/[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 305 by ph10, Sun Jan 20 20:07:32 2008 UTC revision 835 by ph10, Wed Dec 28 16:10:09 2011 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2008 University of Cambridge             Copyright (c) 1997-2011 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 53  supporting internal functions that are n Line 53  supporting internal functions that are n
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56  /* When DEBUG is defined, we need the pcre_printint() function, which is also  /* When PCRE_DEBUG is defined, we need the pcre_printint() function, which is
57  used by pcretest. DEBUG is not defined when building a production library. */  also used by pcretest. PCRE_DEBUG is not defined when building a production
58    library. */
59    
60  #ifdef DEBUG  #ifdef PCRE_DEBUG
61  #include "pcre_printint.src"  #include "pcre_printint.src"
62  #endif  #endif
63    
# Line 91  is 4 there is plenty of room. */ Line 92  is 4 there is plenty of room. */
92    
93  #define COMPILE_WORK_SIZE (4096)  #define COMPILE_WORK_SIZE (4096)
94    
95    /* The overrun tests check for a slightly smaller size so that they detect the
96    overrun before it actually does run off the end of the data block. */
97    
98    #define WORK_SIZE_CHECK (COMPILE_WORK_SIZE - 100)
99    
100    
101  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
102  are simple data values; negative values are for special things like \d and so  are simple data values; negative values are for special things like \d and so
103  on. Zero means further processing is needed (for things like \x), or the escape  on. Zero means further processing is needed (for things like \x), or the escape
104  is invalid. */  is invalid. */
105    
106  #ifndef EBCDIC  /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC
107    
108    /* This is the "normal" table for ASCII systems or for EBCDIC systems running
109    in UTF-8 mode. */
110    
111  static const short int escapes[] = {  static const short int escapes[] = {
112       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,                       0,
113       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,                       0,
114     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */       0,                       0,
115  -ESC_H,      0,      0, -ESC_K,      0,      0,      0,      0,   /* H - O */       0,                       0,
116  -ESC_P, -ESC_Q, -ESC_R, -ESC_S,      0,      0, -ESC_V, -ESC_W,   /* P - W */       0,                       0,
117  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */       CHAR_COLON,              CHAR_SEMICOLON,
118     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
119  -ESC_h,      0,      0, -ESC_k,      0,      0,  ESC_n,      0,   /* h - o */       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
120  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0, -ESC_v, -ESC_w,   /* p - w */       CHAR_COMMERCIAL_AT,      -ESC_A,
121       0,      0, -ESC_z                                            /* x - z */       -ESC_B,                  -ESC_C,
122         -ESC_D,                  -ESC_E,
123         0,                       -ESC_G,
124         -ESC_H,                  0,
125         0,                       -ESC_K,
126         0,                       0,
127         -ESC_N,                  0,
128         -ESC_P,                  -ESC_Q,
129         -ESC_R,                  -ESC_S,
130         0,                       0,
131         -ESC_V,                  -ESC_W,
132         -ESC_X,                  0,
133         -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,
134         CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,
135         CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,
136         CHAR_GRAVE_ACCENT,       7,
137         -ESC_b,                  0,
138         -ESC_d,                  ESC_e,
139         ESC_f,                   0,
140         -ESC_h,                  0,
141         0,                       -ESC_k,
142         0,                       0,
143         ESC_n,                   0,
144         -ESC_p,                  0,
145         ESC_r,                   -ESC_s,
146         ESC_tee,                 0,
147         -ESC_v,                  -ESC_w,
148         0,                       0,
149         -ESC_z
150  };  };
151    
152  #else           /* This is the "abnormal" table for EBCDIC systems */  #else
153    
154    /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
155    
156  static const short int escapes[] = {  static const short int escapes[] = {
157  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
158  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 130  static const short int escapes[] = { Line 171  static const short int escapes[] = {
171  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
172  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,
173  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
174  /*  D0 */   '}',     0, -ESC_K,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,-ESC_N,      0, -ESC_P,
175  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
176  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
177  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
# Line 142  static const short int escapes[] = { Line 183  static const short int escapes[] = {
183    
184  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
185  searched linearly. Put all the names into a single string, in order to reduce  searched linearly. Put all the names into a single string, in order to reduce
186  the number of relocations when a shared library is dynamically linked. */  the number of relocations when a shared library is dynamically linked. The
187    string is built from string macros so that it works in UTF-8 mode on EBCDIC
188    platforms. */
189    
190  typedef struct verbitem {  typedef struct verbitem {
191    int   len;    int   len;                 /* Length of verb name */
192    int   op;    int   op;                  /* Op when no arg, or -1 if arg mandatory */
193      int   op_arg;              /* Op when arg present, or -1 if not allowed */
194  } verbitem;  } verbitem;
195    
196  static const char verbnames[] =  static const char verbnames[] =
197    "ACCEPT\0"    "\0"                       /* Empty name is a shorthand for MARK */
198    "COMMIT\0"    STRING_MARK0
199    "F\0"    STRING_ACCEPT0
200    "FAIL\0"    STRING_COMMIT0
201    "PRUNE\0"    STRING_F0
202    "SKIP\0"    STRING_FAIL0
203    "THEN";    STRING_PRUNE0
204      STRING_SKIP0
205  static verbitem verbs[] = {    STRING_THEN;
206    { 6, OP_ACCEPT },  
207    { 6, OP_COMMIT },  static const verbitem verbs[] = {
208    { 1, OP_FAIL },    { 0, -1,        OP_MARK },
209    { 4, OP_FAIL },    { 4, -1,        OP_MARK },
210    { 5, OP_PRUNE },    { 6, OP_ACCEPT, -1 },
211    { 4, OP_SKIP  },    { 6, OP_COMMIT, -1 },
212    { 4, OP_THEN  }    { 1, OP_FAIL,   -1 },
213      { 4, OP_FAIL,   -1 },
214      { 5, OP_PRUNE,  OP_PRUNE_ARG },
215      { 4, OP_SKIP,   OP_SKIP_ARG  },
216      { 4, OP_THEN,   OP_THEN_ARG  }
217  };  };
218    
219  static int verbcount = sizeof(verbs)/sizeof(verbitem);  static const int verbcount = sizeof(verbs)/sizeof(verbitem);
220    
221    
222  /* Tables of names of POSIX character classes and their lengths. The names are  /* Tables of names of POSIX character classes and their lengths. The names are
# Line 178  length entry. The first three must be al Line 226  length entry. The first three must be al
226  for handling case independence. */  for handling case independence. */
227    
228  static const char posix_names[] =  static const char posix_names[] =
229    "alpha\0"  "lower\0"  "upper\0"  "alnum\0"  "ascii\0"  "blank\0"    STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
230    "cntrl\0"  "digit\0"  "graph\0"  "print\0"  "punct\0"  "space\0"    STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
231    "word\0"   "xdigit";    STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
232      STRING_word0  STRING_xdigit;
233    
234  static const uschar posix_name_lengths[] = {  static const uschar posix_name_lengths[] = {
235    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
# Line 212  static const int posix_class_maps[] = { Line 261  static const int posix_class_maps[] = {
261    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
262  };  };
263    
264    /* Table of substitutes for \d etc when PCRE_UCP is set. The POSIX class
265    substitutes must be in the order of the names, defined above, and there are
266    both positive and negative cases. NULL means no substitute. */
267    
268    #ifdef SUPPORT_UCP
269    static const uschar *substitutes[] = {
270      (uschar *)"\\P{Nd}",    /* \D */
271      (uschar *)"\\p{Nd}",    /* \d */
272      (uschar *)"\\P{Xsp}",   /* \S */       /* NOTE: Xsp is Perl space */
273      (uschar *)"\\p{Xsp}",   /* \s */
274      (uschar *)"\\P{Xwd}",   /* \W */
275      (uschar *)"\\p{Xwd}"    /* \w */
276    };
277    
278    static const uschar *posix_substitutes[] = {
279      (uschar *)"\\p{L}",     /* alpha */
280      (uschar *)"\\p{Ll}",    /* lower */
281      (uschar *)"\\p{Lu}",    /* upper */
282      (uschar *)"\\p{Xan}",   /* alnum */
283      NULL,                   /* ascii */
284      (uschar *)"\\h",        /* blank */
285      NULL,                   /* cntrl */
286      (uschar *)"\\p{Nd}",    /* digit */
287      NULL,                   /* graph */
288      NULL,                   /* print */
289      NULL,                   /* punct */
290      (uschar *)"\\p{Xps}",   /* space */    /* NOTE: Xps is POSIX space */
291      (uschar *)"\\p{Xwd}",   /* word */
292      NULL,                   /* xdigit */
293      /* Negated cases */
294      (uschar *)"\\P{L}",     /* ^alpha */
295      (uschar *)"\\P{Ll}",    /* ^lower */
296      (uschar *)"\\P{Lu}",    /* ^upper */
297      (uschar *)"\\P{Xan}",   /* ^alnum */
298      NULL,                   /* ^ascii */
299      (uschar *)"\\H",        /* ^blank */
300      NULL,                   /* ^cntrl */
301      (uschar *)"\\P{Nd}",    /* ^digit */
302      NULL,                   /* ^graph */
303      NULL,                   /* ^print */
304      NULL,                   /* ^punct */
305      (uschar *)"\\P{Xps}",   /* ^space */   /* NOTE: Xps is POSIX space */
306      (uschar *)"\\P{Xwd}",   /* ^word */
307      NULL                    /* ^xdigit */
308    };
309    #define POSIX_SUBSIZE (sizeof(posix_substitutes)/sizeof(uschar *))
310    #endif
311    
312  #define STRING(a)  # a  #define STRING(a)  # a
313  #define XSTRING(s) STRING(s)  #define XSTRING(s) STRING(s)
# Line 224  the number of relocations needed when a Line 320  the number of relocations needed when a
320  it is now one long string. We cannot use a table of offsets, because the  it is now one long string. We cannot use a table of offsets, because the
321  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
322  simply count through to the one we want - this isn't a performance issue  simply count through to the one we want - this isn't a performance issue
323  because these strings are used only when there is a compilation error. */  because these strings are used only when there is a compilation error.
324    
325    Each substring ends with \0 to insert a null character. This includes the final
326    substring, so that the whole string ends with \0\0, which can be detected when
327    counting through. */
328    
329  static const char error_texts[] =  static const char error_texts[] =
330    "no error\0"    "no error\0"
# Line 271  static const char error_texts[] = Line 371  static const char error_texts[] =
371    /* 35 */    /* 35 */
372    "invalid condition (?(0)\0"    "invalid condition (?(0)\0"
373    "\\C not allowed in lookbehind assertion\0"    "\\C not allowed in lookbehind assertion\0"
374    "PCRE does not support \\L, \\l, \\N, \\U, or \\u\0"    "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
375    "number after (?C is > 255\0"    "number after (?C is > 255\0"
376    "closing ) for (?C expected\0"    "closing ) for (?C expected\0"
377    /* 40 */    /* 40 */
# Line 293  static const char error_texts[] = Line 393  static const char error_texts[] =
393    "internal error: previously-checked referenced subpattern not found\0"    "internal error: previously-checked referenced subpattern not found\0"
394    "DEFINE group contains more than one branch\0"    "DEFINE group contains more than one branch\0"
395    /* 55 */    /* 55 */
396    "repeating a DEFINE group is not allowed\0"    "repeating a DEFINE group is not allowed\0"  /** DEAD **/
397    "inconsistent NEWLINE options\0"    "inconsistent NEWLINE options\0"
398    "\\g is not followed by a braced name or an optionally braced non-zero number\0"    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
399    "(?+ or (?- or (?(+ or (?(- must be followed by a non-zero number\0"    "a numbered reference must not be zero\0"
400    "(*VERB) with an argument is not supported\0"    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
401    /* 60 */    /* 60 */
402    "(*VERB) not recognized\0"    "(*VERB) not recognized\0"
403    "number is too big\0"    "number is too big\0"
404    "subpattern name expected\0"    "subpattern name expected\0"
405    "digit expected after (?+";    "digit expected after (?+\0"
406      "] is an invalid data character in JavaScript compatibility mode\0"
407      /* 65 */
408      "different names for subpatterns of the same number are not allowed\0"
409      "(*MARK) must have an argument\0"
410      "this version of PCRE is not compiled with PCRE_UCP support\0"
411      "\\c must be followed by an ASCII character\0"
412      "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
413      /* 70 */
414      "internal error: unknown opcode in find_fixedlength()\0"
415      ;
416    
417  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
418  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 430  For convenience, we use the same bit def
430    
431  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
432    
433  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC
434    
435    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
436    UTF-8 mode. */
437    
438  static const unsigned char digitab[] =  static const unsigned char digitab[] =
439    {    {
440    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 470  static const unsigned char digitab[] =
470    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
471    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
472    
473  #else           /* This is the "abnormal" case, for EBCDIC systems */  #else
474    
475    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
476    
477  static const unsigned char digitab[] =  static const unsigned char digitab[] =
478    {    {
479    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 548  static const unsigned char ebcdic_charta
548  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
549    
550  static BOOL  static BOOL
551    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,    compile_regex(int, uschar **, const uschar **, int *, BOOL, BOOL, int, int,
552      int *, int *, branch_chain *, compile_data *, int *);      int *, int *, branch_chain *, compile_data *, int *);
553    
554    
# Line 454  static const char * Line 570  static const char *
570  find_error_text(int n)  find_error_text(int n)
571  {  {
572  const char *s = error_texts;  const char *s = error_texts;
573  for (; n > 0; n--) while (*s++ != 0);  for (; n > 0; n--)
574      {
575      while (*s++ != 0) {};
576      if (*s == 0) return "Error text not found (please report)";
577      }
578  return s;  return s;
579  }  }
580    
581    
582  /*************************************************  /*************************************************
583    *            Check for counted repeat            *
584    *************************************************/
585    
586    /* This function is called when a '{' is encountered in a place where it might
587    start a quantifier. It looks ahead to see if it really is a quantifier or not.
588    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
589    where the ddds are digits.
590    
591    Arguments:
592      p         pointer to the first char after '{'
593    
594    Returns:    TRUE or FALSE
595    */
596    
597    static BOOL
598    is_counted_repeat(const uschar *p)
599    {
600    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
601    while ((digitab[*p] & ctype_digit) != 0) p++;
602    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
603    
604    if (*p++ != CHAR_COMMA) return FALSE;
605    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
606    
607    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
608    while ((digitab[*p] & ctype_digit) != 0) p++;
609    
610    return (*p == CHAR_RIGHT_CURLY_BRACKET);
611    }
612    
613    
614    
615    /*************************************************
616  *            Handle escapes                      *  *            Handle escapes                      *
617  *************************************************/  *************************************************/
618    
# Line 502  if (c == 0) *errorcodeptr = ERR1; Line 655  if (c == 0) *errorcodeptr = ERR1;
655  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.
656  Otherwise further processing may be required. */  Otherwise further processing may be required. */
657    
658  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
659  else if (c < '0' || c > 'z') {}                           /* Not alphanumeric */  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */
660  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0) c = i;
661    
662  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
663  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 676  else
676      /* 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
677      error. */      error. */
678    
679      case 'l':      case CHAR_l:
680      case 'L':      case CHAR_L:
     case 'N':  
     case 'u':  
     case 'U':  
681      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
682      break;      break;
683    
684      /* \g must be followed by a number, either plain or braced. If positive, it      case CHAR_u:
685      is an absolute backreference. If negative, it is a relative backreference.      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
686      This is a Perl 5.10 feature. Perl 5.10 also supports \g{name} as a        {
687      reference to a named group. This is part of Perl's movement towards a        /* In JavaScript, \u must be followed by four hexadecimal numbers.
688      unified syntax for back references. As this is synonymous with \k{name}, we        Otherwise it is a lowercase u letter. */
689      fudge it up by pretending it really was \k. */        if ((digitab[ptr[1]] & ctype_xdigit) != 0 && (digitab[ptr[2]] & ctype_xdigit) != 0
690               && (digitab[ptr[3]] & ctype_xdigit) != 0 && (digitab[ptr[4]] & ctype_xdigit) != 0)
691            {
692            c = 0;
693            for (i = 0; i < 4; ++i)
694              {
695              register int cc = *(++ptr);
696    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
697              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
698              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
699    #else           /* EBCDIC coding */
700              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
701              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
702    #endif
703              }
704            }
705          }
706        else
707          *errorcodeptr = ERR37;
708        break;
709    
710        case CHAR_U:
711        /* In JavaScript, \U is an uppercase U letter. */
712        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
713        break;
714    
715        /* In a character class, \g is just a literal "g". Outside a character
716        class, \g must be followed by one of a number of specific things:
717    
718        (1) A number, either plain or braced. If positive, it is an absolute
719        backreference. If negative, it is a relative backreference. This is a Perl
720        5.10 feature.
721    
722        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
723        is part of Perl's movement towards a unified syntax for back references. As
724        this is synonymous with \k{name}, we fudge it up by pretending it really
725        was \k.
726    
727        (3) For Oniguruma compatibility we also support \g followed by a name or a
728        number either in angle brackets or in single quotes. However, these are
729        (possibly recursive) subroutine calls, _not_ backreferences. Just return
730        the -ESC_g code (cf \k). */
731    
732        case CHAR_g:
733        if (isclass) break;
734        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
735          {
736          c = -ESC_g;
737          break;
738          }
739    
740        /* Handle the Perl-compatible cases */
741    
742      case 'g':      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
     if (ptr[1] == '{')  
743        {        {
744        const uschar *p;        const uschar *p;
745        for (p = ptr+2; *p != 0 && *p != '}'; p++)        for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
746          if (*p != '-' && (digitab[*p] & ctype_digit) == 0) break;          if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
747        if (*p != 0 && *p != '}')        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
748          {          {
749          c = -ESC_k;          c = -ESC_k;
750          break;          break;
# Line 554  else Line 754  else
754        }        }
755      else braced = FALSE;      else braced = FALSE;
756    
757      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
758        {        {
759        negated = TRUE;        negated = TRUE;
760        ptr++;        ptr++;
# Line 563  else Line 763  else
763    
764      c = 0;      c = 0;
765      while ((digitab[ptr[1]] & ctype_digit) != 0)      while ((digitab[ptr[1]] & ctype_digit) != 0)
766        c = c * 10 + *(++ptr) - '0';        c = c * 10 + *(++ptr) - CHAR_0;
767    
768      if (c < 0)      if (c < 0)   /* Integer overflow */
769        {        {
770        *errorcodeptr = ERR61;        *errorcodeptr = ERR61;
771        break;        break;
772        }        }
773    
774      if (c == 0 || (braced && *(++ptr) != '}'))      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
775        {        {
776        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
777        break;        break;
778        }        }
779    
780        if (c == 0)
781          {
782          *errorcodeptr = ERR58;
783          break;
784          }
785    
786      if (negated)      if (negated)
787        {        {
788        if (c > bracount)        if (c > bracount)
# Line 602  else Line 808  else
808      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
809      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
810    
811      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:
812      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
813    
814      if (!isclass)      if (!isclass)
815        {        {
816        oldptr = ptr;        oldptr = ptr;
817        c -= '0';        c -= CHAR_0;
818        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
819          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - CHAR_0;
820        if (c < 0)        if (c < 0)    /* Integer overflow */
821          {          {
822          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
823          break;          break;
# Line 628  else Line 834  else
834      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.
835      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
836    
837      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
838        {        {
839        ptr--;        ptr--;
840        c = 0;        c = 0;
# Line 641  else Line 847  else
847      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
848      than 3 octal digits. */      than 3 octal digits. */
849    
850      case '0':      case CHAR_0:
851      c -= '0';      c -= CHAR_0;
852      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
853          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
854      if (!utf8 && c > 255) *errorcodeptr = ERR51;      if (!utf8 && c > 255) *errorcodeptr = ERR51;
855      break;      break;
856    
# Line 652  else Line 858  else
858      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
859      treated as a data character. */      treated as a data character. */
860    
861      case 'x':      case CHAR_x:
862      if (ptr[1] == '{')      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
863          {
864          /* In JavaScript, \x must be followed by two hexadecimal numbers.
865          Otherwise it is a lowercase x letter. */
866          if ((digitab[ptr[1]] & ctype_xdigit) != 0 && (digitab[ptr[2]] & ctype_xdigit) != 0)
867            {
868            c = 0;
869            for (i = 0; i < 2; ++i)
870              {
871              register int cc = *(++ptr);
872    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
873              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
874              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
875    #else           /* EBCDIC coding */
876              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
877              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
878    #endif
879              }
880            }
881          break;
882          }
883    
884        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
885        {        {
886        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
887        int count = 0;        int count = 0;
# Line 662  else Line 890  else
890        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
891          {          {
892          register int cc = *pt++;          register int cc = *pt++;
893          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
894          count++;          count++;
895    
896  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
897          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
898          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
899  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
900          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
901          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
902  #endif  #endif
903          }          }
904    
905        if (*pt == '}')        if (*pt == CHAR_RIGHT_CURLY_BRACKET)
906          {          {
907          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
908          ptr = pt;          ptr = pt;
# Line 690  else Line 918  else
918      c = 0;      c = 0;
919      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
920        {        {
921        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
922        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
923  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
924        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
925        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
926  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
927        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
928        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
929  #endif  #endif
930        }        }
931      break;      break;
932    
933      /* 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.
934      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
935        coding is ASCII-specific, but then the whole concept of \cx is
936      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
937    
938      case 'c':      case CHAR_c:
939      c = *(++ptr);      c = *(++ptr);
940      if (c == 0)      if (c == 0)
941        {        {
942        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
943        break;        break;
944        }        }
945    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
946  #ifndef EBCDIC  /* ASCII coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
947      if (c >= 'a' && c <= 'z') c -= 32;        {
948          *errorcodeptr = ERR68;
949          break;
950          }
951        if (c >= CHAR_a && c <= CHAR_z) c -= 32;
952      c ^= 0x40;      c ^= 0x40;
953  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
954      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
955      c ^= 0xC0;      c ^= 0xC0;
956  #endif  #endif
957      break;      break;
# Line 740  else Line 973  else
973      }      }
974    }    }
975    
976    /* Perl supports \N{name} for character names, as well as plain \N for "not
977    newline". PCRE does not support \N{name}. However, it does support
978    quantification such as \N{2,3}. */
979    
980    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
981         !is_counted_repeat(ptr+2))
982      *errorcodeptr = ERR37;
983    
984    /* If PCRE_UCP is set, we change the values for \d etc. */
985    
986    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
987      c -= (ESC_DU - ESC_D);
988    
989    /* Set the pointer to the final character before returning. */
990    
991  *ptrptr = ptr;  *ptrptr = ptr;
992  return c;  return c;
993  }  }
# Line 780  if (c == 0) goto ERROR_RETURN; Line 1028  if (c == 0) goto ERROR_RETURN;
1028  /* \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
1029  negation. */  negation. */
1030    
1031  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
1032    {    {
1033    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1034      {      {
1035      *negptr = TRUE;      *negptr = TRUE;
1036      ptr++;      ptr++;
# Line 791  if (c == '{') Line 1039  if (c == '{')
1039      {      {
1040      c = *(++ptr);      c = *(++ptr);
1041      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
1042      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1043      name[i] = c;      name[i] = c;
1044      }      }
1045    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
1046    name[i] = 0;    name[i] = 0;
1047    }    }
1048    
# Line 840  return -1; Line 1088  return -1;
1088    
1089    
1090  /*************************************************  /*************************************************
 *            Check for counted repeat            *  
 *************************************************/  
   
