/[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 300 by ph10, Mon Jan 14 19:43:19 2008 UTC revision 629 by ph10, Fri Jul 22 09:18:11 2011 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2007 University of Cambridge             Copyright (c) 1997-2011 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 53  supporting internal functions that are n Line 53  supporting internal functions that are n
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56  /* When DEBUG is defined, we need the pcre_printint() function, which is also  /* When PCRE_DEBUG is defined, we need the pcre_printint() function, which is
57  used by pcretest. DEBUG is not defined when building a production library. */  also used by pcretest. PCRE_DEBUG is not defined when building a production
58    library. */
59    
60  #ifdef DEBUG  #ifdef PCRE_DEBUG
61  #include "pcre_printint.src"  #include "pcre_printint.src"
62  #endif  #endif
63    
# Line 91  is 4 there is plenty of room. */ Line 92  is 4 there is plenty of room. */
92    
93  #define COMPILE_WORK_SIZE (4096)  #define COMPILE_WORK_SIZE (4096)
94    
95    /* The overrun tests check for a slightly smaller size so that they detect the
96    overrun before it actually does run off the end of the data block. */
97    
98    #define WORK_SIZE_CHECK (COMPILE_WORK_SIZE - 100)
99    
100    
101  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
102  are simple data values; negative values are for special things like \d and so  are simple data values; negative values are for special things like \d and so
103  on. Zero means further processing is needed (for things like \x), or the escape  on. Zero means further processing is needed (for things like \x), or the escape
104  is invalid. */  is invalid. */
105    
106  #ifndef EBCDIC  /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC
107    
108    /* This is the "normal" table for ASCII systems or for EBCDIC systems running
109    in UTF-8 mode. */
110    
111  static const short int escapes[] = {  static const short int escapes[] = {
112       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,                       0,
113       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,                       0,
114     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */       0,                       0,
115  -ESC_H,      0,      0, -ESC_K,      0,      0,      0,      0,   /* H - O */       0,                       0,
116  -ESC_P, -ESC_Q, -ESC_R, -ESC_S,      0,      0, -ESC_V, -ESC_W,   /* P - W */       0,                       0,
117  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */       CHAR_COLON,              CHAR_SEMICOLON,
118     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
119  -ESC_h,      0,      0, -ESC_k,      0,      0,  ESC_n,      0,   /* h - o */       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
120  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0, -ESC_v, -ESC_w,   /* p - w */       CHAR_COMMERCIAL_AT,      -ESC_A,
121       0,      0, -ESC_z                                            /* x - z */       -ESC_B,                  -ESC_C,
122         -ESC_D,                  -ESC_E,
123         0,                       -ESC_G,
124         -ESC_H,                  0,
125         0,                       -ESC_K,
126         0,                       0,
127         -ESC_N,                  0,
128         -ESC_P,                  -ESC_Q,
129         -ESC_R,                  -ESC_S,
130         0,                       0,
131         -ESC_V,                  -ESC_W,
132         -ESC_X,                  0,
133         -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,
134         CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,
135         CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,
136         CHAR_GRAVE_ACCENT,       7,
137         -ESC_b,                  0,
138         -ESC_d,                  ESC_e,
139         ESC_f,                   0,
140         -ESC_h,                  0,
141         0,                       -ESC_k,
142         0,                       0,
143         ESC_n,                   0,
144         -ESC_p,                  0,
145         ESC_r,                   -ESC_s,
146         ESC_tee,                 0,
147         -ESC_v,                  -ESC_w,
148         0,                       0,
149         -ESC_z
150  };  };
151    
152  #else           /* This is the "abnormal" table for EBCDIC systems */  #else
153    
154    /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
155    
156  static const short int escapes[] = {  static const short int escapes[] = {
157  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
158  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 130  static const short int escapes[] = { Line 171  static const short int escapes[] = {
171  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
172  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,
173  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
174  /*  D0 */   '}',     0, -ESC_K,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,-ESC_N,      0, -ESC_P,
175  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
176  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
177  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
# Line 142  static const short int escapes[] = { Line 183  static const short int escapes[] = {
183    
184  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
185  searched linearly. Put all the names into a single string, in order to reduce  searched linearly. Put all the names into a single string, in order to reduce
186  the number of relocations when a shared library is dynamically linked. */  the number of relocations when a shared library is dynamically linked. The
187    string is built from string macros so that it works in UTF-8 mode on EBCDIC
188    platforms. */
189    
190  typedef struct verbitem {  typedef struct verbitem {
191    int   len;    int   len;                 /* Length of verb name */
192    int   op;    int   op;                  /* Op when no arg, or -1 if arg mandatory */
193      int   op_arg;              /* Op when arg present, or -1 if not allowed */
194  } verbitem;  } verbitem;
195    
196  static const char verbnames[] =  static const char verbnames[] =
197    "ACCEPT\0"    "\0"                       /* Empty name is a shorthand for MARK */
198    "COMMIT\0"    STRING_MARK0
199    "F\0"    STRING_ACCEPT0
200    "FAIL\0"    STRING_COMMIT0
201    "PRUNE\0"    STRING_F0
202    "SKIP\0"    STRING_FAIL0
203    "THEN";    STRING_PRUNE0
204      STRING_SKIP0
205  static verbitem verbs[] = {    STRING_THEN;
206    { 6, OP_ACCEPT },  
207    { 6, OP_COMMIT },  static const verbitem verbs[] = {
208    { 1, OP_FAIL },    { 0, -1,        OP_MARK },
209    { 4, OP_FAIL },    { 4, -1,        OP_MARK },
210    { 5, OP_PRUNE },    { 6, OP_ACCEPT, -1 },
211    { 4, OP_SKIP  },    { 6, OP_COMMIT, -1 },
212    { 4, OP_THEN  }    { 1, OP_FAIL,   -1 },
213      { 4, OP_FAIL,   -1 },
214      { 5, OP_PRUNE,  OP_PRUNE_ARG },
215      { 4, OP_SKIP,   OP_SKIP_ARG  },
216      { 4, OP_THEN,   OP_THEN_ARG  }
217  };  };
218    
219  static int verbcount = sizeof(verbs)/sizeof(verbitem);  static const int verbcount = sizeof(verbs)/sizeof(verbitem);
220    
221    
222  /* Tables of names of POSIX character classes and their lengths. The names are  /* Tables of names of POSIX character classes and their lengths. The names are
# Line 178  length entry. The first three must be al Line 226  length entry. The first three must be al
226  for handling case independence. */  for handling case independence. */
227    
228  static const char posix_names[] =  static const char posix_names[] =
229    "alpha\0"  "lower\0"  "upper\0"  "alnum\0"  "ascii\0"  "blank\0"    STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
230    "cntrl\0"  "digit\0"  "graph\0"  "print\0"  "punct\0"  "space\0"    STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
231    "word\0"   "xdigit";    STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
232      STRING_word0  STRING_xdigit;
233    
234  static const uschar posix_name_lengths[] = {  static const uschar posix_name_lengths[] = {
235    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
# Line 212  static const int posix_class_maps[] = { Line 261  static const int posix_class_maps[] = {
261    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
262  };  };
263    
264    /* Table of substitutes for \d etc when PCRE_UCP is set. The POSIX class
265    substitutes must be in the order of the names, defined above, and there are
266    both positive and negative cases. NULL means no substitute. */
267    
268    #ifdef SUPPORT_UCP
269    static const uschar *substitutes[] = {
270      (uschar *)"\\P{Nd}",    /* \D */
271      (uschar *)"\\p{Nd}",    /* \d */
272      (uschar *)"\\P{Xsp}",   /* \S */       /* NOTE: Xsp is Perl space */
273      (uschar *)"\\p{Xsp}",   /* \s */
274      (uschar *)"\\P{Xwd}",   /* \W */
275      (uschar *)"\\p{Xwd}"    /* \w */
276    };
277    
278    static const uschar *posix_substitutes[] = {
279      (uschar *)"\\p{L}",     /* alpha */
280      (uschar *)"\\p{Ll}",    /* lower */
281      (uschar *)"\\p{Lu}",    /* upper */
282      (uschar *)"\\p{Xan}",   /* alnum */
283      NULL,                   /* ascii */
284      (uschar *)"\\h",        /* blank */
285      NULL,                   /* cntrl */
286      (uschar *)"\\p{Nd}",    /* digit */
287      NULL,                   /* graph */
288      NULL,                   /* print */
289      NULL,                   /* punct */
290      (uschar *)"\\p{Xps}",   /* space */    /* NOTE: Xps is POSIX space */
291      (uschar *)"\\p{Xwd}",   /* word */
292      NULL,                   /* xdigit */
293      /* Negated cases */
294      (uschar *)"\\P{L}",     /* ^alpha */
295      (uschar *)"\\P{Ll}",    /* ^lower */
296      (uschar *)"\\P{Lu}",    /* ^upper */
297      (uschar *)"\\P{Xan}",   /* ^alnum */
298      NULL,                   /* ^ascii */
299      (uschar *)"\\H",        /* ^blank */
300      NULL,                   /* ^cntrl */
301      (uschar *)"\\P{Nd}",    /* ^digit */
302      NULL,                   /* ^graph */
303      NULL,                   /* ^print */
304      NULL,                   /* ^punct */
305      (uschar *)"\\P{Xps}",   /* ^space */   /* NOTE: Xps is POSIX space */
306      (uschar *)"\\P{Xwd}",   /* ^word */
307      NULL                    /* ^xdigit */
308    };
309    #define POSIX_SUBSIZE (sizeof(posix_substitutes)/sizeof(uschar *))
310    #endif
311    
312  #define STRING(a)  # a  #define STRING(a)  # a
313  #define XSTRING(s) STRING(s)  #define XSTRING(s) STRING(s)
# Line 224  the number of relocations needed when a Line 320  the number of relocations needed when a
320  it is now one long string. We cannot use a table of offsets, because the  it is now one long string. We cannot use a table of offsets, because the
321  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
322  simply count through to the one we want - this isn't a performance issue  simply count through to the one we want - this isn't a performance issue
323  because these strings are used only when there is a compilation error. */  because these strings are used only when there is a compilation error.
324    
325    Each substring ends with \0 to insert a null character. This includes the final
326    substring, so that the whole string ends with \0\0, which can be detected when
327    counting through. */
328    
329  static const char error_texts[] =  static const char error_texts[] =
330    "no error\0"    "no error\0"
# Line 271  static const char error_texts[] = Line 371  static const char error_texts[] =
371    /* 35 */    /* 35 */
372    "invalid condition (?(0)\0"    "invalid condition (?(0)\0"
373    "\\C not allowed in lookbehind assertion\0"    "\\C not allowed in lookbehind assertion\0"
374    "PCRE does not support \\L, \\l, \\N, \\U, or \\u\0"    "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
375    "number after (?C is > 255\0"    "number after (?C is > 255\0"
376    "closing ) for (?C expected\0"    "closing ) for (?C expected\0"
377    /* 40 */    /* 40 */
# Line 295  static const char error_texts[] = Line 395  static const char error_texts[] =
395    /* 55 */    /* 55 */
396    "repeating a DEFINE group is not allowed\0"    "repeating a DEFINE group is not allowed\0"
397    "inconsistent NEWLINE options\0"    "inconsistent NEWLINE options\0"
398    "\\g is not followed by a braced name or an optionally braced non-zero number\0"    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
399    "(?+ or (?- or (?(+ or (?(- must be followed by a non-zero number\0"    "a numbered reference must not be zero\0"
400    "(*VERB) with an argument is not supported\0"    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
401    /* 60 */    /* 60 */
402    "(*VERB) not recognized\0"    "(*VERB) not recognized\0"
403    "number is too big\0"    "number is too big\0"
404    "subpattern name expected\0"    "subpattern name expected\0"
405    "digit expected after (?+";    "digit expected after (?+\0"
406      "] is an invalid data character in JavaScript compatibility mode\0"
407      /* 65 */
408      "different names for subpatterns of the same number are not allowed\0"
409      "(*MARK) must have an argument\0"
410      "this version of PCRE is not compiled with PCRE_UCP support\0"
411      "\\c must be followed by an ASCII character\0"
412      "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
413      ;
414    
415  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
416  patterns. Note that the tables in chartables are dependent on the locale, and  patterns. Note that the tables in chartables are dependent on the locale, and
# Line 321  For convenience, we use the same bit def Line 428  For convenience, we use the same bit def
428    
429  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
430    
431  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC
432    
433    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
434    UTF-8 mode. */
435    
436  static const unsigned char digitab[] =  static const unsigned char digitab[] =
437    {    {
438    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
# Line 357  static const unsigned char digitab[] = Line 468  static const unsigned char digitab[] =
468    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
469    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
470    
471  #else           /* This is the "abnormal" case, for EBCDIC systems */  #else
472    
473    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
474    
475  static const unsigned char digitab[] =  static const unsigned char digitab[] =
476    {    {
477    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
# Line 432  static const unsigned char ebcdic_charta Line 546  static const unsigned char ebcdic_charta
546  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
547    
548  static BOOL  static BOOL
549    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,    compile_regex(int, uschar **, const uschar **, int *, BOOL, BOOL, int, int *,
550      int *, int *, branch_chain *, compile_data *, int *);      int *, branch_chain *, compile_data *, int *);
551    
552    
553    
# Line 454  static const char * Line 568  static const char *
568  find_error_text(int n)  find_error_text(int n)
569  {  {
570  const char *s = error_texts;  const char *s = error_texts;
571  for (; n > 0; n--) while (*s++ != 0);  for (; n > 0; n--)
572      {
573      while (*s++ != 0) {};
574      if (*s == 0) return "Error text not found (please report)";
575      }
576  return s;  return s;
577  }  }
578    
# Line 502  if (c == 0) *errorcodeptr = ERR1; Line 620  if (c == 0) *errorcodeptr = ERR1;
620  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.
621  Otherwise further processing may be required. */  Otherwise further processing may be required. */
622    
623  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
624  else if (c < '0' || c > 'z') {}                           /* Not alphanumeric */  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */
625  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0) c = i;
626    
627  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
628  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 641  else
641      /* 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
642      error. */      error. */
643    
644      case 'l':      case CHAR_l:
645      case 'L':      case CHAR_L:
646      case 'N':      case CHAR_u:
647      case 'u':      case CHAR_U:
     case 'U':  
648      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
649      break;      break;
650    
651      /* \g must be followed by a number, either plain or braced. If positive, it      /* \g must be followed by one of a number of specific things:
652      is an absolute backreference. If negative, it is a relative backreference.  
653      This is a Perl 5.10 feature. Perl 5.10 also supports \g{name} as a      (1) A number, either plain or braced. If positive, it is an absolute
654      reference to a named group. This is part of Perl's movement towards a      backreference. If negative, it is a relative backreference. This is a Perl
655      unified syntax for back references. As this is synonymous with \k{name}, we      5.10 feature.
656      fudge it up by pretending it really was \k. */  
657        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
658        is part of Perl's movement towards a unified syntax for back references. As
659        this is synonymous with \k{name}, we fudge it up by pretending it really
660        was \k.
