 0)
888        {        {
889        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
890        break;        break;
891        }        }
892    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
893  #ifndef EBCDIC  /* ASCII coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
894      if (c >= 'a' && c <= 'z') c -= 32;        {
895          *errorcodeptr = ERR68;
896          break;
897          }
898        if (c >= CHAR_a && c <= CHAR_z) c -= 32;
899      c ^= 0x40;      c ^= 0x40;
900  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
901      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
902      c ^= 0xC0;      c ^= 0xC0;
903  #endif  #endif
904      break;      break;
# Line 763  else Line 920  else
920      }      }
921    }    }
922    
923    /* Perl supports \N{name} for character names, as well as plain \N for "not
924    newline". PCRE does not support \N{name}. However, it does support
925    quantification such as \N{2,3}. */
926    
927    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
928         !is_counted_repeat(ptr+2))
929      *errorcodeptr = ERR37;
930    
931    /* If PCRE_UCP is set, we change the values for \d etc. */
932    
933    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
934      c -= (ESC_DU - ESC_D);
935    
936    /* Set the pointer to the final character before returning. */
937    
938  *ptrptr = ptr;  *ptrptr = ptr;
939  return c;  return c;
940  }  }
# Line 803  if (c == 0) goto ERROR_RETURN; Line 975  if (c == 0) goto ERROR_RETURN;
975  /* \P or \p can be followed by a name in {}, optionally preceded by ^ for  /* \P or \p can be followed by a name in {}, optionally preceded by ^ for
976  negation. */  negation. */
977    
978  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
979    {    {
980    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
981      {      {
982      *negptr = TRUE;      *negptr = TRUE;
983      ptr++;      ptr++;
# Line 814  if (c == '{') Line 986  if (c == '{')
986      {      {
987      c = *(++ptr);      c = *(++ptr);
988      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
989      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
990      name[i] = c;      name[i] = c;
991      }      }
992    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
993    name[i] = 0;    name[i] = 0;
994    }    }
995    
# Line 863  return -1; Line 1035  return -1;
1035    
1036    
1037  /*************************************************  /*************************************************
 *            Check for counted repeat            *  
 *************************************************/  
   
 /* This function is called when a '{' is encountered in a place where it might  
 start a quantifier. It looks ahead to see if it really is a quantifier or not.  
 It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}  
 where the ddds are digits.  
   
 Arguments:  
   p         pointer to the first char after '{'  
   
 Returns:    TRUE or FALSE  
 */  
   
 static BOOL  
 is_counted_repeat(const uschar *p)  
 {  
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
 if (*p == '}') return TRUE;  
   
 if (*p++ != ',') return FALSE;  
 if (*p == '}') return TRUE;  
   
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
   
 return (*p == '}');  
 }  
   
   
   