 /* This function is called when a '{' is encountered in a place where it might  
 start a quantifier. It looks ahead to see if it really is a quantifier or not.  
 It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}  
 where the ddds are digits.  
   
 Arguments:  
   p         pointer to the first char after '{'  
   
 Returns:    TRUE or FALSE  
 */  
   
 static BOOL  
 is_counted_repeat(const uschar *p)  
 {  
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
 if (*p == '}') return TRUE;  
   
 if (*p++ != ',') return FALSE;  
 if (*p == '}') return TRUE;  
   
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
   
 return (*p == '}');  
 }  
   
   
   
 /*************************************************  
1091  *         Read repeat counts                     *  *         Read repeat counts                     *
1092  *************************************************/  *************************************************/
1093    
# Line 900  int max = -1; Line 1115  int max = -1;
1115  /* 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
1116  an integer overflow. */  an integer overflow. */
1117    
1118  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;
1119  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1120    {    {
1121    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 910  if (min < 0 || min > 65535) Line 1125  if (min < 0 || min > 65535)
1125  /* 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.
1126  Also, max must not be less than min. */  Also, max must not be less than min. */
1127    
1128  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1129    {    {
1130    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1131      {      {
1132      max = 0;      max = 0;
1133      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;
1134      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1135        {        {
1136        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 940  return p; Line 1155  return p;
1155    
1156    
1157  /*************************************************  /*************************************************
1158  *       Find forward referenced subpattern       *  *  Subroutine for finding forward reference      *
1159  *************************************************/  *************************************************/
1160    
1161  /* This function scans along a pattern's text looking for capturing  /* This recursive function is called only from find_parens() below. The
1162    top-level call starts at the beginning of the pattern. All other calls must
1163    start at a parenthesis. It scans along a pattern's text looking for capturing
1164  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
1165  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
1166  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
1167  references to subpatterns. We know that if (?P< is encountered, the name will  track of subpatterns that reset the capturing group numbers - the (?| feature.
1168  be terminated by '>' because that is checked in the first pass.  
1169    This function was originally called only from the second pass, in which we know
1170    that if (?< or (?' or (?P< is encountered, the name will be correctly
1171    terminated because that is checked in the first pass. There is now one call to
1172    this function in the first pass, to check for a recursive back reference by
1173    name (so that we can make the whole group atomic). In this case, we need check
1174    only up to the current position in the pattern, and that is still OK because
1175    and previous occurrences will have been checked. To make this work, the test
1176    for "end of pattern" is a check against cd->end_pattern in the main loop,
1177    instead of looking for a binary zero. This means that the special first-pass
1178    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1179    processing items within the loop are OK, because afterwards the main loop will
1180    terminate.)
1181    
1182  Arguments:  Arguments:
1183    ptr          current position in the pattern    ptrptr       address of the current character pointer (updated)
1184    count        current count of capturing parens so far encountered    cd           compile background data
1185    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1186    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1187    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1188      utf8         TRUE if we are in UTF-8 mode
1189      count        pointer to the current capturing subpattern number (updated)
1190    
1191  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1192  */  */
1193    
1194  static int  static int
1195  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,
1196    BOOL xmode)    BOOL xmode, BOOL utf8, int *count)
1197  {  {
1198  const uschar *thisname;  uschar *ptr = *ptrptr;
1199    int start_count = *count;
1200    int hwm_count = start_count;
1201    BOOL dup_parens = FALSE;
1202    
1203    /* If the first character is a parenthesis, check on the type of group we are
1204    dealing with. The very first call may not start with a parenthesis. */
1205    
1206  for (; *ptr != 0; ptr++)  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1207    {    {
1208    int term;    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1209    
1210      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1211    
1212      /* Handle a normal, unnamed capturing parenthesis. */
1213    
1214      else if (ptr[1] != CHAR_QUESTION_MARK)
1215        {
1216        *count += 1;
1217        if (name == NULL && *count == lorn) return *count;
1218        ptr++;
1219        }
1220    
1221      /* All cases now have (? at the start. Remember when we are in a group
1222      where the parenthesis numbers are duplicated. */
1223    
1224      else if (ptr[2] == CHAR_VERTICAL_LINE)
1225        {
1226        ptr += 3;
1227        dup_parens = TRUE;
1228        }
1229    
1230      /* Handle comments; all characters are allowed until a ket is reached. */
1231    
1232      else if (ptr[2] == CHAR_NUMBER_SIGN)
1233        {
1234        for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1235        goto FAIL_EXIT;
1236        }
1237    
1238      /* Handle a condition. If it is an assertion, just carry on so that it
1239      is processed as normal. If not, skip to the closing parenthesis of the
1240      condition (there can't be any nested parens). */
1241    
1242      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1243        {
1244        ptr += 2;
1245        if (ptr[1] != CHAR_QUESTION_MARK)
1246          {
1247          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1248          if (*ptr != 0) ptr++;
1249          }
1250        }
1251    
1252      /* Start with (? but not a condition. */
1253    
1254      else
1255        {
1256        ptr += 2;
1257        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1258    
1259        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1260    
1261        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1262            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1263          {
1264          int term;
1265          const uschar *thisname;
1266          *count += 1;
1267          if (name == NULL && *count == lorn) return *count;
1268          term = *ptr++;
1269          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1270          thisname = ptr;
1271          while (*ptr != term) ptr++;
1272          if (name != NULL && lorn == ptr - thisname &&
1273              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1274            return *count;
1275          term++;
1276          }
1277        }
1278      }
1279    
1280    /* Past any initial parenthesis handling, scan for parentheses or vertical
1281    bars. Stop if we get to cd->end_pattern. Note that this is important for the
1282    first-pass call when this value is temporarily adjusted to stop at the current
1283    position. So DO NOT change this to a test for binary zero. */
1284    
1285    for (; ptr < cd->end_pattern; ptr++)
1286      {
1287    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1288    
1289    if (*ptr == '\\')    if (*ptr == CHAR_BACKSLASH)
1290      {      {
1291      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1292      if (*ptr == 'Q') for (;;)      if (*ptr == CHAR_Q) for (;;)
1293        {        {
1294        while (*(++ptr) != 0 && *ptr != '\\');        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1295        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1296        if (*(++ptr) == 'E') break;        if (*(++ptr) == CHAR_E) break;
1297        }        }
1298      continue;      continue;
1299      }      }
1300    
1301    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1302      are handled for real. If the first character is '^', skip it. Also, if the
1303      first few characters (either before or after ^) are \Q\E or \E we skip them
1304      too. This makes for compatibility with Perl. Note the use of STR macros to
1305      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1306    
1307    if (*ptr == '[')    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1308      {      {
1309      while (*(++ptr) != ']')      BOOL negate_class = FALSE;
1310        for (;;)
1311          {
1312          if (ptr[1] == CHAR_BACKSLASH)
1313            {
1314            if (ptr[2] == CHAR_E)
1315              ptr+= 2;
1316            else if (strncmp((const char *)ptr+2,
1317                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1318              ptr += 4;
1319            else
1320              break;
1321            }
1322          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1323            {
1324            negate_class = TRUE;
1325            ptr++;
1326            }
1327          else break;
1328          }
1329    
1330        /* If the next character is ']', it is a data character that must be
1331        skipped, except in JavaScript compatibility mode. */
1332    
1333        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1334            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1335          ptr++;
1336    
1337        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1338        {        {
1339        if (*ptr == 0) return -1;        if (*ptr == 0) return -1;
1340        if (*ptr == '\\')        if (*ptr == CHAR_BACKSLASH)
1341          {          {
1342          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) goto FAIL_EXIT;
1343          if (*ptr == 'Q') for (;;)          if (*ptr == CHAR_Q) for (;;)
1344            {            {
1345            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1346            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1347            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1348            }            }
1349          continue;          continue;
1350          }          }
# Line 1008  for (; *ptr != 0; ptr++) Line 1354  for (; *ptr != 0; ptr++)
1354    
1355    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1356    
1357    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1358      {      {
1359      while (*(++ptr) != 0 && *ptr != '\n');      ptr++;
1360      if (*ptr == 0) return -1;      while (*ptr != 0)
1361          {
1362          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1363          ptr++;
1364    #ifdef SUPPORT_UTF8
1365          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
1366    #endif
1367          }
1368        if (*ptr == 0) goto FAIL_EXIT;
1369      continue;      continue;
1370      }      }
1371    
1372    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1373    
1374    if (*ptr != '(') continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != '?' && ptr[1] != '*')  
1375      {      {
1376      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, count);
1377      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1378      continue;      if (*ptr == 0) goto FAIL_EXIT;
1379      }      }
1380    
1381    ptr += 2;    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1382    if (*ptr == 'P') ptr++;                      /* Allow optional P */      {
1383        if (dup_parens && *count < hwm_count) *count = hwm_count;
1384        goto FAIL_EXIT;
1385        }
1386    
1387      else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1388        {
1389        if (*count > hwm_count) hwm_count = *count;
1390        *count = start_count;
1391        }
1392      }
1393    
1394    FAIL_EXIT:
1395    *ptrptr = ptr;
1396    return -1;
1397    }
1398    
   /* We have to disambiguate (?<! and (?<= from (?<name> */  
1399    
   if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  
        *ptr != '\'')  
     continue;  
1400    
   count++;  
1401    
1402    if (name == NULL && count == lorn) return count;  /*************************************************
1403    term = *ptr++;  *       Find forward referenced subpattern       *
1404    if (term == '<') term = '>';  *************************************************/
1405    thisname = ptr;  
1406    while (*ptr != term) ptr++;  /* This function scans along a pattern's text looking for capturing
1407    if (name != NULL && lorn == ptr - thisname &&  subpatterns, and counting them. If it finds a named pattern that matches the
1408        strncmp((const char *)name, (const char *)thisname, lorn) == 0)  name it is given, it returns its number. Alternatively, if the name is NULL, it
1409      return count;  returns when it reaches a given numbered subpattern. This is used for forward
1410    references to subpatterns. We used to be able to start this scan from the
1411    current compiling point, using the current count value from cd->bracount, and
1412    do it all in a single loop, but the addition of the possibility of duplicate
1413    subpattern numbers means that we have to scan from the very start, in order to
1414    take account of such duplicates, and to use a recursive function to keep track
1415    of the different types of group.
1416    
1417    Arguments:
1418      cd           compile background data
1419      name         name to seek, or NULL if seeking a numbered subpattern
1420      lorn         name length, or subpattern number if name is NULL
1421      xmode        TRUE if we are in /x mode
1422      utf8         TRUE if we are in UTF-8 mode
1423    
1424    Returns:       the number of the found subpattern, or -1 if not found
1425    */
1426    
1427    static int
1428    find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode,
1429      BOOL utf8)
1430    {
1431    uschar *ptr = (uschar *)cd->start_pattern;
1432    int count = 0;
1433    int rc;
1434    
1435    /* If the pattern does not start with an opening parenthesis, the first call
1436    to find_parens_sub() will scan right to the end (if necessary). However, if it
1437    does start with a parenthesis, find_parens_sub() will return when it hits the
1438    matching closing parens. That is why we have to have a loop. */
1439    
1440    for (;;)
1441      {
1442      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, &count);
1443      if (rc > 0 || *ptr++ == 0) break;
1444    }    }
1445    
1446  return -1;  return rc;
1447  }  }
1448    
1449    
1450    
1451    
1452  /*************************************************  /*************************************************
1453  *      Find first significant op code            *  *      Find first significant op code            *
1454  *************************************************/  *************************************************/
1455    
1456  /* This is called by several functions that scan a compiled expression looking  /* This is called by several functions that scan a compiled expression looking
1457  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
1458  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
1459  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
1460  assertions, and also the \b assertion; for others it does not.  does not.
1461    
1462  Arguments:  Arguments:
1463    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  
1464    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1465    
1466  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1467  */  */
1468    
1469  static const uschar*  static const uschar*
1470  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const uschar *code, BOOL skipassert)
   BOOL skipassert)  
1471  {  {
1472  for (;;)  for (;;)
1473    {    {
1474    switch ((int)*code)    switch ((int)*code)
1475      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1476      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1477      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1478      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
# Line 1100  for (;;) Line 1488  for (;;)
1488    
1489      case OP_CALLOUT:      case OP_CALLOUT:
1490      case OP_CREF:      case OP_CREF:
1491        case OP_NCREF:
1492      case OP_RREF:      case OP_RREF:
1493        case OP_NRREF:
1494      case OP_DEF:      case OP_DEF:
1495      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1496      break;      break;
# Line 1116  for (;;) Line 1506  for (;;)
1506    
1507    
1508  /*************************************************  /*************************************************
1509  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1510  *************************************************/  *************************************************/
1511    
1512  /* 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,
1513  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.
1514  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
1515    temporarily terminated with OP_END when this function is called.
1516    
1517    This function is called when a backward assertion is encountered, so that if it
1518    fails, the error message can point to the correct place in the pattern.
1519    However, we cannot do this when the assertion contains subroutine calls,
1520    because they can be forward references. We solve this by remembering this case
1521    and doing the check at the end; a flag specifies which mode we are running in.
1522    
1523  Arguments:  Arguments:
1524    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1525    options  the compiling options    utf8     TRUE in UTF-8 mode
1526      atend    TRUE if called when the pattern is complete
1527  Returns:   the fixed length, or -1 if there is no fixed length,    cd       the "compile data" structure
1528               or -2 if \C was encountered  
1529    Returns:   the fixed length,
1530                 or -1 if there is no fixed length,
1531                 or -2 if \C was encountered (in UTF-8 mode only)
1532                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1533                 or -4 if an unknown opcode was encountered (internal error)
1534  */  */
1535    
1536  static int  static int
1537  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, BOOL utf8, BOOL atend, compile_data *cd)
1538  {  {
1539  int length = -1;  int length = -1;
1540    
# Line 1145  branch, check the length against that of Line 1547  branch, check the length against that of
1547  for (;;)  for (;;)
1548    {    {
1549    int d;    int d;
1550      uschar *ce, *cs;
1551    register int op = *cc;    register int op = *cc;
1552    switch (op)    switch (op)
1553      {      {
1554        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1555        OP_BRA (normal non-capturing bracket) because the other variants of these
1556        opcodes are all concerned with unlimited repeated groups, which of course
1557        are not of fixed length. */
1558    
1559      case OP_CBRA:      case OP_CBRA:
1560      case OP_BRA:      case OP_BRA:
1561      case OP_ONCE:      case OP_ONCE:
1562        case OP_ONCE_NC:
1563      case OP_COND:      case OP_COND:
1564      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), utf8, atend, cd);
1565      if (d < 0) return d;      if (d < 0) return d;
1566      branchlength += d;      branchlength += d;
1567      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1568      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1569      break;      break;
1570    
1571      /* 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 call.
1572      call. If it's ALT it is an alternation in a nested call. If it is      If it's ALT it is an alternation in a nested call. An ACCEPT is effectively
1573      END it's the end of the outer call. All can be handled by the same code. */      an ALT. If it is END it's the end of the outer call. All can be handled by
1574        the same code. Note that we must not include the OP_KETRxxx opcodes here,
1575        because they all imply an unlimited repeat. */
1576    
1577      case OP_ALT:      case OP_ALT:
1578      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1579      case OP_END:      case OP_END:
1580        case OP_ACCEPT:
1581        case OP_ASSERT_ACCEPT:
1582      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1583        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
1584      if (*cc != OP_ALT) return length;      if (*cc != OP_ALT) return length;
# Line 1175  for (;;) Line 1586  for (;;)
1586      branchlength = 0;      branchlength = 0;
1587      break;      break;
1588    
1589        /* A true recursion implies not fixed length, but a subroutine call may
1590        be OK. If the subroutine is a forward reference, we can't deal with
1591        it until the end of the pattern, so return -3. */
1592    
1593        case OP_RECURSE:
1594        if (!atend) return -3;
1595        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1596        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1597        if (cc > cs && cc < ce) return -1;                /* Recursion */
1598        d = find_fixedlength(cs + 2, utf8, atend, cd);
1599        if (d < 0) return d;
1600        branchlength += d;
1601        cc += 1 + LINK_SIZE;
1602        break;
1603    
1604      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1605    
1606      case OP_ASSERT:      case OP_ASSERT:
# Line 1186  for (;;) Line 1612  for (;;)
1612    
1613      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1614    
1615      case OP_REVERSE:      case OP_MARK:
1616        case OP_PRUNE_ARG:
1617        case OP_SKIP_ARG:
1618        case OP_THEN_ARG:
1619        cc += cc[1] + _pcre_OP_lengths[*cc];
1620        break;
1621    
1622        case OP_CALLOUT:
1623        case OP_CIRC:
1624        case OP_CIRCM:
1625        case OP_CLOSE:
1626        case OP_COMMIT:
1627      case OP_CREF:      case OP_CREF:
     case OP_RREF:  
1628      case OP_DEF:      case OP_DEF:
1629      case OP_OPT:      case OP_DOLL:
1630      case OP_CALLOUT:      case OP_DOLLM:
     case OP_SOD:  
     case OP_SOM:  
1631      case OP_EOD:      case OP_EOD:
1632      case OP_EODN:      case OP_EODN:
1633      case OP_CIRC:      case OP_FAIL:
1634      case OP_DOLL:      case OP_NCREF:
1635        case OP_NRREF:
1636      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1637        case OP_PRUNE:
1638        case OP_REVERSE:
1639        case OP_RREF:
1640        case OP_SET_SOM:
1641        case OP_SKIP:
1642        case OP_SOD:
1643        case OP_SOM:
1644        case OP_THEN:
1645      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1646      cc += _pcre_OP_lengths[*cc];      cc += _pcre_OP_lengths[*cc];
1647      break;      break;
# Line 1206  for (;;) Line 1649  for (;;)
1649      /* Handle literal characters */      /* Handle literal characters */
1650    
1651      case OP_CHAR:      case OP_CHAR:
1652      case OP_CHARNC:      case OP_CHARI:
1653      case OP_NOT:      case OP_NOT:
1654        case OP_NOTI:
1655      branchlength++;      branchlength++;
1656      cc += 2;      cc += 2;
1657  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1658      if ((options & PCRE_UTF8) != 0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       {  
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1659  #endif  #endif
1660      break;      break;
1661    
# Line 1222  for (;;) Line 1663  for (;;)
1663      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1664    
1665      case OP_EXACT:      case OP_EXACT:
1666        case OP_EXACTI:
1667        case OP_NOTEXACT:
1668        case OP_NOTEXACTI:
1669      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1670      cc += 4;      cc += 4;
1671  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1672      if ((options & PCRE_UTF8) != 0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       {  
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1673  #endif  #endif
1674      break;      break;
1675    
# Line 1245  for (;;) Line 1686  for (;;)
1686      cc += 2;      cc += 2;
1687      /* Fall through */      /* Fall through */
1688    
1689        case OP_HSPACE:
1690        case OP_VSPACE:
1691        case OP_NOT_HSPACE:
1692        case OP_NOT_VSPACE:
1693      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
1694      case OP_DIGIT:      case OP_DIGIT:
1695      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
# Line 1252  for (;;) Line 1697  for (;;)
1697      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1698      case OP_WORDCHAR:      case OP_WORDCHAR:
1699      case OP_ANY:      case OP_ANY:
1700        case OP_ALLANY:
1701      branchlength++;      branchlength++;
1702      cc++;      cc++;
1703      break;      break;
1704    
1705      /* The single-byte matcher isn't allowed */      /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1706        otherwise \C is coded as OP_ALLANY. */
1707    
1708      case OP_ANYBYTE:      case OP_ANYBYTE:
1709      return -2;      return -2;
# Line 1275  for (;;) Line 1722  for (;;)
1722    
1723      switch (*cc)      switch (*cc)
1724        {        {
1725          case OP_CRPLUS:
1726          case OP_CRMINPLUS:
1727        case OP_CRSTAR:        case OP_CRSTAR:
1728        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1729        case OP_CRQUERY:        case OP_CRQUERY:
# Line 1293  for (;;) Line 1742  for (;;)
1742        }        }
1743      break;      break;
1744    
1745      /* Anything else is variable length */      /* Anything else is variable length */
1746    
1747        case OP_ANYNL:
1748        case OP_BRAMINZERO:
1749        case OP_BRAPOS:
1750        case OP_BRAPOSZERO:
1751        case OP_BRAZERO:
1752        case OP_CBRAPOS:
1753        case OP_EXTUNI:
1754        case OP_KETRMAX:
1755        case OP_KETRMIN:
1756        case OP_KETRPOS:
1757        case OP_MINPLUS:
1758        case OP_MINPLUSI:
1759        case OP_MINQUERY:
1760        case OP_MINQUERYI:
1761        case OP_MINSTAR:
1762        case OP_MINSTARI:
1763        case OP_MINUPTO:
1764        case OP_MINUPTOI:
1765        case OP_NOTMINPLUS:
1766        case OP_NOTMINPLUSI:
1767        case OP_NOTMINQUERY:
1768        case OP_NOTMINQUERYI:
1769        case OP_NOTMINSTAR:
