/[pcre]/code/branches/pcre16/pcre_compile.c
ViewVC logotype

Diff of /code/branches/pcre16/pcre_compile.c

Parent Directory Parent Directory | Revision Log Revision Log | View Patch Patch

revision 335 by ph10, Sat Apr 12 14:36:14 2008 UTC revision 749 by ph10, Fri Nov 18 10:36:45 2011 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2008 University of Cambridge             Copyright (c) 1997-2011 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 53  supporting internal functions that are n Line 53  supporting internal functions that are n
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56  /* When DEBUG is defined, we need the pcre_printint() function, which is also  /* When PCRE_DEBUG is defined, we need the pcre_printint() function, which is
57  used by pcretest. DEBUG is not defined when building a production library. */  also used by pcretest. PCRE_DEBUG is not defined when building a production
58    library. */
59    
60  #ifdef DEBUG  #ifdef PCRE_DEBUG
61  #include "pcre_printint.src"  #include "pcre_printint.src"
62  #endif  #endif
63    
# Line 91  is 4 there is plenty of room. */ Line 92  is 4 there is plenty of room. */
92    
93  #define COMPILE_WORK_SIZE (4096)  #define COMPILE_WORK_SIZE (4096)
94    
95    /* The overrun tests check for a slightly smaller size so that they detect the
96    overrun before it actually does run off the end of the data block. */
97    
98    #define WORK_SIZE_CHECK (COMPILE_WORK_SIZE - 100)
99    
100    
101  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
102  are simple data values; negative values are for special things like \d and so  are simple data values; negative values are for special things like \d and so
103  on. Zero means further processing is needed (for things like \x), or the escape  on. Zero means further processing is needed (for things like \x), or the escape
104  is invalid. */  is invalid. */
105    
106  #ifndef EBCDIC  /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC
107    
108    /* This is the "normal" table for ASCII systems or for EBCDIC systems running
109    in UTF-8 mode. */
110    
111  static const short int escapes[] = {  static const short int escapes[] = {
112       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,                       0,
113       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,                       0,
114     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */       0,                       0,
115  -ESC_H,      0,      0, -ESC_K,      0,      0,      0,      0,   /* H - O */       0,                       0,
116  -ESC_P, -ESC_Q, -ESC_R, -ESC_S,      0,      0, -ESC_V, -ESC_W,   /* P - W */       0,                       0,
117  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */       CHAR_COLON,              CHAR_SEMICOLON,
118     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
119  -ESC_h,      0,      0, -ESC_k,      0,      0,  ESC_n,      0,   /* h - o */       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
120  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0, -ESC_v, -ESC_w,   /* p - w */       CHAR_COMMERCIAL_AT,      -ESC_A,
121       0,      0, -ESC_z                                            /* x - z */       -ESC_B,                  -ESC_C,
122         -ESC_D,                  -ESC_E,
123         0,                       -ESC_G,
124         -ESC_H,                  0,
125         0,                       -ESC_K,
126         0,                       0,
127         -ESC_N,                  0,
128         -ESC_P,                  -ESC_Q,
129         -ESC_R,                  -ESC_S,
130         0,                       0,
131         -ESC_V,                  -ESC_W,
132         -ESC_X,                  0,
133         -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,
134         CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,
135         CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,
136         CHAR_GRAVE_ACCENT,       7,
137         -ESC_b,                  0,
138         -ESC_d,                  ESC_e,
139         ESC_f,                   0,
140         -ESC_h,                  0,
141         0,                       -ESC_k,
142         0,                       0,
143         ESC_n,                   0,
144         -ESC_p,                  0,
145         ESC_r,                   -ESC_s,
146         ESC_tee,                 0,
147         -ESC_v,                  -ESC_w,
148         0,                       0,
149         -ESC_z
150  };  };
151    
152  #else           /* This is the "abnormal" table for EBCDIC systems */  #else
153    
154    /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