 /*************************************************  
1038  *         Read repeat counts                     *  *         Read repeat counts                     *
1039  *************************************************/  *************************************************/
1040    
# Line 923  int max = -1; Line 1062  int max = -1;
1062  /* Read the minimum value and do a paranoid check: a negative value indicates  /* Read the minimum value and do a paranoid check: a negative value indicates
1063  an integer overflow. */  an integer overflow. */
1064    
1065  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;
1066  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1067    {    {
1068    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 933  if (min < 0 || min > 65535) Line 1072  if (min < 0 || min > 65535)
1072  /* Read the maximum value if there is one, and again do a paranoid on its size.  /* Read the maximum value if there is one, and again do a paranoid on its size.
1073  Also, max must not be less than min. */  Also, max must not be less than min. */
1074    
1075  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1076    {    {
1077    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1078      {      {
1079      max = 0;      max = 0;
1080      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;
1081      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1082        {        {
1083        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 963  return p; Line 1102  return p;
1102    
1103    
1104  /*************************************************  /*************************************************
1105  *       Find forward referenced subpattern       *  *  Subroutine for finding forward reference      *
1106  *************************************************/  *************************************************/
1107    
1108  /* This function scans along a pattern's text looking for capturing  /* This recursive function is called only from find_parens() below. The
1109    top-level call starts at the beginning of the pattern. All other calls must
1110    start at a parenthesis. It scans along a pattern's text looking for capturing
1111  subpatterns, and counting them. If it finds a named pattern that matches the  subpatterns, and counting them. If it finds a named pattern that matches the
1112  name it is given, it returns its number. Alternatively, if the name is NULL, it  name it is given, it returns its number. Alternatively, if the name is NULL, it
1113  returns when it reaches a given numbered subpattern. This is used for forward  returns when it reaches a given numbered subpattern. Recursion is used to keep
1114  references to subpatterns. We know that if (?P< is encountered, the name will  track of subpatterns that reset the capturing group numbers - the (?| feature.
1115  be terminated by '>' because that is checked in the first pass.  
1116    This function was originally called only from the second pass, in which we know
1117    that if (?< or (?' or (?P< is encountered, the name will be correctly
1118    terminated because that is checked in the first pass. There is now one call to
1119    this function in the first pass, to check for a recursive back reference by
1120    name (so that we can make the whole group atomic). In this case, we need check
1121    only up to the current position in the pattern, and that is still OK because
1122    and previous occurrences will have been checked. To make this work, the test
1123    for "end of pattern" is a check against cd->end_pattern in the main loop,
1124    instead of looking for a binary zero. This means that the special first-pass
1125    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1126    processing items within the loop are OK, because afterwards the main loop will
1127    terminate.)
1128    
1129  Arguments:  Arguments:
1130    ptr          current position in the pattern    ptrptr       address of the current character pointer (updated)
1131    count        current count of capturing parens so far encountered    cd           compile background data
1132    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1133    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1134    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1135      utf8         TRUE if we are in UTF-8 mode
1136      count        pointer to the current capturing subpattern number (updated)
1137    
1138  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1139  */  */
1140    
1141  static int  static int
1142  find_parens(const uschar *ptr, int count, const uschar *name, int lorn,  find_parens_sub(uschar **ptrptr, compile_data *cd, const uschar *name, int lorn,
1143    BOOL xmode)    BOOL xmode, BOOL utf8, int *count)
1144  {  {
1145  const uschar *thisname;  uschar *ptr = *ptrptr;
1146    int start_count = *count;
1147    int hwm_count = start_count;
1148    BOOL dup_parens = FALSE;
1149    
1150    /* If the first character is a parenthesis, check on the type of group we are
1151    dealing with. The very first call may not start with a parenthesis. */
1152    
1153  for (; *ptr != 0; ptr++)  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1154    {    {
1155    int term;    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1156    
1157      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1158    
1159      /* Handle a normal, unnamed capturing parenthesis. */
1160    
1161      else if (ptr[1] != CHAR_QUESTION_MARK)
1162        {
1163        *count += 1;
1164        if (name == NULL && *count == lorn) return *count;
1165        ptr++;
1166        }
1167    
1168      /* All cases now have (? at the start. Remember when we are in a group
1169      where the parenthesis numbers are duplicated. */
1170    
1171      else if (ptr[2] == CHAR_VERTICAL_LINE)
1172        {
1173        ptr += 3;
1174        dup_parens = TRUE;
1175        }
1176    
1177      /* Handle comments; all characters are allowed until a ket is reached. */
1178    
1179      else if (ptr[2] == CHAR_NUMBER_SIGN)
1180        {
1181        for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1182        goto FAIL_EXIT;
1183        }
1184    
1185      /* Handle a condition. If it is an assertion, just carry on so that it
1186      is processed as normal. If not, skip to the closing parenthesis of the
1187      condition (there can't be any nested parens). */
1188    
1189      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1190        {
1191        ptr += 2;
1192        if (ptr[1] != CHAR_QUESTION_MARK)
1193          {
1194          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1195          if (*ptr != 0) ptr++;
1196          }
1197        }
1198    
1199      /* Start with (? but not a condition. */
1200    
1201      else
1202        {
1203        ptr += 2;
1204        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1205    
1206        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1207    
1208        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1209            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1210          {
1211          int term;
1212          const uschar *thisname;
1213          *count += 1;
1214          if (name == NULL && *count == lorn) return *count;
1215          term = *ptr++;
1216          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1217          thisname = ptr;
1218          while (*ptr != term) ptr++;
1219          if (name != NULL && lorn == ptr - thisname &&
1220              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1221            return *count;
1222          term++;
1223          }
1224        }
1225      }
1226    
1227    /* Past any initial parenthesis handling, scan for parentheses or vertical
1228    bars. Stop if we get to cd->end_pattern. Note that this is important for the
1229    first-pass call when this value is temporarily adjusted to stop at the current
1230    position. So DO NOT change this to a test for binary zero. */
1231    
1232    for (; ptr < cd->end_pattern; ptr++)
1233      {
1234    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1235    
1236    if (*ptr == '\\')    if (*ptr == CHAR_BACKSLASH)
1237      {      {
1238      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1239      if (*ptr == 'Q') for (;;)      if (*ptr == CHAR_Q) for (;;)
1240        {        {
1241        while (*(++ptr) != 0 && *ptr != '\\');        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1242        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1243        if (*(++ptr) == 'E') break;        if (*(++ptr) == CHAR_E) break;
1244        }        }
1245      continue;      continue;
1246      }      }
1247    
1248    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1249      are handled for real. If the first character is '^', skip it. Also, if the
1250      first few characters (either before or after ^) are \Q\E or \E we skip them
1251      too. This makes for compatibility with Perl. Note the use of STR macros to
1252      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1253    
1254    if (*ptr == '[')    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1255      {      {
1256      while (*(++ptr) != ']')      BOOL negate_class = FALSE;
1257        for (;;)
1258          {
1259          if (ptr[1] == CHAR_BACKSLASH)
1260            {
1261            if (ptr[2] == CHAR_E)
1262              ptr+= 2;
1263            else if (strncmp((const char *)ptr+2,
1264                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1265              ptr += 4;
1266            else
1267              break;
1268            }
1269          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1270            {
1271            negate_class = TRUE;
1272            ptr++;
1273            }
1274          else break;
1275          }
1276    
1277        /* If the next character is ']', it is a data character that must be
1278        skipped, except in JavaScript compatibility mode. */
1279    
1280        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1281            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1282          ptr++;
1283    
1284        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1285        {        {
1286        if (*ptr == 0) return -1;        if (*ptr == 0) return -1;
1287        if (*ptr == '\\')        if (*ptr == CHAR_BACKSLASH)
1288          {          {
1289          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) goto FAIL_EXIT;
1290          if (*ptr == 'Q') for (;;)          if (*ptr == CHAR_Q) for (;;)
1291            {            {
1292            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1293            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1294            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1295            }            }
1296          continue;          continue;
1297          }          }
# Line 1031  for (; *ptr != 0; ptr++) Line 1301  for (; *ptr != 0; ptr++)
1301    
1302    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1303    
1304    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1305      {      {
1306      while (*(++ptr) != 0 && *ptr != '\n');      ptr++;
1307      if (*ptr == 0) return -1;      while (*ptr != 0)
1308          {
1309          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1310          ptr++;
1311    #ifdef SUPPORT_UTF8
1312          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
1313    #endif
1314          }
1315        if (*ptr == 0) goto FAIL_EXIT;
1316      continue;      continue;
1317      }      }
1318    
1319    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1320    
1321    if (*ptr != '(') continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != '?' && ptr[1] != '*')  
1322      {      {
1323      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, count);
1324      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1325      continue;      if (*ptr == 0) goto FAIL_EXIT;
1326      }      }
1327    
1328    ptr += 2;    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1329    if (*ptr == 'P') ptr++;                      /* Allow optional P */      {
1330        if (dup_parens && *count < hwm_count) *count = hwm_count;
1331        goto FAIL_EXIT;
1332        }
1333    
1334      else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1335        {
1336        if (*count > hwm_count) hwm_count = *count;
1337        *count = start_count;
1338        }
1339      }
1340    
1341    FAIL_EXIT:
1342    *ptrptr = ptr;
1343    return -1;
1344    }
1345    
1346    
   /* We have to disambiguate (?<! and (?<= from (?<name> */  
1347    
   if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  
        *ptr != '\'')  
     continue;  
1348    
1349    count++;  /*************************************************
1350    *       Find forward referenced subpattern       *
1351    *************************************************/
1352    
1353    if (name == NULL && count == lorn) return count;  /* This function scans along a pattern's text looking for capturing
1354    term = *ptr++;  subpatterns, and counting them. If it finds a named pattern that matches the
1355    if (term == '<') term = '>';  name it is given, it returns its number. Alternatively, if the name is NULL, it
1356    thisname = ptr;  returns when it reaches a given numbered subpattern. This is used for forward
1357    while (*ptr != term) ptr++;  references to subpatterns. We used to be able to start this scan from the
1358    if (name != NULL && lorn == ptr - thisname &&  current compiling point, using the current count value from cd->bracount, and
1359        strncmp((const char *)name, (const char *)thisname, lorn) == 0)  do it all in a single loop, but the addition of the possibility of duplicate
1360      return count;  subpattern numbers means that we have to scan from the very start, in order to
1361    take account of such duplicates, and to use a recursive function to keep track
1362    of the different types of group.
1363    
1364    Arguments:
1365      cd           compile background data
1366      name         name to seek, or NULL if seeking a numbered subpattern
1367      lorn         name length, or subpattern number if name is NULL
1368      xmode        TRUE if we are in /x mode
1369      utf8         TRUE if we are in UTF-8 mode
1370    
1371    Returns:       the number of the found subpattern, or -1 if not found
1372    */
1373    
1374    static int
1375    find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode,
1376      BOOL utf8)
1377    {
1378    uschar *ptr = (uschar *)cd->start_pattern;
1379    int count = 0;
1380    int rc;
1381    
1382    /* If the pattern does not start with an opening parenthesis, the first call
1383    to find_parens_sub() will scan right to the end (if necessary). However, if it
1384    does start with a parenthesis, find_parens_sub() will return when it hits the
1385    matching closing parens. That is why we have to have a loop. */
1386    
1387    for (;;)
1388      {
1389      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, &count);
1390      if (rc > 0 || *ptr++ == 0) break;
1391    }    }
1392    
1393  return -1;  return rc;
1394  }  }
1395    
1396    
1397    
1398    
1399  /*************************************************  /*************************************************
1400  *      Find first significant op code            *  *      Find first significant op code            *
1401  *************************************************/  *************************************************/
1402    
1403  /* This is called by several functions that scan a compiled expression looking  /* This is called by several functions that scan a compiled expression looking
1404  for a fixed first character, or an anchoring op code etc. It skips over things  for a fixed first character, or an anchoring op code etc. It skips over things
1405  that do not influence this. For some calls, a change of option is important.  that do not influence this. For some calls, it makes sense to skip negative
1406  For some calls, it makes sense to skip negative forward and all backward  forward and all backward assertions, and also the \b assertion; for others it
1407  assertions, and also the \b assertion; for others it does not.  does not.
1408    
1409  Arguments:  Arguments:
1410    code         pointer to the start of the group    code         pointer to the start of the group
   options      pointer to external options  
   optbit       the option bit whose changing is significant, or  
                  zero if none are  
1411    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1412    
1413  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1414  */  */
1415    
1416  static const uschar*  static const uschar*
1417  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const uschar *code, BOOL skipassert)
   BOOL skipassert)  
1418  {  {
1419  for (;;)  for (;;)
1420    {    {
1421    switch ((int)*code)    switch ((int)*code)
1422      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1423      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1424      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1425      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
# Line 1123  for (;;) Line 1435  for (;;)
1435    
1436      case OP_CALLOUT:      case OP_CALLOUT:
1437      case OP_CREF:      case OP_CREF:
1438        case OP_NCREF:
1439      case OP_RREF:      case OP_RREF:
1440        case OP_NRREF:
1441      case OP_DEF:      case OP_DEF:
1442      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1443      break;      break;
# Line 1139  for (;;) Line 1453  for (;;)
1453    
1454    
1455  /*************************************************  /*************************************************
1456  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1457  *************************************************/  *************************************************/
1458    
1459  /* Scan a pattern and compute the fixed length of subject that will match it,  /* Scan a branch and compute the fixed length of subject that will match it,
1460  if the length is fixed. This is needed for dealing with backward assertions.  if the length is fixed. This is needed for dealing with backward assertions.
1461  In UTF8 mode, the result is in characters rather than bytes.  In UTF8 mode, the result is in characters rather than bytes. The branch is
1462    temporarily terminated with OP_END when this function is called.
1463    
1464    This function is called when a backward assertion is encountered, so that if it
1465    fails, the error message can point to the correct place in the pattern.
1466    However, we cannot do this when the assertion contains subroutine calls,
1467    because they can be forward references. We solve this by remembering this case
1468    and doing the check at the end; a flag specifies which mode we are running in.
1469    
1470  Arguments:  Arguments:
1471    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1472    options  the compiling options    utf8     TRUE in UTF-8 mode
1473      atend    TRUE if called when the pattern is complete
1474      cd       the "compile data" structure
1475    
1476  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1477                 or -1 if there is no fixed length,
1478               or -2 if \C was encountered               or -2 if \C was encountered
1479                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1480  */  */
1481    
1482  static int  static int
1483  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, BOOL utf8, BOOL atend, compile_data *cd)
1484  {  {
1485  int length = -1;  int length = -1;
1486    
# Line 1168  branch, check the length against that of Line 1493  branch, check the length against that of
1493  for (;;)  for (;;)
1494    {    {
1495    int d;    int d;
1496      uschar *ce, *cs;
1497    register int op = *cc;    register int op = *cc;
1498    switch (op)    switch (op)
1499      {      {
1500        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1501        OP_BRA (normal non-capturing bracket) because the other variants of these
1502        opcodes are all concerned with unlimited repeated groups, which of course
1503        are not of fixed length. They will cause a -1 response from the default
1504        case of this switch. */
1505    
1506      case OP_CBRA:      case OP_CBRA:
1507      case OP_BRA:      case OP_BRA:
1508      case OP_ONCE:      case OP_ONCE:
1509      case OP_COND:      case OP_COND:
1510      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), utf8, atend, cd);
1511      if (d < 0) return d;      if (d < 0) return d;
1512      branchlength += d;      branchlength += d;
1513      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1184  for (;;) Line 1516  for (;;)
1516    
1517      /* Reached end of a branch; if it's a ket it is the end of a nested      /* Reached end of a branch; if it's a ket it is the end of a nested
1518      call. If it's ALT it is an alternation in a nested call. If it is      call. If it's ALT it is an alternation in a nested call. If it is
1519      END it's the end of the outer call. All can be handled by the same code. */      END it's the end of the outer call. All can be handled by the same code.
1520        Note that we must not include the OP_KETRxxx opcodes here, because they
1521        all imply an unlimited repeat. */
1522    
1523      case OP_ALT:      case OP_ALT:
1524      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1525      case OP_END:      case OP_END:
1526      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1527        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
# Line 1198  for (;;) Line 1530  for (;;)
1530      branchlength = 0;      branchlength = 0;
1531      break;      break;
1532    
1533        /* A true recursion implies not fixed length, but a subroutine call may
1534        be OK. If the subroutine is a forward reference, we can't deal with
1535        it until the end of the pattern, so return -3. */
1536    
1537        case OP_RECURSE:
1538        if (!atend) return -3;
1539        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1540        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1541        if (cc > cs && cc < ce) return -1;                /* Recursion */
1542        d = find_fixedlength(cs + 2, utf8, atend, cd);
1543        if (d < 0) return d;
1544        branchlength += d;
1545        cc += 1 + LINK_SIZE;
1546        break;
1547    
1548      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1549    
1550      case OP_ASSERT:      case OP_ASSERT:
# Line 1211  for (;;) Line 1558  for (;;)
1558    
1559      case OP_REVERSE:      case OP_REVERSE:
1560      case OP_CREF:      case OP_CREF:
1561        case OP_NCREF:
1562      case OP_RREF:      case OP_RREF:
1563        case OP_NRREF:
1564      case OP_DEF:      case OP_DEF:
     case OP_OPT:  
1565      case OP_CALLOUT:      case OP_CALLOUT:
1566      case OP_SOD:      case OP_SOD:
1567      case OP_SOM:      case OP_SOM:
1568        case OP_SET_SOM:
1569      case OP_EOD:      case OP_EOD:
1570      case OP_EODN:      case OP_EODN:
1571      case OP_CIRC:      case OP_CIRC:
1572        case OP_CIRCM:
1573      case OP_DOLL:      case OP_DOLL:
1574        case OP_DOLLM:
1575      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1576      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1577      cc += _pcre_OP_lengths[*cc];      cc += _pcre_OP_lengths[*cc];
# Line 1229  for (;;) Line 1580  for (;;)
1580      /* Handle literal characters */      /* Handle literal characters */
1581    
1582      case OP_CHAR:      case OP_CHAR:
1583      case OP_CHARNC:      case OP_CHARI:
1584      case OP_NOT:      case OP_NOT:
1585        case OP_NOTI:
1586      branchlength++;      branchlength++;
1587      cc += 2;      cc += 2;
1588  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1589      if ((options & PCRE_UTF8) != 0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       {  
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1590  #endif  #endif
1591      break;      break;
1592    
# Line 1248  for (;;) Line 1597  for (;;)
1597      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1598      cc += 4;      cc += 4;
1599  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1600      if ((options & PCRE_UTF8) != 0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       {  
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1601  #endif  #endif
1602      break;      break;
1603    
# Line 1275  for (;;) Line 1621  for (;;)
1621      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1622      case OP_WORDCHAR:      case OP_WORDCHAR:
1623      case OP_ANY:      case OP_ANY:
1624        case OP_ALLANY:
1625      branchlength++;      branchlength++;
1626      cc++;      cc++;
1627      break;      break;
# Line 1329  for (;;) Line 1676  for (;;)
1676    
1677    
1678  /*************************************************  /*************************************************
1679  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1680  *************************************************/  *************************************************/
1681    
1682  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1683  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1684    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1685    so that it can be called from pcre_study() when finding the minimum matching
1686    length.
1687    
1688  Arguments:  Arguments:
1689    code        points to start of expression    code        points to start of expression
1690    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1691    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1692    
1693  Returns:      pointer to the opcode for the bracket, or NULL if not found  Returns:      pointer to the opcode for the bracket, or NULL if not found
1694  */  */
1695    
1696  static const uschar *  const uschar *
1697  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1698  {  {
1699  for (;;)  for (;;)
1700    {    {
1701    register int c = *code;    register int c = *code;
1702    
1703    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1704    
1705    /* XCLASS is used for classes that cannot be represented just by a bit    /* XCLASS is used for classes that cannot be represented just by a bit
# Line 1357  for (;;) Line 1708  for (;;)
1708    
1709    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1710    
1711      /* Handle recursion */
1712    
1713      else if (c == OP_REVERSE)
1714        {
1715        if (number < 0) return (uschar *)code;
1716        code += _pcre_OP_lengths[c];
1717        }
1718    
1719    /* Handle capturing bracket */    /* Handle capturing bracket */
1720    
1721    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
1722               c == OP_CBRAPOS || c == OP_SCBRAPOS)
1723      {      {
1724      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
1725      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
# Line 1368  for (;;) Line 1728  for (;;)
1728    
1729    /* Otherwise, we can get the item's length from the table, except that for    /* Otherwise, we can get the item's length from the table, except that for
1730    repeated character types, we have to test for \p and \P, which have an extra    repeated character types, we have to test for \p and \P, which have an extra
1731    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1732      must add in its length. */
1733    
1734    else    else
1735      {      {
# Line 1392  for (;;) Line 1753  for (;;)
1753        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
1754        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1755        break;        break;
1756    
1757          case OP_MARK:
1758          case OP_PRUNE_ARG:
1759          case OP_SKIP_ARG:
1760          code += code[1];
1761          break;
1762    
1763          case OP_THEN_ARG:
1764          code += code[1];
1765          break;
1766        }        }
1767    
1768      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1406  for (;;) Line 1777  for (;;)
1777      if (utf8) switch(c)      if (utf8) switch(c)
1778        {        {
1779        case OP_CHAR:        case OP_CHAR:
1780        case OP_CHARNC:        case OP_CHARI:
1781        case OP_EXACT:        case OP_EXACT:
1782          case OP_EXACTI:
1783        case OP_UPTO:        case OP_UPTO:
1784          case OP_UPTOI:
1785        case OP_MINUPTO:        case OP_MINUPTO:
1786          case OP_MINUPTOI:
1787        case OP_POSUPTO:        case OP_POSUPTO:
1788          case OP_POSUPTOI:
1789        case OP_STAR:        case OP_STAR:
1790          case OP_STARI:
1791        case OP_MINSTAR:        case OP_MINSTAR:
1792          case OP_MINSTARI:
1793        case OP_POSSTAR:        case OP_POSSTAR:
1794          case OP_POSSTARI:
1795        case OP_PLUS:        case OP_PLUS:
1796          case OP_PLUSI:
1797        case OP_MINPLUS:        case OP_MINPLUS:
1798          case OP_MINPLUSI:
1799        case OP_POSPLUS:        case OP_POSPLUS:
1800          case OP_POSPLUSI:
1801        case OP_QUERY:        case OP_QUERY:
1802          case OP_QUERYI:
1803        case OP_MINQUERY:        case OP_MINQUERY:
1804          case OP_MINQUERYI:
1805        case OP_POSQUERY:        case OP_POSQUERY:
1806          case OP_POSQUERYI:
1807        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1808        break;        break;
1809        }        }
1810    #else
1811        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1812  #endif  #endif
1813      }      }
1814    }    }
# Line 1461  for (;;) Line 1847  for (;;)
1847    
1848    /* Otherwise, we can get the item's length from the table, except that for    /* Otherwise, we can get the item's length from the table, except that for
1849    repeated character types, we have to test for \p and \P, which have an extra    repeated character types, we have to test for \p and \P, which have an extra
1850    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1851      must add in its length. */
1852    
1853    else    else
1854      {      {
# Line 1485  for (;;) Line 1872  for (;;)
1872        case OP_TYPEEXACT:        case OP_TYPEEXACT:
1873        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1874        break;        break;
1875    
1876          case OP_MARK:
1877          case OP_PRUNE_ARG:
1878          case OP_SKIP_ARG:
1879          code += code[1];
1880          break;
1881    
1882          case OP_THEN_ARG:
1883          code += code[1];
1884          break;
1885        }        }
1886    
1887      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1499  for (;;) Line 1896  for (;;)
1896      if (utf8) switch(c)      if (utf8) switch(c)
1897        {        {
1898        case OP_CHAR:        case OP_CHAR:
1899        case OP_CHARNC:        case OP_CHARI:
1900        case OP_EXACT:        case OP_EXACT:
1901          case OP_EXACTI:
1902        case OP_UPTO:        case OP_UPTO:
1903          case OP_UPTOI:
1904        case OP_MINUPTO:        case OP_MINUPTO:
1905          case OP_MINUPTOI:
1906        case OP_POSUPTO:        case OP_POSUPTO:
1907          case OP_POSUPTOI:
1908        case OP_STAR:        case OP_STAR:
1909          case OP_STARI:
1910        case OP_MINSTAR:        case OP_MINSTAR:
1911          case OP_MINSTARI:
1912        case OP_POSSTAR:        case OP_POSSTAR:
1913          case OP_POSSTARI:
1914        case OP_PLUS:        case OP_PLUS:
1915          case OP_PLUSI:
1916        case OP_MINPLUS:        case OP_MINPLUS:
1917          case OP_MINPLUSI:
1918        case OP_POSPLUS:        case OP_POSPLUS:
1919          case OP_POSPLUSI:
1920        case OP_QUERY:        case OP_QUERY:
1921          case OP_QUERYI:
1922        case OP_MINQUERY:        case OP_MINQUERY:
1923          case OP_MINQUERYI:
1924        case OP_POSQUERY:        case OP_POSQUERY:
1925          case OP_POSQUERYI:
1926        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1927        break;        break;
1928        }        }
1929    #else
1930        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1931  #endif  #endif
1932      }      }
1933    }    }
# Line 1539  Arguments: Line 1951  Arguments:
1951    code        points to start of search    code        points to start of search
1952    endcode     points to where to stop    endcode     points to where to stop
1953    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
1954      cd          contains pointers to tables etc.
1955    
1956  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
1957  */  */
1958    
1959  static BOOL  static BOOL
1960  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
1961      compile_data *cd)
1962  {  {
1963  register int c;  register int c;
1964  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], TRUE);
1965       code < endcode;       code < endcode;
1966       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], TRUE))
1967    {    {
1968    const uschar *ccode;    const uschar *ccode;
1969    
# Line 1565  for (code = first_significant_code(code Line 1979  for (code = first_significant_code(code
1979      continue;      continue;
1980      }      }
1981    
1982      /* For a recursion/subroutine call, if its end has been reached, which
1983      implies a backward reference subroutine call, we can scan it. If it's a
1984      forward reference subroutine call, we can't. To detect forward reference
1985      we have to scan up the list that is kept in the workspace. This function is
1986      called only when doing the real compile, not during the pre-compile that
1987      measures the size of the compiled pattern. */
1988    
1989      if (c == OP_RECURSE)
1990        {
1991        const uschar *scode;
1992        BOOL empty_branch;
1993    
1994        /* Test for forward reference */
1995    
1996        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
1997          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
1998    
1999        /* Not a forward reference, test for completed backward reference */
2000    
2001        empty_branch = FALSE;
2002        scode = cd->start_code + GET(code, 1);
2003        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2004    
2005        /* Completed backwards reference */
2006    
2007        do
2008          {
2009          if (could_be_empty_branch(scode, endcode, utf8, cd))
2010            {
2011            empty_branch = TRUE;
2012            break;
2013            }
2014          scode += GET(scode, 1);
2015          }
2016        while (*scode == OP_ALT);
2017    
2018        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2019        continue;
2020        }
2021    
2022    /* Groups with zero repeats can of course be empty; skip them. */    /* Groups with zero repeats can of course be empty; skip them. */
2023    
2024    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2025          c == OP_BRAPOSZERO)
2026      {      {
2027      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
2028      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
# Line 1575  for (code = first_significant_code(code Line 2030  for (code = first_significant_code(code
2030      continue;      continue;
2031      }      }
2032    
2033      /* A nested group that is already marked as "could be empty" can just be
2034      skipped. */
2035    
2036      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2037          c == OP_SCBRA || c == OP_SCBRAPOS)
2038        {
2039        do code += GET(code, 1); while (*code == OP_ALT);
2040        c = *code;
2041        continue;
2042        }
2043    
2044    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
2045    
2046    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_BRA  || c == OP_BRAPOS ||
2047          c == OP_CBRA || c == OP_CBRAPOS ||
2048          c == OP_ONCE || c == OP_COND)
2049      {      {
2050      BOOL empty_branch;      BOOL empty_branch;
2051      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
2052    
2053      /* Scan a closed bracket */      /* If a conditional group has only one branch, there is a second, implied,
2054        empty branch, so just skip over the conditional, because it could be empty.
2055        Otherwise, scan the individual branches of the group. */
2056    
2057      empty_branch = FALSE;      if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
     do  
       {  
       if (!empty_branch && could_be_empty_branch(code, endcode, utf8))  
         empty_branch = TRUE;  
2058        code += GET(code, 1);        code += GET(code, 1);
2059        else
2060          {
2061          empty_branch = FALSE;
2062          do
2063            {
2064            if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
2065              empty_branch = TRUE;
2066            code += GET(code, 1);
2067            }
2068          while (*code == OP_ALT);
2069          if (!empty_branch) return FALSE;   /* All branches are non-empty */
2070        }        }
2071      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
2072      c = *code;      c = *code;
2073      continue;      continue;
2074      }      }
# Line 1653  for (code = first_significant_code(code Line 2129  for (code = first_significant_code(code
2129      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2130      case OP_WORDCHAR:      case OP_WORDCHAR:
2131      case OP_ANY:      case OP_ANY:
2132        case OP_ALLANY:
2133      case OP_ANYBYTE:      case OP_ANYBYTE:
2134      case OP_CHAR:      case OP_CHAR:
2135      case OP_CHARNC:      case OP_CHARI:
2136      case OP_NOT:      case OP_NOT:
2137        case OP_NOTI:
2138      case OP_PLUS:      case OP_PLUS:
2139      case OP_MINPLUS:      case OP_MINPLUS:
2140      case OP_POSPLUS:      case OP_POSPLUS:
# Line 1696  for (code = first_significant_code(code Line 2174  for (code = first_significant_code(code
2174      case OP_KET:      case OP_KET:
2175      case OP_KETRMAX:      case OP_KETRMAX:
2176      case OP_KETRMIN:      case OP_KETRMIN:
2177        case OP_KETRPOS:
2178      case OP_ALT:      case OP_ALT:
2179      return TRUE;      return TRUE;
2180    
# Line 1704  for (code = first_significant_code(code Line 2183  for (code = first_significant_code(code
2183    
2184  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2185      case OP_STAR:      case OP_STAR:
2186        case OP_STARI:
2187      case OP_MINSTAR:      case OP_MINSTAR:
2188        case OP_MINSTARI:
2189      case OP_POSSTAR:      case OP_POSSTAR:
2190        case OP_POSSTARI:
2191      case OP_QUERY:      case OP_QUERY:
2192        case OP_QUERYI:
2193      case OP_MINQUERY:      case OP_MINQUERY:
2194        case OP_MINQUERYI:
2195      case OP_POSQUERY:      case OP_POSQUERY:
2196        case OP_POSQUERYI:
2197        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2198        break;
2199    
2200      case OP_UPTO:      case OP_UPTO:
2201        case OP_UPTOI:
2202      case OP_MINUPTO:      case OP_MINUPTO:
2203        case OP_MINUPTOI:
2204      case OP_POSUPTO:      case OP_POSUPTO:
2205      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_POSUPTOI:
2206        if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2207      break;      break;
2208  #endif  #endif
2209    
2210        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2211        string. */
2212    
2213        case OP_MARK:
2214        case OP_PRUNE_ARG:
2215        case OP_SKIP_ARG:
2216        code += code[1];
2217        break;
2218    
2219        case OP_THEN_ARG:
2220        code += code[1];
2221        break;
2222    
2223        /* None of the remaining opcodes are required to match a character. */
2224    
2225        default:
2226        break;
2227      }      }
2228    }    }
2229    
# Line 1731  return TRUE; Line 2240  return TRUE;
2240  the current branch of the current pattern to see if it could match the empty  the current branch of the current pattern to see if it could match the empty
2241  string. If it could, we must look outwards for branches at other levels,  string. If it could, we must look outwards for branches at other levels,
2242  stopping when we pass beyond the bracket which is the subject of the recursion.  stopping when we pass beyond the bracket which is the subject of the recursion.
2243    This function is called only during the real compile, not during the
2244    pre-compile.
2245    
2246  Arguments:  Arguments:
2247    code        points to start of the recursion    code        points to start of the recursion
2248    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2249    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2250    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2251      cd          pointers to tables etc
2252    
2253  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2254  */  */
2255    
2256  static BOOL  static BOOL
2257  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2258    BOOL utf8)    BOOL utf8, compile_data *cd)
2259  {  {
2260  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2261    {    {
2262    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2263        return FALSE;
2264    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2265    }    }
2266  return TRUE;  return TRUE;
# Line 1779  where Perl recognizes it as the POSIX cl Line 2292  where Perl recognizes it as the POSIX cl
2292  "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,  "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2293  I think.  I think.
2294    
2295    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2296    It seems that the appearance of a nested POSIX class supersedes an apparent
2297    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2298    a digit.
2299    
2300    In Perl, unescaped square brackets may also appear as part of class names. For
2301    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2302    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2303    seem right at all. PCRE does not allow closing square brackets in POSIX class
2304    names.
2305    
2306  Arguments:  Arguments:
2307    ptr      pointer to the initial [    ptr      pointer to the initial [
2308    endptr   where to return the end pointer    endptr   where to return the end pointer
# Line 1793  int terminator; /* Don't combin Line 2317  int terminator; /* Don't combin
2317  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2318  for (++ptr; *ptr != 0; ptr++)  for (++ptr; *ptr != 0; ptr++)
2319    {    {
2320    if (*ptr == '\\' && ptr[1] == ']') ptr++; else    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2321        ptr++;
2322      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2323      else
2324      {      {
2325      if (*ptr == ']') return FALSE;      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
     if (*ptr == terminator && ptr[1] == ']')  
2326        {        {
2327        *endptr = ptr;        *endptr = ptr;
2328        return TRUE;        return TRUE;
2329        }        }
2330        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2331             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2332              ptr[1] == CHAR_EQUALS_SIGN) &&
2333            check_posix_syntax(ptr, endptr))
2334          return FALSE;
2335      }      }
2336    }    }
2337  return FALSE;  return FALSE;
# Line 1847  return -1; Line 2378  return -1;
2378  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2379  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2380  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
2381  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
2382  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
2383  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
2384  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
2385  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2386    OP_END.
2387    
2388  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2389  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 1928  auto_callout(uschar *code, const uschar Line 2460  auto_callout(uschar *code, const uschar
2460  {  {
2461  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2462  *code++ = 255;  *code++ = 255;
2463  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2464  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2465  return code + 2*LINK_SIZE;  return code + 2*LINK_SIZE;
2466  }  }
2467    
# Line 1954  Returns: nothing Line 2486  Returns: nothing
2486  static void  static void
2487  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)
2488  {  {
2489  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2490  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2491  }  }
2492    
# Line 1986  get_othercase_range(unsigned int *cptr, Line 2518  get_othercase_range(unsigned int *cptr,
2518  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2519    
2520  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2521    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2522    
2523  if (c > d) return FALSE;  if (c > d) return FALSE;
2524    
# Line 1995  next = othercase + 1; Line 2527  next = othercase + 1;
2527    
2528  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2529    {    {
2530    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2531    next++;    next++;
2532    }    }
2533    
# Line 2004  for (++c; c <= d; c++) Line 2536  for (++c; c <= d; c++)
2536    
2537  return TRUE;  return TRUE;
2538  }  }
2539    
2540    
2541    
2542    /*************************************************
2543    *        Check a character and a property        *
2544    *************************************************/
2545    
2546    /* This function is called by check_auto_possessive() when a property item
2547    is adjacent to a fixed character.
2548    
2549    Arguments:
2550      c            the character
2551      ptype        the property type
2552      pdata        the data for the type
2553      negated      TRUE if it's a negated property (\P or \p{^)
2554    
2555    Returns:       TRUE if auto-possessifying is OK
2556    */
2557    
2558    static BOOL
2559    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2560    {
2561    const ucd_record *prop = GET_UCD(c);
2562    switch(ptype)
2563      {
2564      case PT_LAMP:
2565      return (prop->chartype == ucp_Lu ||
2566              prop->chartype == ucp_Ll ||
2567              prop->chartype == ucp_Lt) == negated;
2568    
2569      case PT_GC:
2570      return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;
2571    
2572      case PT_PC:
2573      return (pdata == prop->chartype) == negated;
2574    
2575      case PT_SC:
2576      return (pdata == prop->script) == negated;
2577    
2578      /* These are specials */
2579    
2580      case PT_ALNUM:
2581      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2582              _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;
2583    
2584      case PT_SPACE:    /* Perl space */
2585      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2586              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2587              == negated;
2588    
2589      case PT_PXSPACE:  /* POSIX space */
2590      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2591              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2592              c == CHAR_FF || c == CHAR_CR)
2593              == negated;
2594    
2595      case PT_WORD:
2596      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2597              _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2598              c == CHAR_UNDERSCORE) == negated;
2599      }
2600    return FALSE;
2601    }
2602  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2603    
2604    
# Line 2017  whether the next thing could possibly ma Line 2612  whether the next thing could possibly ma
2612  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
2613    
2614  Arguments:  Arguments:
2615    op_code       the repeated op code    previous      pointer to the repeated opcode
   this          data for this item, depends on the opcode  
2616    utf8          TRUE in UTF-8 mode    utf8          TRUE in UTF-8 mode
   utf8_char     used for utf8 character bytes, NULL if not relevant  
2617    ptr           next character in pattern    ptr           next character in pattern
2618    options       options bits    options       options bits
2619    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 2029  Returns: TRUE if possessifying is Line 2622  Returns: TRUE if possessifying is
2622  */  */
2623    
2624  static BOOL  static BOOL
2625  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
2626    const uschar *ptr, int options, compile_data *cd)    int options, compile_data *cd)
2627  {  {
2628  int next;  int c, next;
2629    int op_code = *previous++;
2630    
2631  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
2632    
# Line 2041  if ((options & PCRE_EXTENDED) != 0) Line 2635  if ((options & PCRE_EXTENDED) != 0)
2635    for (;;)    for (;;)
2636      {      {
2637      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2638      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2639        {        {
2640        while (*(++ptr) != 0)        ptr++;
2641          while (*ptr != 0)
2642            {
2643          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2644            ptr++;
2645    #ifdef SUPPORT_UTF8
2646            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2647    #endif
2648            }
2649        }        }
2650      else break;      else break;
2651      }      }
# Line 2053  if ((options & PCRE_EXTENDED) != 0) Line 2654  if ((options & PCRE_EXTENDED) != 0)
2654  /* 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
2655  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2656    
2657  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2658    {    {
2659    int temperrorcode = 0;    int temperrorcode = 0;
2660    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
# Line 2078  if ((options & PCRE_EXTENDED) != 0) Line 2679  if ((options & PCRE_EXTENDED) != 0)
2679    for (;;)    for (;;)
2680      {      {
2681      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2682      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2683        {        {
2684        while (*(++ptr) != 0)        ptr++;
2685          while (*ptr != 0)
2686            {
2687          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2688            ptr++;
2689    #ifdef SUPPORT_UTF8
2690            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2691    #endif
2692            }
2693        }        }
2694      else break;      else break;
2695      }      }
# Line 2089  if ((options & PCRE_EXTENDED) != 0) Line 2697  if ((options & PCRE_EXTENDED) != 0)
2697    
2698  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
2699    
2700  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2701    return FALSE;    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2702        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. */  
   