1770        case OP_NOTMINSTARI:
1771        case OP_NOTMINUPTO:
1772        case OP_NOTMINUPTOI:
1773        case OP_NOTPLUS:
1774        case OP_NOTPLUSI:
1775        case OP_NOTPOSPLUS:
1776        case OP_NOTPOSPLUSI:
1777        case OP_NOTPOSQUERY:
1778        case OP_NOTPOSQUERYI:
1779        case OP_NOTPOSSTAR:
1780        case OP_NOTPOSSTARI:
1781        case OP_NOTPOSUPTO:
1782        case OP_NOTPOSUPTOI:
1783        case OP_NOTQUERY:
1784        case OP_NOTQUERYI:
1785        case OP_NOTSTAR:
1786        case OP_NOTSTARI:
1787        case OP_NOTUPTO:
1788        case OP_NOTUPTOI:
1789        case OP_PLUS:
1790        case OP_PLUSI:
1791        case OP_POSPLUS:
1792        case OP_POSPLUSI:
1793        case OP_POSQUERY:
1794        case OP_POSQUERYI:
1795        case OP_POSSTAR:
1796        case OP_POSSTARI:
1797        case OP_POSUPTO:
1798        case OP_POSUPTOI:
1799        case OP_QUERY:
1800        case OP_QUERYI:
1801        case OP_REF:
1802        case OP_REFI:
1803        case OP_SBRA:
1804        case OP_SBRAPOS:
1805        case OP_SCBRA:
1806        case OP_SCBRAPOS:
1807        case OP_SCOND:
1808        case OP_SKIPZERO:
1809        case OP_STAR:
1810        case OP_STARI:
1811        case OP_TYPEMINPLUS:
1812        case OP_TYPEMINQUERY:
1813        case OP_TYPEMINSTAR:
1814        case OP_TYPEMINUPTO:
1815        case OP_TYPEPLUS:
1816        case OP_TYPEPOSPLUS:
1817        case OP_TYPEPOSQUERY:
1818        case OP_TYPEPOSSTAR:
1819        case OP_TYPEPOSUPTO:
1820        case OP_TYPEQUERY:
1821        case OP_TYPESTAR:
1822        case OP_TYPEUPTO:
1823        case OP_UPTO:
1824        case OP_UPTOI:
1825        return -1;
1826    
1827        /* Catch unrecognized opcodes so that when new ones are added they
1828        are not forgotten, as has happened in the past. */
1829    
1830      default:      default:
1831      return -1;      return -4;
1832      }      }
1833    }    }
1834  /* Control never gets here */  /* Control never gets here */
# Line 1306  for (;;) Line 1838  for (;;)
1838    
1839    
1840  /*************************************************  /*************************************************
1841  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1842  *************************************************/  *************************************************/
1843    
1844  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1845  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1846    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1847    so that it can be called from pcre_study() when finding the minimum matching
1848    length.
1849    
1850  Arguments:  Arguments:
1851    code        points to start of expression    code        points to start of expression
1852    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1853    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1854    
1855  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
1856  */  */
1857    
1858  static const uschar *  const uschar *
1859  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1860  {  {
1861  for (;;)  for (;;)
1862    {    {
1863    register int c = *code;    register int c = *code;
1864    
1865    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1866    
1867    /* 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 1870  for (;;)
1870    
1871    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1872    
1873      /* Handle recursion */
1874    
1875      else if (c == OP_REVERSE)
1876        {
1877        if (number < 0) return (uschar *)code;
1878        code += _pcre_OP_lengths[c];
1879        }
1880    
1881    /* Handle capturing bracket */    /* Handle capturing bracket */
1882    
1883    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
1884               c == OP_CBRAPOS || c == OP_SCBRAPOS)
1885      {      {
1886      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
1887      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
# Line 1345  for (;;) Line 1890  for (;;)
1890    
1891    /* 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
1892    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
1893    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1894      must add in its length. */
1895    
1896    else    else
1897      {      {
# Line 1369  for (;;) Line 1915  for (;;)
1915        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
1916        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1917        break;        break;
1918    
1919          case OP_MARK:
1920          case OP_PRUNE_ARG:
1921          case OP_SKIP_ARG:
1922          code += code[1];
1923          break;
1924    
1925          case OP_THEN_ARG:
1926          code += code[1];
1927          break;
1928        }        }
1929    
1930      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1383  for (;;) Line 1939  for (;;)
1939      if (utf8) switch(c)      if (utf8) switch(c)
1940        {        {
1941        case OP_CHAR:        case OP_CHAR:
1942        case OP_CHARNC:        case OP_CHARI:
1943        case OP_EXACT:        case OP_EXACT:
1944          case OP_EXACTI:
1945        case OP_UPTO:        case OP_UPTO:
1946          case OP_UPTOI:
1947        case OP_MINUPTO:        case OP_MINUPTO:
1948          case OP_MINUPTOI:
1949        case OP_POSUPTO:        case OP_POSUPTO:
1950          case OP_POSUPTOI:
1951        case OP_STAR:        case OP_STAR:
1952          case OP_STARI:
1953        case OP_MINSTAR:        case OP_MINSTAR:
1954          case OP_MINSTARI:
1955        case OP_POSSTAR:        case OP_POSSTAR:
1956          case OP_POSSTARI:
1957        case OP_PLUS:        case OP_PLUS:
1958          case OP_PLUSI:
1959        case OP_MINPLUS:        case OP_MINPLUS:
1960          case OP_MINPLUSI:
1961        case OP_POSPLUS:        case OP_POSPLUS:
1962          case OP_POSPLUSI:
1963        case OP_QUERY:        case OP_QUERY:
1964          case OP_QUERYI:
1965        case OP_MINQUERY:        case OP_MINQUERY:
1966          case OP_MINQUERYI:
1967        case OP_POSQUERY:        case OP_POSQUERY:
1968          case OP_POSQUERYI:
1969        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1970        break;        break;
1971        }        }
1972    #else
1973        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1974  #endif  #endif
1975      }      }
1976    }    }
# Line 1438  for (;;) Line 2009  for (;;)
2009    
2010    /* 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
2011    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
2012    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2013      must add in its length. */
2014    
2015    else    else
2016      {      {
# Line 1462  for (;;) Line 2034  for (;;)
2034        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2035        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
2036        break;        break;
2037    
2038          case OP_MARK:
2039          case OP_PRUNE_ARG:
2040          case OP_SKIP_ARG:
2041          code += code[1];
2042          break;
2043    
2044          case OP_THEN_ARG:
2045          code += code[1];
2046          break;
2047        }        }
2048    
2049      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1476  for (;;) Line 2058  for (;;)
2058      if (utf8) switch(c)      if (utf8) switch(c)
2059        {        {
2060        case OP_CHAR:        case OP_CHAR:
2061        case OP_CHARNC:        case OP_CHARI:
2062        case OP_EXACT:        case OP_EXACT:
2063          case OP_EXACTI:
2064        case OP_UPTO:        case OP_UPTO:
2065          case OP_UPTOI:
2066        case OP_MINUPTO:        case OP_MINUPTO:
2067          case OP_MINUPTOI:
2068        case OP_POSUPTO:        case OP_POSUPTO:
2069          case OP_POSUPTOI:
2070        case OP_STAR:        case OP_STAR:
2071          case OP_STARI:
2072        case OP_MINSTAR:        case OP_MINSTAR:
2073          case OP_MINSTARI:
2074        case OP_POSSTAR:        case OP_POSSTAR:
2075          case OP_POSSTARI:
2076        case OP_PLUS:        case OP_PLUS:
2077          case OP_PLUSI:
2078        case OP_MINPLUS:        case OP_MINPLUS:
2079          case OP_MINPLUSI:
2080        case OP_POSPLUS:        case OP_POSPLUS:
2081          case OP_POSPLUSI:
2082        case OP_QUERY:        case OP_QUERY:
2083          case OP_QUERYI:
2084        case OP_MINQUERY:        case OP_MINQUERY:
2085          case OP_MINQUERYI:
2086        case OP_POSQUERY:        case OP_POSQUERY:
2087          case OP_POSQUERYI:
2088        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
2089        break;        break;
2090        }        }
2091    #else
2092        (void)(utf8);  /* Keep compiler happy by referencing function argument */
2093  #endif  #endif
2094      }      }
2095    }    }
# Line 1516  Arguments: Line 2113  Arguments:
2113    code        points to start of search    code        points to start of search
2114    endcode     points to where to stop    endcode     points to where to stop
2115    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
2116      cd          contains pointers to tables etc.
2117    
2118  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2119  */  */
2120    
2121  static BOOL  static BOOL
2122  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
2123      compile_data *cd)
2124  {  {
2125  register int c;  register int c;
2126  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], TRUE);
2127       code < endcode;       code < endcode;
2128       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], TRUE))
2129    {    {
2130    const uschar *ccode;    const uschar *ccode;
2131    
# Line 1542  for (code = first_significant_code(code Line 2141  for (code = first_significant_code(code
2141      continue;      continue;
2142      }      }
2143    
2144      /* For a recursion/subroutine call, if its end has been reached, which
2145      implies a backward reference subroutine call, we can scan it. If it's a
2146      forward reference subroutine call, we can't. To detect forward reference
2147      we have to scan up the list that is kept in the workspace. This function is
2148      called only when doing the real compile, not during the pre-compile that
2149      measures the size of the compiled pattern. */
2150    
2151      if (c == OP_RECURSE)
2152        {
2153        const uschar *scode;
2154        BOOL empty_branch;
2155    
2156        /* Test for forward reference */
2157    
2158        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
2159          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
2160    
2161        /* Not a forward reference, test for completed backward reference */
2162    
2163        empty_branch = FALSE;
2164        scode = cd->start_code + GET(code, 1);
2165        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2166    
2167        /* Completed backwards reference */
2168    
2169        do
2170          {
2171          if (could_be_empty_branch(scode, endcode, utf8, cd))
2172            {
2173            empty_branch = TRUE;
2174            break;
2175            }
2176          scode += GET(scode, 1);
2177          }
2178        while (*scode == OP_ALT);
2179    
2180        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2181        continue;
2182        }
2183    
2184    /* Groups with zero repeats can of course be empty; skip them. */    /* Groups with zero repeats can of course be empty; skip them. */
2185    
2186    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2187          c == OP_BRAPOSZERO)
2188      {      {
2189      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
2190      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 2192  for (code = first_significant_code(code
2192      continue;      continue;
2193      }      }
2194    
2195      /* A nested group that is already marked as "could be empty" can just be
2196      skipped. */
2197    
2198      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2199          c == OP_SCBRA || c == OP_SCBRAPOS)
2200        {
2201        do code += GET(code, 1); while (*code == OP_ALT);
2202        c = *code;
2203        continue;
2204        }
2205    
2206    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
2207    
2208    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_BRA  || c == OP_BRAPOS ||
2209          c == OP_CBRA || c == OP_CBRAPOS ||
2210          c == OP_ONCE || c == OP_ONCE_NC ||
2211          c == OP_COND)
2212      {      {
2213      BOOL empty_branch;      BOOL empty_branch;
2214      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
2215    
2216      /* Scan a closed bracket */      /* If a conditional group has only one branch, there is a second, implied,
2217        empty branch, so just skip over the conditional, because it could be empty.
2218        Otherwise, scan the individual branches of the group. */
2219    
2220      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;  
2221        code += GET(code, 1);        code += GET(code, 1);
2222        else
2223          {
2224          empty_branch = FALSE;
2225          do
2226            {
2227            if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
2228              empty_branch = TRUE;
2229            code += GET(code, 1);
2230            }
2231          while (*code == OP_ALT);
2232          if (!empty_branch) return FALSE;   /* All branches are non-empty */
2233        }        }
2234      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
2235      c = *code;      c = *code;
2236      continue;      continue;
2237      }      }
# Line 1630  for (code = first_significant_code(code Line 2292  for (code = first_significant_code(code
2292      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2293      case OP_WORDCHAR:      case OP_WORDCHAR:
2294      case OP_ANY:      case OP_ANY:
2295        case OP_ALLANY:
2296      case OP_ANYBYTE:      case OP_ANYBYTE:
2297      case OP_CHAR:      case OP_CHAR:
2298      case OP_CHARNC:      case OP_CHARI:
2299      case OP_NOT:      case OP_NOT:
2300        case OP_NOTI:
2301      case OP_PLUS:      case OP_PLUS:
2302      case OP_MINPLUS:      case OP_MINPLUS:
2303      case OP_POSPLUS:      case OP_POSPLUS:
# Line 1673  for (code = first_significant_code(code Line 2337  for (code = first_significant_code(code
2337      case OP_KET:      case OP_KET:
2338      case OP_KETRMAX:      case OP_KETRMAX:
2339      case OP_KETRMIN:      case OP_KETRMIN:
2340        case OP_KETRPOS:
2341      case OP_ALT:      case OP_ALT:
2342      return TRUE;      return TRUE;
2343    
# Line 1681  for (code = first_significant_code(code Line 2346  for (code = first_significant_code(code
2346    
2347  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2348      case OP_STAR:      case OP_STAR:
2349        case OP_STARI:
2350      case OP_MINSTAR:      case OP_MINSTAR:
2351        case OP_MINSTARI:
2352      case OP_POSSTAR:      case OP_POSSTAR:
2353        case OP_POSSTARI:
2354      case OP_QUERY:      case OP_QUERY:
2355        case OP_QUERYI:
2356      case OP_MINQUERY:      case OP_MINQUERY:
2357        case OP_MINQUERYI:
2358      case OP_POSQUERY:      case OP_POSQUERY:
2359        case OP_POSQUERYI:
2360        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2361        break;
2362    
2363      case OP_UPTO:      case OP_UPTO:
2364        case OP_UPTOI:
2365      case OP_MINUPTO:      case OP_MINUPTO:
2366        case OP_MINUPTOI:
2367      case OP_POSUPTO:      case OP_POSUPTO:
2368      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_POSUPTOI:
2369        if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2370      break;      break;
2371  #endif  #endif
2372    
2373        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2374        string. */
2375    
2376        case OP_MARK:
2377        case OP_PRUNE_ARG:
2378        case OP_SKIP_ARG:
2379        code += code[1];
2380        break;
2381    
2382        case OP_THEN_ARG:
2383        code += code[1];
2384        break;
2385    
2386        /* None of the remaining opcodes are required to match a character. */
2387    
2388        default:
2389        break;
2390      }      }
2391    }    }
2392    
# Line 1708  return TRUE; Line 2403  return TRUE;
2403  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
2404  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,
2405  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.
2406    This function is called only during the real compile, not during the
2407    pre-compile.
2408    
2409  Arguments:  Arguments:
2410    code        points to start of the recursion    code        points to start of the recursion
2411    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2412    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2413    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2414      cd          pointers to tables etc
2415    
2416  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2417  */  */
2418    
2419  static BOOL  static BOOL
2420  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2421    BOOL utf8)    BOOL utf8, compile_data *cd)
2422  {  {
2423  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2424    {    {
2425    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2426        return FALSE;
2427    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2428    }    }
2429  return TRUE;  return TRUE;
# Line 1756  where Perl recognizes it as the POSIX cl Line 2455  where Perl recognizes it as the POSIX cl
2455  "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,  "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2456  I think.  I think.
2457    
2458    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2459    It seems that the appearance of a nested POSIX class supersedes an apparent
2460    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2461    a digit.
2462    
2463    In Perl, unescaped square brackets may also appear as part of class names. For
2464    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2465    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2466    seem right at all. PCRE does not allow closing square brackets in POSIX class
2467    names.
2468    
2469  Arguments:  Arguments:
2470    ptr      pointer to the initial [    ptr      pointer to the initial [
2471    endptr   where to return the end pointer    endptr   where to return the end pointer
# Line 1770  int terminator; /* Don't combin Line 2480  int terminator; /* Don't combin
2480  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2481  for (++ptr; *ptr != 0; ptr++)  for (++ptr; *ptr != 0; ptr++)
2482    {    {
2483    if (*ptr == '\\' && ptr[1] == ']') ptr++; else    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2484        ptr++;
2485      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2486      else
2487      {      {
2488      if (*ptr == ']') return FALSE;      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
     if (*ptr == terminator && ptr[1] == ']')  
2489        {        {
2490        *endptr = ptr;        *endptr = ptr;
2491        return TRUE;        return TRUE;
2492        }        }
2493        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2494             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2495              ptr[1] == CHAR_EQUALS_SIGN) &&
2496            check_posix_syntax(ptr, endptr))
2497          return FALSE;
2498      }      }
2499    }    }
2500  return FALSE;  return FALSE;
# Line 1824  return -1; Line 2541  return -1;
2541  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2542  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2543  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
2544  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
2545  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
2546  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
2547  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
2548  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2549    OP_END.
2550    
2551  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2552  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 2623  auto_callout(uschar *code, const uschar
2623  {  {
2624  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2625  *code++ = 255;  *code++ = 255;
2626  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2627  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2628  return code + 2*LINK_SIZE;  return code + 2*LINK_SIZE;
2629  }  }
2630    
# Line 1931  Returns: nothing Line 2649  Returns: nothing
2649  static void  static void
2650  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)
2651  {  {
2652  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2653  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2654  }  }
2655    
# Line 1963  get_othercase_range(unsigned int *cptr, Line 2681  get_othercase_range(unsigned int *cptr,
2681  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2682    
2683  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2684    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2685    
2686  if (c > d) return FALSE;  if (c > d) return FALSE;
2687    
# Line 1972  next = othercase + 1; Line 2690  next = othercase + 1;
2690    
2691  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2692    {    {
2693    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2694    next++;    next++;
2695    }    }
2696    
# Line 1981  for (++c; c <= d; c++) Line 2699  for (++c; c <= d; c++)
2699    
2700  return TRUE;  return TRUE;
2701  }  }
2702    
2703    
2704    
2705    /*************************************************
2706    *        Check a character and a property        *
2707    *************************************************/
2708    
2709    /* This function is called by check_auto_possessive() when a property item
2710    is adjacent to a fixed character.
2711    
2712    Arguments:
2713      c            the character
2714      ptype        the property type
2715      pdata        the data for the type
2716      negated      TRUE if it's a negated property (\P or \p{^)
2717    
2718    Returns:       TRUE if auto-possessifying is OK
2719    */
2720    
2721    static BOOL
2722    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2723    {
2724    const ucd_record *prop = GET_UCD(c);
2725    switch(ptype)
2726      {
2727      case PT_LAMP:
2728      return (prop->chartype == ucp_Lu ||
2729              prop->chartype == ucp_Ll ||
2730              prop->chartype == ucp_Lt) == negated;
2731    
2732      case PT_GC:
2733      return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;
2734    
2735      case PT_PC:
2736      return (pdata == prop->chartype) == negated;
2737    
2738      case PT_SC:
2739      return (pdata == prop->script) == negated;
2740    
2741      /* These are specials */
2742    
2743      case PT_ALNUM:
2744      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2745              _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;
2746    
2747      case PT_SPACE:    /* Perl space */
2748      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2749              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2750              == negated;
2751    
2752      case PT_PXSPACE:  /* POSIX space */
2753      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2754              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2755              c == CHAR_FF || c == CHAR_CR)
2756              == negated;
2757    
2758      case PT_WORD:
2759      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2760              _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2761              c == CHAR_UNDERSCORE) == negated;
2762      }
2763    return FALSE;
2764    }
2765  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2766    
2767    
# Line 1994  whether the next thing could possibly ma Line 2775  whether the next thing could possibly ma
2775  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
2776    
2777  Arguments:  Arguments:
2778    op_code       the repeated op code    previous      pointer to the repeated opcode
   this          data for this item, depends on the opcode  
2779    utf8          TRUE in UTF-8 mode    utf8          TRUE in UTF-8 mode
   utf8_char     used for utf8 character bytes, NULL if not relevant  
2780    ptr           next character in pattern    ptr           next character in pattern
2781    options       options bits    options       options bits
2782    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 2006  Returns: TRUE if possessifying is Line 2785  Returns: TRUE if possessifying is
2785  */  */
2786    
2787  static BOOL  static BOOL
2788  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
2789    const uschar *ptr, int options, compile_data *cd)    int options, compile_data *cd)
2790  {  {
2791  int next;  int c, next;
2792    int op_code = *previous++;
2793    
2794  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
2795    
# Line 2018  if ((options & PCRE_EXTENDED) != 0) Line 2798  if ((options & PCRE_EXTENDED) != 0)
2798    for (;;)    for (;;)
2799      {      {
2800      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2801      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2802        {        {
2803        while (*(++ptr) != 0)        ptr++;
2804          while (*ptr != 0)
2805            {
2806          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2807            ptr++;
2808    #ifdef SUPPORT_UTF8
2809            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2810    #endif
2811            }
2812        }        }
2813      else break;      else break;
2814      }      }
# Line 2030  if ((options & PCRE_EXTENDED) != 0) Line 2817  if ((options & PCRE_EXTENDED) != 0)
2817  /* 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
2818  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2819    
2820  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2821    {    {
2822    int temperrorcode = 0;    int temperrorcode = 0;
2823    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 2842  if ((options & PCRE_EXTENDED) != 0)
2842    for (;;)    for (;;)
2843      {      {
2844      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2845      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2846        {        {
2847        while (*(++ptr) != 0)        ptr++;
2848          while (*ptr != 0)
2849            {
2850          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2851            ptr++;
2852    #ifdef SUPPORT_UTF8
2853            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2854    #endif
2855            }
2856        }        }
2857      else break;      else break;
2858      }      }
# Line 2066  if ((options & PCRE_EXTENDED) != 0) Line 2860  if ((options & PCRE_EXTENDED) != 0)
2860    
2861  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
2862    
2863  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2864    return FALSE;    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2865        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. */  
   