155    
156  static const short int escapes[] = {  static const short int escapes[] = {
157  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
158  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 130  static const short int escapes[] = { Line 171  static const short int escapes[] = {
171  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
172  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,
173  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
174  /*  D0 */   '}',     0, -ESC_K,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,-ESC_N,      0, -ESC_P,
175  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
176  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
177  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
# Line 142  static const short int escapes[] = { Line 183  static const short int escapes[] = {
183    
184  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
185  searched linearly. Put all the names into a single string, in order to reduce  searched linearly. Put all the names into a single string, in order to reduce
186  the number of relocations when a shared library is dynamically linked. */  the number of relocations when a shared library is dynamically linked. The
187    string is built from string macros so that it works in UTF-8 mode on EBCDIC
188    platforms. */
189    
190  typedef struct verbitem {  typedef struct verbitem {
191    int   len;    int   len;                 /* Length of verb name */
192    int   op;    int   op;                  /* Op when no arg, or -1 if arg mandatory */
193      int   op_arg;              /* Op when arg present, or -1 if not allowed */
194  } verbitem;  } verbitem;
195    
196  static const char verbnames[] =  static const char verbnames[] =
197    "ACCEPT\0"    "\0"                       /* Empty name is a shorthand for MARK */
198    "COMMIT\0"    STRING_MARK0
199    "F\0"    STRING_ACCEPT0
200    "FAIL\0"    STRING_COMMIT0
201    "PRUNE\0"    STRING_F0
202    "SKIP\0"    STRING_FAIL0
203    "THEN";    STRING_PRUNE0
204      STRING_SKIP0
205      STRING_THEN;
206    
207  static const verbitem verbs[] = {  static const verbitem verbs[] = {
208    { 6, OP_ACCEPT },    { 0, -1,        OP_MARK },
209    { 6, OP_COMMIT },    { 4, -1,        OP_MARK },
210    { 1, OP_FAIL },    { 6, OP_ACCEPT, -1 },
211    { 4, OP_FAIL },    { 6, OP_COMMIT, -1 },
212    { 5, OP_PRUNE },    { 1, OP_FAIL,   -1 },
213    { 4, OP_SKIP  },    { 4, OP_FAIL,   -1 },
214    { 4, OP_THEN  }    { 5, OP_PRUNE,  OP_PRUNE_ARG },
215      { 4, OP_SKIP,   OP_SKIP_ARG  },
216      { 4, OP_THEN,   OP_THEN_ARG  }
217  };  };
218    
219  static const int verbcount = sizeof(verbs)/sizeof(verbitem);  static const int verbcount = sizeof(verbs)/sizeof(verbitem);
# Line 178  length entry. The first three must be al Line 226  length entry. The first three must be al
226  for handling case independence. */  for handling case independence. */
227    
228  static const char posix_names[] =  static const char posix_names[] =
229    "alpha\0"  "lower\0"  "upper\0"  "alnum\0"  "ascii\0"  "blank\0"    STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
230    "cntrl\0"  "digit\0"  "graph\0"  "print\0"  "punct\0"  "space\0"    STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
231    "word\0"   "xdigit";    STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
232      STRING_word0  STRING_xdigit;
233    
234  static const uschar posix_name_lengths[] = {  static const uschar posix_name_lengths[] = {
235    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
# Line 212  static const int posix_class_maps[] = { Line 261  static const int posix_class_maps[] = {
261    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
262  };  };
263    
264    /* Table of substitutes for \d etc when PCRE_UCP is set. The POSIX class
265    substitutes must be in the order of the names, defined above, and there are
266    both positive and negative cases. NULL means no substitute. */
267    
268    #ifdef SUPPORT_UCP
269    static const uschar *substitutes[] = {
270      (uschar *)"\\P{Nd}",    /* \D */
271      (uschar *)"\\p{Nd}",    /* \d */
272      (uschar *)"\\P{Xsp}",   /* \S */       /* NOTE: Xsp is Perl space */
273      (uschar *)"\\p{Xsp}",   /* \s */
274      (uschar *)"\\P{Xwd}",   /* \W */
275      (uschar *)"\\p{Xwd}"    /* \w */
276    };
277    
278    static const uschar *posix_substitutes[] = {
279      (uschar *)"\\p{L}",     /* alpha */
280      (uschar *)"\\p{Ll}",    /* lower */
281      (uschar *)"\\p{Lu}",    /* upper */
282      (uschar *)"\\p{Xan}",   /* alnum */
283      NULL,                   /* ascii */