2703    
2704  /* Handle cases when the next item is a character. */  /* Now compare the next item with the previous opcode. First, handle cases when
2705    the next item is a character. */
2706    
2707  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
2708    {    {
2709    case OP_CHAR:    case OP_CHAR:
2710  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2711    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2712    #else
2713      c = *previous;
2714  #endif  #endif
2715    return item != next;    return c != next;
2716    
2717    /* 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
2718    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
2719    high-valued characters. */    high-valued characters. */
2720    
2721    case OP_CHARNC:    case OP_CHARI:
2722  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2723    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2724    #else
2725      c = *previous;
2726  #endif  #endif
2727    if (item == next) return FALSE;    if (c == next) return FALSE;
2728  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2729    if (utf8)    if (utf8)
2730      {      {
2731      unsigned int othercase;      unsigned int othercase;
2732      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2733  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2734      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
2735  #else  #else
2736      othercase = NOTACHAR;      othercase = NOTACHAR;
2737  #endif  #endif
2738      return (unsigned int)item != othercase;      return (unsigned int)c != othercase;
2739      }      }
2740    else    else
2741  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2742    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2743    
2744    /* 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
2745      opcodes are not used for multi-byte characters, because they are coded using
2746      an XCLASS instead. */
2747    
2748    case OP_NOT:    case OP_NOT:
2749    if (item == next) return TRUE;    return (c = *previous) == next;
2750    if ((options & PCRE_CASELESS) == 0) return FALSE;  
2751      case OP_NOTI:
2752      if ((c = *previous) == next) return TRUE;
2753  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2754    if (utf8)    if (utf8)
2755      {      {
2756      unsigned int othercase;      unsigned int othercase;
2757      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2758  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2759      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
2760  #else  #else
2761      othercase = NOTACHAR;      othercase = NOTACHAR;
2762  #endif  #endif
2763      return (unsigned int)item == othercase;      return (unsigned int)c == othercase;
2764      }      }
2765    else    else
2766  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2767    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2768    
2769      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2770      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
2771    
2772    case OP_DIGIT:    case OP_DIGIT:
2773    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
# Line 2195  if (next >= 0) switch(op_code) Line 2810  if (next >= 0) switch(op_code)
2810      case 0x202f:      case 0x202f:
2811      case 0x205f:      case 0x205f:
2812      case 0x3000:      case 0x3000:
2813      return op_code != OP_HSPACE;      return op_code == OP_NOT_HSPACE;
2814      default:      default:
2815      return op_code == OP_HSPACE;      return op_code != OP_NOT_HSPACE;
2816      }      }
2817    
2818      case OP_ANYNL:
2819    case OP_VSPACE:    case OP_VSPACE:
2820    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
2821    switch(next)    switch(next)
# Line 2211  if (next >= 0) switch(op_code) Line 2827  if (next >= 0) switch(op_code)
2827      case 0x85:      case 0x85:
2828      case 0x2028:      case 0x2028:
2829      case 0x2029:      case 0x2029:
2830      return op_code != OP_VSPACE;      return op_code == OP_NOT_VSPACE;
2831      default:      default:
2832      return op_code == OP_VSPACE;      return op_code != OP_NOT_VSPACE;
2833      }      }
2834    
2835    #ifdef SUPPORT_UCP
2836      case OP_PROP:
2837      return check_char_prop(next, previous[0], previous[1], FALSE);
2838    
2839      case OP_NOTPROP:
2840      return check_char_prop(next, previous[0], previous[1], TRUE);
2841    #endif
2842    
2843    default:    default:
2844    return FALSE;    return FALSE;
2845    }    }
2846    
2847    
2848  /* 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
2849    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
2850    generated only when PCRE_UCP is *not* set, that is, when only ASCII
2851    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
2852    replaced by OP_PROP codes when PCRE_UCP is set. */
2853    
2854  switch(op_code)  switch(op_code)
2855    {    {
2856    case OP_CHAR:    case OP_CHAR:
2857    case OP_CHARNC:    case OP_CHARI:
2858  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2859    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2860    #else
2861      c = *previous;
2862  #endif  #endif
2863    switch(-next)    switch(-next)
2864      {      {
2865      case ESC_d:      case ESC_d:
2866      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
2867    
2868      case ESC_D:      case ESC_D:
2869      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
2870    
2871      case ESC_s:      case ESC_s:
2872      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
2873    
2874      case ESC_S:      case ESC_S:
2875      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
2876    
2877      case ESC_w:      case ESC_w:
2878      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
2879    
2880      case ESC_W:      case ESC_W:
2881      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
2882    
2883      case ESC_h:      case ESC_h:
2884      case ESC_H:      case ESC_H:
2885      switch(item)      switch(c)
2886        {        {
2887        case 0x09:        case 0x09:
2888        case 0x20:        case 0x20:
# Line 2280  switch(op_code) Line 2910  switch(op_code)
2910    
2911      case ESC_v:      case ESC_v:
2912      case ESC_V:      case ESC_V:
2913      switch(item)      switch(c)
2914        {        {
2915        case 0x0a:        case 0x0a:
2916        case 0x0b:        case 0x0b:
# Line 2294  switch(op_code) Line 2924  switch(op_code)
2924        return -next == ESC_v;        return -next == ESC_v;
2925        }        }
2926    
2927        /* When PCRE_UCP is set, these values get generated for \d etc. Find
2928        their substitutions and process them. The result will always be either
2929        -ESC_p or -ESC_P. Then fall through to process those values. */
2930    
2931    #ifdef SUPPORT_UCP
2932        case ESC_du:
2933        case ESC_DU:
2934        case ESC_wu:
2935        case ESC_WU:
2936        case ESC_su:
2937        case ESC_SU:
2938          {
2939          int temperrorcode = 0;
2940          ptr = substitutes[-next - ESC_DU];
2941          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
2942          if (temperrorcode != 0) return FALSE;
2943          ptr++;    /* For compatibility */
2944          }
2945        /* Fall through */
2946    
2947        case ESC_p:
2948        case ESC_P:
2949          {
2950          int ptype, pdata, errorcodeptr;
2951          BOOL negated;
2952    
2953          ptr--;      /* Make ptr point at the p or P */
2954          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
2955          if (ptype < 0) return FALSE;
2956          ptr++;      /* Point past the final curly ket */
2957    
2958          /* If the property item is optional, we have to give up. (When generated
2959          from \d etc by PCRE_UCP, this test will have been applied much earlier,
2960          to the original \d etc. At this point, ptr will point to a zero byte. */
2961    
2962          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2963            strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2964              return FALSE;
2965    
2966          /* Do the property check. */
2967    
2968          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
2969          }
2970    #endif
2971    
2972      default:      default:
2973      return FALSE;      return FALSE;
2974      }      }
2975    
2976      /* In principle, support for Unicode properties should be integrated here as
2977      well. It means re-organizing the above code so as to get hold of the property
2978      values before switching on the op-code. However, I wonder how many patterns
2979      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
2980      these op-codes are never generated.) */
2981    
2982    case OP_DIGIT:    case OP_DIGIT:
2983    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2984           next == -ESC_h || next == -ESC_v;           next == -ESC_h || next == -ESC_v || next == -ESC_R;
2985    
2986    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
2987    return next == -ESC_d;    return next == -ESC_d;
2988    
2989    case OP_WHITESPACE:    case OP_WHITESPACE:
2990    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
2991    
2992    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
2993    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2994    
2995    case OP_HSPACE:    case OP_HSPACE:
2996    return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
2997             next == -ESC_w || next == -ESC_v || next == -ESC_R;
2998    
2999    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
3000    return next == -ESC_h;    return next == -ESC_h;
3001    
3002    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
3003      case OP_ANYNL:
3004    case OP_VSPACE:    case OP_VSPACE:
3005    return next == -ESC_V || next == -ESC_d || next == -ESC_w;    return next == -ESC_V || next == -ESC_d || next == -ESC_w;
3006    
3007    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
3008    return next == -ESC_v;    return next == -ESC_v || next == -ESC_R;
3009    
3010    case OP_WORDCHAR:    case OP_WORDCHAR:
3011    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
3012             next == -ESC_v || next == -ESC_R;
3013    
3014    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
3015    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2357  Arguments: Line 3041  Arguments:
3041    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
3042    reqbyteptr     set to the last literal character required, else < 0    reqbyteptr     set to the last literal character required, else < 0
3043    bcptr          points to current branch chain    bcptr          points to current branch chain
3044      cond_depth     conditional nesting depth
3045    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
3046    lengthptr      NULL during the real compile phase    lengthptr      NULL during the real compile phase
3047                   points to length accumulator during pre-compile phase                   points to length accumulator during pre-compile phase
# Line 2368  Returns: TRUE on success Line 3053  Returns: TRUE on success
3053  static BOOL  static BOOL
3054  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
3055    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
3056    compile_data *cd, int *lengthptr)    int cond_depth, compile_data *cd, int *lengthptr)
3057  {  {
3058  int repeat_type, op_type;  int repeat_type, op_type;
3059  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
# Line 2377  int greedy_default, greedy_non_default; Line 3062  int greedy_default, greedy_non_default;
3062  int firstbyte, reqbyte;  int firstbyte, reqbyte;
3063  int zeroreqbyte, zerofirstbyte;  int zeroreqbyte, zerofirstbyte;
3064  int req_caseopt, reqvary, tempreqvary;  int req_caseopt, reqvary, tempreqvary;
3065  int options = *optionsptr;  int options = *optionsptr;               /* May change dynamically */
3066  int after_manual_callout = 0;  int after_manual_callout = 0;
3067  int length_prevgroup = 0;  int length_prevgroup = 0;
3068  register int c;  register int c;
# Line 2389  BOOL inescq = FALSE; Line 3074  BOOL inescq = FALSE;
3074  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
3075  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
3076  const uschar *tempptr;  const uschar *tempptr;
3077    const uschar *nestptr = NULL;
3078  uschar *previous = NULL;  uschar *previous = NULL;
3079  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
3080  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
3081  uschar classbits[32];  uschar classbits[32];
3082    
3083    /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3084    must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3085    dynamically as we process the pattern. */
3086    
3087  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3088  BOOL class_utf8;  BOOL class_utf8;
3089  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
# Line 2402  uschar *class_utf8data_base; Line 3092  uschar *class_utf8data_base;
3092  uschar utf8_char[6];  uschar utf8_char[6];
3093  #else  #else
3094  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
 uschar *utf8_char = NULL;  
3095  #endif  #endif
3096    
3097  #ifdef DEBUG  #ifdef PCRE_DEBUG
3098  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
3099  #endif  #endif
3100    
# Line 2459  for (;; ptr++) Line 3148  for (;; ptr++)
3148    
3149    c = *ptr;    c = *ptr;
3150    
3151      /* If we are at the end of a nested substitution, revert to the outer level
3152      string. Nesting only happens one level deep. */
3153    
3154      if (c == 0 && nestptr != NULL)
3155        {
3156        ptr = nestptr;
3157        nestptr = NULL;
3158        c = *ptr;
3159        }
3160    
3161    /* 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
3162    previous cycle of this loop. */    previous cycle of this loop. */
3163    
3164    if (lengthptr != NULL)    if (lengthptr != NULL)
3165      {      {
3166  #ifdef DEBUG  #ifdef PCRE_DEBUG
3167      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3168  #endif  #endif
3169      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
3170        {        {
3171        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3172        goto FAILED;        goto FAILED;
# Line 2489  for (;; ptr++) Line 3188  for (;; ptr++)
3188        goto FAILED;        goto FAILED;
3189        }        }
3190    
3191      *lengthptr += code - last_code;      *lengthptr += (int)(code - last_code);
3192      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));
3193    
3194      /* 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 2516  for (;; ptr++) Line 3215  for (;; ptr++)
3215    /* 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
3216    reference list. */    reference list. */
3217    
3218    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
3219      {      {
3220      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3221      goto FAILED;      goto FAILED;
# Line 2526  for (;; ptr++) Line 3225  for (;; ptr++)
3225    
3226    if (inescq && c != 0)    if (inescq && c != 0)
3227      {      {
3228      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3229        {        {
3230        inescq = FALSE;        inescq = FALSE;
3231        ptr++;        ptr++;
# Line 2552  for (;; ptr++) Line 3251  for (;; ptr++)
3251    /* 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
3252    a quantifier. */    a quantifier. */
3253    
3254    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
3255      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
3256        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
3257    
3258    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
3259         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2563  for (;; ptr++) Line 3263  for (;; ptr++)
3263      previous_callout = NULL;      previous_callout = NULL;
3264      }      }
3265    
3266    /* In extended mode, skip white space and comments */    /* In extended mode, skip white space and comments. */
3267    
3268    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3269      {      {
3270      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
3271      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
3272        {        {
3273        while (*(++ptr) != 0)        ptr++;
3274          while (*ptr != 0)
3275          {          {
3276          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3277            ptr++;
3278    #ifdef SUPPORT_UTF8
3279            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
3280    #endif
3281          }          }
3282        if (*ptr != 0) continue;        if (*ptr != 0) continue;
3283    
# Line 2593  for (;; ptr++) Line 3298  for (;; ptr++)
3298      {      {
3299      /* ===================================================================*/      /* ===================================================================*/
3300      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
3301      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
3302      case ')':      case CHAR_RIGHT_PARENTHESIS:
3303      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
3304      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
3305      *codeptr = code;      *codeptr = code;
# Line 2606  for (;; ptr++) Line 3311  for (;; ptr++)
3311          *errorcodeptr = ERR20;          *errorcodeptr = ERR20;
3312          goto FAILED;          goto FAILED;
3313          }          }
3314        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += (int)(code - last_code);   /* To include callout length */
3315        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
3316        }        }
3317      return TRUE;      return TRUE;
# Line 2616  for (;; ptr++) Line 3321  for (;; ptr++)
3321      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
3322      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3323    
3324      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
3325        previous = NULL;
3326      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3327        {        {
3328        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3329          *code++ = OP_CIRCM;
3330        }        }
3331      previous = NULL;      else *code++ = OP_CIRC;
     *code++ = OP_CIRC;  
3332      break;      break;
3333    
3334      case '$':      case CHAR_DOLLAR_SIGN:
3335      previous = NULL;      previous = NULL;
3336      *code++ = OP_DOLL;      *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
3337      break;      break;
3338    
3339      /* 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
3340      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
3341    
3342      case '.':      case CHAR_DOT:
3343      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3344      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
3345      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3346      previous = code;      previous = code;
3347      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
3348      break;      break;
3349    
3350    
# Line 2653  for (;; ptr++) Line 3359  for (;; ptr++)
3359      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,
3360      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
3361      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.
     */  
3362    
3363      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
3364        default (Perl) mode, it is treated as a data character. */
3365    
3366        case CHAR_RIGHT_SQUARE_BRACKET:
3367        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3368          {
3369          *errorcodeptr = ERR64;
3370          goto FAILED;
3371          }
3372        goto NORMAL_CHAR;
3373    
3374        case CHAR_LEFT_SQUARE_BRACKET:
3375      previous = code;      previous = code;
3376    
3377      /* 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
3378      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. */
3379    
3380      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3381             ptr[1] == CHAR_EQUALS_SIGN) &&
3382          check_posix_syntax(ptr, &tempptr))          check_posix_syntax(ptr, &tempptr))
3383        {        {
3384        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
3385        goto FAILED;        goto FAILED;
3386        }        }
3387    
# Line 2676  for (;; ptr++) Line 3393  for (;; ptr++)
3393      for (;;)      for (;;)
3394        {        {
3395        c = *(++ptr);        c = *(++ptr);
3396        if (c == '\\')        if (c == CHAR_BACKSLASH)
3397          {          {
3398          if (ptr[1] == 'E') ptr++;          if (ptr[1] == CHAR_E)
3399            else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;            ptr++;
3400              else break;          else if (strncmp((const char *)ptr+1,
3401                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
3402              ptr += 3;
3403            else
3404              break;
3405          }          }
3406        else if (!negate_class && c == '^')        else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
3407          negate_class = TRUE;          negate_class = TRUE;
3408        else break;        else break;
3409        }        }
3410    
3411        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
3412        an initial ']' is taken as a data character -- the code below handles
3413        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3414        [^] must match any character, so generate OP_ALLANY. */
3415    
3416        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3417            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3418          {
3419          *code++ = negate_class? OP_ALLANY : OP_FAIL;
3420          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3421          zerofirstbyte = firstbyte;
3422          break;
3423          }
3424    
3425      /* 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
3426      negation flag at the end, so that support for characters > 255 works      negation flag at the end, so that support for characters > 255 works
3427      correctly (they are all included in the class). */      correctly (they are all included in the class). */
# Line 2744  for (;; ptr++) Line 3479  for (;; ptr++)
3479    
3480        if (inescq)        if (inescq)
3481          {          {
3482          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3483            {            {
3484            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
3485            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2759  for (;; ptr++) Line 3494  for (;; ptr++)
3494        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3495        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3496    
3497        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3498            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3499            check_posix_syntax(ptr, &tempptr))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3500          {          {
3501          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3502          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3503          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3504          uschar pbits[32];          uschar pbits[32];
3505    
3506          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3507            {            {
3508            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3509            goto FAILED;            goto FAILED;
3510            }            }
3511    
3512          ptr += 2;          ptr += 2;
3513          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3514            {            {
3515            local_negate = TRUE;            local_negate = TRUE;
3516            should_flip_negation = TRUE;  /* Note negative special */            should_flip_negation = TRUE;  /* Note negative special */
3517            ptr++;            ptr++;
3518            }            }
3519    
3520          posix_class = check_posix_name(ptr, tempptr - ptr);          posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3521          if (posix_class < 0)          if (posix_class < 0)
3522            {            {
3523            *errorcodeptr = ERR30;            *errorcodeptr = ERR30;
# Line 2796  for (;; ptr++) Line 3531  for (;; ptr++)
3531          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3532            posix_class = 0;            posix_class = 0;
3533    
3534          /* 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
3535          because we may be adding and subtracting from it, and we don't want to          different escape sequences that use Unicode properties. */
3536          subtract bits that may be in the main map already. At the end we or the  
3537          result into the bit map that is being built. */  #ifdef SUPPORT_UCP
3538            if ((options & PCRE_UCP) != 0)
3539              {
3540              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3541              if (posix_substitutes[pc] != NULL)
3542                {
3543                nestptr = tempptr + 1;
3544                ptr = posix_substitutes[pc] - 1;
3545                continue;
3546                }
3547              }
3548    #endif
3549            /* In the non-UCP case, we build the bit map for the POSIX class in a
3550            chunk of local store because we may be adding and subtracting from it,
3551            and we don't want to subtract bits that may be in the main map already.
3552            At the end we or the result into the bit map that is being built. */
3553    
3554          posix_class *= 3;          posix_class *= 3;
3555    
# Line 2843  for (;; ptr++) Line 3593  for (;; ptr++)
3593    
3594        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3595        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
3596        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
3597        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
3598        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
3599        character in them, so set class_charcount bigger than one. */        are either treated as literal characters (by default), or are faulted if
3600          PCRE_EXTRA is set. */
3601    
3602        if (c == '\\')        if (c == CHAR_BACKSLASH)
3603          {          {
3604          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3605          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3606    
3607          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 */  
3608          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3609            {            {
3610            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3611              {              {
3612              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3613              }              }
# Line 2872  for (;; ptr++) Line 3621  for (;; ptr++)
3621            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
3622            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3623    
3624            /* Save time by not doing this in the pre-compile phase. */            switch (-c)
   