2866    
2867  /* Handle cases when the next item is a character. */  /* Now compare the next item with the previous opcode. First, handle cases when
2868    the next item is a character. */
2869    
2870  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
2871    {    {
2872    case OP_CHAR:    case OP_CHAR:
2873  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2874    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2875    #else
2876      c = *previous;
2877  #endif  #endif
2878    return item != next;    return c != next;
2879    
2880    /* 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
2881    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
2882    high-valued characters. */    high-valued characters. */
2883    
2884    case OP_CHARNC:    case OP_CHARI:
2885  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2886    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2887    #else
2888      c = *previous;
2889  #endif  #endif
2890    if (item == next) return FALSE;    if (c == next) return FALSE;
2891  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2892    if (utf8)    if (utf8)
2893      {      {
2894      unsigned int othercase;      unsigned int othercase;
2895      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2896  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2897      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
2898  #else  #else
2899      othercase = NOTACHAR;      othercase = NOTACHAR;
2900  #endif  #endif
2901      return (unsigned int)item != othercase;      return (unsigned int)c != othercase;
2902      }      }
2903    else    else
2904  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2905    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2906    
2907    /* 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
2908      opcodes are not used for multi-byte characters, because they are coded using
2909      an XCLASS instead. */
2910    
2911    case OP_NOT:    case OP_NOT:
2912    if (next < 0) return FALSE;  /* Not a character */    return (c = *previous) == next;
2913    if (item == next) return TRUE;  
2914    if ((options & PCRE_CASELESS) == 0) return FALSE;    case OP_NOTI:
2915      if ((c = *previous) == next) return TRUE;
2916  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2917    if (utf8)    if (utf8)
2918      {      {
2919      unsigned int othercase;      unsigned int othercase;
2920      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2921  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2922      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
2923  #else  #else
2924      othercase = NOTACHAR;      othercase = NOTACHAR;
2925  #endif  #endif
2926      return (unsigned int)item == othercase;      return (unsigned int)c == othercase;
2927      }      }
2928    else    else
2929  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2930    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2931    
2932      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2933      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
2934    
2935    case OP_DIGIT:    case OP_DIGIT:
2936    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 2973  if (next >= 0) switch(op_code)
2973      case 0x202f:      case 0x202f:
2974      case 0x205f:      case 0x205f:
2975      case 0x3000:      case 0x3000:
2976      return op_code != OP_HSPACE;      return op_code == OP_NOT_HSPACE;
2977      default:      default:
2978      return op_code == OP_HSPACE;      return op_code != OP_NOT_HSPACE;
2979      }      }
2980    
2981      case OP_ANYNL:
2982    case OP_VSPACE:    case OP_VSPACE:
2983    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
2984    switch(next)    switch(next)
# Line 2189  if (next >= 0) switch(op_code) Line 2990  if (next >= 0) switch(op_code)
2990      case 0x85:      case 0x85:
2991      case 0x2028:      case 0x2028:
2992      case 0x2029:      case 0x2029:
2993      return op_code != OP_VSPACE;      return op_code == OP_NOT_VSPACE;
2994      default:      default:
2995      return op_code == OP_VSPACE;      return op_code != OP_NOT_VSPACE;
2996      }      }
2997    
2998    #ifdef SUPPORT_UCP
2999      case OP_PROP:
3000      return check_char_prop(next, previous[0], previous[1], FALSE);
3001    
3002      case OP_NOTPROP:
3003      return check_char_prop(next, previous[0], previous[1], TRUE);
3004    #endif
3005    
3006    default:    default:
3007    return FALSE;    return FALSE;
3008    }    }
3009    
3010    
3011  /* 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
3012    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
3013    generated only when PCRE_UCP is *not* set, that is, when only ASCII
3014    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
3015    replaced by OP_PROP codes when PCRE_UCP is set. */
3016    
3017  switch(op_code)  switch(op_code)
3018    {    {
3019    case OP_CHAR:    case OP_CHAR:
3020    case OP_CHARNC:    case OP_CHARI:
3021  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3022    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
3023    #else
3024      c = *previous;
3025  #endif  #endif
3026    switch(-next)    switch(-next)
3027      {      {
3028      case ESC_d:      case ESC_d:
3029      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
3030    
3031      case ESC_D:      case ESC_D:
3032      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
3033    
3034      case ESC_s:      case ESC_s:
3035      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
3036    
3037      case ESC_S:      case ESC_S:
3038      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
3039    
3040      case ESC_w:      case ESC_w:
3041      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
3042    
3043      case ESC_W:      case ESC_W:
3044      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
3045    
3046      case ESC_h:      case ESC_h:
3047      case ESC_H:      case ESC_H:
3048      switch(item)      switch(c)
3049        {        {
3050        case 0x09:        case 0x09:
3051        case 0x20:        case 0x20:
# Line 2258  switch(op_code) Line 3073  switch(op_code)
3073    
3074      case ESC_v:      case ESC_v:
3075      case ESC_V:      case ESC_V:
3076      switch(item)      switch(c)
3077        {        {
3078        case 0x0a:        case 0x0a:
3079        case 0x0b:        case 0x0b:
# Line 2272  switch(op_code) Line 3087  switch(op_code)
3087        return -next == ESC_v;        return -next == ESC_v;
3088        }        }
3089    
3090        /* When PCRE_UCP is set, these values get generated for \d etc. Find
3091        their substitutions and process them. The result will always be either
3092        -ESC_p or -ESC_P. Then fall through to process those values. */
3093    
3094    #ifdef SUPPORT_UCP
3095        case ESC_du:
3096        case ESC_DU:
3097        case ESC_wu:
3098        case ESC_WU:
3099        case ESC_su:
3100        case ESC_SU:
3101          {
3102          int temperrorcode = 0;
3103          ptr = substitutes[-next - ESC_DU];
3104          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
3105          if (temperrorcode != 0) return FALSE;
3106          ptr++;    /* For compatibility */
3107          }
3108        /* Fall through */
3109    
3110        case ESC_p:
3111        case ESC_P:
3112          {
3113          int ptype, pdata, errorcodeptr;
3114          BOOL negated;
3115    
3116          ptr--;      /* Make ptr point at the p or P */
3117          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
3118          if (ptype < 0) return FALSE;
3119          ptr++;      /* Point past the final curly ket */
3120    
3121          /* If the property item is optional, we have to give up. (When generated
3122          from \d etc by PCRE_UCP, this test will have been applied much earlier,
3123          to the original \d etc. At this point, ptr will point to a zero byte. */
3124    
3125          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3126            strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3127              return FALSE;
3128    
3129          /* Do the property check. */
3130    
3131          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
3132          }
3133    #endif
3134    
3135      default:      default:
3136      return FALSE;      return FALSE;
3137      }      }
3138    
3139      /* In principle, support for Unicode properties should be integrated here as
3140      well. It means re-organizing the above code so as to get hold of the property
3141      values before switching on the op-code. However, I wonder how many patterns
3142      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
3143      these op-codes are never generated.) */
3144    
3145    case OP_DIGIT:    case OP_DIGIT:
3146    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
3147           next == -ESC_h || next == -ESC_v;           next == -ESC_h || next == -ESC_v || next == -ESC_R;
3148    
3149    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
3150    return next == -ESC_d;    return next == -ESC_d;
3151    
3152    case OP_WHITESPACE:    case OP_WHITESPACE:
3153    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
3154    
3155    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
3156    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
3157    
3158    case OP_HSPACE:    case OP_HSPACE:
3159    return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
3160             next == -ESC_w || next == -ESC_v || next == -ESC_R;
3161    
3162    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
3163    return next == -ESC_h;    return next == -ESC_h;
3164    
3165    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
3166      case OP_ANYNL:
3167    case OP_VSPACE:    case OP_VSPACE:
3168    return next == -ESC_V || next == -ESC_d || next == -ESC_w;    return next == -ESC_V || next == -ESC_d || next == -ESC_w;
3169    
3170    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
3171    return next == -ESC_v;    return next == -ESC_v || next == -ESC_R;
3172    
3173    case OP_WORDCHAR:    case OP_WORDCHAR:
3174    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
3175             next == -ESC_v || next == -ESC_R;
3176    
3177    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
3178    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2335  Arguments: Line 3204  Arguments:
3204    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
3205    reqbyteptr     set to the last literal character required, else < 0    reqbyteptr     set to the last literal character required, else < 0
3206    bcptr          points to current branch chain    bcptr          points to current branch chain
3207      cond_depth     conditional nesting depth
3208    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
3209    lengthptr      NULL during the real compile phase    lengthptr      NULL during the real compile phase
3210                   points to length accumulator during pre-compile phase                   points to length accumulator during pre-compile phase
# Line 2346  Returns: TRUE on success Line 3216  Returns: TRUE on success
3216  static BOOL  static BOOL
3217  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
3218    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
3219    compile_data *cd, int *lengthptr)    int cond_depth, compile_data *cd, int *lengthptr)
3220  {  {
3221  int repeat_type, op_type;  int repeat_type, op_type;
3222  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
# Line 2355  int greedy_default, greedy_non_default; Line 3225  int greedy_default, greedy_non_default;
3225  int firstbyte, reqbyte;  int firstbyte, reqbyte;
3226  int zeroreqbyte, zerofirstbyte;  int zeroreqbyte, zerofirstbyte;
3227  int req_caseopt, reqvary, tempreqvary;  int req_caseopt, reqvary, tempreqvary;
3228  int options = *optionsptr;  int options = *optionsptr;               /* May change dynamically */
3229  int after_manual_callout = 0;  int after_manual_callout = 0;
3230  int length_prevgroup = 0;  int length_prevgroup = 0;
3231  register int c;  register int c;
# Line 2367  BOOL inescq = FALSE; Line 3237  BOOL inescq = FALSE;
3237  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
3238  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
3239  const uschar *tempptr;  const uschar *tempptr;
3240    const uschar *nestptr = NULL;
3241  uschar *previous = NULL;  uschar *previous = NULL;
3242  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
3243  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
3244  uschar classbits[32];  uschar classbits[32];
3245    
3246    /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3247    must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3248    dynamically as we process the pattern. */
3249    
3250  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3251  BOOL class_utf8;  BOOL class_utf8;
3252  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
# Line 2380  uschar *class_utf8data_base; Line 3255  uschar *class_utf8data_base;
3255  uschar utf8_char[6];  uschar utf8_char[6];
3256  #else  #else
3257  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
 uschar *utf8_char = NULL;  
3258  #endif  #endif
3259    
3260  #ifdef DEBUG  #ifdef PCRE_DEBUG
3261  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
3262  #endif  #endif
3263    
# Line 2431  for (;; ptr++) Line 3305  for (;; ptr++)
3305    int subfirstbyte;    int subfirstbyte;
3306    int terminator;    int terminator;
3307    int mclength;    int mclength;
3308      int tempbracount;
3309    uschar mcbuffer[8];    uschar mcbuffer[8];
3310    
3311    /* Get next byte in the pattern */    /* Get next byte in the pattern */
3312    
3313    c = *ptr;    c = *ptr;
3314    
3315      /* If we are at the end of a nested substitution, revert to the outer level
3316      string. Nesting only happens one level deep. */
3317    
3318      if (c == 0 && nestptr != NULL)
3319        {
3320        ptr = nestptr;
3321        nestptr = NULL;
3322        c = *ptr;
3323        }
3324    
3325    /* 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
3326    previous cycle of this loop. */    previous cycle of this loop. */
3327    
3328    if (lengthptr != NULL)    if (lengthptr != NULL)
3329      {      {
3330  #ifdef DEBUG  #ifdef PCRE_DEBUG
3331      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3332  #endif  #endif
3333      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
3334        {        {
3335        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3336        goto FAILED;        goto FAILED;
# Line 2467  for (;; ptr++) Line 3352  for (;; ptr++)
3352        goto FAILED;        goto FAILED;
3353        }        }
3354    
3355      *lengthptr += code - last_code;      *lengthptr += (int)(code - last_code);
3356      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, (int)(code - last_code),
3357          c));
3358    
3359      /* 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
3360      it back to there, in order to avoid filling up the work space. Otherwise,      it back to there, in order to avoid filling up the work space. Otherwise,
# Line 2494  for (;; ptr++) Line 3380  for (;; ptr++)
3380    /* 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
3381    reference list. */    reference list. */
3382    
3383    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
3384      {      {
3385      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3386      goto FAILED;      goto FAILED;
# Line 2504  for (;; ptr++) Line 3390  for (;; ptr++)
3390    
3391    if (inescq && c != 0)    if (inescq && c != 0)
3392      {      {
3393      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3394        {        {
3395        inescq = FALSE;        inescq = FALSE;
3396        ptr++;        ptr++;
# Line 2530  for (;; ptr++) Line 3416  for (;; ptr++)
3416    /* 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
3417    a quantifier. */    a quantifier. */
3418    
3419    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
3420      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
3421        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
3422    
3423    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
3424         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2541  for (;; ptr++) Line 3428  for (;; ptr++)
3428      previous_callout = NULL;      previous_callout = NULL;
3429      }      }
3430    
3431    /* In extended mode, skip white space and comments */    /* In extended mode, skip white space and comments. */
3432    
3433    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3434      {      {
3435      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
3436      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
3437        {        {
3438        while (*(++ptr) != 0)        ptr++;
3439          while (*ptr != 0)
3440          {          {
3441          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3442            ptr++;
3443    #ifdef SUPPORT_UTF8
3444            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
3445    #endif
3446          }          }
3447        if (*ptr != 0) continue;        if (*ptr != 0) continue;
3448    
# Line 2571  for (;; ptr++) Line 3463  for (;; ptr++)
3463      {      {
3464      /* ===================================================================*/      /* ===================================================================*/
3465      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
3466      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
3467      case ')':      case CHAR_RIGHT_PARENTHESIS:
3468      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
3469      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
3470      *codeptr = code;      *codeptr = code;
# Line 2584  for (;; ptr++) Line 3476  for (;; ptr++)
3476          *errorcodeptr = ERR20;          *errorcodeptr = ERR20;
3477          goto FAILED;          goto FAILED;
3478          }          }
3479        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += (int)(code - last_code);   /* To include callout length */
3480        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
3481        }        }
3482      return TRUE;      return TRUE;
# Line 2594  for (;; ptr++) Line 3486  for (;; ptr++)
3486      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
3487      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3488    
3489      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
3490        previous = NULL;
3491      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3492        {        {
3493        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3494          *code++ = OP_CIRCM;
3495        }        }
3496      previous = NULL;      else *code++ = OP_CIRC;
     *code++ = OP_CIRC;  
3497      break;      break;
3498    
3499      case '$':      case CHAR_DOLLAR_SIGN:
3500      previous = NULL;      previous = NULL;
3501      *code++ = OP_DOLL;      *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
3502      break;      break;
3503    
3504      /* 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
3505      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
3506    
3507      case '.':      case CHAR_DOT:
3508      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3509      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
3510      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3511      previous = code;      previous = code;
3512      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
3513      break;      break;
3514    
3515    
# Line 2631  for (;; ptr++) Line 3524  for (;; ptr++)
3524      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,
3525      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
3526      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.
     */  
3527    
3528      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
3529        default (Perl) mode, it is treated as a data character. */
3530    
3531        case CHAR_RIGHT_SQUARE_BRACKET:
3532        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3533          {
3534          *errorcodeptr = ERR64;
3535          goto FAILED;
3536          }
3537        goto NORMAL_CHAR;
3538    
3539        case CHAR_LEFT_SQUARE_BRACKET:
3540      previous = code;      previous = code;
3541    
3542      /* 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
3543      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. */
3544    
3545      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3546             ptr[1] == CHAR_EQUALS_SIGN) &&
3547          check_posix_syntax(ptr, &tempptr))          check_posix_syntax(ptr, &tempptr))
3548        {        {
3549        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
3550        goto FAILED;        goto FAILED;
3551        }        }
3552    
# Line 2654  for (;; ptr++) Line 3558  for (;; ptr++)
3558      for (;;)      for (;;)
3559        {        {
3560        c = *(++ptr);        c = *(++ptr);
3561        if (c == '\\')        if (c == CHAR_BACKSLASH)
3562          {          {
3563          if (ptr[1] == 'E') ptr++;          if (ptr[1] == CHAR_E)
3564            else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;            ptr++;
3565              else break;          else if (strncmp((const char *)ptr+1,
3566                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
3567              ptr += 3;
3568            else
3569              break;
3570          }          }
3571        else if (!negate_class && c == '^')        else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
3572          negate_class = TRUE;          negate_class = TRUE;
3573        else break;        else break;
3574        }        }
3575    
3576        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
3577        an initial ']' is taken as a data character -- the code below handles
3578        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3579        [^] must match any character, so generate OP_ALLANY. */
3580    
3581        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3582            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3583          {
3584          *code++ = negate_class? OP_ALLANY : OP_FAIL;
3585          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3586          zerofirstbyte = firstbyte;
3587          break;
3588          }
3589    
3590      /* 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
3591      negation flag at the end, so that support for characters > 255 works      negation flag at the end, so that support for characters > 255 works
3592      correctly (they are all included in the class). */      correctly (they are all included in the class). */
# Line 2688  for (;; ptr++) Line 3610  for (;; ptr++)
3610  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3611      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
3612      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
3613      class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */      class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
3614  #endif  #endif
3615    
3616      /* 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 3626  for (;; ptr++)
3626          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3627          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3628          }          }
3629    
3630        /* 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
3631        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
3632        contain a zillion UTF-8 characters no longer overwrite the work space        contain a zillion UTF-8 characters no longer overwrite the work space
3633        (which is on the stack). */        (which is on the stack). */
3634    
3635        if (lengthptr != NULL)        if (lengthptr != NULL)
3636          {          {
3637          *lengthptr += class_utf8data - class_utf8data_base;          *lengthptr += class_utf8data - class_utf8data_base;
3638          class_utf8data = class_utf8data_base;          class_utf8data = class_utf8data_base;
3639          }          }
3640    
3641  #endif  #endif
3642    
3643        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
3644    
3645        if (inescq)        if (inescq)
3646          {          {
3647          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3648            {            {
3649            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
3650            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2737  for (;; ptr++) Line 3659  for (;; ptr++)
3659        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3660        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3661    
3662        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3663            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3664            check_posix_syntax(ptr, &tempptr))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3665          {          {
3666          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3667          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3668          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3669          uschar pbits[32];          uschar pbits[32];
3670    
3671          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3672            {            {
3673            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3674            goto FAILED;            goto FAILED;
3675            }            }
3676    
3677          ptr += 2;          ptr += 2;
3678          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3679            {            {
3680            local_negate = TRUE;            local_negate = TRUE;
3681            should_flip_negation = TRUE;  /* Note negative special */            should_flip_negation = TRUE;  /* Note negative special */
3682            ptr++;            ptr++;
3683            }            }
3684    
3685          posix_class = check_posix_name(ptr, tempptr - ptr);          posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3686          if (posix_class < 0)          if (posix_class < 0)
3687            {            {
3688            *errorcodeptr = ERR30;            *errorcodeptr = ERR30;
# Line 2774  for (;; ptr++) Line 3696  for (;; ptr++)
3696          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3697            posix_class = 0;            posix_class = 0;
3698    
3699          /* 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
3700          because we may be adding and subtracting from it, and we don't want to          different escape sequences that use Unicode properties. */
3701          subtract bits that may be in the main map already. At the end we or the  
3702          result into the bit map that is being built. */  #ifdef SUPPORT_UCP
3703            if ((options & PCRE_UCP) != 0)
3704              {
3705              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3706              if (posix_substitutes[pc] != NULL)
3707                {
3708                nestptr = tempptr + 1;
3709                ptr = posix_substitutes[pc] - 1;
3710                continue;
3711                }
3712              }
3713    #endif
3714            /* In the non-UCP case, we build the bit map for the POSIX class in a
3715            chunk of local store because we may be adding and subtracting from it,
3716            and we don't want to subtract bits that may be in the main map already.
3717            At the end we or the result into the bit map that is being built. */
3718    
3719          posix_class *= 3;          posix_class *= 3;
3720    
# Line 2821  for (;; ptr++) Line 3758  for (;; ptr++)
3758    
3759        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3760        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
3761        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
3762        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
3763        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
3764        character in them, so set class_charcount bigger than one. */        are either treated as literal characters (by default), or are faulted if
3765          PCRE_EXTRA is set. */
3766    
3767        if (c == '\\')        if (c == CHAR_BACKSLASH)
3768          {          {
3769          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3770          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3771    
3772          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 */  
3773          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3774            {            {
3775            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3776              {              {
3777              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3778              }              }
# Line 2850  for (;; ptr++) Line 3786  for (;; ptr++)
3786            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
3787            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3788    
3789            /* Save time by not doing this in the pre-compile phase. */            switch (-c)
   