661    
662        (3) For Oniguruma compatibility we also support \g followed by a name or a
663        number either in angle brackets or in single quotes. However, these are
664        (possibly recursive) subroutine calls, _not_ backreferences. Just return
665        the -ESC_g code (cf \k). */
666    
667      case 'g':      case CHAR_g:
668      if (ptr[1] == '{')      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
669          {
670          c = -ESC_g;
671          break;
672          }
673    
674        /* Handle the Perl-compatible cases */
675    
676        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
677        {        {
678        const uschar *p;        const uschar *p;
679        for (p = ptr+2; *p != 0 && *p != '}'; p++)        for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
680          if (*p != '-' && (digitab[*p] & ctype_digit) == 0) break;          if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
681        if (*p != 0 && *p != '}')        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
682          {          {
683          c = -ESC_k;          c = -ESC_k;
684          break;          break;
# Line 554  else Line 688  else
688        }        }
689      else braced = FALSE;      else braced = FALSE;
690    
691      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
692        {        {
693        negated = TRUE;        negated = TRUE;
694        ptr++;        ptr++;
# Line 563  else Line 697  else
697    
698      c = 0;      c = 0;
699      while ((digitab[ptr[1]] & ctype_digit) != 0)      while ((digitab[ptr[1]] & ctype_digit) != 0)
700        c = c * 10 + *(++ptr) - '0';        c = c * 10 + *(++ptr) - CHAR_0;
701    
702      if (c < 0)      if (c < 0)   /* Integer overflow */
703        {        {
704        *errorcodeptr = ERR61;        *errorcodeptr = ERR61;
705        break;        break;
706        }        }
707    
708      if (c == 0 || (braced && *(++ptr) != '}'))      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
709        {        {
710        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
711        break;        break;
712        }        }
713    
714        if (c == 0)
715          {
716          *errorcodeptr = ERR58;
717          break;
718          }
719    
720      if (negated)      if (negated)
721        {        {
722        if (c > bracount)        if (c > bracount)
# Line 602  else Line 742  else
742      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
743      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
744    
745      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:
746      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
747    
748      if (!isclass)      if (!isclass)
749        {        {
750        oldptr = ptr;        oldptr = ptr;
751        c -= '0';        c -= CHAR_0;
752        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
753          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - CHAR_0;
754        if (c < 0)        if (c < 0)    /* Integer overflow */
755          {          {
756          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
757          break;          break;
# Line 628  else Line 768  else
768      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.
769      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
770    
771      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
772        {        {
773        ptr--;        ptr--;
774        c = 0;        c = 0;
# Line 641  else Line 781  else
781      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
782      than 3 octal digits. */      than 3 octal digits. */
783    
784      case '0':      case CHAR_0:
785      c -= '0';      c -= CHAR_0;
786      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
787          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
788      if (!utf8 && c > 255) *errorcodeptr = ERR51;      if (!utf8 && c > 255) *errorcodeptr = ERR51;
789      break;      break;
790    
# Line 652  else Line 792  else
792      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
793      treated as a data character. */      treated as a data character. */
794    
795      case 'x':      case CHAR_x:
796      if (ptr[1] == '{')      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
797        {        {
798        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
799        int count = 0;        int count = 0;
# Line 662  else Line 802  else
802        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
803          {          {
804          register int cc = *pt++;          register int cc = *pt++;
805          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
806          count++;          count++;
807    
808  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
809          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
810          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
811  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
812          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
813          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
814  #endif  #endif
815          }          }
816    
817        if (*pt == '}')        if (*pt == CHAR_RIGHT_CURLY_BRACKET)
818          {          {
819          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
820          ptr = pt;          ptr = pt;
# Line 690  else Line 830  else
830      c = 0;      c = 0;
831      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
832        {        {
833        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
834        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
835  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
836        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
837        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
838  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
839        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
840        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
841  #endif  #endif
842        }        }
843      break;      break;
844    
845      /* 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.
846      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
847        coding is ASCII-specific, but then the whole concept of \cx is
848      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
849    
850      case 'c':      case CHAR_c:
851      c = *(++ptr);      c = *(++ptr);
852      if (c == 0)      if (c == 0)
853        {        {
854        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
855        break;        break;
856        }        }
857    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
858  #ifndef EBCDIC  /* ASCII coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
859      if (c >= 'a' && c <= 'z') c -= 32;        {
860          *errorcodeptr = ERR68;
861          break;
862          }
863        if (c >= CHAR_a && c <= CHAR_z) c -= 32;
864      c ^= 0x40;      c ^= 0x40;
865  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
866      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
867      c ^= 0xC0;      c ^= 0xC0;
868  #endif  #endif
869      break;      break;
# Line 740  else Line 885  else
885      }      }
886    }    }
887    
888    /* Perl supports \N{name} for character names, as well as plain \N for "not
889    newline". PCRE does not support \N{name}. */
890    
891    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET)
892      *errorcodeptr = ERR37;
893    
894    /* If PCRE_UCP is set, we change the values for \d etc. */
895    
896    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
897      c -= (ESC_DU - ESC_D);
898    
899    /* Set the pointer to the final character before returning. */
900    
901  *ptrptr = ptr;  *ptrptr = ptr;
902  return c;  return c;
903  }  }
# Line 780  if (c == 0) goto ERROR_RETURN; Line 938  if (c == 0) goto ERROR_RETURN;
938  /* \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
939  negation. */  negation. */
940    
941  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
942    {    {
943    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
944      {      {
945      *negptr = TRUE;      *negptr = TRUE;
946      ptr++;      ptr++;
# Line 791  if (c == '{') Line 949  if (c == '{')
949      {      {
950      c = *(++ptr);      c = *(++ptr);
951      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
952      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
953      name[i] = c;      name[i] = c;
954      }      }
955    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
956    name[i] = 0;    name[i] = 0;
957    }    }
958    
# Line 859  is_counted_repeat(const uschar *p) Line 1017  is_counted_repeat(const uschar *p)
1017  {  {
1018  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
1019  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
1020  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
1021    
1022  if (*p++ != ',') return FALSE;  if (*p++ != CHAR_COMMA) return FALSE;
1023  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
1024    
1025  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
1026  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
1027    
1028  return (*p == '}');  return (*p == CHAR_RIGHT_CURLY_BRACKET);
1029  }  }
1030    
1031    
# Line 900  int max = -1; Line 1058  int max = -1;
1058  /* 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
1059  an integer overflow. */  an integer overflow. */
1060    
1061  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;
1062  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1063    {    {
1064    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 910  if (min < 0 || min > 65535) Line 1068  if (min < 0 || min > 65535)
1068  /* 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.
1069  Also, max must not be less than min. */  Also, max must not be less than min. */
1070    
1071  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1072    {    {
1073    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1074      {      {
1075      max = 0;      max = 0;
1076      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;
1077      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1078        {        {
1079        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 940  return p; Line 1098  return p;
1098    
1099    
1100  /*************************************************  /*************************************************
1101  *       Find forward referenced subpattern       *  *  Subroutine for finding forward reference      *
1102  *************************************************/  *************************************************/
1103    
1104  /* This function scans along a pattern's text looking for capturing  /* This recursive function is called only from find_parens() below. The
1105    top-level call starts at the beginning of the pattern. All other calls must
1106    start at a parenthesis. It scans along a pattern's text looking for capturing
1107  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
1108  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
1109  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
1110  references to subpatterns. We know that if (?P< is encountered, the name will  track of subpatterns that reset the capturing group numbers - the (?| feature.
1111  be terminated by '>' because that is checked in the first pass.  
1112    This function was originally called only from the second pass, in which we know
1113    that if (?< or (?' or (?P< is encountered, the name will be correctly
1114    terminated because that is checked in the first pass. There is now one call to
1115    this function in the first pass, to check for a recursive back reference by
1116    name (so that we can make the whole group atomic). In this case, we need check
1117    only up to the current position in the pattern, and that is still OK because
1118    and previous occurrences will have been checked. To make this work, the test
1119    for "end of pattern" is a check against cd->end_pattern in the main loop,
1120    instead of looking for a binary zero. This means that the special first-pass
1121    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1122    processing items within the loop are OK, because afterwards the main loop will
1123    terminate.)
1124    
1125  Arguments:  Arguments:
1126    ptr          current position in the pattern    ptrptr       address of the current character pointer (updated)
1127    count        current count of capturing parens so far encountered    cd           compile background data
1128    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1129    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1130    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1131      utf8         TRUE if we are in UTF-8 mode
1132      count        pointer to the current capturing subpattern number (updated)
1133    
1134  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1135  */  */
1136    
1137  static int  static int
1138  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,
1139    BOOL xmode)    BOOL xmode, BOOL utf8, int *count)
1140  {  {
1141  const uschar *thisname;  uschar *ptr = *ptrptr;
1142    int start_count = *count;
1143    int hwm_count = start_count;
1144    BOOL dup_parens = FALSE;
1145    
1146    /* If the first character is a parenthesis, check on the type of group we are
1147    dealing with. The very first call may not start with a parenthesis. */
1148    
1149  for (; *ptr != 0; ptr++)  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1150    {    {
1151    int term;    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1152    
1153      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1154    
1155      /* Handle a normal, unnamed capturing parenthesis. */
1156    
1157      else if (ptr[1] != CHAR_QUESTION_MARK)
1158        {
1159        *count += 1;
1160        if (name == NULL && *count == lorn) return *count;
1161        ptr++;
1162        }
1163    
1164      /* All cases now have (? at the start. Remember when we are in a group
1165      where the parenthesis numbers are duplicated. */
1166    
1167      else if (ptr[2] == CHAR_VERTICAL_LINE)
1168        {
1169        ptr += 3;
1170        dup_parens = TRUE;
1171        }
1172    
1173      /* Handle comments; all characters are allowed until a ket is reached. */
1174    
1175      else if (ptr[2] == CHAR_NUMBER_SIGN)
1176        {
1177        for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1178        goto FAIL_EXIT;
1179        }
1180    
1181      /* Handle a condition. If it is an assertion, just carry on so that it
1182      is processed as normal. If not, skip to the closing parenthesis of the
1183      condition (there can't be any nested parens). */
1184    
1185      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1186        {
1187        ptr += 2;
1188        if (ptr[1] != CHAR_QUESTION_MARK)
1189          {
1190          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1191          if (*ptr != 0) ptr++;
1192          }
1193        }
1194    
1195      /* Start with (? but not a condition. */
1196    
1197      else
1198        {
1199        ptr += 2;
1200        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1201    
1202        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1203    
1204        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1205            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1206          {
1207          int term;
1208          const uschar *thisname;
1209          *count += 1;
1210          if (name == NULL && *count == lorn) return *count;
1211          term = *ptr++;
1212          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1213          thisname = ptr;
1214          while (*ptr != term) ptr++;
1215          if (name != NULL && lorn == ptr - thisname &&
1216              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1217            return *count;
1218          term++;
1219          }
1220        }
1221      }
1222    
1223    /* Past any initial parenthesis handling, scan for parentheses or vertical
1224    bars. Stop if we get to cd->end_pattern. Note that this is important for the
1225    first-pass call when this value is temporarily adjusted to stop at the current
1226    position. So DO NOT change this to a test for binary zero. */
1227    
1228    for (; ptr < cd->end_pattern; ptr++)
1229      {
1230    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1231    
1232    if (*ptr == '\\')    if (*ptr == CHAR_BACKSLASH)
1233      {      {
1234      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1235      if (*ptr == 'Q') for (;;)      if (*ptr == CHAR_Q) for (;;)
1236        {        {
1237        while (*(++ptr) != 0 && *ptr != '\\');        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1238        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1239        if (*(++ptr) == 'E') break;        if (*(++ptr) == CHAR_E) break;
1240        }        }
1241      continue;      continue;
1242      }      }
1243    
1244    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1245      are handled for real. If the first character is '^', skip it. Also, if the
1246      first few characters (either before or after ^) are \Q\E or \E we skip them
1247      too. This makes for compatibility with Perl. Note the use of STR macros to
1248      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1249    
1250    if (*ptr == '[')    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1251      {      {
1252      while (*(++ptr) != ']')      BOOL negate_class = FALSE;
1253        for (;;)
1254          {
1255          if (ptr[1] == CHAR_BACKSLASH)
1256            {
1257            if (ptr[2] == CHAR_E)
1258              ptr+= 2;
1259            else if (strncmp((const char *)ptr+2,
1260                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1261              ptr += 4;
1262            else
1263              break;
1264            }
1265          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1266            {
1267            negate_class = TRUE;
1268            ptr++;
1269            }
1270          else break;
1271          }
1272    
1273        /* If the next character is ']', it is a data character that must be
1274        skipped, except in JavaScript compatibility mode. */
1275    
1276        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1277            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1278          ptr++;
1279    
1280        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1281        {        {
1282        if (*ptr == 0) return -1;        if (*ptr == 0) return -1;
1283        if (*ptr == '\\')        if (*ptr == CHAR_BACKSLASH)
1284          {          {
1285          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) goto FAIL_EXIT;
1286          if (*ptr == 'Q') for (;;)          if (*ptr == CHAR_Q) for (;;)
1287            {            {
1288            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1289            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1290            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1291            }            }
1292          continue;          continue;
1293          }          }
# Line 1008  for (; *ptr != 0; ptr++) Line 1297  for (; *ptr != 0; ptr++)
1297    
1298    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1299    
1300    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1301      {      {
1302      while (*(++ptr) != 0 && *ptr != '\n');      ptr++;
1303      if (*ptr == 0) return -1;      while (*ptr != 0)
1304          {
1305          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1306          ptr++;
1307    #ifdef SUPPORT_UTF8
1308          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
1309    #endif
1310          }
1311        if (*ptr == 0) goto FAIL_EXIT;
1312      continue;      continue;
1313      }      }
1314    
1315    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1316    
1317    if (*ptr != '(') continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != '?' && ptr[1] != '*')  
1318      {      {
1319      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, count);
1320      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1321      continue;      if (*ptr == 0) goto FAIL_EXIT;
1322      }      }
1323    
1324    ptr += 2;    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1325    if (*ptr == 'P') ptr++;                      /* Allow optional P */      {
1326        if (dup_parens && *count < hwm_count) *count = hwm_count;
1327        goto FAIL_EXIT;
1328        }
1329    
1330    /* We have to disambiguate (?<! and (?<= from (?<name> */    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1331        {
1332        if (*count > hwm_count) hwm_count = *count;
1333        *count = start_count;
1334        }
1335      }
1336    
1337    if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  FAIL_EXIT:
1338         *ptr != '\'')  *ptrptr = ptr;
1339      continue;  return -1;
1340    }
1341    
   count++;  
1342    
1343    if (name == NULL && count == lorn) return count;  
1344    term = *ptr++;  
1345    if (term == '<') term = '>';  /*************************************************
1346    thisname = ptr;  *       Find forward referenced subpattern       *
1347    while (*ptr != term) ptr++;  *************************************************/
1348    if (name != NULL && lorn == ptr - thisname &&  
1349        strncmp((const char *)name, (const char *)thisname, lorn) == 0)  /* This function scans along a pattern's text looking for capturing
1350      return count;  subpatterns, and counting them. If it finds a named pattern that matches the
1351    name it is given, it returns its number. Alternatively, if the name is NULL, it
1352    returns when it reaches a given numbered subpattern. This is used for forward
1353    references to subpatterns. We used to be able to start this scan from the
1354    current compiling point, using the current count value from cd->bracount, and
1355    do it all in a single loop, but the addition of the possibility of duplicate
1356    subpattern numbers means that we have to scan from the very start, in order to
1357    take account of such duplicates, and to use a recursive function to keep track
1358    of the different types of group.
1359    
1360    Arguments:
1361      cd           compile background data
1362      name         name to seek, or NULL if seeking a numbered subpattern
1363      lorn         name length, or subpattern number if name is NULL
1364      xmode        TRUE if we are in /x mode
1365      utf8         TRUE if we are in UTF-8 mode
1366    
1367    Returns:       the number of the found subpattern, or -1 if not found
1368    */
1369    
1370    static int
1371    find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode,
1372      BOOL utf8)
1373    {
1374    uschar *ptr = (uschar *)cd->start_pattern;
1375    int count = 0;
1376    int rc;
1377    
1378    /* If the pattern does not start with an opening parenthesis, the first call
1379    to find_parens_sub() will scan right to the end (if necessary). However, if it
1380    does start with a parenthesis, find_parens_sub() will return when it hits the
1381    matching closing parens. That is why we have to have a loop. */
1382    
1383    for (;;)
1384      {
1385      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, &count);
1386      if (rc > 0 || *ptr++ == 0) break;
1387    }    }
1388    
1389  return -1;  return rc;
1390  }  }
1391    
1392    
1393    
1394    
1395  /*************************************************  /*************************************************
1396  *      Find first significant op code            *  *      Find first significant op code            *
1397  *************************************************/  *************************************************/
1398    
1399  /* This is called by several functions that scan a compiled expression looking  /* This is called by several functions that scan a compiled expression looking
1400  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
1401  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
1402  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
1403  assertions, and also the \b assertion; for others it does not.  does not.