           if (lengthptr == NULL) switch (-c)  
3625              {              {
3626    #ifdef SUPPORT_UCP
3627                case ESC_du:     /* These are the values given for \d etc */
3628                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3629                case ESC_wu:     /* escape sequence with an appropriate \p */
3630                case ESC_WU:     /* or \P to test Unicode properties instead */
3631                case ESC_su:     /* of the default ASCII testing. */
3632                case ESC_SU:
3633                nestptr = ptr;
3634                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3635                class_charcount -= 2;                /* Undo! */
3636                continue;
3637    #endif
3638              case ESC_d:              case ESC_d:
3639              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3640              continue;              continue;
# Line 2894  for (;; ptr++) Line 3653  for (;; ptr++)
3653              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3654              continue;              continue;
3655    
3656                /* Perl 5.004 onwards omits VT from \s, but we must preserve it
3657                if it was previously set by something earlier in the character
3658                class. */
3659    
3660              case ESC_s:              case ESC_s:
3661              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];              classbits[0] |= cbits[cbit_space];
3662              classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= cbits[cbit_space+1] & ~0x08;
3663                for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3664              continue;              continue;
3665    
3666              case ESC_S:              case ESC_S:
# Line 2905  for (;; ptr++) Line 3669  for (;; ptr++)
3669              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3670              continue;              continue;
3671    
3672              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)  
             {  
3673              SETBIT(classbits, 0x09); /* VT */              SETBIT(classbits, 0x09); /* VT */
3674              SETBIT(classbits, 0x20); /* SPACE */              SETBIT(classbits, 0x20); /* SPACE */
3675              SETBIT(classbits, 0xa0); /* NSBP */              SETBIT(classbits, 0xa0); /* NSBP */
# Line 2942  for (;; ptr++) Line 3693  for (;; ptr++)
3693                }                }
3694  #endif  #endif
3695              continue;              continue;
             }  
3696    
3697            if (-c == ESC_H)              case ESC_H:
             {  
3698              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3699                {                {
3700                int x = 0xff;                int x = 0xff;
# Line 2987  for (;; ptr++) Line 3736  for (;; ptr++)
3736                }                }
3737  #endif  #endif
3738              continue;              continue;
             }  
3739    
3740            if (-c == ESC_v)              case ESC_v:
             {  
3741              SETBIT(classbits, 0x0a); /* LF */              SETBIT(classbits, 0x0a); /* LF */
3742              SETBIT(classbits, 0x0b); /* VT */              SETBIT(classbits, 0x0b); /* VT */
3743              SETBIT(classbits, 0x0c); /* FF */              SETBIT(classbits, 0x0c); /* FF */
# Line 3006  for (;; ptr++) Line 3753  for (;; ptr++)
3753                }                }
3754  #endif  #endif
3755              continue;              continue;
             }  
3756    
3757            if (-c == ESC_V)              case ESC_V:
             {  
3758              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3759                {                {
3760                int x = 0xff;                int x = 0xff;
# Line 3039  for (;; ptr++) Line 3784  for (;; ptr++)
3784                }                }
3785  #endif  #endif
3786              continue;              continue;
             }  
   