           if (lengthptr == NULL) switch (-c)  
3790              {              {
3791    #ifdef SUPPORT_UCP
3792                case ESC_du:     /* These are the values given for \d etc */
3793                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3794                case ESC_wu:     /* escape sequence with an appropriate \p */
3795                case ESC_WU:     /* or \P to test Unicode properties instead */
3796                case ESC_su:     /* of the default ASCII testing. */
3797                case ESC_SU:
3798                nestptr = ptr;
3799                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3800                class_charcount -= 2;                /* Undo! */
3801                continue;
3802    #endif
3803              case ESC_d:              case ESC_d:
3804              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3805              continue;              continue;
# Line 2872  for (;; ptr++) Line 3818  for (;; ptr++)
3818              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3819              continue;              continue;
3820    
3821                /* Perl 5.004 onwards omits VT from \s, but we must preserve it
3822                if it was previously set by something earlier in the character
3823                class. */
3824    
3825              case ESC_s:              case ESC_s:
3826              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];              classbits[0] |= cbits[cbit_space];
3827              classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= cbits[cbit_space+1] & ~0x08;
3828                for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3829              continue;              continue;
3830    
3831              case ESC_S:              case ESC_S:
# Line 2883  for (;; ptr++) Line 3834  for (;; ptr++)
3834              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3835              continue;              continue;
3836    
3837              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)  
             {  
3838              SETBIT(classbits, 0x09); /* VT */              SETBIT(classbits, 0x09); /* VT */
3839              SETBIT(classbits, 0x20); /* SPACE */              SETBIT(classbits, 0x20); /* SPACE */
3840              SETBIT(classbits, 0xa0); /* NSBP */              SETBIT(classbits, 0xa0); /* NSBP */
# Line 2920  for (;; ptr++) Line 3858  for (;; ptr++)
3858                }                }
3859  #endif  #endif
3860              continue;              continue;
             }  
3861    
3862            if (-c == ESC_H)              case ESC_H:
             {  
3863              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3864                {                {
3865                int x = 0xff;                int x = 0xff;
# Line 2965  for (;; ptr++) Line 3901  for (;; ptr++)
3901                }                }
3902  #endif  #endif
3903              continue;              continue;
             }  
3904    
3905            if (-c == ESC_v)              case ESC_v:
             {  
3906              SETBIT(classbits, 0x0a); /* LF */              SETBIT(classbits, 0x0a); /* LF */
3907              SETBIT(classbits, 0x0b); /* VT */              SETBIT(classbits, 0x0b); /* VT */
3908              SETBIT(classbits, 0x0c); /* FF */              SETBIT(classbits, 0x0c); /* FF */
# Line 2984  for (;; ptr++) Line 3918  for (;; ptr++)
3918                }                }
3919  #endif  #endif
3920              continue;              continue;
             }  
3921    
3922            if (-c == ESC_V)              case ESC_V:
             {  
3923              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3924                {                {
3925                int x = 0xff;                int x = 0xff;
# Line 3017  for (;; ptr++) Line 3949  for (;; ptr++)
3949                }                }
3950  #endif  #endif
3951              continue;              continue;
             }  
   