1404    
1405  Arguments:  Arguments:
1406    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  
1407    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1408    
1409  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1410  */  */
1411    
1412  static const uschar*  static const uschar*
1413  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const uschar *code, BOOL skipassert)
   BOOL skipassert)  
1414  {  {
1415  for (;;)  for (;;)
1416    {    {
1417    switch ((int)*code)    switch ((int)*code)
1418      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1419      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1420      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1421      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
# Line 1100  for (;;) Line 1431  for (;;)
1431    
1432      case OP_CALLOUT:      case OP_CALLOUT:
1433      case OP_CREF:      case OP_CREF:
1434        case OP_NCREF:
1435      case OP_RREF:      case OP_RREF:
1436        case OP_NRREF:
1437      case OP_DEF:      case OP_DEF:
1438      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1439      break;      break;
# Line 1116  for (;;) Line 1449  for (;;)
1449    
1450    
1451  /*************************************************  /*************************************************
1452  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1453  *************************************************/  *************************************************/
1454    
1455  /* 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,
1456  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.
1457  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
1458    temporarily terminated with OP_END when this function is called.
1459    
1460    This function is called when a backward assertion is encountered, so that if it
1461    fails, the error message can point to the correct place in the pattern.
1462    However, we cannot do this when the assertion contains subroutine calls,
1463    because they can be forward references. We solve this by remembering this case
1464    and doing the check at the end; a flag specifies which mode we are running in.
1465    
1466  Arguments:  Arguments:
1467    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1468    options  the compiling options    utf8     TRUE in UTF-8 mode
1469      atend    TRUE if called when the pattern is complete
1470      cd       the "compile data" structure
1471    
1472  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1473                 or -1 if there is no fixed length,
1474               or -2 if \C was encountered               or -2 if \C was encountered
1475                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1476  */  */
1477    
1478  static int  static int
1479  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, BOOL utf8, BOOL atend, compile_data *cd)
1480  {  {
1481  int length = -1;  int length = -1;
1482    
# Line 1145  branch, check the length against that of Line 1489  branch, check the length against that of
1489  for (;;)  for (;;)
1490    {    {
1491    int d;    int d;
1492      uschar *ce, *cs;
1493    register int op = *cc;    register int op = *cc;
1494    switch (op)    switch (op)
1495      {      {
1496        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1497        OP_BRA (normal non-capturing bracket) because the other variants of these
1498        opcodes are all concerned with unlimited repeated groups, which of course
1499        are not of fixed length. They will cause a -1 response from the default
1500        case of this switch. */
1501    
1502      case OP_CBRA:      case OP_CBRA:
1503      case OP_BRA:      case OP_BRA:
1504      case OP_ONCE:      case OP_ONCE:
1505      case OP_COND:      case OP_COND:
1506      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), utf8, atend, cd);
1507      if (d < 0) return d;      if (d < 0) return d;
1508      branchlength += d;      branchlength += d;
1509      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1161  for (;;) Line 1512  for (;;)
1512    
1513      /* 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
1514      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
1515      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.
1516        Note that we must not include the OP_KETRxxx opcodes here, because they
1517        all imply an unlimited repeat. */
1518    
1519      case OP_ALT:      case OP_ALT:
1520      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1521      case OP_END:      case OP_END:
1522      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1523        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
# Line 1175  for (;;) Line 1526  for (;;)
1526      branchlength = 0;      branchlength = 0;
1527      break;      break;
1528    
1529        /* A true recursion implies not fixed length, but a subroutine call may
1530        be OK. If the subroutine is a forward reference, we can't deal with
1531        it until the end of the pattern, so return -3. */
1532    
1533        case OP_RECURSE:
1534        if (!atend) return -3;
1535        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1536        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1537        if (cc > cs && cc < ce) return -1;                /* Recursion */
1538        d = find_fixedlength(cs + 2, utf8, atend, cd);
1539        if (d < 0) return d;
1540        branchlength += d;
1541        cc += 1 + LINK_SIZE;
1542        break;
1543    
1544      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1545    
1546      case OP_ASSERT:      case OP_ASSERT:
# Line 1188  for (;;) Line 1554  for (;;)
1554    
1555      case OP_REVERSE:      case OP_REVERSE:
1556      case OP_CREF:      case OP_CREF:
1557        case OP_NCREF:
1558      case OP_RREF:      case OP_RREF:
1559        case OP_NRREF:
1560      case OP_DEF:      case OP_DEF:
     case OP_OPT:  
1561      case OP_CALLOUT:      case OP_CALLOUT:
1562      case OP_SOD:      case OP_SOD:
1563      case OP_SOM:      case OP_SOM:
1564        case OP_SET_SOM:
1565      case OP_EOD:      case OP_EOD:
1566      case OP_EODN:      case OP_EODN:
1567      case OP_CIRC:      case OP_CIRC:
1568        case OP_CIRCM:
1569      case OP_DOLL:      case OP_DOLL:
1570        case OP_DOLLM:
1571      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1572      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1573      cc += _pcre_OP_lengths[*cc];      cc += _pcre_OP_lengths[*cc];
# Line 1206  for (;;) Line 1576  for (;;)
1576      /* Handle literal characters */      /* Handle literal characters */
1577    
1578      case OP_CHAR:      case OP_CHAR:
1579      case OP_CHARNC:      case OP_CHARI:
1580      case OP_NOT:      case OP_NOT:
1581        case OP_NOTI:
1582      branchlength++;      branchlength++;
1583      cc += 2;      cc += 2;
1584  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1585      if ((options & PCRE_UTF8) != 0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       {  
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1586  #endif  #endif
1587      break;      break;
1588    
# Line 1225  for (;;) Line 1593  for (;;)
1593      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1594      cc += 4;      cc += 4;
1595  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1596      if ((options & PCRE_UTF8) != 0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       {  
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1597  #endif  #endif
1598      break;      break;
1599    
# Line 1252  for (;;) Line 1617  for (;;)
1617      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1618      case OP_WORDCHAR:      case OP_WORDCHAR:
1619      case OP_ANY:      case OP_ANY:
1620        case OP_ALLANY:
1621      branchlength++;      branchlength++;
1622      cc++;      cc++;
1623      break;      break;
# Line 1306  for (;;) Line 1672  for (;;)
1672    
1673    
1674  /*************************************************  /*************************************************
1675  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1676  *************************************************/  *************************************************/
1677    
1678  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1679  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1680    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1681    so that it can be called from pcre_study() when finding the minimum matching
1682    length.
1683    
1684  Arguments:  Arguments:
1685    code        points to start of expression    code        points to start of expression
1686    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1687    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1688    
1689  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
1690  */  */
1691    
1692  static const uschar *  const uschar *
1693  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1694  {  {
1695  for (;;)  for (;;)
1696    {    {
1697    register int c = *code;    register int c = *code;
1698    
1699    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1700    
1701    /* XCLASS is used for classes that cannot be represented just by a bit    /* XCLASS is used for classes that cannot be represented just by a bit
# Line 1334  for (;;) Line 1704  for (;;)
1704    
1705    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1706    
1707      /* Handle recursion */
1708    
1709      else if (c == OP_REVERSE)
1710        {
1711        if (number < 0) return (uschar *)code;
1712        code += _pcre_OP_lengths[c];
1713        }
1714    
1715    /* Handle capturing bracket */    /* Handle capturing bracket */
1716    
1717    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
1718               c == OP_CBRAPOS || c == OP_SCBRAPOS)
1719      {      {
1720      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
1721      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
# Line 1345  for (;;) Line 1724  for (;;)
1724    
1725    /* 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
1726    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
1727    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1728      must add in its length. */
1729    
1730    else    else
1731      {      {
# Line 1369  for (;;) Line 1749  for (;;)
1749        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
1750        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1751        break;        break;
1752    
1753          case OP_MARK:
1754          case OP_PRUNE_ARG:
1755          case OP_SKIP_ARG:
1756          code += code[1];
1757          break;
1758    
1759          case OP_THEN_ARG:
1760          code += code[1+LINK_SIZE];
1761          break;
1762        }        }
1763    
1764      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1383  for (;;) Line 1773  for (;;)
1773      if (utf8) switch(c)      if (utf8) switch(c)
1774        {        {
1775        case OP_CHAR:        case OP_CHAR:
1776        case OP_CHARNC:        case OP_CHARI:
1777        case OP_EXACT:        case OP_EXACT:
1778          case OP_EXACTI:
1779        case OP_UPTO:        case OP_UPTO:
1780          case OP_UPTOI:
1781        case OP_MINUPTO:        case OP_MINUPTO:
1782          case OP_MINUPTOI:
1783        case OP_POSUPTO:        case OP_POSUPTO:
1784          case OP_POSUPTOI:
1785        case OP_STAR:        case OP_STAR:
1786          case OP_STARI:
1787        case OP_MINSTAR:        case OP_MINSTAR:
1788          case OP_MINSTARI:
1789        case OP_POSSTAR:        case OP_POSSTAR:
1790          case OP_POSSTARI:
1791        case OP_PLUS:        case OP_PLUS:
1792          case OP_PLUSI:
1793        case OP_MINPLUS:        case OP_MINPLUS:
1794          case OP_MINPLUSI:
1795        case OP_POSPLUS:        case OP_POSPLUS:
1796          case OP_POSPLUSI:
1797        case OP_QUERY:        case OP_QUERY:
1798          case OP_QUERYI:
1799        case OP_MINQUERY:        case OP_MINQUERY:
1800          case OP_MINQUERYI:
1801        case OP_POSQUERY:        case OP_POSQUERY:
1802          case OP_POSQUERYI:
1803        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1804        break;        break;
1805        }        }
1806    #else
1807        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1808  #endif  #endif
1809      }      }
1810    }    }
# Line 1438  for (;;) Line 1843  for (;;)
1843    
1844    /* 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
1845    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
1846    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1847      must add in its length. */
1848    
1849    else    else
1850      {      {
# Line 1462  for (;;) Line 1868  for (;;)
1868        case OP_TYPEEXACT:        case OP_TYPEEXACT:
1869        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1870        break;        break;
1871    
1872          case OP_MARK:
1873          case OP_PRUNE_ARG:
1874          case OP_SKIP_ARG:
1875          code += code[1];
1876          break;
1877    
1878          case OP_THEN_ARG:
1879          code += code[1+LINK_SIZE];
1880          break;
1881        }        }
1882    
1883      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1476  for (;;) Line 1892  for (;;)
1892      if (utf8) switch(c)      if (utf8) switch(c)
1893        {        {
1894        case OP_CHAR:        case OP_CHAR:
1895        case OP_CHARNC:        case OP_CHARI:
1896        case OP_EXACT:        case OP_EXACT:
1897          case OP_EXACTI:
1898        case OP_UPTO:        case OP_UPTO:
1899          case OP_UPTOI:
1900        case OP_MINUPTO:        case OP_MINUPTO:
1901          case OP_MINUPTOI:
1902        case OP_POSUPTO:        case OP_POSUPTO:
1903          case OP_POSUPTOI:
1904        case OP_STAR:        case OP_STAR:
1905          case OP_STARI:
1906        case OP_MINSTAR:        case OP_MINSTAR:
1907          case OP_MINSTARI:
1908        case OP_POSSTAR:        case OP_POSSTAR:
1909          case OP_POSSTARI:
1910        case OP_PLUS:        case OP_PLUS:
1911          case OP_PLUSI:
1912        case OP_MINPLUS:        case OP_MINPLUS:
1913          case OP_MINPLUSI:
1914        case OP_POSPLUS:        case OP_POSPLUS:
1915          case OP_POSPLUSI:
1916        case OP_QUERY:        case OP_QUERY:
1917          case OP_QUERYI:
1918        case OP_MINQUERY:        case OP_MINQUERY:
1919          case OP_MINQUERYI:
1920        case OP_POSQUERY:        case OP_POSQUERY:
1921          case OP_POSQUERYI:
1922        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1923        break;        break;
1924        }        }
1925    #else
1926        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1927  #endif  #endif
1928      }      }
1929    }    }
# Line 1516  Arguments: Line 1947  Arguments:
1947    code        points to start of search    code        points to start of search
1948    endcode     points to where to stop    endcode     points to where to stop
1949    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
1950      cd          contains pointers to tables etc.
1951    
1952  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
1953  */  */
1954    
1955  static BOOL  static BOOL
1956  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
1957      compile_data *cd)
1958  {  {
1959  register int c;  register int c;
1960  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], TRUE);
1961       code < endcode;       code < endcode;
1962       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], TRUE))
1963    {    {
1964    const uschar *ccode;    const uschar *ccode;
1965    
# Line 1542  for (code = first_significant_code(code Line 1975  for (code = first_significant_code(code
1975      continue;      continue;
1976      }      }
1977    
1978      /* For a recursion/subroutine call, if its end has been reached, which
1979      implies a backward reference subroutine call, we can scan it. If it's a
1980      forward reference subroutine call, we can't. To detect forward reference
1981      we have to scan up the list that is kept in the workspace. This function is
1982      called only when doing the real compile, not during the pre-compile that
1983      measures the size of the compiled pattern. */
1984    
1985      if (c == OP_RECURSE)
1986        {
1987        const uschar *scode;
1988        BOOL empty_branch;
1989    
1990        /* Test for forward reference */
1991    
1992        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
1993          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
1994    
1995        /* Not a forward reference, test for completed backward reference */
1996    
1997        empty_branch = FALSE;
1998        scode = cd->start_code + GET(code, 1);
1999        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2000    
2001        /* Completed backwards reference */
2002    
2003        do
2004          {
2005          if (could_be_empty_branch(scode, endcode, utf8, cd))
2006            {
2007            empty_branch = TRUE;
2008            break;
2009            }
2010          scode += GET(scode, 1);
2011          }
2012        while (*scode == OP_ALT);
2013    
2014        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2015        continue;
2016        }
2017    
2018    /* Groups with zero repeats can of course be empty; skip them. */    /* Groups with zero repeats can of course be empty; skip them. */
2019    
2020    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2021          c == OP_BRAPOSZERO)
2022      {      {
2023      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
2024      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
# Line 1552  for (code = first_significant_code(code Line 2026  for (code = first_significant_code(code
2026      continue;      continue;
2027      }      }
2028    
2029      /* A nested group that is already marked as "could be empty" can just be
2030      skipped. */
2031    
2032      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2033          c == OP_SCBRA || c == OP_SCBRAPOS)
2034        {
2035        do code += GET(code, 1); while (*code == OP_ALT);
2036        c = *code;
2037        continue;
2038        }
2039    
2040    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
2041    
2042    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_BRA  || c == OP_BRAPOS ||
2043          c == OP_CBRA || c == OP_CBRAPOS ||
2044          c == OP_ONCE || c == OP_COND)
2045      {      {
2046      BOOL empty_branch;      BOOL empty_branch;
2047      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
2048    
2049      /* Scan a closed bracket */      /* If a conditional group has only one branch, there is a second, implied,
2050        empty branch, so just skip over the conditional, because it could be empty.