           /* We need to deal with \P and \p in both phases. */  
3787    
3788  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3789            if (-c == ESC_p || -c == ESC_P)              case ESC_p:
3790              {              case ESC_P:
3791              BOOL negated;                {
3792              int pdata;                BOOL negated;
3793              int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                int pdata;
3794              if (ptype < 0) goto FAILED;                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3795              class_utf8 = TRUE;                if (ptype < 0) goto FAILED;
3796              *class_utf8data++ = ((-c == ESC_p) != negated)?                class_utf8 = TRUE;
3797                XCL_PROP : XCL_NOTPROP;                *class_utf8data++ = ((-c == ESC_p) != negated)?
3798              *class_utf8data++ = ptype;                  XCL_PROP : XCL_NOTPROP;
3799              *class_utf8data++ = pdata;                *class_utf8data++ = ptype;
3800              class_charcount -= 2;   /* Not a < 256 character */                *class_utf8data++ = pdata;
3801              continue;                class_charcount -= 2;   /* Not a < 256 character */
3802              }                continue;
3803                  }
3804  #endif  #endif
3805            /* Unrecognized escapes are faulted if PCRE is running in its              /* Unrecognized escapes are faulted if PCRE is running in its
3806            strict mode. By default, for compatibility with Perl, they are              strict mode. By default, for compatibility with Perl, they are
3807            treated as literals. */              treated as literals. */
3808    
3809            if ((options & PCRE_EXTRA) != 0)              default:
3810              {              if ((options & PCRE_EXTRA) != 0)
3811              *errorcodeptr = ERR7;                {
3812              goto FAILED;                *errorcodeptr = ERR7;
3813                  goto FAILED;
3814                  }
3815                class_charcount -= 2;  /* Undo the default count from above */
3816                c = *ptr;              /* Get the final character and fall through */
3817                break;
3818              }              }
   