284      (uschar *)"\\h",        /* blank */
285      NULL,                   /* cntrl */
286      (uschar *)"\\p{Nd}",    /* digit */
287      NULL,                   /* graph */
288      NULL,                   /* print */
289      NULL,                   /* punct */
290      (uschar *)"\\p{Xps}",   /* space */    /* NOTE: Xps is POSIX space */
291      (uschar *)"\\p{Xwd}",   /* word */
292      NULL,                   /* xdigit */
293      /* Negated cases */
294      (uschar *)"\\P{L}",     /* ^alpha */
295      (uschar *)"\\P{Ll}",    /* ^lower */
296      (uschar *)"\\P{Lu}",    /* ^upper */
297      (uschar *)"\\P{Xan}",   /* ^alnum */
298      NULL,                   /* ^ascii */
299      (uschar *)"\\H",        /* ^blank */
300      NULL,                   /* ^cntrl */
301      (uschar *)"\\P{Nd}",    /* ^digit */
302      NULL,                   /* ^graph */
303      NULL,                   /* ^print */
304      NULL,                   /* ^punct */
305      (uschar *)"\\P{Xps}",   /* ^space */   /* NOTE: Xps is POSIX space */
306      (uschar *)"\\P{Xwd}",   /* ^word */
307      NULL                    /* ^xdigit */
308    };
309    #define POSIX_SUBSIZE (sizeof(posix_substitutes)/sizeof(uschar *))
310    #endif
311    
312  #define STRING(a)  # a  #define STRING(a)  # a
313  #define XSTRING(s) STRING(s)  #define XSTRING(s) STRING(s)
# Line 224  the number of relocations needed when a Line 320  the number of relocations needed when a
320  it is now one long string. We cannot use a table of offsets, because the  it is now one long string. We cannot use a table of offsets, because the
321  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
322  simply count through to the one we want - this isn't a performance issue  simply count through to the one we want - this isn't a performance issue
323  because these strings are used only when there is a compilation error. */  because these strings are used only when there is a compilation error.
324    
325    Each substring ends with \0 to insert a null character. This includes the final
326    substring, so that the whole string ends with \0\0, which can be detected when
327    counting through. */
328    
329  static const char error_texts[] =  static const char error_texts[] =
330    "no error\0"    "no error\0"
# Line 271  static const char error_texts[] = Line 371  static const char error_texts[] =
371    /* 35 */    /* 35 */
372    "invalid condition (?(0)\0"    "invalid condition (?(0)\0"
373    "\\C not allowed in lookbehind assertion\0"    "\\C not allowed in lookbehind assertion\0"
374    "PCRE does not support \\L, \\l, \\N, \\U, or \\u\0"    "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
375    "number after (?C is > 255\0"    "number after (?C is > 255\0"
376    "closing ) for (?C expected\0"    "closing ) for (?C expected\0"
377    /* 40 */    /* 40 */
# Line 293  static const char error_texts[] = Line 393  static const char error_texts[] =
393    "internal error: previously-checked referenced subpattern not found\0"    "internal error: previously-checked referenced subpattern not found\0"
394    "DEFINE group contains more than one branch\0"    "DEFINE group contains more than one branch\0"
395    /* 55 */    /* 55 */
396    "repeating a DEFINE group is not allowed\0"    "repeating a DEFINE group is not allowed\0"  /** DEAD **/
397    "inconsistent NEWLINE options\0"    "inconsistent NEWLINE options\0"
398    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
399    "a numbered reference must not be zero\0"    "a numbered reference must not be zero\0"
400    "(*VERB) with an argument is not supported\0"    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
401    /* 60 */    /* 60 */
402    "(*VERB) not recognized\0"    "(*VERB) not recognized\0"
403    "number is too big\0"    "number is too big\0"
404    "subpattern name expected\0"    "subpattern name expected\0"
405    "digit expected after (?+";    "digit expected after (?+\0"
406      "] is an invalid data character in JavaScript compatibility mode\0"
407      /* 65 */
408      "different names for subpatterns of the same number are not allowed\0"
409      "(*MARK) must have an argument\0"
410      "this version of PCRE is not compiled with PCRE_UCP support\0"
411      "\\c must be followed by an ASCII character\0"
412      "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
413      /* 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 one of a number of specific things:      case CHAR_u:
685        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
686          {
687          /* In JavaScript, \u must be followed by four hexadecimal numbers.