2051        Otherwise, scan the individual branches of the group. */
2052    
2053      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;  
2054        code += GET(code, 1);        code += GET(code, 1);
2055        else
2056          {
2057          empty_branch = FALSE;
2058          do
2059            {
2060            if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
2061              empty_branch = TRUE;
2062            code += GET(code, 1);
2063            }
2064          while (*code == OP_ALT);
2065          if (!empty_branch) return FALSE;   /* All branches are non-empty */
2066        }        }
2067      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
2068      c = *code;      c = *code;
2069      continue;      continue;
2070      }      }
# Line 1630  for (code = first_significant_code(code Line 2125  for (code = first_significant_code(code
2125      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2126      case OP_WORDCHAR:      case OP_WORDCHAR:
2127      case OP_ANY:      case OP_ANY:
2128        case OP_ALLANY:
2129      case OP_ANYBYTE:      case OP_ANYBYTE:
2130      case OP_CHAR:      case OP_CHAR:
2131      case OP_CHARNC:      case OP_CHARI:
2132      case OP_NOT:      case OP_NOT:
2133        case OP_NOTI:
2134      case OP_PLUS:      case OP_PLUS:
2135      case OP_MINPLUS:      case OP_MINPLUS:
2136      case OP_POSPLUS:      case OP_POSPLUS:
# Line 1673  for (code = first_significant_code(code Line 2170  for (code = first_significant_code(code
2170      case OP_KET:      case OP_KET:
2171      case OP_KETRMAX:      case OP_KETRMAX:
2172      case OP_KETRMIN:      case OP_KETRMIN:
2173        case OP_KETRPOS:
2174      case OP_ALT:      case OP_ALT:
2175      return TRUE;      return TRUE;
2176    
# Line 1681  for (code = first_significant_code(code Line 2179  for (code = first_significant_code(code
2179    
2180  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2181      case OP_STAR:      case OP_STAR:
2182        case OP_STARI:
2183      case OP_MINSTAR:      case OP_MINSTAR:
2184        case OP_MINSTARI:
2185      case OP_POSSTAR:      case OP_POSSTAR:
2186        case OP_POSSTARI:
2187      case OP_QUERY:      case OP_QUERY:
2188        case OP_QUERYI:
2189      case OP_MINQUERY:      case OP_MINQUERY:
2190        case OP_MINQUERYI:
2191      case OP_POSQUERY:      case OP_POSQUERY:
2192        case OP_POSQUERYI:
2193        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2194        break;
2195    
2196      case OP_UPTO:      case OP_UPTO:
2197        case OP_UPTOI:
2198      case OP_MINUPTO:      case OP_MINUPTO:
2199        case OP_MINUPTOI:
2200      case OP_POSUPTO:      case OP_POSUPTO:
2201      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_POSUPTOI:
2202        if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2203      break;      break;
2204  #endif  #endif
2205    
2206        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2207        string. */
2208    
2209        case OP_MARK:
2210        case OP_PRUNE_ARG:
2211        case OP_SKIP_ARG:
2212        code += code[1];
2213        break;
2214    
2215        case OP_THEN_ARG:
2216        code += code[1+LINK_SIZE];
2217        break;
2218    
2219        /* None of the remaining opcodes are required to match a character. */
2220    
2221        default:
2222        break;
2223      }      }
2224    }    }
2225    
# Line 1708  return TRUE; Line 2236  return TRUE;
2236  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
2237  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,
2238  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.
2239    This function is called only during the real compile, not during the
2240    pre-compile.
2241    
2242  Arguments:  Arguments:
2243    code        points to start of the recursion    code        points to start of the recursion
2244    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2245    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2246    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2247      cd          pointers to tables etc
2248    
2249  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2250  */  */
2251    
2252  static BOOL  static BOOL
2253  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2254    BOOL utf8)    BOOL utf8, compile_data *cd)
2255  {  {
2256  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2257    {    {
2258    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2259        return FALSE;
2260    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2261    }    }
2262  return TRUE;  return TRUE;
# Line 1770  int terminator;          /* Don't combin Line 2302  int terminator;          /* Don't combin
2302  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2303  for (++ptr; *ptr != 0; ptr++)  for (++ptr; *ptr != 0; ptr++)
2304    {    {
2305    if (*ptr == '\\' && ptr[1] == ']') ptr++; else    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else
2306      {      {
2307      if (*ptr == ']') return FALSE;      if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2308      if (*ptr == terminator && ptr[1] == ']')      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2309        {        {
2310        *endptr = ptr;        *endptr = ptr;
2311        return TRUE;        return TRUE;
# Line 1824  return -1; Line 2356  return -1;
2356  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2357  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2358  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
2359  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
2360  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
2361  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
2362  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
2363  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2364    OP_END.
2365    
2366  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2367  recursions and subroutine calls. It must also check the list of such references  recursions and subroutine calls. It must also check the list of such references
# Line 1905  auto_callout(uschar *code, const uschar Line 2438  auto_callout(uschar *code, const uschar
2438  {  {
2439  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2440  *code++ = 255;  *code++ = 255;
2441  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2442  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2443  return code + 2*LINK_SIZE;  return code + 2*LINK_SIZE;
2444  }  }
2445    
# Line 1931  Returns:             nothing Line 2464  Returns:             nothing
2464  static void  static void
2465  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)
2466  {  {
2467  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2468  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2469  }  }
2470    
# Line 1963  get_othercase_range(unsigned int *cptr, Line 2496  get_othercase_range(unsigned int *cptr,
2496  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2497    
2498  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2499    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2500    
2501  if (c > d) return FALSE;  if (c > d) return FALSE;
2502    
# Line 1972  next = othercase + 1; Line 2505  next = othercase + 1;
2505    
2506  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2507    {    {
2508    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2509    next++;    next++;
2510    }    }
2511    
# Line 1981  for (++c; c <= d; c++) Line 2514  for (++c; c <= d; c++)
2514    
2515  return TRUE;  return TRUE;
2516  }  }
2517    
2518    
2519    
2520    /*************************************************
2521    *        Check a character and a property        *
2522    *************************************************/
2523    
2524    /* This function is called by check_auto_possessive() when a property item
2525    is adjacent to a fixed character.
2526    
2527    Arguments:
2528      c            the character
2529      ptype        the property type
2530      pdata        the data for the type
2531      negated      TRUE if it's a negated property (\P or \p{^)
2532    
2533    Returns:       TRUE if auto-possessifying is OK
2534    */
2535    
2536    static BOOL
2537    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2538    {
2539    const ucd_record *prop = GET_UCD(c);
2540    switch(ptype)
2541      {
2542      case PT_LAMP:
2543      return (prop->chartype == ucp_Lu ||
2544              prop->chartype == ucp_Ll ||
2545              prop->chartype == ucp_Lt) == negated;
2546    
2547      case PT_GC:
2548      return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;
2549    
2550      case PT_PC:
2551      return (pdata == prop->chartype) == negated;
2552    
2553      case PT_SC:
2554      return (pdata == prop->script) == negated;
2555    
2556      /* These are specials */
2557    
2558      case PT_ALNUM:
2559      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2560              _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;
2561    
2562      case PT_SPACE:    /* Perl space */
2563      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2564              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2565              == negated;
2566    
2567      case PT_PXSPACE:  /* POSIX space */
2568      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2569              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2570              c == CHAR_FF || c == CHAR_CR)
2571              == negated;
2572    
2573      case PT_WORD:
2574      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2575              _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2576              c == CHAR_UNDERSCORE) == negated;
2577      }
2578    return FALSE;
2579    }
2580  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2581    
2582    
# Line 1994  whether the next thing could possibly ma Line 2590  whether the next thing could possibly ma
2590  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
2591    
2592  Arguments:  Arguments:
2593    op_code       the repeated op code    previous      pointer to the repeated opcode
   this          data for this item, depends on the opcode  
2594    utf8          TRUE in UTF-8 mode    utf8          TRUE in UTF-8 mode
   utf8_char     used for utf8 character bytes, NULL if not relevant  
2595    ptr           next character in pattern    ptr           next character in pattern
2596    options       options bits    options       options bits
2597    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 2006  Returns:        TRUE if possessifying is Line 2600  Returns:        TRUE if possessifying is
2600  */  */
2601    
2602  static BOOL  static BOOL
2603  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
2604    const uschar *ptr, int options, compile_data *cd)    int options, compile_data *cd)
2605  {  {
2606  int next;  int c, next;
2607    int op_code = *previous++;
2608    
2609  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
2610    
# Line 2018  if ((options & PCRE_EXTENDED) != 0) Line 2613  if ((options & PCRE_EXTENDED) != 0)
2613    for (;;)    for (;;)
2614      {      {
2615      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2616      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2617        {        {
2618        while (*(++ptr) != 0)        ptr++;
2619          while (*ptr != 0)
2620            {
2621          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2622            ptr++;
2623    #ifdef SUPPORT_UTF8
2624            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2625    #endif
2626            }
2627        }        }
2628      else break;      else break;
2629      }      }
# Line 2030  if ((options & PCRE_EXTENDED) != 0) Line 2632  if ((options & PCRE_EXTENDED) != 0)
2632  /* 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
2633  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2634    
2635  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2636    {    {
2637    int temperrorcode = 0;    int temperrorcode = 0;
2638    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
# Line 2055  if ((options & PCRE_EXTENDED) != 0) Line 2657  if ((options & PCRE_EXTENDED) != 0)
2657    for (;;)    for (;;)
2658      {      {
2659      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2660      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2661        {        {
2662        while (*(++ptr) != 0)        ptr++;
2663          while (*ptr != 0)
2664            {
2665          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2666            ptr++;
2667    #ifdef SUPPORT_UTF8
2668            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2669    #endif
2670            }
2671        }        }
2672      else break;      else break;
2673      }      }
# Line 2066  if ((options & PCRE_EXTENDED) != 0) Line 2675  if ((options & PCRE_EXTENDED) != 0)
2675    
2676  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
2677    
2678  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2679    return FALSE;    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2680        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. */  
   
2681    
2682  /* Handle cases when the next item is a character. */  /* Now compare the next item with the previous opcode. First, handle cases when
2683    the next item is a character. */
2684    
2685  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
2686    {    {
2687    case OP_CHAR:    case OP_CHAR:
2688  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2689    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2690    #else
2691      c = *previous;
2692  #endif  #endif
2693    return item != next;    return c != next;
2694    
2695    /* 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
2696    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
2697    high-valued characters. */    high-valued characters. */
2698    
2699    case OP_CHARNC:    case OP_CHARI:
2700  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2701    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2702    #else
2703      c = *previous;
2704  #endif  #endif
2705    if (item == next) return FALSE;    if (c == next) return FALSE;
2706  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2707    if (utf8)    if (utf8)
2708      {      {
2709      unsigned int othercase;      unsigned int othercase;
2710      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2711  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2712      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
2713  #else  #else
2714      othercase = NOTACHAR;      othercase = NOTACHAR;
2715  #endif  #endif
2716      return (unsigned int)item != othercase;      return (unsigned int)c != othercase;
2717      }      }
2718    else    else
2719  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2720    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2721    
2722    /* 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
2723      opcodes are not used for multi-byte characters, because they are coded using
2724      an XCLASS instead. */
2725    
2726    case OP_NOT:    case OP_NOT:
2727    if (next < 0) return FALSE;  /* Not a character */    return (c = *previous) == next;
2728    if (item == next) return TRUE;  
2729    if ((options & PCRE_CASELESS) == 0) return FALSE;    case OP_NOTI:
2730      if ((c = *previous) == next) return TRUE;
2731  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2732    if (utf8)    if (utf8)
2733      {      {
2734      unsigned int othercase;      unsigned int othercase;
2735      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2736  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2737      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
2738  #else  #else
2739      othercase = NOTACHAR;      othercase = NOTACHAR;
2740  #endif  #endif
2741      return (unsigned int)item == othercase;      return (unsigned int)c == othercase;
2742      }      }
2743    else    else
2744  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2745    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2746    
2747      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2748      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
2749    
2750    case OP_DIGIT:    case OP_DIGIT:
2751    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
# Line 2173  if (next >= 0) switch(op_code) Line 2788  if (next >= 0) switch(op_code)
2788      case 0x202f:      case 0x202f:
2789      case 0x205f:      case 0x205f:
2790      case 0x3000:      case 0x3000:
2791      return op_code != OP_HSPACE;      return op_code == OP_NOT_HSPACE;
2792      default:      default:
2793      return op_code == OP_HSPACE;      return op_code != OP_NOT_HSPACE;
2794      }      }
2795    
2796      case OP_ANYNL:
2797    case OP_VSPACE:    case OP_VSPACE:
2798    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
2799    switch(next)    switch(next)
# Line 2189  if (next >= 0) switch(op_code) Line 2805  if (next >= 0) switch(op_code)
2805      case 0x85:      case 0x85:
2806      case 0x2028:      case 0x2028:
2807      case 0x2029:      case 0x2029:
2808      return op_code != OP_VSPACE;      return op_code == OP_NOT_VSPACE;
2809      default:      default:
2810      return op_code == OP_VSPACE;      return op_code != OP_NOT_VSPACE;
2811      }      }
2812    
2813    #ifdef SUPPORT_UCP
2814      case OP_PROP:
2815      return check_char_prop(next, previous[0], previous[1], FALSE);
2816    
2817      case OP_NOTPROP:
2818      return check_char_prop(next, previous[0], previous[1], TRUE);
2819    #endif
2820    
2821    default:    default:
2822    return FALSE;    return FALSE;
2823    }    }
2824    
2825    
2826  /* 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
2827    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
2828    generated only when PCRE_UCP is *not* set, that is, when only ASCII
2829    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
2830    replaced by OP_PROP codes when PCRE_UCP is set. */
2831    
2832  switch(op_code)  switch(op_code)
2833    {    {
2834    case OP_CHAR:    case OP_CHAR:
2835    case OP_CHARNC:    case OP_CHARI:
2836  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2837    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2838    #else
2839      c = *previous;
2840  #endif  #endif
2841    switch(-next)    switch(-next)
2842      {      {
2843      case ESC_d:      case ESC_d:
2844      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
2845    
2846      case ESC_D:      case ESC_D:
2847      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
2848    
2849      case ESC_s:      case ESC_s:
2850      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
2851    
2852      case ESC_S:      case ESC_S:
2853      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
2854    
2855      case ESC_w:      case ESC_w:
2856      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
2857    
2858      case ESC_W:      case ESC_W:
2859      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
2860    
2861      case ESC_h:      case ESC_h:
2862      case ESC_H:      case ESC_H:
2863      switch(item)      switch(c)
2864        {        {
2865        case 0x09:        case 0x09:
2866        case 0x20:        case 0x20:
# Line 2258  switch(op_code) Line 2888  switch(op_code)
2888    
2889      case ESC_v:      case ESC_v:
2890      case ESC_V:      case ESC_V:
2891      switch(item)      switch(c)
2892        {        {
2893        case 0x0a:        case 0x0a:
2894        case 0x0b:        case 0x0b:
# Line 2272  switch(op_code) Line 2902  switch(op_code)
2902        return -next == ESC_v;        return -next == ESC_v;
2903        }        }
2904    
2905        /* When PCRE_UCP is set, these values get generated for \d etc. Find
2906        their substitutions and process them. The result will always be either
2907        -ESC_p or -ESC_P. Then fall through to process those values. */
2908    
2909    #ifdef SUPPORT_UCP
2910        case ESC_du:
2911        case ESC_DU:
2912        case ESC_wu:
2913        case ESC_WU:
2914        case ESC_su:
2915        case ESC_SU:
2916          {
2917          int temperrorcode = 0;
2918          ptr = substitutes[-next - ESC_DU];
2919          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
2920          if (temperrorcode != 0) return FALSE;
2921          ptr++;    /* For compatibility */
2922          }
2923        /* Fall through */
2924    
2925        case ESC_p:
2926        case ESC_P:
2927          {
2928          int ptype, pdata, errorcodeptr;
2929          BOOL negated;
2930    
2931          ptr--;      /* Make ptr point at the p or P */
2932          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
2933          if (ptype < 0) return FALSE;
2934          ptr++;      /* Point past the final curly ket */
2935    
2936          /* If the property item is optional, we have to give up. (When generated
2937          from \d etc by PCRE_UCP, this test will have been applied much earlier,
2938          to the original \d etc. At this point, ptr will point to a zero byte. */
2939    
2940          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2941            strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2942              return FALSE;
2943    
2944          /* Do the property check. */
2945    
2946          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
2947          }
2948    #endif
2949    
2950      default:      default:
2951      return FALSE;      return FALSE;
2952      }      }
2953    
2954      /* In principle, support for Unicode properties should be integrated here as
2955      well. It means re-organizing the above code so as to get hold of the property
2956      values before switching on the op-code. However, I wonder how many patterns
2957      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
2958      these op-codes are never generated.) */
2959    
2960    case OP_DIGIT:    case OP_DIGIT:
2961    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2962           next == -ESC_h || next == -ESC_v;           next == -ESC_h || next == -ESC_v || next == -ESC_R;
2963    