           class_charcount -= 2;  /* Undo the default count from above */  
           c = *ptr;              /* Get the final character and fall through */  
3819            }            }
3820    
3821          /* 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 3084  for (;; ptr++) Line 3829  for (;; ptr++)
3829        entirely. The code for handling \Q and \E is messy. */        entirely. The code for handling \Q and \E is messy. */
3830    
3831        CHECK_RANGE:        CHECK_RANGE:
3832        while (ptr[1] == '\\' && ptr[2] == 'E')        while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3833          {          {
3834          inescq = FALSE;          inescq = FALSE;
3835          ptr += 2;          ptr += 2;
# Line 3094  for (;; ptr++) Line 3839  for (;; ptr++)
3839    
3840        /* Remember \r or \n */        /* Remember \r or \n */
3841    
3842        if (c == '\r' || c == '\n') cd->external_flags |= PCRE_HASCRORLF;        if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3843    
3844        /* Check for range */        /* Check for range */
3845    
3846        if (!inescq && ptr[1] == '-')        if (!inescq && ptr[1] == CHAR_MINUS)
3847          {          {
3848          int d;          int d;
3849          ptr += 2;          ptr += 2;
3850          while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3851    
3852          /* 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
3853          mode. */          mode. */
3854    
3855          while (*ptr == '\\' && ptr[1] == 'Q')          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3856            {            {
3857            ptr += 2;            ptr += 2;
3858            if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }            if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3859                { ptr += 2; continue; }
3860            inescq = TRUE;            inescq = TRUE;
3861            break;            break;
3862            }            }
3863    
3864          if (*ptr == 0 || (!inescq && *ptr == ']'))          if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3865            {            {
3866            ptr = oldptr;            ptr = oldptr;
3867            goto LONE_SINGLE_CHARACTER;            goto LONE_SINGLE_CHARACTER;
# Line 3134  for (;; ptr++) Line 3880  for (;; ptr++)
3880          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
3881          in such circumstances. */          in such circumstances. */
3882    
3883          if (!inescq && d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3884            {            {
3885            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3886            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3887    
3888            /* \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 */  
3889    
3890            if (d < 0)            if (d < 0)
3891              {              {
3892              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  
3893                {                {
3894                ptr = oldptr;                ptr = oldptr;
3895                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 3167  for (;; ptr++) Line 3910  for (;; ptr++)
3910    
3911          /* Remember \r or \n */          /* Remember \r or \n */
3912    
3913          if (d == '\r' || d == '\n') cd->external_flags |= PCRE_HASCRORLF;          if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3914    
3915          /* 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
3916          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 3287  for (;; ptr++) Line 4030  for (;; ptr++)
4030          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
4031            {            {
4032            unsigned int othercase;            unsigned int othercase;
4033            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)            if ((othercase = UCD_OTHERCASE(c)) != c)
4034              {              {
4035              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
4036              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 3312  for (;; ptr++) Line 4055  for (;; ptr++)
4055          }          }
4056        }        }
4057    
4058      /* 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.
4059        If we are at the end of an internal nested string, revert to the outer
4060        string. */
4061    
4062        while (((c = *(++ptr)) != 0 ||
4063               (nestptr != NULL &&
4064                 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != 0)) &&
4065               (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
4066    
4067      while ((c = *(++ptr)) != 0 && (c != ']' || inescq));      /* Check for missing terminating ']' */
4068    
4069      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)
4070        {        {
4071        *errorcodeptr = ERR6;        *errorcodeptr = ERR6;
4072        goto FAILED;        goto FAILED;
4073        }        }
4074    
   