           /* We need to deal with \P and \p in both phases. */  
3952    
3953  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3954            if (-c == ESC_p || -c == ESC_P)              case ESC_p:
3955              {              case ESC_P:
3956              BOOL negated;                {
3957              int pdata;                BOOL negated;
3958              int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                int pdata;
3959              if (ptype < 0) goto FAILED;                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3960              class_utf8 = TRUE;                if (ptype < 0) goto FAILED;
3961              *class_utf8data++ = ((-c == ESC_p) != negated)?                class_utf8 = TRUE;
3962                XCL_PROP : XCL_NOTPROP;                *class_utf8data++ = ((-c == ESC_p) != negated)?
3963              *class_utf8data++ = ptype;                  XCL_PROP : XCL_NOTPROP;
3964              *class_utf8data++ = pdata;                *class_utf8data++ = ptype;
3965              class_charcount -= 2;   /* Not a < 256 character */                *class_utf8data++ = pdata;
3966              continue;                class_charcount -= 2;   /* Not a < 256 character */
3967              }                continue;
3968                  }
3969  #endif  #endif
3970            /* Unrecognized escapes are faulted if PCRE is running in its              /* Unrecognized escapes are faulted if PCRE is running in its
3971            strict mode. By default, for compatibility with Perl, they are              strict mode. By default, for compatibility with Perl, they are
3972            treated as literals. */              treated as literals. */
3973    
3974            if ((options & PCRE_EXTRA) != 0)              default:
3975              {              if ((options & PCRE_EXTRA) != 0)
3976              *errorcodeptr = ERR7;                {
3977              goto FAILED;                *errorcodeptr = ERR7;
3978                  goto FAILED;
3979                  }
3980                class_charcount -= 2;  /* Undo the default count from above */
3981                c = *ptr;              /* Get the final character and fall through */
3982                break;
3983              }              }
   