2964    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
2965    return next == -ESC_d;    return next == -ESC_d;
2966    
2967    case OP_WHITESPACE:    case OP_WHITESPACE:
2968    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
2969    
2970    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
2971    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2972    
2973    case OP_HSPACE:    case OP_HSPACE:
2974    return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
2975             next == -ESC_w || next == -ESC_v || next == -ESC_R;
2976    
2977    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
2978    return next == -ESC_h;    return next == -ESC_h;
2979    
2980    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
2981      case OP_ANYNL:
2982    case OP_VSPACE:    case OP_VSPACE:
2983    return next == -ESC_V || next == -ESC_d || next == -ESC_w;    return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2984    
2985    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
2986    return next == -ESC_v;    return next == -ESC_v || next == -ESC_R;
2987    
2988    case OP_WORDCHAR:    case OP_WORDCHAR:
2989    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
2990             next == -ESC_v || next == -ESC_R;
2991    
2992    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
2993    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2367  BOOL inescq = FALSE; Line 3051  BOOL inescq = FALSE;
3051  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
3052  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
3053  const uschar *tempptr;  const uschar *tempptr;
3054    const uschar *nestptr = NULL;
3055  uschar *previous = NULL;  uschar *previous = NULL;
3056  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
3057  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
# Line 2383  BOOL utf8 = FALSE; Line 3068  BOOL utf8 = FALSE;
3068  uschar *utf8_char = NULL;  uschar *utf8_char = NULL;
3069  #endif  #endif
3070    
3071  #ifdef DEBUG  #ifdef PCRE_DEBUG
3072  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
3073  #endif  #endif
3074    
# Line 2437  for (;; ptr++) Line 3122  for (;; ptr++)
3122    
3123    c = *ptr;    c = *ptr;
3124    
3125      /* If we are at the end of a nested substitution, revert to the outer level
3126      string. Nesting only happens one level deep. */
3127    
3128      if (c == 0 && nestptr != NULL)
3129        {
3130        ptr = nestptr;
3131        nestptr = NULL;
3132        c = *ptr;
3133        }
3134    
3135    /* 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
3136    previous cycle of this loop. */    previous cycle of this loop. */
3137    
3138    if (lengthptr != NULL)    if (lengthptr != NULL)
3139      {      {
3140  #ifdef DEBUG  #ifdef PCRE_DEBUG
3141      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3142  #endif  #endif
3143      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
3144        {        {
3145        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3146        goto FAILED;        goto FAILED;
# Line 2467  for (;; ptr++) Line 3162  for (;; ptr++)
3162        goto FAILED;        goto FAILED;
3163        }        }
3164    
3165      *lengthptr += code - last_code;      *lengthptr += (int)(code - last_code);
3166      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));
3167    
3168      /* 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 2494  for (;; ptr++) Line 3189  for (;; ptr++)
3189    /* 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
3190    reference list. */    reference list. */
3191    
3192    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
3193      {      {
3194      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3195      goto FAILED;      goto FAILED;
# Line 2504  for (;; ptr++) Line 3199  for (;; ptr++)
3199    
3200    if (inescq && c != 0)    if (inescq && c != 0)
3201      {      {
3202      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3203        {        {
3204        inescq = FALSE;        inescq = FALSE;
3205        ptr++;        ptr++;
# Line 2530  for (;; ptr++) Line 3225  for (;; ptr++)
3225    /* 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
3226    a quantifier. */    a quantifier. */
3227    
3228    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
3229      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
3230        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
3231    
3232    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
3233         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2546  for (;; ptr++) Line 3242  for (;; ptr++)
3242    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3243      {      {
3244      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
3245      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
3246        {        {
3247        while (*(++ptr) != 0)        ptr++;
3248          while (*ptr != 0)
3249          {          {
3250          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3251            ptr++;
3252    #ifdef SUPPORT_UTF8
3253            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
3254    #endif
3255          }          }
3256        if (*ptr != 0) continue;        if (*ptr != 0) continue;
3257    
# Line 2571  for (;; ptr++) Line 3272  for (;; ptr++)
3272      {      {
3273      /* ===================================================================*/      /* ===================================================================*/
3274      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
3275      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
3276      case ')':      case CHAR_RIGHT_PARENTHESIS:
3277      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
3278      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
3279      *codeptr = code;      *codeptr = code;
# Line 2584  for (;; ptr++) Line 3285  for (;; ptr++)
3285          *errorcodeptr = ERR20;          *errorcodeptr = ERR20;
3286          goto FAILED;          goto FAILED;
3287          }          }
3288        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += (int)(code - last_code);   /* To include callout length */
3289        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
3290        }        }
3291      return TRUE;      return TRUE;
# Line 2594  for (;; ptr++) Line 3295  for (;; ptr++)
3295      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
3296      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3297    
3298      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
3299        previous = NULL;
3300      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3301        {        {
3302        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3303          *code++ = OP_CIRCM;
3304        }        }
3305      previous = NULL;      else *code++ = OP_CIRC;
     *code++ = OP_CIRC;  
3306      break;      break;
3307    
3308      case '$':      case CHAR_DOLLAR_SIGN:
3309      previous = NULL;      previous = NULL;
3310      *code++ = OP_DOLL;      *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
3311      break;      break;
3312    
3313      /* 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
3314      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
3315    
3316      case '.':      case CHAR_DOT:
3317      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3318      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
3319      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3320      previous = code;      previous = code;
3321      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
3322      break;      break;
3323    
3324    
# Line 2631  for (;; ptr++) Line 3333  for (;; ptr++)
3333      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,
3334      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
3335      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.
     */  
3336    
3337      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
3338        default (Perl) mode, it is treated as a data character. */
3339    
3340        case CHAR_RIGHT_SQUARE_BRACKET:
3341        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3342          {
3343          *errorcodeptr = ERR64;
3344          goto FAILED;
3345          }
3346        goto NORMAL_CHAR;
3347    
3348        case CHAR_LEFT_SQUARE_BRACKET:
3349      previous = code;      previous = code;
3350    
3351      /* 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
3352      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. */
3353    
3354      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3355             ptr[1] == CHAR_EQUALS_SIGN) &&
3356          check_posix_syntax(ptr, &tempptr))          check_posix_syntax(ptr, &tempptr))
3357        {        {
3358        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
3359        goto FAILED;        goto FAILED;
3360        }        }
3361    
# Line 2654  for (;; ptr++) Line 3367  for (;; ptr++)
3367      for (;;)      for (;;)
3368        {        {
3369        c = *(++ptr);        c = *(++ptr);
3370        if (c == '\\')        if (c == CHAR_BACKSLASH)
3371          {          {
3372          if (ptr[1] == 'E') ptr++;          if (ptr[1] == CHAR_E)
3373            else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;            ptr++;
3374              else break;          else if (strncmp((const char *)ptr+1,
3375                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
3376              ptr += 3;
3377            else
3378              break;
3379          }          }
3380        else if (!negate_class && c == '^')        else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
3381          negate_class = TRUE;          negate_class = TRUE;
3382        else break;        else break;
3383        }        }
3384    
3385        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
3386        an initial ']' is taken as a data character -- the code below handles
3387        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3388        [^] must match any character, so generate OP_ALLANY. */
3389    
3390        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3391            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3392          {
3393          *code++ = negate_class? OP_ALLANY : OP_FAIL;
3394          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3395          zerofirstbyte = firstbyte;
3396          break;
3397          }
3398    
3399      /* 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
3400      negation flag at the end, so that support for characters > 255 works      negation flag at the end, so that support for characters > 255 works
3401      correctly (they are all included in the class). */      correctly (they are all included in the class). */
# Line 2688  for (;; ptr++) Line 3419  for (;; ptr++)
3419  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3420      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
3421      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
3422      class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */      class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
3423  #endif  #endif
3424    
3425      /* Process characters until ] is reached. By writing this as a "do" it      /* Process characters until ] is reached. By writing this as a "do" it
# Line 2704  for (;; ptr++) Line 3435  for (;; ptr++)
3435          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3436          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3437          }          }
3438    
3439        /* In the pre-compile phase, accumulate the length of any UTF-8 extra        /* In the pre-compile phase, accumulate the length of any UTF-8 extra
3440        data and reset the pointer. This is so that very large classes that        data and reset the pointer. This is so that very large classes that
3441        contain a zillion UTF-8 characters no longer overwrite the work space        contain a zillion UTF-8 characters no longer overwrite the work space
3442        (which is on the stack). */        (which is on the stack). */
3443    
3444        if (lengthptr != NULL)        if (lengthptr != NULL)
3445          {          {
3446          *lengthptr += class_utf8data - class_utf8data_base;          *lengthptr += class_utf8data - class_utf8data_base;
3447          class_utf8data = class_utf8data_base;          class_utf8data = class_utf8data_base;
3448          }          }
3449    
3450  #endif  #endif
3451    
3452        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
3453    
3454        if (inescq)        if (inescq)
3455          {          {
3456          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3457            {            {
3458            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
3459            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2737  for (;; ptr++) Line 3468  for (;; ptr++)
3468        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3469        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3470    
3471        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3472            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3473            check_posix_syntax(ptr, &tempptr))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3474          {          {
3475          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3476          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3477          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3478          uschar pbits[32];          uschar pbits[32];
3479    
3480          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3481            {            {
3482            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3483            goto FAILED;            goto FAILED;
3484            }            }
3485    
3486          ptr += 2;          ptr += 2;
3487          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3488            {            {
3489            local_negate = TRUE;            local_negate = TRUE;
3490            should_flip_negation = TRUE;  /* Note negative special */            should_flip_negation = TRUE;  /* Note negative special */
3491            ptr++;            ptr++;
3492            }            }
3493    
3494          posix_class = check_posix_name(ptr, tempptr - ptr);          posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3495          if (posix_class < 0)          if (posix_class < 0)
3496            {            {
3497            *errorcodeptr = ERR30;            *errorcodeptr = ERR30;
# Line 2774  for (;; ptr++) Line 3505  for (;; ptr++)
3505          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3506            posix_class = 0;            posix_class = 0;
3507    
3508          /* 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
3509          because we may be adding and subtracting from it, and we don't want to          different escape sequences that use Unicode properties. */
3510          subtract bits that may be in the main map already. At the end we or the  
3511          result into the bit map that is being built. */  #ifdef SUPPORT_UCP
3512            if ((options & PCRE_UCP) != 0)
3513              {
3514              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3515              if (posix_substitutes[pc] != NULL)
3516                {
3517                nestptr = tempptr + 1;
3518                ptr = posix_substitutes[pc] - 1;
3519                continue;
3520                }
3521              }
3522    #endif
3523            /* In the non-UCP case, we build the bit map for the POSIX class in a
3524            chunk of local store because we may be adding and subtracting from it,
3525            and we don't want to subtract bits that may be in the main map already.
3526            At the end we or the result into the bit map that is being built. */
3527    
3528          posix_class *= 3;          posix_class *= 3;
3529    
# Line 2821  for (;; ptr++) Line 3567  for (;; ptr++)
3567    
3568        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3569        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
3570        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
3571        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
3572        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
3573        character in them, so set class_charcount bigger than one. */        are either treated as literal characters (by default), or are faulted if
3574          PCRE_EXTRA is set. */
3575    
3576        if (c == '\\')        if (c == CHAR_BACKSLASH)
3577          {          {
3578          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3579          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3580    
3581          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 */  
3582          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3583            {            {
3584            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3585              {              {
3586              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3587              }              }
# Line 2850  for (;; ptr++) Line 3595  for (;; ptr++)
3595            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
3596            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3597    
3598            /* Save time by not doing this in the pre-compile phase. */            switch (-c)
   
           if (lengthptr == NULL) switch (-c)  
3599              {              {
3600    #ifdef SUPPORT_UCP
3601                case ESC_du:     /* These are the values given for \d etc */
3602                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3603                case ESC_wu:     /* escape sequence with an appropriate \p */
3604                case ESC_WU:     /* or \P to test Unicode properties instead */
3605                case ESC_su:     /* of the default ASCII testing. */
3606                case ESC_SU:
3607                nestptr = ptr;
3608                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3609                class_charcount -= 2;                /* Undo! */
3610                continue;
3611    #endif
3612              case ESC_d:              case ESC_d:
3613              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3614              continue;              continue;
# Line 2872  for (;; ptr++) Line 3627  for (;; ptr++)
3627              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3628              continue;              continue;
3629    
3630                /* Perl 5.004 onwards omits VT from \s, but we must preserve it
3631                if it was previously set by something earlier in the character
3632                class. */
3633    
3634              case ESC_s:              case ESC_s:
3635              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];              classbits[0] |= cbits[cbit_space];
3636              classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= cbits[cbit_space+1] & ~0x08;
3637                for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3638              continue;              continue;
3639    
3640              case ESC_S:              case ESC_S:
# Line 2883  for (;; ptr++) Line 3643  for (;; ptr++)
3643              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3644              continue;              continue;
3645    
3646              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)  
             {  
3647              SETBIT(classbits, 0x09); /* VT */              SETBIT(classbits, 0x09); /* VT */
3648              SETBIT(classbits, 0x20); /* SPACE */              SETBIT(classbits, 0x20); /* SPACE */
3649              SETBIT(classbits, 0xa0); /* NSBP */              SETBIT(classbits, 0xa0); /* NSBP */
# Line 2920  for (;; ptr++) Line 3667  for (;; ptr++)
3667                }                }
3668  #endif  #endif
3669              continue;              continue;
             }  
3670    
3671            if (-c == ESC_H)              case ESC_H:
             {  
3672              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3673                {                {
3674                int x = 0xff;                int x = 0xff;
# Line 2965  for (;; ptr++) Line 3710  for (;; ptr++)
3710                }                }
3711  #endif  #endif
3712              continue;              continue;
             }  
3713    
3714            if (-c == ESC_v)              case ESC_v:
             {  
3715              SETBIT(classbits, 0x0a); /* LF */              SETBIT(classbits, 0x0a); /* LF */
3716              SETBIT(classbits, 0x0b); /* VT */              SETBIT(classbits, 0x0b); /* VT */
3717              SETBIT(classbits, 0x0c); /* FF */              SETBIT(classbits, 0x0c); /* FF */
# Line 2984  for (;; ptr++) Line 3727  for (;; ptr++)
3727                }                }
3728  #endif  #endif
3729              continue;              continue;
             }  
3730    
3731            if (-c == ESC_V)              case ESC_V:
             {  
3732              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3733                {                {
3734                int x = 0xff;                int x = 0xff;
# Line 3017  for (;; ptr++) Line 3758  for (;; ptr++)
3758                }                }
3759  #endif  #endif
3760              continue;              continue;
             }  
   
           /* We need to deal with \P and \p in both phases. */  
3761    
3762  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3763            if (-c == ESC_p || -c == ESC_P)              case ESC_p:
3764              {              case ESC_P:
3765              BOOL negated;                {
3766              int pdata;                BOOL negated;
3767              int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                int pdata;
3768              if (ptype < 0) goto FAILED;                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3769              class_utf8 = TRUE;                if (ptype < 0) goto FAILED;
3770              *class_utf8data++ = ((-c == ESC_p) != negated)?                class_utf8 = TRUE;