688          Otherwise it is a lowercase u letter. */
689          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      (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      backreference. If negative, it is a relative backreference. This is a Perl
720      5.10 feature.      5.10 feature.
721    
722      (2) Perl 5.10 also supports \g{name} as a reference to a named group. This      (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
723      is part of Perl's movement towards a unified syntax for back references. As      is part of Perl's movement towards a unified syntax for back references. As
724      this is synonymous with \k{name}, we fudge it up by pretending it really      this is synonymous with \k{name}, we fudge it up by pretending it really
725      was \k.      was \k.
726    
727      (3) For Oniguruma compatibility we also support \g followed by a name or a      (3) For Oniguruma compatibility we also support \g followed by a name or a
728      number either in angle brackets or in single quotes. However, these are      number either in angle brackets or in single quotes. However, these are
729      (possibly recursive) subroutine calls, _not_ backreferences. Just return      (possibly recursive) subroutine calls, _not_ backreferences. Just return
730      the -ESC_g code (cf \k). */      the -ESC_g code (cf \k). */
731    
732      case 'g':      case CHAR_g:
733      if (ptr[1] == '<' || ptr[1] == '\'')      if (isclass) break;
734        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
735        {        {
736        c = -ESC_g;        c = -ESC_g;
737        break;        break;
738        }        }
739    
740      /* Handle the Perl-compatible cases */      /* Handle the Perl-compatible cases */
741    
742      if (ptr[1] == '{')      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
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 571  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 580  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)   /* Integer overflow */      if (c < 0)   /* Integer overflow */
769        {        {
770        *errorcodeptr = ERR61;        *errorcodeptr = ERR61;
771        break;        break;
772        }        }
773    
774      if (braced && *(++ptr) != '}')      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
775        {        {
776        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
777        break;        break;
778        }        }
779    
780      if (c == 0)      if (c == 0)
781        {        {
782        *errorcodeptr = ERR58;        *errorcodeptr = ERR58;
783        break;        break;
784        }        }
785    
786      if (negated)      if (negated)
787        {        {
# Line 625  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)    /* Integer overflow */        if (c < 0)    /* Integer overflow */
821          {          {
822          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
# Line 651  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 664  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 675  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 685  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 713  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 763  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 803  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 814  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 863  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 923  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 933  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 963  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 1031  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      else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1382        {
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    ptr += 2;  FAIL_EXIT:
1395    if (*ptr == 'P') ptr++;                      /* Allow optional P */  *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 1123  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 1139  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      cd       the "compile data" structure
1528    
1529  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1530                 or -1 if there is no fixed length,
1531               or -2 if \C was encountered               or -2 if \C was encountered
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 1168  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 1198  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 1209  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 1229  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 1245  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 1268  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 1275  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;
# Line 1298  for (;;) Line 1721  for (;;)
1721    
1722      switch (*cc)      switch (*cc)
1723        {        {
1724          case OP_CRPLUS:
1725          case OP_CRMINPLUS:
1726        case OP_CRSTAR:        case OP_CRSTAR:
1727        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1728        case OP_CRQUERY:        case OP_CRQUERY:
# Line 1316  for (;;) Line 1741  for (;;)
1741        }        }
1742      break;      break;
1743    
1744      /* Anything else is variable length */      /* Anything else is variable length */
1745    
1746        case OP_ANYNL:
1747        case OP_BRAMINZERO:
1748        case OP_BRAPOS:
1749        case OP_BRAPOSZERO:
1750        case OP_BRAZERO:
1751        case OP_CBRAPOS:
1752        case OP_EXTUNI:
1753        case OP_KETRMAX:
1754        case OP_KETRMIN:
1755        case OP_KETRPOS:
1756        case OP_MINPLUS:
1757        case OP_MINPLUSI:
1758        case OP_MINQUERY:
1759        case OP_MINQUERYI:
1760        case OP_MINSTAR:
1761        case OP_MINSTARI:
1762        case OP_MINUPTO:
1763        case OP_MINUPTOI:
1764        case OP_NOTMINPLUS:
1765        case OP_NOTMINPLUSI:
1766        case OP_NOTMINQUERY:
1767        case OP_NOTMINQUERYI:
1768        case OP_NOTMINSTAR:
1769        case OP_NOTMINSTARI:
1770        case OP_NOTMINUPTO:
1771        case OP_NOTMINUPTOI:
1772        case OP_NOTPLUS:
1773        case OP_NOTPLUSI:
1774        case OP_NOTPOSPLUS:
1775        case OP_NOTPOSPLUSI:
1776        case OP_NOTPOSQUERY:
1777        case OP_NOTPOSQUERYI:
1778        case OP_NOTPOSSTAR:
1779        case OP_NOTPOSSTARI:
1780        case OP_NOTPOSUPTO:
1781        case OP_NOTPOSUPTOI:
1782        case OP_NOTQUERY:
1783        case OP_NOTQUERYI:
1784        case OP_NOTSTAR:
1785        case OP_NOTSTARI:
1786        case OP_NOTUPTO:
1787        case OP_NOTUPTOI:
1788        case OP_PLUS:
1789        case OP_PLUSI:
1790        case OP_POSPLUS:
1791        case OP_POSPLUSI:
1792        case OP_POSQUERY:
1793        case OP_POSQUERYI:
1794        case OP_POSSTAR:
1795        case OP_POSSTARI:
1796        case OP_POSUPTO:
1797        case OP_POSUPTOI:
1798        case OP_QUERY:
1799        case OP_QUERYI:
1800        case OP_REF:
1801        case OP_REFI:
1802        case OP_SBRA:
1803        case OP_SBRAPOS:
1804        case OP_SCBRA:
1805        case OP_SCBRAPOS:
1806        case OP_SCOND:
1807        case OP_SKIPZERO:
1808        case OP_STAR:
1809        case OP_STARI:
1810        case OP_TYPEMINPLUS:
1811        case OP_TYPEMINQUERY:
1812        case OP_TYPEMINSTAR:
1813        case OP_TYPEMINUPTO:
1814        case OP_TYPEPLUS:
1815        case OP_TYPEPOSPLUS:
1816        case OP_TYPEPOSQUERY:
1817        case OP_TYPEPOSSTAR:
1818        case OP_TYPEPOSUPTO:
1819        case OP_TYPEQUERY:
1820        case OP_TYPESTAR:
1821        case OP_TYPEUPTO:
1822        case OP_UPTO:
1823        case OP_UPTOI:
1824        return -1;
1825    
1826        /* Catch unrecognized opcodes so that when new ones are added they
1827        are not forgotten, as has happened in the past. */
1828    
1829      default:      default:
1830      return -1;      return -4;
1831      }      }
1832    }    }
1833  /* Control never gets here */  /* Control never gets here */
# Line 1329  for (;;) Line 1837  for (;;)
1837    
1838    
1839  /*************************************************  /*************************************************
1840  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1841  *************************************************/  *************************************************/
1842    
1843  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1844  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1845    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1846    so that it can be called from pcre_study() when finding the minimum matching
1847    length.
1848    
1849  Arguments:  Arguments:
1850    code        points to start of expression    code        points to start of expression
1851    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1852    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1853    
1854  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
1855  */  */
1856    
1857  static const uschar *  const uschar *
1858  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1859  {  {
1860  for (;;)  for (;;)
1861    {    {
1862    register int c = *code;    register int c = *code;
1863    
1864    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1865    
1866    /* XCLASS is used for classes that cannot be represented just by a bit    /* XCLASS is used for classes that cannot be represented just by a bit
# Line 1357  for (;;) Line 1869  for (;;)
1869    
1870    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1871    
1872      /* Handle recursion */
1873    
1874      else if (c == OP_REVERSE)
1875        {
1876        if (number < 0) return (uschar *)code;
1877        code += _pcre_OP_lengths[c];
1878        }
1879    
1880    /* Handle capturing bracket */    /* Handle capturing bracket */
1881    
1882    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
1883               c == OP_CBRAPOS || c == OP_SCBRAPOS)
1884      {      {
1885      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
1886      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
# Line 1368  for (;;) Line 1889  for (;;)
1889    