           class_charcount -= 2;  /* Undo the default count from above */  
           c = *ptr;              /* Get the final character and fall through */  
3984            }            }
3985    
3986          /* 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 3994  for (;; ptr++)
3994        entirely. The code for handling \Q and \E is messy. */        entirely. The code for handling \Q and \E is messy. */
3995    
3996        CHECK_RANGE:        CHECK_RANGE:
3997        while (ptr[1] == '\\' && ptr[2] == 'E')        while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3998          {          {
3999          inescq = FALSE;          inescq = FALSE;
4000          ptr += 2;          ptr += 2;
# Line 3072  for (;; ptr++) Line 4004  for (;; ptr++)
4004    
4005        /* Remember \r or \n */        /* Remember \r or \n */
4006    
4007        if (c == '\r' || c == '\n') cd->external_flags |= PCRE_HASCRORLF;        if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
4008    
4009        /* Check for range */        /* Check for range */
4010    
4011        if (!inescq && ptr[1] == '-')        if (!inescq && ptr[1] == CHAR_MINUS)
4012          {          {
4013          int d;          int d;
4014          ptr += 2;          ptr += 2;
4015          while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
4016    
4017          /* 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
4018          mode. */          mode. */
4019    
4020          while (*ptr == '\\' && ptr[1] == 'Q')          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
4021            {            {
4022            ptr += 2;            ptr += 2;
4023            if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }            if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
4024                { ptr += 2; continue; }
4025            inescq = TRUE;            inescq = TRUE;
4026            break;            break;
4027            }            }
4028    
4029          if (*ptr == 0 || (!inescq && *ptr == ']'))          if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
4030            {            {
4031            ptr = oldptr;            ptr = oldptr;
4032            goto LONE_SINGLE_CHARACTER;            goto LONE_SINGLE_CHARACTER;
# Line 3112  for (;; ptr++) Line 4045  for (;; ptr++)
4045          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
4046          in such circumstances. */          in such circumstances. */
4047    
4048          if (!inescq && d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
4049            {            {
4050            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
4051            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
4052    
4053            /* \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 */  
4054    
4055            if (d < 0)            if (d < 0)
4056              {              {
4057              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  
4058                {                {
4059                ptr = oldptr;                ptr = oldptr;
4060                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 3145  for (;; ptr++) Line 4075  for (;; ptr++)
4075    
4076          /* Remember \r or \n */          /* Remember \r or \n */
4077    
4078          if (d == '\r' || d == '\n') cd->external_flags |= PCRE_HASCRORLF;          if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
4079    
4080          /* 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
4081          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 4195  for (;; ptr++)
4195          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
4196            {            {
4197            unsigned int othercase;            unsigned int othercase;
4198            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)            if ((othercase = UCD_OTHERCASE(c)) != c)
4199              {              {
4200              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
4201              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 3290  for (;; ptr++) Line 4220  for (;; ptr++)
4220          }          }
4221        }        }
4222    
4223      /* 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.
4224        If we are at the end of an internal nested string, revert to the outer
4225        string. */
4226    
4227        while (((c = *(++ptr)) != 0 ||
4228               (nestptr != NULL &&
4229                 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != 0)) &&
4230               (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
4231    
4232      while ((c = *(++ptr)) != 0 && (c != ']' || inescq));      /* Check for missing terminating ']' */
4233    
4234      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)
4235        {        {
4236        *errorcodeptr = ERR6;        *errorcodeptr = ERR6;
4237        goto FAILED;        goto FAILED;
4238        }        }
4239    
   
 /* 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  
   
   
4240      /* 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
4241      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
4242      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 4244  we set the flag only if there is a liter
4244    
4245      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
4246      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
4247      operate on single-bytes only. This is an historical hangover. Maybe one day      operate on single-bytes characters only. This is an historical hangover.
4248      we can tidy these opcodes to handle multi-byte characters.      Maybe one day we can tidy these opcodes to handle multi-byte characters.
4249    
4250      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
4251      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.
4252      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
4253      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
4254      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
4255      reqbyte, save the previous value for reinstating. */      of reqbyte, save the previous value for reinstating. */
4256    
4257  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4258      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 4263  we set the flag only if there is a liter
4263        {        {
4264        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
4265    
4266        /* The OP_NOT opcode works on one-byte characters only. */        /* The OP_NOT[I] opcodes work on one-byte characters only. */
4267    
4268        if (negate_class)        if (negate_class)
4269          {          {
4270          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
4271          zerofirstbyte = firstbyte;          zerofirstbyte = firstbyte;
4272          *code++ = OP_NOT;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
4273          *code++ = class_lastchar;          *code++ = class_lastchar;
4274          break;          break;
4275          }          }
# Line 3382  we set the flag only if there is a liter Line 4300  we set the flag only if there is a liter
4300    
4301      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
4302      extended class, with its own opcode, unless there was a negated special      extended class, with its own opcode, unless there was a negated special
4303      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
4304      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
4305      (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
4306      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
4307        actual compiled code. */
4308    
4309  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4310      if (class_utf8 && !should_flip_negation)      if (class_utf8 && (!should_flip_negation || (options & PCRE_UCP) != 0))
4311        {        {
4312        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
4313        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 3414  we set the flag only if there is a liter Line 4333  we set the flag only if there is a liter
4333        }        }
4334  #endif  #endif
4335    
4336      /* 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
4337      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
4338      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
4339      map into the code vector, negating it if necessary. */      (non-UCP) in the class. Then copy the 32-byte map into the code vector,
4340        negating it if necessary. */
4341    
4342      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
4343      if (negate_class)      if (negate_class)
# Line 3437  we set the flag only if there is a liter Line 4357  we set the flag only if there is a liter
4357      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
4358      has been tested above. */      has been tested above. */
4359    
4360      case '{':      case CHAR_LEFT_CURLY_BRACKET:
4361      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
4362      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
4363      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
4364      goto REPEAT;      goto REPEAT;
4365    
4366      case '*':      case CHAR_ASTERISK:
4367      repeat_min = 0;      repeat_min = 0;
4368      repeat_max = -1;      repeat_max = -1;
4369      goto REPEAT;      goto REPEAT;
4370    
4371      case '+':      case CHAR_PLUS:
4372      repeat_min = 1;      repeat_min = 1;
4373      repeat_max = -1;      repeat_max = -1;
4374      goto REPEAT;      goto REPEAT;
4375    
4376      case '?':      case CHAR_QUESTION_MARK:
4377      repeat_min = 0;      repeat_min = 0;
4378      repeat_max = 1;      repeat_max = 1;
4379    
# Line 3477  we set the flag only if there is a liter Line 4397  we set the flag only if there is a liter
4397      op_type = 0;                    /* Default single-char op codes */      op_type = 0;                    /* Default single-char op codes */
4398      possessive_quantifier = FALSE;  /* Default not possessive quantifier */      possessive_quantifier = FALSE;  /* Default not possessive quantifier */
4399    
4400      /* Save start of previous item, in case we have to move it up to make space      /* Save start of previous item, in case we have to move it up in order to
4401      for an inserted OP_ONCE for the additional '+' extension. */      insert something before it. */
4402    
4403      tempcode = previous;      tempcode = previous;
4404    
# Line 3488  we set the flag only if there is a liter Line 4408  we set the flag only if there is a liter
4408      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
4409      repeat type to the non-default. */      repeat type to the non-default. */
4410    
4411      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
4412        {        {
4413        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
4414        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
4415        ptr++;        ptr++;
4416        }        }
4417      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
4418        {        {
4419        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
4420        ptr++;        ptr++;
4421        }        }
4422      else repeat_type = greedy_default;      else repeat_type = greedy_default;
4423    
4424        /* If previous was a recursion call, wrap it in atomic brackets so that
4425        previous becomes the atomic group. All recursions were so wrapped in the
4426        past, but it no longer happens for non-repeated recursions. In fact, the
4427        repeated ones could be re-implemented independently so as not to need this,
4428        but for the moment we rely on the code for repeating groups. */
4429    
4430        if (*previous == OP_RECURSE)
4431          {
4432          memmove(previous + 1 + LINK_SIZE, previous, 1 + LINK_SIZE);
4433          *previous = OP_ONCE;
4434          PUT(previous, 1, 2 + 2*LINK_SIZE);
4435          previous[2 + 2*LINK_SIZE] = OP_KET;
4436          PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
4437          code += 2 + 2 * LINK_SIZE;
4438          length_prevgroup = 3 + 3*LINK_SIZE;
4439    
4440          /* When actually compiling, we need to check whether this was a forward
4441          reference, and if so, adjust the offset. */
4442    
4443          if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
4444            {
4445            int offset = GET(cd->hwm, -LINK_SIZE);
4446            if (offset == previous + 1 - cd->start_code)
4447              PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
4448            }
4449          }
4450    
4451        /* Now handle repetition for the different types of item. */
4452    
4453      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
4454      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
4455      that it is set in reqbyte - it might not be if a sequence such as x{3} is      that it is set in reqbyte - it might not be if a sequence such as x{3} is
4456      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
4457      instead.  */      instead.  */
4458    
4459      if (*previous == OP_CHAR || *previous == OP_CHARNC)      if (*previous == OP_CHAR || *previous == OP_CHARI)
4460        {        {
4461          op_type = (*previous == OP_CHAR)? 0 : OP_STARI - OP_STAR;
4462    
4463        /* 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
4464        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
4465        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 4492  we set the flag only if there is a liter
4492    
4493        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4494            repeat_max < 0 &&            repeat_max < 0 &&
4495            check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
             options, cd))  
4496          {          {
4497          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4498          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3554  we set the flag only if there is a liter Line 4504  we set the flag only if there is a liter
4504      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
4505      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-
4506      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
4507      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
4508      currently used only for single-byte chars. */      are currently used only for single-byte chars. */
4509    
4510      else if (*previous == OP_NOT)      else if (*previous == OP_NOT || *previous == OP_NOTI)
4511        {        {
4512        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = ((*previous == OP_NOT)? OP_NOTSTAR : OP_NOTSTARI) - OP_STAR;
4513        c = previous[1];        c = previous[1];
4514        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4515            repeat_max < 0 &&            repeat_max < 0 &&
4516            check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4517          {          {
4518          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4519          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3587  we set the flag only if there is a liter Line 4537  we set the flag only if there is a liter
4537    
4538        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4539            repeat_max < 0 &&            repeat_max < 0 &&
4540            check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4541          {          {
4542          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4543          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3609  we set the flag only if there is a liter Line 4559  we set the flag only if there is a liter
4559    
4560        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
4561    
4562          /*--------------------------------------------------------------------*/
4563          /* This code is obsolete from release 8.00; the restriction was finally
4564          removed: */
4565    
4566        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4567        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4568    
4569        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4570          /*--------------------------------------------------------------------*/
4571    
4572        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
4573    
# Line 3751  we set the flag only if there is a liter Line 4706  we set the flag only if there is a liter
4706  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4707               *previous == OP_XCLASS ||               *previous == OP_XCLASS ||
4708  #endif  #endif
4709               *previous == OP_REF)               *previous == OP_REF ||
4710                 *previous == OP_REFI)
4711        {        {
4712        if (repeat_max == 0)        if (repeat_max == 0)
4713          {          {
# Line 3759  we set the flag only if there is a liter Line 4715  we set the flag only if there is a liter
4715          goto END_REPEAT;          goto END_REPEAT;
4716          }          }
4717    
4718          /*--------------------------------------------------------------------*/
4719          /* This code is obsolete from release 8.00; the restriction was finally
4720          removed: */
4721    
4722        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4723        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4724    
4725        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4726          /*--------------------------------------------------------------------*/
4727    
4728        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4729          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 3780  we set the flag only if there is a liter Line 4741  we set the flag only if there is a liter
4741        }        }
4742    
4743      /* 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
4744      cases. */      cases. Note that at this point we can encounter only the "basic" bracket
4745        opcodes such as BRA and CBRA, as this is the place where they get converted
4746        into the more special varieties such as BRAPOS and SBRA. A test for >=
4747        OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
4748        ASSERTBACK_NOT, ONCE, BRA, CBRA, and COND. Originally, PCRE did not allow
4749        repetition of assertions, but now it does, for Perl compatibility. */
4750    
4751      else if (*previous == OP_BRA  || *previous == OP_CBRA ||      else if (*previous >= OP_ASSERT && *previous <= OP_COND)
              *previous == OP_ONCE || *previous == OP_COND)  
4752        {        {
4753        register int i;        register int i;
4754        int ketoffset = 0;        int len = (int)(code - previous);
       int len = code - previous;  
4755        uschar *bralink = NULL;        uschar *bralink = NULL;
4756          uschar *brazeroptr = NULL;
4757    
4758        /* Repeating a DEFINE group is pointless */        /* Repeating a DEFINE group is pointless, but Perl allows the syntax, so
4759          we just ignore the repeat. */
4760    
4761        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
4762          {          goto END_REPEAT;
         *errorcodeptr = ERR55;  
         goto FAILED;  
         }  
4763    
4764        /* If the maximum repeat count is unlimited, find the end of the bracket        /* There is no sense in actually repeating assertions. The only potential
4765        by scanning through from the start, and compute the offset back to it        use of repetition is in cases when the assertion is optional. Therefore,
4766        from the current code pointer. There may be an OP_OPT setting following        if the minimum is greater than zero, just ignore the repeat. If the
4767        the final KET, so we can't find the end just by going back from the code        maximum is not not zero or one, set it to 1. */
4768        pointer. */  
4769          if (*previous < OP_ONCE)    /* Assertion */
4770        if (repeat_max == -1)          {
4771          {          if (repeat_min > 0) goto END_REPEAT;
4772          register uschar *ket = previous;          if (repeat_max < 0 || repeat_max > 1) repeat_max = 1;
         do ket += GET(ket, 1); while (*ket != OP_KET);  
         ketoffset = code - ket;  
4773          }          }
4774    
4775        /* The case of a zero minimum is special because of the need to stick        /* The case of a zero minimum is special because of the need to stick
# Line 3819  we set the flag only if there is a liter Line 4780  we set the flag only if there is a liter
4780        minimum is zero. */        minimum is zero. */
4781    
4782        if (repeat_min == 0)        if (repeat_min == 0)
4783          {          {
4784          /* 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
4785          altogether. */          output altogether, like this:
   