3771                XCL_PROP : XCL_NOTPROP;                *class_utf8data++ = ((-c == ESC_p) != negated)?
3772              *class_utf8data++ = ptype;                  XCL_PROP : XCL_NOTPROP;
3773              *class_utf8data++ = pdata;                *class_utf8data++ = ptype;
3774              class_charcount -= 2;   /* Not a < 256 character */                *class_utf8data++ = pdata;
3775              continue;                class_charcount -= 2;   /* Not a < 256 character */
3776              }                continue;
3777                  }
3778  #endif  #endif
3779            /* Unrecognized escapes are faulted if PCRE is running in its              /* Unrecognized escapes are faulted if PCRE is running in its
3780            strict mode. By default, for compatibility with Perl, they are              strict mode. By default, for compatibility with Perl, they are
3781            treated as literals. */              treated as literals. */
3782    
3783            if ((options & PCRE_EXTRA) != 0)              default:
3784              {              if ((options & PCRE_EXTRA) != 0)
3785              *errorcodeptr = ERR7;                {
3786              goto FAILED;                *errorcodeptr = ERR7;
3787                  goto FAILED;
3788                  }
3789                class_charcount -= 2;  /* Undo the default count from above */
3790                c = *ptr;              /* Get the final character and fall through */
3791                break;
3792              }              }
   
           class_charcount -= 2;  /* Undo the default count from above */  
           c = *ptr;              /* Get the final character and fall through */  
3793            }            }
3794    
3795          /* 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 3062  for (;; ptr++) Line 3803  for (;; ptr++)
3803        entirely. The code for handling \Q and \E is messy. */        entirely. The code for handling \Q and \E is messy. */
3804    
3805        CHECK_RANGE:        CHECK_RANGE:
3806        while (ptr[1] == '\\' && ptr[2] == 'E')        while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3807          {          {
3808          inescq = FALSE;          inescq = FALSE;
3809          ptr += 2;          ptr += 2;
# Line 3072  for (;; ptr++) Line 3813  for (;; ptr++)
3813    
3814        /* Remember \r or \n */        /* Remember \r or \n */
3815    
3816        if (c == '\r' || c == '\n') cd->external_flags |= PCRE_HASCRORLF;        if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3817    
3818        /* Check for range */        /* Check for range */
3819    
3820        if (!inescq && ptr[1] == '-')        if (!inescq && ptr[1] == CHAR_MINUS)
3821          {          {
3822          int d;          int d;
3823          ptr += 2;          ptr += 2;
3824          while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3825    
3826          /* 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
3827          mode. */          mode. */
3828    
3829          while (*ptr == '\\' && ptr[1] == 'Q')          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3830            {            {
3831            ptr += 2;            ptr += 2;
3832            if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }            if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3833                { ptr += 2; continue; }
3834            inescq = TRUE;            inescq = TRUE;
3835            break;            break;
3836            }            }
3837    
3838          if (*ptr == 0 || (!inescq && *ptr == ']'))          if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3839            {            {
3840            ptr = oldptr;            ptr = oldptr;
3841            goto LONE_SINGLE_CHARACTER;            goto LONE_SINGLE_CHARACTER;
# Line 3112  for (;; ptr++) Line 3854  for (;; ptr++)
3854          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
3855          in such circumstances. */          in such circumstances. */
3856    
3857          if (!inescq && d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3858            {            {
3859            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3860            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3861    
3862            /* \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 */  
3863    
3864            if (d < 0)            if (d < 0)
3865              {              {
3866              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  
3867                {                {
3868                ptr = oldptr;                ptr = oldptr;
3869                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 3145  for (;; ptr++) Line 3884  for (;; ptr++)
3884    
3885          /* Remember \r or \n */          /* Remember \r or \n */
3886    
3887          if (d == '\r' || d == '\n') cd->external_flags |= PCRE_HASCRORLF;          if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3888    
3889          /* 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
3890          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 3265  for (;; ptr++) Line 4004  for (;; ptr++)
4004          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
4005            {            {
4006            unsigned int othercase;            unsigned int othercase;
4007            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)            if ((othercase = UCD_OTHERCASE(c)) != c)
4008              {              {
4009              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
4010              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 3290  for (;; ptr++) Line 4029  for (;; ptr++)
4029          }          }
4030        }        }
4031    
4032      /* 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.
4033        If we are at the end of an internal nested string, revert to the outer
4034        string. */
4035    
4036        while (((c = *(++ptr)) != 0 ||
4037               (nestptr != NULL &&
4038                 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != 0)) &&
4039               (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
4040    
4041      while ((c = *(++ptr)) != 0 && (c != ']' || inescq));      /* Check for missing terminating ']' */
4042    
4043      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)
4044        {        {
4045        *errorcodeptr = ERR6;        *errorcodeptr = ERR6;
4046        goto FAILED;        goto FAILED;
4047        }        }
4048    
   
 /* 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  
   
   
4049      /* 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
4050      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
4051      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 3326  we set the flag only if there is a liter Line 4053  we set the flag only if there is a liter
4053    
4054      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
4055      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
4056      operate on single-bytes only. This is an historical hangover. Maybe one day      operate on single-bytes characters only. This is an historical hangover.
4057      we can tidy these opcodes to handle multi-byte characters.      Maybe one day we can tidy these opcodes to handle multi-byte characters.
4058    
4059      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
4060      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.
4061      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
4062      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
4063      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
4064      reqbyte, save the previous value for reinstating. */      of reqbyte, save the previous value for reinstating. */
4065    
4066  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4067      if (class_charcount == 1 && !class_utf8 &&      if (class_charcount == 1 && !class_utf8 &&
# Line 3345  we set the flag only if there is a liter Line 4072  we set the flag only if there is a liter
4072        {        {
4073        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
4074    
4075        /* The OP_NOT opcode works on one-byte characters only. */        /* The OP_NOT[I] opcodes work on one-byte characters only. */
4076    
4077        if (negate_class)        if (negate_class)
4078          {          {
4079          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
4080          zerofirstbyte = firstbyte;          zerofirstbyte = firstbyte;
4081          *code++ = OP_NOT;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
4082          *code++ = class_lastchar;          *code++ = class_lastchar;
4083          break;          break;
4084          }          }
# Line 3382  we set the flag only if there is a liter Line 4109  we set the flag only if there is a liter
4109    
4110      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
4111      extended class, with its own opcode, unless there was a negated special      extended class, with its own opcode, unless there was a negated special
4112      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
4113      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
4114      (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
4115      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
4116        actual compiled code. */
4117    
4118  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4119      if (class_utf8 && !should_flip_negation)      if (class_utf8 && (!should_flip_negation || (options & PCRE_UCP) != 0))
4120        {        {
4121        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
4122        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 3414  we set the flag only if there is a liter Line 4142  we set the flag only if there is a liter
4142        }        }
4143  #endif  #endif
4144    
4145      /* 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
4146      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
4147      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
4148      map into the code vector, negating it if necessary. */      (non-UCP) in the class. Then copy the 32-byte map into the code vector,
4149        negating it if necessary. */
4150    
4151      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
4152      if (negate_class)      if (negate_class)
# Line 3437  we set the flag only if there is a liter Line 4166  we set the flag only if there is a liter
4166      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
4167      has been tested above. */      has been tested above. */
4168    
4169      case '{':      case CHAR_LEFT_CURLY_BRACKET:
4170      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
4171      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
4172      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
4173      goto REPEAT;      goto REPEAT;
4174    
4175      case '*':      case CHAR_ASTERISK:
4176      repeat_min = 0;      repeat_min = 0;
4177      repeat_max = -1;      repeat_max = -1;
4178      goto REPEAT;      goto REPEAT;
4179    
4180      case '+':      case CHAR_PLUS:
4181      repeat_min = 1;      repeat_min = 1;
4182      repeat_max = -1;      repeat_max = -1;
4183      goto REPEAT;      goto REPEAT;
4184    
4185      case '?':      case CHAR_QUESTION_MARK:
4186      repeat_min = 0;      repeat_min = 0;
4187      repeat_max = 1;      repeat_max = 1;
4188    
# Line 3488  we set the flag only if there is a liter Line 4217  we set the flag only if there is a liter
4217      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
4218      repeat type to the non-default. */      repeat type to the non-default. */
4219    
4220      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
4221        {        {
4222        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
4223        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
4224        ptr++;        ptr++;
4225        }        }
4226      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
4227        {        {
4228        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
4229        ptr++;        ptr++;
4230        }        }
4231      else repeat_type = greedy_default;      else repeat_type = greedy_default;
4232    
4233        /* If previous was a recursion call, wrap it in atomic brackets so that
4234        previous becomes the atomic group. All recursions were so wrapped in the
4235        past, but it no longer happens for non-repeated recursions. In fact, the
4236        repeated ones could be re-implemented independently so as not to need this,
4237        but for the moment we rely on the code for repeating groups. */
4238    
4239        if (*previous == OP_RECURSE)
4240          {
4241          memmove(previous + 1 + LINK_SIZE, previous, 1 + LINK_SIZE);
4242          *previous = OP_ONCE;
4243          PUT(previous, 1, 2 + 2*LINK_SIZE);
4244          previous[2 + 2*LINK_SIZE] = OP_KET;
4245          PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
4246          code += 2 + 2 * LINK_SIZE;
4247          length_prevgroup = 3 + 3*LINK_SIZE;
4248    
4249          /* When actually compiling, we need to check whether this was a forward
4250          reference, and if so, adjust the offset. */
4251    
4252          if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
4253            {
4254            int offset = GET(cd->hwm, -LINK_SIZE);
4255            if (offset == previous + 1 - cd->start_code)
4256              PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
4257            }
4258          }
4259    
4260        /* Now handle repetition for the different types of item. */
4261    
4262      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
4263      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 3507  we set the flag only if there is a liter Line 4265  we set the flag only if there is a liter
4265      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
4266      instead.  */      instead.  */
4267    
4268      if (*previous == OP_CHAR || *previous == OP_CHARNC)      if (*previous == OP_CHAR || *previous == OP_CHARI)
4269        {        {
4270          op_type = (*previous == OP_CHAR)? 0 : OP_STARI - OP_STAR;
4271    
4272        /* 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
4273        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
4274        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 3541  we set the flag only if there is a liter Line 4301  we set the flag only if there is a liter
4301    
4302        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4303            repeat_max < 0 &&            repeat_max < 0 &&
4304            check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
             options, cd))  
4305          {          {
4306          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4307          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3554  we set the flag only if there is a liter Line 4313  we set the flag only if there is a liter
4313      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
4314      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-
4315      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
4316      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
4317      currently used only for single-byte chars. */      are currently used only for single-byte chars. */
4318    
4319      else if (*previous == OP_NOT)      else if (*previous == OP_NOT || *previous == OP_NOTI)
4320        {        {
4321        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = ((*previous == OP_NOT)? OP_NOTSTAR : OP_NOTSTARI) - OP_STAR;
4322        c = previous[1];        c = previous[1];
4323        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4324            repeat_max < 0 &&            repeat_max < 0 &&
4325            check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4326          {          {
4327          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4328          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3587  we set the flag only if there is a liter Line 4346  we set the flag only if there is a liter
4346    
4347        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4348            repeat_max < 0 &&            repeat_max < 0 &&
4349            check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4350          {          {
4351          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4352          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3609  we set the flag only if there is a liter Line 4368  we set the flag only if there is a liter
4368    
4369        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
4370    
4371          /*--------------------------------------------------------------------*/
4372          /* This code is obsolete from release 8.00; the restriction was finally
4373          removed: */
4374    
4375        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4376        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4377    
4378        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4379          /*--------------------------------------------------------------------*/
4380    
4381        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
4382    
# Line 3751  we set the flag only if there is a liter Line 4515  we set the flag only if there is a liter
4515  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4516               *previous == OP_XCLASS ||               *previous == OP_XCLASS ||
4517  #endif  #endif
4518               *previous == OP_REF)               *previous == OP_REF ||
4519                 *previous == OP_REFI)
4520        {        {
4521        if (repeat_max == 0)        if (repeat_max == 0)
4522          {          {
# Line 3759  we set the flag only if there is a liter Line 4524  we set the flag only if there is a liter
4524          goto END_REPEAT;          goto END_REPEAT;
4525          }          }
4526    
4527          /*--------------------------------------------------------------------*/
4528          /* This code is obsolete from release 8.00; the restriction was finally
4529          removed: */
4530    
4531        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4532        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4533    
4534        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4535          /*--------------------------------------------------------------------*/
4536    
4537        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4538          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 3780  we set the flag only if there is a liter Line 4550  we set the flag only if there is a liter
4550        }        }
4551    
4552      /* 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
4553      cases. */      cases. Note that at this point we can encounter only the "basic" BRA and
4554        KET opcodes, as this is the place where they get converted into the more
4555        special varieties. */
4556    
4557      else if (*previous == OP_BRA  || *previous == OP_CBRA ||      else if (*previous == OP_BRA  || *previous == OP_CBRA ||
4558               *previous == OP_ONCE || *previous == OP_COND)               *previous == OP_ONCE || *previous == OP_COND)
4559        {        {
4560        register int i;        register int i;
4561        int ketoffset = 0;        int len = (int)(code - previous);
       int len = code - previous;  
4562        uschar *bralink = NULL;        uschar *bralink = NULL;
4563          uschar *brazeroptr = NULL;
4564    
4565        /* Repeating a DEFINE group is pointless */        /* Repeating a DEFINE group is pointless */
4566    
4567        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
# Line 3798  we set the flag only if there is a liter Line 4570  we set the flag only if there is a liter
4570          goto FAILED;          goto FAILED;
4571          }          }
4572    
       /* If the maximum repeat count is unlimited, find the end of the bracket  
       by scanning through from the start, and compute the offset back to it  
       from the current code pointer. There may be an OP_OPT setting following  
       the final KET, so we can't find the end just by going back from the code  
       pointer. */  
   
       if (repeat_max == -1)  
         {  
         register uschar *ket = previous;  
         do ket += GET(ket, 1); while (*ket != OP_KET);  
         ketoffset = code - ket;  
         }  
   
4573        /* 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
4574        OP_BRAZERO in front of it, and because the group appears once in the        OP_BRAZERO in front of it, and because the group appears once in the
4575        data, whereas in other cases it appears the minimum number of times. For        data, whereas in other cases it appears the minimum number of times. For
# Line 3820  we set the flag only if there is a liter Line 4579  we set the flag only if there is a liter
4579    
4580        if (repeat_min == 0)        if (repeat_min == 0)
4581          {          {
4582          /* 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
4583          altogether. */          output altogether, like this:
   
         if (repeat_max == 0)  
           {  
           code = previous;  
           goto END_REPEAT;  
           }  
4584    
4585          /* If the maximum is 1 or unlimited, we just have to stick in the          ** if (repeat_max == 0)
4586          BRAZERO and do no more at this point. However, we do need to adjust          **   {
4587          any OP_RECURSE calls inside the group that refer to the group itself or          **   code = previous;
4588          any internal or forward referenced group, because the offset is from          **   goto END_REPEAT;
4589          the start of the whole regex. Temporarily terminate the pattern while          **   }
4590          doing this. */  
4591            However, that fails when a group is referenced as a subroutine from
4592            elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it
4593            so that it is skipped on execution. As we don't have a list of which
4594            groups are referenced, we cannot do this selectively.