 /* 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  
   
   
4075      /* 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
4076      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
4077      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 3348  we set the flag only if there is a liter Line 4079  we set the flag only if there is a liter
4079    
4080      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
4081      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
4082      operate on single-bytes only. This is an historical hangover. Maybe one day      operate on single-bytes characters only. This is an historical hangover.
4083      we can tidy these opcodes to handle multi-byte characters.      Maybe one day we can tidy these opcodes to handle multi-byte characters.
4084    
4085      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
4086      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.
4087      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
4088      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
4089      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
4090      reqbyte, save the previous value for reinstating. */      of reqbyte, save the previous value for reinstating. */
4091    
4092  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4093      if (class_charcount == 1 && !class_utf8 &&      if (class_charcount == 1 && !class_utf8 &&
# Line 3367  we set the flag only if there is a liter Line 4098  we set the flag only if there is a liter
4098        {        {
4099        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
4100    
4101        /* The OP_NOT opcode works on one-byte characters only. */        /* The OP_NOT[I] opcodes work on one-byte characters only. */
4102    
4103        if (negate_class)        if (negate_class)
4104          {          {
4105          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
4106          zerofirstbyte = firstbyte;          zerofirstbyte = firstbyte;
4107          *code++ = OP_NOT;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
4108          *code++ = class_lastchar;          *code++ = class_lastchar;
4109          break;          break;
4110          }          }
# Line 3404  we set the flag only if there is a liter Line 4135  we set the flag only if there is a liter
4135    
4136      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
4137      extended class, with its own opcode, unless there was a negated special      extended class, with its own opcode, unless there was a negated special
4138      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
4139      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
4140      (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
4141      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
4142        actual compiled code. */
4143    
4144  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4145      if (class_utf8 && !should_flip_negation)      if (class_utf8 && (!should_flip_negation || (options & PCRE_UCP) != 0))
4146        {        {
4147        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
4148        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 3436  we set the flag only if there is a liter Line 4168  we set the flag only if there is a liter
4168        }        }
4169  #endif  #endif
4170    
4171      /* 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
4172      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
4173      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
4174      map into the code vector, negating it if necessary. */      (non-UCP) in the class. Then copy the 32-byte map into the code vector,
4175        negating it if necessary. */
4176    
4177      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
4178      if (negate_class)      if (negate_class)
# Line 3459  we set the flag only if there is a liter Line 4192  we set the flag only if there is a liter
4192      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
4193      has been tested above. */      has been tested above. */
4194    
4195      case '{':      case CHAR_LEFT_CURLY_BRACKET:
4196      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
4197      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
4198      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
4199      goto REPEAT;      goto REPEAT;
4200    
4201      case '*':      case CHAR_ASTERISK:
4202      repeat_min = 0;      repeat_min = 0;
4203      repeat_max = -1;      repeat_max = -1;
4204      goto REPEAT;      goto REPEAT;
4205    
4206      case '+':      case CHAR_PLUS:
4207      repeat_min = 1;      repeat_min = 1;
4208      repeat_max = -1;      repeat_max = -1;
4209      goto REPEAT;      goto REPEAT;
4210    
4211      case '?':      case CHAR_QUESTION_MARK:
4212      repeat_min = 0;      repeat_min = 0;
4213      repeat_max = 1;      repeat_max = 1;
4214    
# Line 3499  we set the flag only if there is a liter Line 4232  we set the flag only if there is a liter
4232      op_type = 0;                    /* Default single-char op codes */      op_type = 0;                    /* Default single-char op codes */
4233      possessive_quantifier = FALSE;  /* Default not possessive quantifier */      possessive_quantifier = FALSE;  /* Default not possessive quantifier */
4234    
4235      /* 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
4236      for an inserted OP_ONCE for the additional '+' extension. */      insert something before it. */
4237    
4238      tempcode = previous;      tempcode = previous;
4239    
# Line 3510  we set the flag only if there is a liter Line 4243  we set the flag only if there is a liter
4243      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
4244      repeat type to the non-default. */      repeat type to the non-default. */
4245    
4246      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
4247        {        {
4248        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
4249        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
4250        ptr++;        ptr++;
4251        }        }
4252      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
4253        {        {
4254        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
4255        ptr++;        ptr++;
4256        }        }
4257      else repeat_type = greedy_default;      else repeat_type = greedy_default;
4258    
4259        /* If previous was a recursion call, wrap it in atomic brackets so that
4260        previous becomes the atomic group. All recursions were so wrapped in the
4261        past, but it no longer happens for non-repeated recursions. In fact, the
4262        repeated ones could be re-implemented independently so as not to need this,
4263        but for the moment we rely on the code for repeating groups. */
4264    
4265        if (*previous == OP_RECURSE)
4266          {
4267          memmove(previous + 1 + LINK_SIZE, previous, 1 + LINK_SIZE);
4268          *previous = OP_ONCE;
4269          PUT(previous, 1, 2 + 2*LINK_SIZE);
4270          previous[2 + 2*LINK_SIZE] = OP_KET;
4271          PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
4272          code += 2 + 2 * LINK_SIZE;
4273          length_prevgroup = 3 + 3*LINK_SIZE;
4274    
4275          /* When actually compiling, we need to check whether this was a forward
4276          reference, and if so, adjust the offset. */
4277    
4278          if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
4279            {
4280            int offset = GET(cd->hwm, -LINK_SIZE);
4281            if (offset == previous + 1 - cd->start_code)
4282              PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
4283            }
4284          }
4285    
4286        /* Now handle repetition for the different types of item. */
4287    
4288      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
4289      repeat item instead. If a char item has a minumum of more than one, ensure      repeat item instead. If a char item has a minumum of more than one, ensure
# Line 3529  we set the flag only if there is a liter Line 4291  we set the flag only if there is a liter
4291      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
4292      instead.  */      instead.  */
4293    
4294      if (*previous == OP_CHAR || *previous == OP_CHARNC)      if (*previous == OP_CHAR || *previous == OP_CHARI)
4295        {        {
4296          op_type = (*previous == OP_CHAR)? 0 : OP_STARI - OP_STAR;
4297    
4298        /* 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
4299        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
4300        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 3563  we set the flag only if there is a liter Line 4327  we set the flag only if there is a liter
4327    
4328        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4329            repeat_max < 0 &&            repeat_max < 0 &&
4330            check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
             options, cd))  
4331          {          {
4332          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4333          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3576  we set the flag only if there is a liter Line 4339  we set the flag only if there is a liter
4339      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
4340      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-
4341      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
4342      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
4343      currently used only for single-byte chars. */      are currently used only for single-byte chars. */
4344    
4345      else if (*previous == OP_NOT)      else if (*previous == OP_NOT || *previous == OP_NOTI)
4346        {        {
4347        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = ((*previous == OP_NOT)? OP_NOTSTAR : OP_NOTSTARI) - OP_STAR;
4348        c = previous[1];        c = previous[1];
4349        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4350            repeat_max < 0 &&            repeat_max < 0 &&
4351            check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4352          {          {
4353          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4354          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3609  we set the flag only if there is a liter Line 4372  we set the flag only if there is a liter
4372    
4373        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4374            repeat_max < 0 &&            repeat_max < 0 &&
4375            check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4376          {          {
4377          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4378          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3631  we set the flag only if there is a liter Line 4394  we set the flag only if there is a liter
4394    
4395        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
4396    
4397          /*--------------------------------------------------------------------*/
4398          /* This code is obsolete from release 8.00; the restriction was finally
4399          removed: */
4400    
4401        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4402        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4403    
4404        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4405          /*--------------------------------------------------------------------*/
4406    
4407        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
4408    
# Line 3773  we set the flag only if there is a liter Line 4541  we set the flag only if there is a liter
4541  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4542               *previous == OP_XCLASS ||               *previous == OP_XCLASS ||
4543  #endif  #endif
4544               *previous == OP_REF)               *previous == OP_REF ||
4545                 *previous == OP_REFI)
4546        {        {
4547        if (repeat_max == 0)        if (repeat_max == 0)
4548          {          {
# Line 3781  we set the flag only if there is a liter Line 4550  we set the flag only if there is a liter
4550          goto END_REPEAT;          goto END_REPEAT;
4551          }          }
4552    
4553          /*--------------------------------------------------------------------*/
4554          /* This code is obsolete from release 8.00; the restriction was finally
4555          removed: */
4556    
4557        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4558        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4559    
4560        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4561          /*--------------------------------------------------------------------*/
4562    
4563        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4564          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 3802  we set the flag only if there is a liter Line 4576  we set the flag only if there is a liter
4576        }        }
4577    
4578      /* 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
4579      cases. */      cases. Note that at this point we can encounter only the "basic" bracket
4580        opcodes such as BRA and CBRA, as this is the place where they get converted
4581        into the more special varieties such as BRAPOS and SBRA. A test for >=
4582        OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
4583        ASSERTBACK_NOT, ONCE, BRA, CBRA, and COND. Originally, PCRE did not allow
4584        repetition of assertions, but now it does, for Perl compatibility. */
4585    
4586      else if (*previous == OP_BRA  || *previous == OP_CBRA ||      else if (*previous >= OP_ASSERT && *previous <= OP_COND)
              *previous == OP_ONCE || *previous == OP_COND)  
4587        {        {
4588        register int i;        register int i;
4589        int ketoffset = 0;        int len = (int)(code - previous);
       int len = code - previous;  
4590        uschar *bralink = NULL;        uschar *bralink = NULL;
4591          uschar *brazeroptr = NULL;
4592    
4593        /* Repeating a DEFINE group is pointless */        /* Repeating a DEFINE group is pointless, but Perl allows the syntax, so
4594          we just ignore the repeat. */
4595    
4596        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
4597          {          goto END_REPEAT;
4598          *errorcodeptr = ERR55;  
4599          goto FAILED;        /* There is no sense in actually repeating assertions. The only potential
4600          }        use of repetition is in cases when the assertion is optional. Therefore,
4601          if the minimum is greater than zero, just ignore the repeat. If the
4602          maximum is not not zero or one, set it to 1. */
4603    
4604        /* If the maximum repeat count is unlimited, find the end of the bracket        if (*previous < OP_ONCE)    /* Assertion */
4605        by scanning through from the start, and compute the offset back to it          {
4606        from the current code pointer. There may be an OP_OPT setting following          if (repeat_min > 0) goto END_REPEAT;
4607        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;  
4608          }          }
4609    
4610        /* 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 3842  we set the flag only if there is a liter Line 4616  we set the flag only if there is a liter
4616    
4617        if (repeat_min == 0)        if (repeat_min == 0)
4618          {          {
4619          /* 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
4620          altogether. */          output altogether, like this:
   