1890    /* 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
1891    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
1892    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1893      must add in its length. */
1894    
1895    else    else
1896      {      {
# Line 1392  for (;;) Line 1914  for (;;)
1914        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
1915        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1916        break;        break;
1917    
1918          case OP_MARK:
1919          case OP_PRUNE_ARG:
1920          case OP_SKIP_ARG:
1921          code += code[1];
1922          break;
1923    
1924          case OP_THEN_ARG:
1925          code += code[1];
1926          break;
1927        }        }
1928    
1929      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1406  for (;;) Line 1938  for (;;)
1938      if (utf8) switch(c)      if (utf8) switch(c)
1939        {        {
1940        case OP_CHAR:        case OP_CHAR:
1941        case OP_CHARNC:        case OP_CHARI:
1942        case OP_EXACT:        case OP_EXACT:
1943          case OP_EXACTI:
1944        case OP_UPTO:        case OP_UPTO:
1945          case OP_UPTOI:
1946        case OP_MINUPTO:        case OP_MINUPTO:
1947          case OP_MINUPTOI:
1948        case OP_POSUPTO:        case OP_POSUPTO:
1949          case OP_POSUPTOI:
1950        case OP_STAR:        case OP_STAR:
1951          case OP_STARI:
1952        case OP_MINSTAR:        case OP_MINSTAR:
1953          case OP_MINSTARI:
1954        case OP_POSSTAR:        case OP_POSSTAR:
1955          case OP_POSSTARI:
1956        case OP_PLUS:        case OP_PLUS:
1957          case OP_PLUSI:
1958        case OP_MINPLUS:        case OP_MINPLUS:
1959          case OP_MINPLUSI:
1960        case OP_POSPLUS:        case OP_POSPLUS:
1961          case OP_POSPLUSI:
1962        case OP_QUERY:        case OP_QUERY:
1963          case OP_QUERYI:
1964        case OP_MINQUERY:        case OP_MINQUERY:
1965          case OP_MINQUERYI:
1966        case OP_POSQUERY:        case OP_POSQUERY:
1967          case OP_POSQUERYI:
1968        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1969        break;        break;
1970        }        }
1971    #else
1972        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1973  #endif  #endif
1974      }      }
1975    }    }
# Line 1461  for (;;) Line 2008  for (;;)
2008    
2009    /* 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
2010    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
2011    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2012      must add in its length. */
2013    
2014    else    else
2015      {      {
# Line 1485  for (;;) Line 2033  for (;;)
2033        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2034        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
2035        break;        break;
2036    
2037          case OP_MARK:
2038          case OP_PRUNE_ARG:
2039          case OP_SKIP_ARG:
2040          code += code[1];
2041          break;
2042    
2043          case OP_THEN_ARG:
2044          code += code[1];
2045          break;
2046        }        }
2047    
2048      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1499  for (;;) Line 2057  for (;;)
2057      if (utf8) switch(c)      if (utf8) switch(c)
2058        {        {
2059        case OP_CHAR:        case OP_CHAR:
2060        case OP_CHARNC:        case OP_CHARI:
2061        case OP_EXACT:        case OP_EXACT:
2062          case OP_EXACTI:
2063        case OP_UPTO:        case OP_UPTO:
2064          case OP_UPTOI:
2065        case OP_MINUPTO:        case OP_MINUPTO:
2066          case OP_MINUPTOI:
2067        case OP_POSUPTO:        case OP_POSUPTO:
2068          case OP_POSUPTOI:
2069        case OP_STAR:        case OP_STAR:
2070          case OP_STARI:
2071        case OP_MINSTAR:        case OP_MINSTAR:
2072          case OP_MINSTARI:
2073        case OP_POSSTAR:        case OP_POSSTAR:
2074          case OP_POSSTARI:
2075        case OP_PLUS:        case OP_PLUS:
2076          case OP_PLUSI:
2077        case OP_MINPLUS:        case OP_MINPLUS:
2078          case OP_MINPLUSI:
2079        case OP_POSPLUS:        case OP_POSPLUS:
2080          case OP_POSPLUSI:
2081        case OP_QUERY:        case OP_QUERY:
2082          case OP_QUERYI:
2083        case OP_MINQUERY:        case OP_MINQUERY:
2084          case OP_MINQUERYI:
2085        case OP_POSQUERY:        case OP_POSQUERY:
2086          case OP_POSQUERYI:
2087        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
2088        break;        break;
2089        }        }
2090    #else
2091        (void)(utf8);  /* Keep compiler happy by referencing function argument */
2092  #endif  #endif
2093      }      }
2094    }    }
# Line 1539  Arguments: Line 2112  Arguments:
2112    code        points to start of search    code        points to start of search
2113    endcode     points to where to stop    endcode     points to where to stop
2114    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
2115      cd          contains pointers to tables etc.