4595    
4596            If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4597            and do no more at this point. However, we do need to adjust any
4598            OP_RECURSE calls inside the group that refer to the group itself or any
4599            internal or forward referenced group, because the offset is from the
4600            start of the whole regex. Temporarily terminate the pattern while doing
4601            this. */
4602    
4603          if (repeat_max <= 1)          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
4604            {            {
4605            *code = OP_END;            *code = OP_END;
4606            adjust_recurse(previous, 1, utf8, cd, save_hwm);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
4607            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
4608            code++;            code++;
4609              if (repeat_max == 0)
4610                {
4611                *previous++ = OP_SKIPZERO;
4612                goto END_REPEAT;
4613                }
4614              brazeroptr = previous;    /* Save for possessive optimizing */
4615            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4616            }            }
4617    
# Line 3866  we set the flag only if there is a liter Line 4636  we set the flag only if there is a liter
4636            /* We chain together the bracket offset fields that have to be            /* We chain together the bracket offset fields that have to be
4637            filled in later when the ends of the brackets are reached. */            filled in later when the ends of the brackets are reached. */
4638    
4639            offset = (bralink == NULL)? 0 : previous - bralink;            offset = (bralink == NULL)? 0 : (int)(previous - bralink);
4640            bralink = previous;            bralink = previous;
4641            PUTINC(previous, 0, offset);            PUTINC(previous, 0, offset);
4642            }            }
# Line 3887  we set the flag only if there is a liter Line 4657  we set the flag only if there is a liter
4657            {            {
4658            /* 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
4659            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
4660            potential integer overflow. */            potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
4661              integer type when available, otherwise double. */
4662    
4663            if (lengthptr != NULL)            if (lengthptr != NULL)
4664              {              {
4665              int delta = (repeat_min - 1)*length_prevgroup;              int delta = (repeat_min - 1)*length_prevgroup;
4666              if ((double)(repeat_min - 1)*(double)length_prevgroup >              if ((INT64_OR_DOUBLE)(repeat_min - 1)*
4667                                                              (double)INT_MAX ||                    (INT64_OR_DOUBLE)length_prevgroup >
4668                        (INT64_OR_DOUBLE)INT_MAX ||
4669                  OFLOW_MAX - *lengthptr < delta)                  OFLOW_MAX - *lengthptr < delta)
4670                {                {
4671                *errorcodeptr = ERR20;                *errorcodeptr = ERR20;
# Line 3939  we set the flag only if there is a liter Line 4711  we set the flag only if there is a liter
4711          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
4712          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
4713          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
4714          paranoid checks to avoid integer overflow. */          paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
4715            a 64-bit integer type when available, otherwise double. */
4716    
4717          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4718            {            {
4719            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4720                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4721            if ((double)repeat_max *            if ((INT64_OR_DOUBLE)repeat_max *
4722                  (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)                  (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4723                    > (double)INT_MAX ||                    > (INT64_OR_DOUBLE)INT_MAX ||
4724                OFLOW_MAX - *lengthptr < delta)                OFLOW_MAX - *lengthptr < delta)
4725              {              {
4726              *errorcodeptr = ERR20;              *errorcodeptr = ERR20;
# Line 3972  we set the flag only if there is a liter Line 4745  we set the flag only if there is a liter
4745              {              {
4746              int offset;              int offset;
4747              *code++ = OP_BRA;              *code++ = OP_BRA;
4748              offset = (bralink == NULL)? 0 : code - bralink;              offset = (bralink == NULL)? 0 : (int)(code - bralink);
4749              bralink = code;              bralink = code;
4750              PUTINC(code, 0, offset);              PUTINC(code, 0, offset);
4751              }              }
# Line 3993  we set the flag only if there is a liter Line 4766  we set the flag only if there is a liter
4766          while (bralink != NULL)          while (bralink != NULL)
4767            {            {
4768            int oldlinkoffset;            int oldlinkoffset;
4769            int offset = code - bralink + 1;            int offset = (int)(code - bralink + 1);
4770            uschar *bra = code - offset;            uschar *bra = code - offset;
4771            oldlinkoffset = GET(bra, 1);            oldlinkoffset = GET(bra, 1);
4772            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;
# Line 4003  we set the flag only if there is a liter Line 4776  we set the flag only if there is a liter
4776            }            }
4777          }          }
4778    
4779        /* 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
4780        can't just offset backwards from the current code point, because we        ONCE brackets, that's all we need to do. However, possessively repeated
4781        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
4782        correct offset was computed above.        behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to
4783          deal with possessive ONCEs specially.
4784    
4785          Otherwise, if the quantifier was possessive, we convert the BRA code to
4786          the POS form, and the KET code to KETRPOS. (It turns out to be convenient
4787          at runtime to detect this kind of subpattern at both the start and at the
4788          end.) The use of special opcodes makes it possible to reduce greatly the
4789          stack usage in pcre_exec(). If the group is preceded by OP_BRAZERO,
4790          convert this to OP_BRAPOSZERO. Then cancel the possessive flag so that
4791          the default action below, of wrapping everything inside atomic brackets,
4792          does not happen.
4793    
4794        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
4795        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
4796        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
4797        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
4798        atomic groups at runtime, but in a different way.] */        runtime, but in a different way.] */
4799    
4800        else        else
4801          {          {
4802          uschar *ketcode = code - ketoffset;          uschar *ketcode = code - 1 - LINK_SIZE;
4803          uschar *bracode = ketcode - GET(ketcode, 1);          uschar *bracode = ketcode - GET(ketcode, 1);
4804          *ketcode = OP_KETRMAX + repeat_type;  
4805          if (lengthptr == NULL && *bracode != OP_ONCE)          if (*bracode == OP_ONCE && possessive_quantifier) *bracode = OP_BRA;
4806            if (*bracode == OP_ONCE)
4807              *ketcode = OP_KETRMAX + repeat_type;
4808            else
4809            {            {
4810            uschar *scode = bracode;            if (possessive_quantifier)
4811            do              {
4812                *bracode += 1;                   /* Switch to xxxPOS opcodes */
4813                *ketcode = OP_KETRPOS;
4814                if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
4815                possessive_quantifier = FALSE;
4816                }
4817              else *ketcode = OP_KETRMAX + repeat_type;
4818    
4819              if (lengthptr == NULL)
4820              {              {
4821              if (could_be_empty_branch(scode, ketcode, utf8))              uschar *scode = bracode;
4822                do
4823                {                {
4824                *bracode += OP_SBRA - OP_BRA;                if (could_be_empty_branch(scode, ketcode, utf8, cd))
4825                break;                  {
4826                    *bracode += OP_SBRA - OP_BRA;
4827                    break;
4828                    }
4829                  scode += GET(scode, 1);
4830                }                }
4831              scode += GET(scode, 1);              while (*scode == OP_ALT);
4832              }              }
           while (*scode == OP_ALT);  
4833            }            }
4834          }          }
4835        }        }
4836    
4837        /* If previous is OP_FAIL, it was generated by an empty class [] in
4838        JavaScript mode. The other ways in which OP_FAIL can be generated, that is
4839        by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
4840        error above. We can just ignore the repeat in JS case. */
4841    
4842        else if (*previous == OP_FAIL) goto END_REPEAT;
4843    
4844      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
4845    
4846      else      else
# Line 4045  we set the flag only if there is a liter Line 4850  we set the flag only if there is a liter
4850        }        }
4851    
4852      /* If the character following a repeat is '+', or if certain optimization      /* If the character following a repeat is '+', or if certain optimization
4853      tests above succeeded, possessive_quantifier is TRUE. For some of the      tests above succeeded, possessive_quantifier is TRUE. For some opcodes,
4854      simpler opcodes, there is an special alternative opcode for this. For      there are special alternative opcodes for this case. For anything else, we
4855      anything else, we wrap the entire repeated item inside OP_ONCE brackets.      wrap the entire repeated item inside OP_ONCE brackets. Logically, the '+'
4856      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
4857      but the special opcodes can optimize it a bit. The repeated item starts at      special opcodes can optimize it.
4858      tempcode, not at previous, which might be the first part of a string whose  
4859      (former) last char we repeated.      Possessively repeated subpatterns have already been handled in the code
4860        just above, so possessive_quantifier is always FALSE for them at this
4861        stage.
4862    
4863        Note that the repeated item starts at tempcode, not at previous, which
4864        might be the first part of a string whose (former) last char we repeated.
4865    
4866      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
4867      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 4060  we set the flag only if there is a liter Line 4870  we set the flag only if there is a liter
4870      if (possessive_quantifier)      if (possessive_quantifier)
4871        {        {
4872        int len;        int len;
4873        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||  
4874            *tempcode == OP_NOTEXACT)        if (*tempcode == OP_TYPEEXACT)
4875          tempcode += _pcre_OP_lengths[*tempcode] +          tempcode += _pcre_OP_lengths[*tempcode] +
4876            ((*tempcode == OP_TYPEEXACT &&            ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
4877               (tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP))? 2:0);  
4878        len = code - tempcode;        else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
4879            {
4880            tempcode += _pcre_OP_lengths[*tempcode];
4881    #ifdef SUPPORT_UTF8
4882            if (utf8 && tempcode[-1] >= 0xc0)
4883              tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];
4884    #endif
4885            }
4886    
4887          len = (int)(code - tempcode);
4888        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4889          {          {
4890          case OP_STAR:  *tempcode = OP_POSSTAR; break;          case OP_STAR:  *tempcode = OP_POSSTAR; break;
# Line 4073  we set the flag only if there is a liter Line 4892  we set the flag only if there is a liter
4892          case OP_QUERY: *tempcode = OP_POSQUERY; break;          case OP_QUERY: *tempcode = OP_POSQUERY; break;
4893          case OP_UPTO:  *tempcode = OP_POSUPTO; break;          case OP_UPTO:  *tempcode = OP_POSUPTO; break;
4894    
4895          case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;          case OP_STARI:  *tempcode = OP_POSSTARI; break;
4896          case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;          case OP_PLUSI:  *tempcode = OP_POSPLUSI; break;
4897          case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;          case OP_QUERYI: *tempcode = OP_POSQUERYI; break;
4898          case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;          case OP_UPTOI:  *tempcode = OP_POSUPTOI; break;
4899    
4900          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;
4901          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;
4902          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4903          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4904    
4905            case OP_NOTSTARI:  *tempcode = OP_NOTPOSSTARI; break;
4906            case OP_NOTPLUSI:  *tempcode = OP_NOTPOSPLUSI; break;
4907            case OP_NOTQUERYI: *tempcode = OP_NOTPOSQUERYI; break;
4908            case OP_NOTUPTOI:  *tempcode = OP_NOTPOSUPTOI; break;
4909    
4910            case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;
4911            case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;
4912            case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
4913            case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;
4914    
4915            /* Because we are moving code along, we must ensure that any
4916            pending recursive references are updated. */
4917    
4918          default:          default:
4919            *code = OP_END;
4920            adjust_recurse(tempcode, 1 + LINK_SIZE, utf8, cd, save_hwm);
4921          memmove(tempcode + 1+LINK_SIZE, tempcode, len);          memmove(tempcode + 1+LINK_SIZE, tempcode, len);
4922          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
4923          len += 1 + LINK_SIZE;          len += 1 + LINK_SIZE;
# Line 4110  we set the flag only if there is a liter Line 4944  we set the flag only if there is a liter
4944      lookbehind or option setting or condition or all the other extended      lookbehind or option setting or condition or all the other extended
4945      parenthesis forms.  */      parenthesis forms.  */
4946    
4947      case '(':      case CHAR_LEFT_PARENTHESIS:
4948      newoptions = options;      newoptions = options;
4949      skipbytes = 0;      skipbytes = 0;
4950      bravalue = OP_CBRA;      bravalue = OP_CBRA;
# Line 4119  we set the flag only if there is a liter Line 4953  we set the flag only if there is a liter
4953    
4954      /* First deal with various "verbs" that can be introduced by '*'. */      /* First deal with various "verbs" that can be introduced by '*'. */
4955    
4956      if (*(++ptr) == '*' && (cd->ctypes[ptr[1]] & ctype_letter) != 0)      if (*(++ptr) == CHAR_ASTERISK &&
4957             ((cd->ctypes[ptr[1]] & ctype_letter) != 0 || ptr[1] == ':'))
4958        {        {
4959        int i, namelen;        int i, namelen;
4960          int arglen = 0;
4961        const char *vn = verbnames;        const char *vn = verbnames;
4962        const uschar *name = ++ptr;        const uschar *name = ptr + 1;
4963          const uschar *arg = NULL;
4964        previous = NULL;        previous = NULL;
4965        while ((cd->ctypes[*++ptr] & ctype_letter) != 0);        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
4966        if (*ptr == ':')        namelen = (int)(ptr - name);
4967    
4968          if (*ptr == CHAR_COLON)
4969          {          {
4970          *errorcodeptr = ERR59;   /* Not supported */          arg = ++ptr;
4971          goto FAILED;          while ((cd->ctypes[*ptr] & (ctype_letter|ctype_digit)) != 0
4972              || *ptr == '_') ptr++;
4973            arglen = (int)(ptr - arg);
4974          }          }
4975        if (*ptr != ')')  
4976          if (*ptr != CHAR_RIGHT_PARENTHESIS)
4977          {          {
4978          *errorcodeptr = ERR60;          *errorcodeptr = ERR60;
4979          goto FAILED;          goto FAILED;
4980          }          }
4981        namelen = ptr - name;  
4982          /* Scan the table of verb names */
4983    
4984        for (i = 0; i < verbcount; i++)        for (i = 0; i < verbcount; i++)
4985          {          {
4986          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
4987              strncmp((char *)name, vn, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
4988            {            {
4989            *code = verbs[i].op;            /* Check for open captures before ACCEPT and convert it to
4990            if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;            ASSERT_ACCEPT if in an assertion. */
4991            break;  
4992              if (verbs[i].op == OP_ACCEPT)
4993                {
4994                open_capitem *oc;
4995                if (arglen != 0)
4996                  {
4997                  *errorcodeptr = ERR59;
4998                  goto FAILED;
4999                  }
5000                cd->had_accept = TRUE;
5001                for (oc = cd->open_caps; oc != NULL; oc = oc->next)
5002                  {
5003                  *code++ = OP_CLOSE;
5004                  PUT2INC(code, 0, oc->number);
5005                  }
5006                *code++ = (cd->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT;
5007                }
5008    
5009              /* Handle other cases with/without an argument */
5010    
5011              else if (arglen == 0)
5012                {
5013                if (verbs[i].op < 0)   /* Argument is mandatory */
5014                  {
5015                  *errorcodeptr = ERR66;
5016                  goto FAILED;
5017                  }
5018                *code = verbs[i].op;
5019                if (*code++ == OP_THEN)
5020                  {
5021                  PUT(code, 0, code - bcptr->current_branch - 1);
5022                  code += LINK_SIZE;
5023                  }
5024                }
5025    
5026              else
5027                {
5028                if (verbs[i].op_arg < 0)   /* Argument is forbidden */
5029                  {
5030                  *errorcodeptr = ERR59;
5031                  goto FAILED;
5032                  }
5033                *code = verbs[i].op_arg;
5034                if (*code++ == OP_THEN_ARG)
5035                  {
5036                  PUT(code, 0, code - bcptr->current_branch - 1);
5037                  code += LINK_SIZE;
5038                  }
5039                *code++ = arglen;
5040                memcpy(code, arg, arglen);
5041                code += arglen;
5042                *code++ = 0;
5043                }
5044    
5045              break;  /* Found verb, exit loop */
5046            }            }
5047    
5048          vn += verbs[i].len + 1;          vn += verbs[i].len + 1;
5049          }          }
5050        if (i < verbcount) continue;  
5051        *errorcodeptr = ERR60;        if (i < verbcount) continue;    /* Successfully handled a verb */
5052          *errorcodeptr = ERR60;          /* Verb not recognized */
5053        goto FAILED;        goto FAILED;
5054        }        }
5055    
5056      /* Deal with the extended parentheses; all are introduced by '?', and the      /* Deal with the extended parentheses; all are introduced by '?', and the
5057      appearance of any of them means that this is not a capturing group. */      appearance of any of them means that this is not a capturing group. */
5058    
5059      else if (*ptr == '?')      else if (*ptr == CHAR_QUESTION_MARK)
5060        {        {
5061        int i, set, unset, namelen;        int i, set, unset, namelen;
5062        int *optset;        int *optset;
# Line 4165  we set the flag only if there is a liter Line 5065  we set the flag only if there is a liter
5065    
5066        switch (*(++ptr))        switch (*(++ptr))
5067          {          {
5068          case '#':                 /* Comment; skip to ket */          case CHAR_NUMBER_SIGN:                 /* Comment; skip to ket */
5069          ptr++;          ptr++;
5070          while (*ptr != 0 && *ptr != ')') ptr++;          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
5071          if (*ptr == 0)          if (*ptr == 0)
5072            {            {
5073            *errorcodeptr = ERR18;            *errorcodeptr = ERR18;
# Line 4177  we set the flag only if there is a liter Line 5077  we set the flag only if there is a liter
5077    
5078    
5079          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5080          case '|':                 /* Reset capture count for each branch */          case CHAR_VERTICAL_LINE:  /* Reset capture count for each branch */
5081          reset_bracount = TRUE;          reset_bracount = TRUE;
5082          /* Fall through */          /* Fall through */
5083    
5084          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5085          case ':':                 /* Non-capturing bracket */          case CHAR_COLON:          /* Non-capturing bracket */
5086          bravalue = OP_BRA;          bravalue = OP_BRA;
5087          ptr++;          ptr++;
5088          break;          break;
5089    
5090    
5091          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5092          case '(':          case CHAR_LEFT_PARENTHESIS:
5093          bravalue = OP_COND;       /* Conditional group */          bravalue = OP_COND;       /* Conditional group */
5094    
5095          /* A condition can be an assertion, a number (referring to a numbered          /* A condition can be an assertion, a number (referring to a numbered
# Line 4209  we set the flag only if there is a liter Line 5109  we set the flag only if there is a liter
5109          the switch. This will take control down to where bracketed groups,          the switch. This will take control down to where bracketed groups,
5110          including assertions, are processed. */          including assertions, are processed. */
5111    
5112          if (ptr[1] == '?' && (ptr[2] == '=' || ptr[2] == '!' || ptr[2] == '<'))          if (ptr[1] == CHAR_QUESTION_MARK && (ptr[2] == CHAR_EQUALS_SIGN ||
5113                ptr[2] == CHAR_EXCLAMATION_MARK || ptr[2] == CHAR_LESS_THAN_SIGN))
5114            break;            break;
5115    
5116          /* Most other conditions use OP_CREF (a couple change to OP_RREF          /* Most other conditions use OP_CREF (a couple change to OP_RREF
# Line 4221  we set the flag only if there is a liter Line 5122  we set the flag only if there is a liter
5122    
5123          /* Check for a test for recursion in a named group. */          /* Check for a test for recursion in a named group. */
5124    
5125          if (ptr[1] == 'R' && ptr[2] == '&')          if (ptr[1] == CHAR_R && ptr[2] == CHAR_AMPERSAND)
5126            {            {
5127            terminator = -1;            terminator = -1;
5128            ptr += 2;            ptr += 2;
# Line 4231  we set the flag only if there is a liter Line 5132  we set the flag only if there is a liter
5132          /* Check for a test for a named group's having been set, using the Perl          /* Check for a test for a named group's having been set, using the Perl
5133          syntax (?(<name>) or (?('name') */          syntax (?(<name>) or (?('name') */
5134    
5135          else if (ptr[1] == '<')          else if (ptr[1] == CHAR_LESS_THAN_SIGN)
5136            {            {
5137            terminator = '>';            terminator = CHAR_GREATER_THAN_SIGN;
5138            ptr++;            ptr++;
5139            }            }
5140          else if (ptr[1] == '\'')          else if (ptr[1] == CHAR_APOSTROPHE)
5141            {            {
5142            terminator = '\'';            terminator = CHAR_APOSTROPHE;
5143            ptr++;            ptr++;
5144            }            }
5145          else          else
5146            {            {
5147            terminator = 0;            terminator = 0;
5148            if (ptr[1] == '-' || ptr[1] == '+') refsign = *(++ptr);            if (ptr[1] == CHAR_MINUS || ptr[1] == CHAR_PLUS) refsign = *(++ptr);
5149            }            }
5150    
5151          /* We now expect to read a name; any thing else is an error */          /* We now expect to read a name; any thing else is an error */
# Line 4264  we set the flag only if there is a liter Line 5165  we set the flag only if there is a liter
5165            {            {
5166            if (recno >= 0)            if (recno >= 0)
5167              recno = ((digitab[*ptr] & ctype_digit) != 0)?              recno = ((digitab[*ptr] & ctype_digit) != 0)?