         if (repeat_max == 0)  
           {  
           code = previous;  
           goto END_REPEAT;  
           }  
4621    
4622          /* If the maximum is 1 or unlimited, we just have to stick in the          ** if (repeat_max == 0)
4623          BRAZERO and do no more at this point. However, we do need to adjust          **   {
4624          any OP_RECURSE calls inside the group that refer to the group itself or          **   code = previous;
4625          any internal or forward referenced group, because the offset is from          **   goto END_REPEAT;
4626          the start of the whole regex. Temporarily terminate the pattern while          **   }
4627          doing this. */  
4628            However, that fails when a group or a subgroup within it is referenced
4629            as a subroutine from elsewhere in the pattern, so now we stick in
4630            OP_SKIPZERO in front of it so that it is skipped on execution. As we
4631            don't have a list of which groups are referenced, we cannot do this
4632            selectively.
4633    
4634            If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4635            and do no more at this point. However, we do need to adjust any
4636            OP_RECURSE calls inside the group that refer to the group itself or any
4637            internal or forward referenced group, because the offset is from the
4638            start of the whole regex. Temporarily terminate the pattern while doing
4639            this. */
4640    
4641          if (repeat_max <= 1)          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
4642            {            {
4643            *code = OP_END;            *code = OP_END;
4644            adjust_recurse(previous, 1, utf8, cd, save_hwm);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
4645            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
4646            code++;            code++;
4647              if (repeat_max == 0)
4648                {
4649                *previous++ = OP_SKIPZERO;
4650                goto END_REPEAT;
4651                }
4652              brazeroptr = previous;    /* Save for possessive optimizing */
4653            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4654            }            }
4655    
# Line 3888  we set the flag only if there is a liter Line 4674  we set the flag only if there is a liter
4674            /* We chain together the bracket offset fields that have to be            /* We chain together the bracket offset fields that have to be
4675            filled in later when the ends of the brackets are reached. */            filled in later when the ends of the brackets are reached. */
4676    
4677            offset = (bralink == NULL)? 0 : previous - bralink;            offset = (bralink == NULL)? 0 : (int)(previous - bralink);
4678            bralink = previous;            bralink = previous;
4679            PUTINC(previous, 0, offset);            PUTINC(previous, 0, offset);
4680            }            }
# Line 3909  we set the flag only if there is a liter Line 4695  we set the flag only if there is a liter
4695            {            {
4696            /* 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
4697            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
4698            potential integer overflow. */            potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
4699              integer type when available, otherwise double. */
4700    
4701            if (lengthptr != NULL)            if (lengthptr != NULL)
4702              {              {
4703              int delta = (repeat_min - 1)*length_prevgroup;              int delta = (repeat_min - 1)*length_prevgroup;
4704              if ((double)(repeat_min - 1)*(double)length_prevgroup >              if ((INT64_OR_DOUBLE)(repeat_min - 1)*
4705                                                              (double)INT_MAX ||                    (INT64_OR_DOUBLE)length_prevgroup >
4706                        (INT64_OR_DOUBLE)INT_MAX ||
4707                  OFLOW_MAX - *lengthptr < delta)                  OFLOW_MAX - *lengthptr < delta)
4708                {                {
4709                *errorcodeptr = ERR20;                *errorcodeptr = ERR20;
# Line 3961  we set the flag only if there is a liter Line 4749  we set the flag only if there is a liter
4749          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
4750          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
4751          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
4752          paranoid checks to avoid integer overflow. */          paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
4753            a 64-bit integer type when available, otherwise double. */
4754    
4755          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4756            {            {
4757            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4758                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4759            if ((double)repeat_max *            if ((INT64_OR_DOUBLE)repeat_max *
4760                  (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)                  (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4761                    > (double)INT_MAX ||                    > (INT64_OR_DOUBLE)INT_MAX ||
4762                OFLOW_MAX - *lengthptr < delta)                OFLOW_MAX - *lengthptr < delta)
4763              {              {
4764              *errorcodeptr = ERR20;              *errorcodeptr = ERR20;
# Line 3994  we set the flag only if there is a liter Line 4783  we set the flag only if there is a liter
4783              {              {
4784              int offset;              int offset;
4785              *code++ = OP_BRA;              *code++ = OP_BRA;
4786              offset = (bralink == NULL)? 0 : code - bralink;              offset = (bralink == NULL)? 0 : (int)(code - bralink);
4787              bralink = code;              bralink = code;
4788              PUTINC(code, 0, offset);              PUTINC(code, 0, offset);
4789              }              }
# Line 4015  we set the flag only if there is a liter Line 4804  we set the flag only if there is a liter
4804          while (bralink != NULL)          while (bralink != NULL)
4805            {            {
4806            int oldlinkoffset;            int oldlinkoffset;
4807            int offset = code - bralink + 1;            int offset = (int)(code - bralink + 1);
4808            uschar *bra = code - offset;            uschar *bra = code - offset;
4809            oldlinkoffset = GET(bra, 1);            oldlinkoffset = GET(bra, 1);
4810            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;
# Line 4025  we set the flag only if there is a liter Line 4814  we set the flag only if there is a liter
4814            }            }
4815          }          }
4816    
4817        /* 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
4818        can't just offset backwards from the current code point, because we        ONCE brackets, that's all we need to do. However, possessively repeated
4819        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
4820        correct offset was computed above.        behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to
4821          deal with possessive ONCEs specially.
4822    
4823          Otherwise, if the quantifier was possessive, we convert the BRA code to
4824          the POS form, and the KET code to KETRPOS. (It turns out to be convenient
4825          at runtime to detect this kind of subpattern at both the start and at the
4826          end.) The use of special opcodes makes it possible to reduce greatly the
4827          stack usage in pcre_exec(). If the group is preceded by OP_BRAZERO,
4828          convert this to OP_BRAPOSZERO. Then cancel the possessive flag so that
4829          the default action below, of wrapping everything inside atomic brackets,
4830          does not happen.
4831    
4832        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
4833        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
4834        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
4835        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
4836        atomic groups at runtime, but in a different way.] */        runtime, but in a different way.] */
4837    
4838        else        else
4839          {          {
4840          uschar *ketcode = code - ketoffset;          uschar *ketcode = code - 1 - LINK_SIZE;
4841          uschar *bracode = ketcode - GET(ketcode, 1);          uschar *bracode = ketcode - GET(ketcode, 1);
4842          *ketcode = OP_KETRMAX + repeat_type;  
4843          if (lengthptr == NULL && *bracode != OP_ONCE)          if (*bracode == OP_ONCE && possessive_quantifier) *bracode = OP_BRA;
4844            if (*bracode == OP_ONCE)
4845              *ketcode = OP_KETRMAX + repeat_type;
4846            else
4847            {            {
4848            uschar *scode = bracode;            if (possessive_quantifier)
4849            do              {
4850                *bracode += 1;                   /* Switch to xxxPOS opcodes */
4851                *ketcode = OP_KETRPOS;
4852                if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
4853                possessive_quantifier = FALSE;
4854                }
4855              else *ketcode = OP_KETRMAX + repeat_type;
4856    
4857              if (lengthptr == NULL)
4858              {              {
4859              if (could_be_empty_branch(scode, ketcode, utf8))              uschar *scode = bracode;
4860                do
4861                {                {
4862                *bracode += OP_SBRA - OP_BRA;                if (could_be_empty_branch(scode, ketcode, utf8, cd))
4863                break;                  {
4864                    *bracode += OP_SBRA - OP_BRA;
4865                    break;
4866                    }
4867                  scode += GET(scode, 1);
4868                }                }
4869              scode += GET(scode, 1);              while (*scode == OP_ALT);
4870              }              }
           while (*scode == OP_ALT);  
4871            }            }
4872          }          }
4873        }        }
4874    
4875        /* If previous is OP_FAIL, it was generated by an empty class [] in
4876        JavaScript mode. The other ways in which OP_FAIL can be generated, that is
4877        by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
4878        error above. We can just ignore the repeat in JS case. */
4879    
4880        else if (*previous == OP_FAIL) goto END_REPEAT;
4881    
4882      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
4883    
4884      else      else
# Line 4067  we set the flag only if there is a liter Line 4888  we set the flag only if there is a liter
4888        }        }
4889    
4890      /* If the character following a repeat is '+', or if certain optimization      /* If the character following a repeat is '+', or if certain optimization
4891      tests above succeeded, possessive_quantifier is TRUE. For some of the      tests above succeeded, possessive_quantifier is TRUE. For some opcodes,
4892      simpler opcodes, there is an special alternative opcode for this. For      there are special alternative opcodes for this case. For anything else, we
4893      anything else, we wrap the entire repeated item inside OP_ONCE brackets.      wrap the entire repeated item inside OP_ONCE brackets. Logically, the '+'
4894      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
4895      but the special opcodes can optimize it a bit. The repeated item starts at      special opcodes can optimize it.
4896      tempcode, not at previous, which might be the first part of a string whose  
4897      (former) last char we repeated.      Possessively repeated subpatterns have already been handled in the code
4898        just above, so possessive_quantifier is always FALSE for them at this
4899        stage.
4900    
4901        Note that the repeated item starts at tempcode, not at previous, which
4902        might be the first part of a string whose (former) last char we repeated.
4903    
4904      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
4905      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 4082  we set the flag only if there is a liter Line 4908  we set the flag only if there is a liter
4908      if (possessive_quantifier)      if (possessive_quantifier)
4909        {        {
4910        int len;        int len;
4911        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||  
4912            *tempcode == OP_NOTEXACT)        if (*tempcode == OP_TYPEEXACT)
4913          tempcode += _pcre_OP_lengths[*tempcode] +          tempcode += _pcre_OP_lengths[*tempcode] +
4914            ((*tempcode == OP_TYPEEXACT &&            ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
4915               (tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP))? 2:0);  
4916        len = code - tempcode;        else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
4917            {
4918            tempcode += _pcre_OP_lengths[*tempcode];
4919    #ifdef SUPPORT_UTF8
4920            if (utf8 && tempcode[-1] >= 0xc0)
4921              tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];
4922    #endif
4923            }
4924    
4925          len = (int)(code - tempcode);
4926        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4927          {          {
4928          case OP_STAR:  *tempcode = OP_POSSTAR; break;          case OP_STAR:  *tempcode = OP_POSSTAR; break;
# Line 4095  we set the flag only if there is a liter Line 4930  we set the flag only if there is a liter
4930          case OP_QUERY: *tempcode = OP_POSQUERY; break;          case OP_QUERY: *tempcode = OP_POSQUERY; break;
4931          case OP_UPTO:  *tempcode = OP_POSUPTO; break;          case OP_UPTO:  *tempcode = OP_POSUPTO; break;
4932    
4933          case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;          case OP_STARI:  *tempcode = OP_POSSTARI; break;
4934          case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;          case OP_PLUSI:  *tempcode = OP_POSPLUSI; break;
4935          case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;          case OP_QUERYI: *tempcode = OP_POSQUERYI; break;
4936          case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;          case OP_UPTOI:  *tempcode = OP_POSUPTOI; break;
4937    
4938          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;
4939          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;
4940          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4941          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4942    
4943            case OP_NOTSTARI:  *tempcode = OP_NOTPOSSTARI; break;
4944            case OP_NOTPLUSI:  *tempcode = OP_NOTPOSPLUSI; break;
4945            case OP_NOTQUERYI: *tempcode = OP_NOTPOSQUERYI; break;
4946            case OP_NOTUPTOI:  *tempcode = OP_NOTPOSUPTOI; break;
4947    
4948            case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;
4949            case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;
4950            case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
4951            case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;
4952    
4953            /* Because we are moving code along, we must ensure that any
4954            pending recursive references are updated. */
4955    
4956          default:          default:
4957            *code = OP_END;
4958            adjust_recurse(tempcode, 1 + LINK_SIZE, utf8, cd, save_hwm);
4959          memmove(tempcode + 1+LINK_SIZE, tempcode, len);          memmove(tempcode + 1+LINK_SIZE, tempcode, len);
4960          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
4961          len += 1 + LINK_SIZE;          len += 1 + LINK_SIZE;
# Line 4132  we set the flag only if there is a liter Line 4982  we set the flag only if there is a liter
4982      lookbehind or option setting or condition or all the other extended      lookbehind or option setting or condition or all the other extended
4983      parenthesis forms.  */      parenthesis forms.  */
4984    
4985      case '(':      case CHAR_LEFT_PARENTHESIS:
4986      newoptions = options;      newoptions = options;
4987      skipbytes = 0;      skipbytes = 0;
4988      bravalue = OP_CBRA;      bravalue = OP_CBRA;
# Line 4141  we set the flag only if there is a liter Line 4991  we set the flag only if there is a liter
4991    
4992      /* First deal with various "verbs" that can be introduced by '*'. */      /* First deal with various "verbs" that can be introduced by '*'. */
4993    
4994      if (*(++ptr) == '*' && (cd->ctypes[ptr[1]] & ctype_letter) != 0)      if (*(++ptr) == CHAR_ASTERISK &&
4995             ((cd->ctypes[ptr[1]] & ctype_letter) != 0 || ptr[1] == ':'))
4996        {        {
4997        int i, namelen;        int i, namelen;
4998          int arglen = 0;
4999        const char *vn = verbnames;        const char *vn = verbnames;
5000        const uschar *name = ++ptr;        const uschar *name = ptr + 1;
5001          const uschar *arg = NULL;
5002        previous = NULL;        previous = NULL;
5003        while ((cd->ctypes[*++ptr] & ctype_letter) != 0);        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
5004        if (*ptr == ':')        namelen = (int)(ptr - name);
5005    
5006          /* It appears that Perl allows any characters whatsoever, other than
5007          a closing parenthesis, to appear in arguments, so we no longer insist on
5008          letters, digits, and underscores. */
5009    
5010          if (*ptr == CHAR_COLON)
5011          {          {
5012          *errorcodeptr = ERR59;   /* Not supported */          arg = ++ptr;
5013          goto FAILED;          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
5014            arglen = (int)(ptr - arg);
5015          }          }
5016        if (*ptr != ')')  
5017          if (*ptr != CHAR_RIGHT_PARENTHESIS)
5018          {          {
5019          *errorcodeptr = ERR60;          *errorcodeptr = ERR60;
5020          goto FAILED;          goto FAILED;
5021          }          }
5022        namelen = ptr - name;  
5023          /* Scan the table of verb names */
5024    
5025        for (i = 0; i < verbcount; i++)        for (i = 0; i < verbcount; i++)
5026          {          {
5027          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
5028              strncmp((char *)name, vn, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
5029            {            {
5030            *code = verbs[i].op;            /* Check for open captures before ACCEPT and convert it to
5031            if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;            ASSERT_ACCEPT if in an assertion. */
5032            break;  
5033              if (verbs[i].op == OP_ACCEPT)
5034                {
5035                open_capitem *oc;
5036                if (arglen != 0)
5037                  {
5038                  *errorcodeptr = ERR59;
5039                  goto FAILED;
5040                  }
5041                cd->had_accept = TRUE;
5042                for (oc = cd->open_caps; oc != NULL; oc = oc->next)
5043                  {
5044                  *code++ = OP_CLOSE;
5045                  PUT2INC(code, 0, oc->number);
5046                  }
5047                *code++ = (cd->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT;
5048    
5049                /* Do not set firstbyte after *ACCEPT */
5050                if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
5051                }
5052    
5053              /* Handle other cases with/without an argument */
5054    
5055              else if (arglen == 0)
5056                {
5057                if (verbs[i].op < 0)   /* Argument is mandatory */
5058                  {
5059                  *errorcodeptr = ERR66;