2116    
2117  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2118  */  */
2119    
2120  static BOOL  static BOOL
2121  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
2122      compile_data *cd)
2123  {  {
2124  register int c;  register int c;
2125  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], TRUE);
2126       code < endcode;       code < endcode;
2127       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], TRUE))
2128    {    {
2129    const uschar *ccode;    const uschar *ccode;
2130    
# Line 1565  for (code = first_significant_code(code Line 2140  for (code = first_significant_code(code
2140      continue;      continue;
2141      }      }
2142    
2143      /* For a recursion/subroutine call, if its end has been reached, which
2144      implies a backward reference subroutine call, we can scan it. If it's a
2145      forward reference subroutine call, we can't. To detect forward reference
2146      we have to scan up the list that is kept in the workspace. This function is
2147      called only when doing the real compile, not during the pre-compile that
2148      measures the size of the compiled pattern. */
2149    
2150      if (c == OP_RECURSE)
2151        {
2152        const uschar *scode;
2153        BOOL empty_branch;
2154    
2155        /* Test for forward reference */
2156    
2157        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
2158          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
2159    
2160        /* Not a forward reference, test for completed backward reference */
2161    
2162        empty_branch = FALSE;
2163        scode = cd->start_code + GET(code, 1);
2164        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2165    
2166        /* Completed backwards reference */
2167    
2168        do
2169          {
2170          if (could_be_empty_branch(scode, endcode, utf8, cd))
2171            {
2172            empty_branch = TRUE;
2173            break;
2174            }
2175          scode += GET(scode, 1);
2176          }
2177        while (*scode == OP_ALT);
2178    
2179        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2180        continue;
2181        }
2182    
2183    /* Groups with zero repeats can of course be empty; skip them. */    /* Groups with zero repeats can of course be empty; skip them. */
2184    
2185    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2186          c == OP_BRAPOSZERO)
2187      {      {
2188      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
2189      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
# Line 1575  for (code = first_significant_code(code Line 2191  for (code = first_significant_code(code
2191      continue;      continue;
2192      }      }
2193    
2194      /* A nested group that is already marked as "could be empty" can just be
2195      skipped. */
2196    
2197      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2198          c == OP_SCBRA || c == OP_SCBRAPOS)
2199        {
2200        do code += GET(code, 1); while (*code == OP_ALT);
2201        c = *code;
2202        continue;
2203        }
2204    
2205    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
2206    
2207    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_BRA  || c == OP_BRAPOS ||
2208          c == OP_CBRA || c == OP_CBRAPOS ||
2209          c == OP_ONCE || c == OP_ONCE_NC ||
2210          c == OP_COND)
2211      {      {
2212      BOOL empty_branch;      BOOL empty_branch;
2213      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
2214    
2215      /* Scan a closed bracket */      /* If a conditional group has only one branch, there is a second, implied,
2216        empty branch, so just skip over the conditional, because it could be empty.