5168                recno * 10 + *ptr - '0' : -1;                recno * 10 + *ptr - CHAR_0 : -1;
5169            ptr++;            ptr++;
5170            }            }
5171          namelen = ptr - name;          namelen = (int)(ptr - name);
5172    
5173          if ((terminator > 0 && *ptr++ != terminator) || *ptr++ != ')')          if ((terminator > 0 && *ptr++ != terminator) ||
5174                *ptr++ != CHAR_RIGHT_PARENTHESIS)
5175            {            {
5176            ptr--;      /* Error offset */            ptr--;      /* Error offset */
5177            *errorcodeptr = ERR26;            *errorcodeptr = ERR26;
# Line 4291  we set the flag only if there is a liter Line 5193  we set the flag only if there is a liter
5193              *errorcodeptr = ERR58;              *errorcodeptr = ERR58;
5194              goto FAILED;              goto FAILED;
5195              }              }
5196            recno = (refsign == '-')?            recno = (refsign == CHAR_MINUS)?
5197              cd->bracount - recno + 1 : recno +cd->bracount;              cd->bracount - recno + 1 : recno +cd->bracount;
5198            if (recno <= 0 || recno > cd->final_bracount)            if (recno <= 0 || recno > cd->final_bracount)
5199              {              {
# Line 4303  we set the flag only if there is a liter Line 5205  we set the flag only if there is a liter
5205            }            }
5206    
5207          /* Otherwise (did not start with "+" or "-"), start by looking for the          /* Otherwise (did not start with "+" or "-"), start by looking for the
5208          name. */          name. If we find a name, add one to the opcode to change OP_CREF or
5209            OP_RREF into OP_NCREF or OP_NRREF. These behave exactly the same,
5210            except they record that the reference was originally to a name. The
5211            information is used to check duplicate names. */
5212    
5213          slot = cd->name_table;          slot = cd->name_table;
5214          for (i = 0; i < cd->names_found; i++)          for (i = 0; i < cd->names_found; i++)
# Line 4318  we set the flag only if there is a liter Line 5223  we set the flag only if there is a liter
5223            {            {
5224            recno = GET2(slot, 0);            recno = GET2(slot, 0);
5225            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
5226              code[1+LINK_SIZE]++;
5227            }            }
5228    
5229          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
5230    
5231          else if ((i = find_parens(ptr, cd->bracount, name, namelen,          else if ((i = find_parens(cd, name, namelen,
5232                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0, utf8)) > 0)
5233            {            {
5234            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
5235              code[1+LINK_SIZE]++;
5236            }            }
5237    
5238          /* If terminator == 0 it means that the name followed directly after          /* If terminator == 0 it means that the name followed directly after
# Line 4343  we set the flag only if there is a liter Line 5250  we set the flag only if there is a liter
5250          /* Check for (?(R) for recursion. Allow digits after R to specify a          /* Check for (?(R) for recursion. Allow digits after R to specify a
5251          specific group number. */          specific group number. */
5252    
5253          else if (*name == 'R')          else if (*name == CHAR_R)
5254            {            {
5255            recno = 0;            recno = 0;
5256            for (i = 1; i < namelen; i++)            for (i = 1; i < namelen; i++)
# Line 4353  we set the flag only if there is a liter Line 5260  we set the flag only if there is a liter
5260                *errorcodeptr = ERR15;                *errorcodeptr = ERR15;
5261                goto FAILED;                goto FAILED;
5262                }                }
5263              recno = recno * 10 + name[i] - '0';              recno = recno * 10 + name[i] - CHAR_0;
5264              }              }
5265            if (recno == 0) recno = RREF_ANY;            if (recno == 0) recno = RREF_ANY;
5266            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */
# Line 4363  we set the flag only if there is a liter Line 5270  we set the flag only if there is a liter
5270          /* Similarly, check for the (?(DEFINE) "condition", which is always          /* Similarly, check for the (?(DEFINE) "condition", which is always
5271          false. */          false. */
5272    
5273          else if (namelen == 6 && strncmp((char *)name, "DEFINE", 6) == 0)          else if (namelen == 6 && strncmp((char *)name, STRING_DEFINE, 6) == 0)
5274            {            {
5275            code[1+LINK_SIZE] = OP_DEF;            code[1+LINK_SIZE] = OP_DEF;
5276            skipbytes = 1;            skipbytes = 1;
# Line 4388  we set the flag only if there is a liter Line 5295  we set the flag only if there is a liter
5295    
5296    
5297          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5298          case '=':                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
5299          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
5300            cd->assert_depth += 1;
5301          ptr++;          ptr++;
5302          break;          break;
5303    
5304    
5305          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5306          case '!':                 /* Negative lookahead */          case CHAR_EXCLAMATION_MARK:            /* Negative lookahead */
5307          ptr++;          ptr++;
5308          if (*ptr == ')')          /* Optimize (?!) */          if (*ptr == CHAR_RIGHT_PARENTHESIS)    /* Optimize (?!) */
5309            {            {
5310            *code++ = OP_FAIL;            *code++ = OP_FAIL;
5311            previous = NULL;            previous = NULL;
5312            continue;            continue;
5313            }            }
5314          bravalue = OP_ASSERT_NOT;          bravalue = OP_ASSERT_NOT;
5315            cd->assert_depth += 1;
5316          break;          break;
5317    
5318    
5319          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5320          case '<':                 /* Lookbehind or named define */          case CHAR_LESS_THAN_SIGN:              /* Lookbehind or named define */
5321          switch (ptr[1])          switch (ptr[1])
5322            {            {
5323            case '=':               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
5324            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
5325              cd->assert_depth += 1;
5326            ptr += 2;            ptr += 2;
5327            break;            break;
5328    
5329            case '!':               /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
5330            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
5331              cd->assert_depth += 1;
5332            ptr += 2;            ptr += 2;
5333            break;            break;
5334    
# Line 4431  we set the flag only if there is a liter Line 5342  we set the flag only if there is a liter
5342    
5343    
5344          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5345          case '>':                 /* One-time brackets */          case CHAR_GREATER_THAN_SIGN:           /* One-time brackets */
5346          bravalue = OP_ONCE;          bravalue = OP_ONCE;
5347          ptr++;          ptr++;
5348          break;          break;
5349    
5350    
5351          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5352          case 'C':                 /* Callout - may be followed by digits; */          case CHAR_C:                 /* Callout - may be followed by digits; */
5353          previous_callout = code;  /* Save for later completion */          previous_callout = code;  /* Save for later completion */
5354          after_manual_callout = 1; /* Skip one item before completing */          after_manual_callout = 1; /* Skip one item before completing */
5355          *code++ = OP_CALLOUT;          *code++ = OP_CALLOUT;
5356            {            {
5357            int n = 0;            int n = 0;
5358            while ((digitab[*(++ptr)] & ctype_digit) != 0)            while ((digitab[*(++ptr)] & ctype_digit) != 0)
5359              n = n * 10 + *ptr - '0';              n = n * 10 + *ptr - CHAR_0;
5360            if (*ptr != ')')            if (*ptr != CHAR_RIGHT_PARENTHESIS)
5361              {              {
5362              *errorcodeptr = ERR39;              *errorcodeptr = ERR39;
5363              goto FAILED;              goto FAILED;
# Line 4457  we set the flag only if there is a liter Line 5368  we set the flag only if there is a liter
5368              goto FAILED;              goto FAILED;
5369              }              }
5370            *code++ = n;            *code++ = n;
5371            PUT(code, 0, ptr - cd->start_pattern + 1);  /* Pattern offset */            PUT(code, 0, (int)(ptr - cd->start_pattern + 1)); /* Pattern offset */
5372            PUT(code, LINK_SIZE, 0);                    /* Default length */            PUT(code, LINK_SIZE, 0);                          /* Default length */
5373            code += 2 * LINK_SIZE;            code += 2 * LINK_SIZE;
5374            }            }
5375          previous = NULL;          previous = NULL;
# Line 4466  we set the flag only if there is a liter Line 5377  we set the flag only if there is a liter
5377    
5378    
5379          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5380          case 'P':                 /* Python-style named subpattern handling */          case CHAR_P:              /* Python-style named subpattern handling */
5381          if (*(++ptr) == '=' || *ptr == '>')  /* Reference or recursion */          if (*(++ptr) == CHAR_EQUALS_SIGN ||
5382                *ptr == CHAR_GREATER_THAN_SIGN)  /* Reference or recursion */
5383            {            {
5384            is_recurse = *ptr == '>';            is_recurse = *ptr == CHAR_GREATER_THAN_SIGN;
5385            terminator = ')';            terminator = CHAR_RIGHT_PARENTHESIS;
5386            goto NAMED_REF_OR_RECURSE;            goto NAMED_REF_OR_RECURSE;
5387            }            }
5388          else if (*ptr != '<')    /* Test for Python-style definition */          else if (*ptr != CHAR_LESS_THAN_SIGN)  /* Test for Python-style defn */
5389            {            {
5390            *errorcodeptr = ERR41;            *errorcodeptr = ERR41;
5391            goto FAILED;            goto FAILED;
# Line 4483  we set the flag only if there is a liter Line 5395  we set the flag only if there is a liter
5395    
5396          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5397          DEFINE_NAME:    /* Come here from (?< handling */          DEFINE_NAME:    /* Come here from (?< handling */
5398          case '\'':          case CHAR_APOSTROPHE:
5399            {            {
5400            terminator = (*ptr == '<')? '>' : '\'';            terminator = (*ptr == CHAR_LESS_THAN_SIGN)?
5401                CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
5402            name = ++ptr;            name = ++ptr;
5403    
5404            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
5405            namelen = ptr - name;            namelen = (int)(ptr - name);
5406    
5407            /* In the pre-compile phase, just do a syntax check. */            /* In the pre-compile phase, just do a syntax check. */
5408    
# Line 4516  we set the flag only if there is a liter Line 5429  we set the flag only if there is a liter
5429                }                }
5430              }              }
5431    
5432            /* In the real compile, create the entry in the table */            /* In the real compile, create the entry in the table, maintaining
5433              alphabetical order. Duplicate names for different numbers are
5434              permitted only if PCRE_DUPNAMES is set. Duplicate names for the same
5435              number are always OK. (An existing number can be re-used if (?|
5436              appears in the pattern.) In either event, a duplicate name results in
5437              a duplicate entry in the table, even if the number is the same. This
5438              is because the number of names, and hence the table size, is computed
5439              in the pre-compile, and it affects various numbers and pointers which
5440              would all have to be modified, and the compiled code moved down, if
5441              duplicates with the same number were omitted from the table. This
5442              doesn't seem worth the hassle. However, *different* names for the
5443              same number are not permitted. */
5444    
5445            else            else
5446              {              {
5447                BOOL dupname = FALSE;
5448              slot = cd->name_table;              slot = cd->name_table;
5449    
5450              for (i = 0; i < cd->names_found; i++)              for (i = 0; i < cd->names_found; i++)
5451                {                {
5452                int crc = memcmp(name, slot+2, namelen);                int crc = memcmp(name, slot+2, namelen);
# Line 4528  we set the flag only if there is a liter Line 5454  we set the flag only if there is a liter
5454                  {                  {
5455                  if (slot[2+namelen] == 0)                  if (slot[2+namelen] == 0)
5456                    {                    {