         if (repeat_max == 0)  
           {  
           code = previous;  
           goto END_REPEAT;  
           }  
4786    
4787          /* If the maximum is 1 or unlimited, we just have to stick in the          ** if (repeat_max == 0)
4788          BRAZERO and do no more at this point. However, we do need to adjust          **   {
4789          any OP_RECURSE calls inside the group that refer to the group itself or          **   code = previous;
4790          any internal or forward referenced group, because the offset is from          **   goto END_REPEAT;
4791          the start of the whole regex. Temporarily terminate the pattern while          **   }
4792          doing this. */  
4793            However, that fails when a group or a subgroup within it is referenced
4794            as a subroutine from elsewhere in the pattern, so now we stick in
4795            OP_SKIPZERO in front of it so that it is skipped on execution. As we
4796            don't have a list of which groups are referenced, we cannot do this
4797            selectively.
4798    
4799            If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4800            and do no more at this point. However, we do need to adjust any
4801            OP_RECURSE calls inside the group that refer to the group itself or any
4802            internal or forward referenced group, because the offset is from the
4803            start of the whole regex. Temporarily terminate the pattern while doing
4804            this. */
4805    
4806          if (repeat_max <= 1)          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
4807            {            {
4808            *code = OP_END;            *code = OP_END;
4809            adjust_recurse(previous, 1, utf8, cd, save_hwm);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
4810            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
4811            code++;            code++;
4812              if (repeat_max == 0)
4813                {
4814                *previous++ = OP_SKIPZERO;
4815                goto END_REPEAT;
4816                }
4817              brazeroptr = previous;    /* Save for possessive optimizing */
4818            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4819            }            }
4820    
# Line 3866  we set the flag only if there is a liter Line 4839  we set the flag only if there is a liter
4839            /* We chain together the bracket offset fields that have to be            /* We chain together the bracket offset fields that have to be
4840            filled in later when the ends of the brackets are reached. */            filled in later when the ends of the brackets are reached. */
4841    
4842            offset = (bralink == NULL)? 0 : previous - bralink;            offset = (bralink == NULL)? 0 : (int)(previous - bralink);
4843            bralink = previous;            bralink = previous;
4844            PUTINC(previous, 0, offset);            PUTINC(previous, 0, offset);
4845            }            }
# Line 3887  we set the flag only if there is a liter Line 4860  we set the flag only if there is a liter
4860            {            {
4861            /* 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
4862            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
4863            potential integer overflow. */            potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
4864              integer type when available, otherwise double. */
4865    
4866            if (lengthptr != NULL)            if (lengthptr != NULL)
4867              {              {
4868              int delta = (repeat_min - 1)*length_prevgroup;              int delta = (repeat_min - 1)*length_prevgroup;
4869              if ((double)(repeat_min - 1)*(double)length_prevgroup >              if ((INT64_OR_DOUBLE)(repeat_min - 1)*
4870                                                              (double)INT_MAX ||                    (INT64_OR_DOUBLE)length_prevgroup >
4871                        (INT64_OR_DOUBLE)INT_MAX ||
4872                  OFLOW_MAX - *lengthptr < delta)                  OFLOW_MAX - *lengthptr < delta)
4873                {                {
4874                *errorcodeptr = ERR20;                *errorcodeptr = ERR20;
# Line 3939  we set the flag only if there is a liter Line 4914  we set the flag only if there is a liter
4914          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
4915          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
4916          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
4917          paranoid checks to avoid integer overflow. */          paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
4918            a 64-bit integer type when available, otherwise double. */
4919    
4920          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4921            {            {
4922            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4923                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4924            if ((double)repeat_max *            if ((INT64_OR_DOUBLE)repeat_max *
4925                  (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)                  (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4926                    > (double)INT_MAX ||                    > (INT64_OR_DOUBLE)INT_MAX ||
4927                OFLOW_MAX - *lengthptr < delta)                OFLOW_MAX - *lengthptr < delta)
4928              {              {
4929              *errorcodeptr = ERR20;              *errorcodeptr = ERR20;
# Line 3972  we set the flag only if there is a liter Line 4948  we set the flag only if there is a liter
4948              {              {
4949              int offset;              int offset;
4950              *code++ = OP_BRA;              *code++ = OP_BRA;
4951              offset = (bralink == NULL)? 0 : code - bralink;              offset = (bralink == NULL)? 0 : (int)(code - bralink);
4952              bralink = code;              bralink = code;
4953              PUTINC(code, 0, offset);              PUTINC(code, 0, offset);
4954              }              }
# Line 3993  we set the flag only if there is a liter Line 4969  we set the flag only if there is a liter
4969          while (bralink != NULL)          while (bralink != NULL)
4970            {            {
4971            int oldlinkoffset;            int oldlinkoffset;
4972            int offset = code - bralink + 1;            int offset = (int)(code - bralink + 1);
4973            uschar *bra = code - offset;            uschar *bra = code - offset;
4974            oldlinkoffset = GET(bra, 1);            oldlinkoffset = GET(bra, 1);
4975            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 4979  we set the flag only if there is a liter
4979            }            }
4980          }          }
4981    
4982        /* 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
4983        can't just offset backwards from the current code point, because we        ONCE brackets, that's all we need to do. However, possessively repeated
4984        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
4985        correct offset was computed above.        behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to
4986          deal with possessive ONCEs specially.
4987        Then, when we are doing the actual compile phase, check to see whether  
4988        this group is a non-atomic one that could match an empty string. If so,        Otherwise, when we are doing the actual compile phase, check to see
4989          whether this group is one that could match an empty string. If so,
4990        convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so        convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so
4991        that runtime checking can be done. [This check is also applied to        that runtime checking can be done. [This check is also applied to ONCE
4992        atomic groups at runtime, but in a different way.] */        groups at runtime, but in a different way.]
4993    
4994          Then, if the quantifier was possessive and the bracket is not a
4995          conditional, we convert the BRA code to the POS form, and the KET code to
4996          KETRPOS. (It turns out to be convenient at runtime to detect this kind of
4997          subpattern at both the start and at the end.) The use of special opcodes
4998          makes it possible to reduce greatly the stack usage in pcre_exec(). If
4999          the group is preceded by OP_BRAZERO, convert this to OP_BRAPOSZERO.
5000    
5001          Then, if the minimum number of matches is 1 or 0, cancel the possessive
5002          flag so that the default action below, of wrapping everything inside
5003          atomic brackets, does not happen. When the minimum is greater than 1,
5004          there will be earlier copies of the group, and so we still have to wrap
5005          the whole thing. */
5006    
5007        else        else
5008          {          {
5009          uschar *ketcode = code - ketoffset;          uschar *ketcode = code - 1 - LINK_SIZE;
5010          uschar *bracode = ketcode - GET(ketcode, 1);          uschar *bracode = ketcode - GET(ketcode, 1);
5011          *ketcode = OP_KETRMAX + repeat_type;  
5012          if (lengthptr == NULL && *bracode != OP_ONCE)          /* Convert possessive ONCE brackets to non-capturing */
5013    
5014            if ((*bracode == OP_ONCE || *bracode == OP_ONCE_NC) &&
5015                possessive_quantifier) *bracode = OP_BRA;
5016    
5017            /* For non-possessive ONCE brackets, all we need to do is to
5018            set the KET. */
5019    
5020            if (*bracode == OP_ONCE || *bracode == OP_ONCE_NC)
5021              *ketcode = OP_KETRMAX + repeat_type;
5022    
5023            /* Handle non-ONCE brackets and possessive ONCEs (which have been
5024            converted to non-capturing above). */
5025    
5026            else
5027            {            {
5028            uschar *scode = bracode;            /* In the compile phase, check for empty string matching. */
5029            do  
5030              if (lengthptr == NULL)
5031              {              {
5032              if (could_be_empty_branch(scode, ketcode, utf8))              uschar *scode = bracode;
5033                do
5034                {                {
5035                *bracode += OP_SBRA - OP_BRA;                if (could_be_empty_branch(scode, ketcode, utf8, cd))
5036                break;                  {
5037                    *bracode += OP_SBRA - OP_BRA;
5038                    break;
5039                    }
5040                  scode += GET(scode, 1);
5041                  }
5042                while (*scode == OP_ALT);
5043                }
5044    
5045              /* Handle possessive quantifiers. */
5046    
5047              if (possessive_quantifier)
5048                {
5049                /* For COND brackets, we wrap the whole thing in a possessively
5050                repeated non-capturing bracket, because we have not invented POS
5051                versions of the COND opcodes. Because we are moving code along, we
5052                must ensure that any pending recursive references are updated. */
5053    
5054                if (*bracode == OP_COND || *bracode == OP_SCOND)
5055                  {
5056                  int nlen = (int)(code - bracode);
5057                  *code = OP_END;
5058                  adjust_recurse(bracode, 1 + LINK_SIZE, utf8, cd, save_hwm);
5059                  memmove(bracode + 1+LINK_SIZE, bracode, nlen);
5060                  code += 1 + LINK_SIZE;
5061                  nlen += 1 + LINK_SIZE;
5062                  *bracode = OP_BRAPOS;
5063                  *code++ = OP_KETRPOS;
5064                  PUTINC(code, 0, nlen);
5065                  PUT(bracode, 1, nlen);
5066                  }
5067    
5068                /* For non-COND brackets, we modify the BRA code and use KETRPOS. */
5069    
5070                else
5071                  {
5072                  *bracode += 1;              /* Switch to xxxPOS opcodes */
5073                  *ketcode = OP_KETRPOS;
5074                }                }
5075              scode += GET(scode, 1);  
5076                /* If the minimum is zero, mark it as possessive, then unset the
5077                possessive flag when the minimum is 0 or 1. */
5078    
5079                if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
5080                if (repeat_min < 2) possessive_quantifier = FALSE;
5081              }              }
5082            while (*scode == OP_ALT);  
5083              /* Non-possessive quantifier */
5084    
5085              else *ketcode = OP_KETRMAX + repeat_type;
5086            }            }
5087          }          }
5088        }        }
5089    
5090        /* If previous is OP_FAIL, it was generated by an empty class [] in
5091        JavaScript mode. The other ways in which OP_FAIL can be generated, that is
5092        by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
5093        error above. We can just ignore the repeat in JS case. */
5094    
5095        else if (*previous == OP_FAIL) goto END_REPEAT;
5096    
5097      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
5098    
5099      else      else
# Line 4045  we set the flag only if there is a liter Line 5103  we set the flag only if there is a liter
5103        }        }
5104    
5105      /* If the character following a repeat is '+', or if certain optimization      /* If the character following a repeat is '+', or if certain optimization
5106      tests above succeeded, possessive_quantifier is TRUE. For some of the      tests above succeeded, possessive_quantifier is TRUE. For some opcodes,
5107      simpler opcodes, there is an special alternative opcode for this. For      there are special alternative opcodes for this case. For anything else, we
5108      anything else, we wrap the entire repeated item inside OP_ONCE brackets.      wrap the entire repeated item inside OP_ONCE brackets. Logically, the '+'
5109      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
5110      but the special opcodes can optimize it a bit. The repeated item starts at      special opcodes can optimize it.
5111      tempcode, not at previous, which might be the first part of a string whose  
5112      (former) last char we repeated.      Some (but not all) possessively repeated subpatterns have already been
5113        completely handled in the code just above. For them, possessive_quantifier
5114        is always FALSE at this stage.
5115    
5116        Note that the repeated item starts at tempcode, not at previous, which
5117        might be the first part of a string whose (former) last char we repeated.
5118    
5119      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
5120      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 5123  we set the flag only if there is a liter
5123      if (possessive_quantifier)      if (possessive_quantifier)
5124        {        {
5125        int len;        int len;
5126        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||  
5127            *tempcode == OP_NOTEXACT)        if (*tempcode == OP_TYPEEXACT)
5128          tempcode += _pcre_OP_lengths[*tempcode] +          tempcode += _pcre_OP_lengths[*tempcode] +
5129            ((*tempcode == OP_TYPEEXACT &&            ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
5130               (tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP))? 2:0);  
5131        len = code - tempcode;        else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
5132            {
5133            tempcode += _pcre_OP_lengths[*tempcode];
5134    #ifdef SUPPORT_UTF8
5135            if (utf8 && tempcode[-1] >= 0xc0)
5136              tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];
5137    #endif
5138            }
5139    
5140          len = (int)(code - tempcode);
5141        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
5142          {          {
5143          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 5145  we set the flag only if there is a liter
5145          case OP_QUERY: *tempcode = OP_POSQUERY; break;          case OP_QUERY: *tempcode = OP_POSQUERY; break;
5146          case OP_UPTO:  *tempcode = OP_POSUPTO; break;          case OP_UPTO:  *tempcode = OP_POSUPTO; break;
5147    
5148          case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;          case OP_STARI:  *tempcode = OP_POSSTARI; break;
5149          case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;          case OP_PLUSI:  *tempcode = OP_POSPLUSI; break;
5150          case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;          case OP_QUERYI: *tempcode = OP_POSQUERYI; break;
5151          case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;          case OP_UPTOI:  *tempcode = OP_POSUPTOI; break;
5152    
5153          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;
5154          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;
5155          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
5156          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
5157