2217        Otherwise, scan the individual branches of the group. */
2218    
2219      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;  
2220        code += GET(code, 1);        code += GET(code, 1);
2221        else
2222          {
2223          empty_branch = FALSE;
2224          do
2225            {
2226            if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
2227              empty_branch = TRUE;
2228            code += GET(code, 1);
2229            }
2230          while (*code == OP_ALT);
2231          if (!empty_branch) return FALSE;   /* All branches are non-empty */
2232        }        }
2233      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
2234      c = *code;      c = *code;
2235      continue;      continue;
2236      }      }
# Line 1653  for (code = first_significant_code(code Line 2291  for (code = first_significant_code(code
2291      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2292      case OP_WORDCHAR:      case OP_WORDCHAR:
2293      case OP_ANY:      case OP_ANY:
2294        case OP_ALLANY:
2295      case OP_ANYBYTE:      case OP_ANYBYTE:
2296      case OP_CHAR:      case OP_CHAR:
2297      case OP_CHARNC:      case OP_CHARI:
2298      case OP_NOT:      case OP_NOT:
2299        case OP_NOTI:
2300      case OP_PLUS:      case OP_PLUS:
2301      case OP_MINPLUS:      case OP_MINPLUS:
2302      case OP_POSPLUS:      case OP_POSPLUS:
# Line 1696  for (code = first_significant_code(code Line 2336  for (code = first_significant_code(code
2336      case OP_KET:      case OP_KET:
2337      case OP_KETRMAX:      case OP_KETRMAX:
2338      case OP_KETRMIN:      case OP_KETRMIN:
2339        case OP_KETRPOS:
2340      case OP_ALT:      case OP_ALT:
2341      return TRUE;      return TRUE;
2342    
# Line 1704  for (code = first_significant_code(code Line 2345  for (code = first_significant_code(code
2345    
2346  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2347      case OP_STAR:      case OP_STAR:
2348        case OP_STARI:
2349      case OP_MINSTAR:      case OP_MINSTAR:
2350        case OP_MINSTARI:
2351      case OP_POSSTAR:      case OP_POSSTAR:
2352        case OP_POSSTARI:
2353      case OP_QUERY:      case OP_QUERY:
2354        case OP_QUERYI:
2355      case OP_MINQUERY:      case OP_MINQUERY:
2356        case OP_MINQUERYI:
2357      case OP_POSQUERY:      case OP_POSQUERY:
2358        case OP_POSQUERYI:
2359        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2360        break;
2361    
2362      case OP_UPTO:      case OP_UPTO:
2363        case OP_UPTOI:
2364      case OP_MINUPTO:      case OP_MINUPTO:
2365        case OP_MINUPTOI:
2366      case OP_POSUPTO:      case OP_POSUPTO:
2367      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_POSUPTOI:
2368        if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2369      break;      break;
2370  #endif  #endif
2371    
2372        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2373        string. */
2374    
2375        case OP_MARK:
2376        case OP_PRUNE_ARG:
2377        case OP_SKIP_ARG:
2378        code += code[1];
2379        break;
2380    
2381        case OP_THEN_ARG:
2382        code += code[1];
2383        break;
2384    
2385        /* None of the remaining opcodes are required to match a character. */
2386    
2387        default:
2388        break;
2389      }      }
2390    }    }
2391    
# Line 1731  return TRUE; Line 2402  return TRUE;
2402  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
2403  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,
2404  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.
2405    This function is called only during the real compile, not during the
2406    pre-compile.
2407    
2408  Arguments:  Arguments:
2409    code        points to start of the recursion    code        points to start of the recursion
2410    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2411    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2412    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2413      cd          pointers to tables etc
2414    
2415  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2416  */  */
2417    
2418  static BOOL  static BOOL
2419  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2420    BOOL utf8)    BOOL utf8, compile_data *cd)
2421  {  {
2422  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2423    {    {
2424    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2425        return FALSE;
2426    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2427    }    }
2428  return TRUE;  return TRUE;
# Line 1779  where Perl recognizes it as the POSIX cl Line 2454  where Perl recognizes it as the POSIX cl
2454  "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,
2455  I think.  I think.
2456    
2457    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2458    It seems that the appearance of a nested POSIX class supersedes an apparent
2459    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2460    a digit.
2461    
2462    In Perl, unescaped square brackets may also appear as part of class names. For
2463    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2464    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2465    seem right at all. PCRE does not allow closing square brackets in POSIX class
2466    names.
2467    
2468  Arguments:  Arguments:
2469    ptr      pointer to the initial [    ptr      pointer to the initial [
2470    endptr   where to return the end pointer    endptr   where to return the end pointer
# Line 1793  int terminator; /* Don't combin Line 2479  int terminator; /* Don't combin
2479  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2480  for (++ptr; *ptr != 0; ptr++)  for (++ptr; *ptr != 0; ptr++)
2481    {    {
2482    if (*ptr == '\\' && ptr[1] == ']') ptr++; else    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2483        ptr++;
2484      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2485      else
2486      {      {
2487      if (*ptr == ']') return FALSE;      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
     if (*ptr == terminator && ptr[1] == ']')  
2488        {        {
2489        *endptr = ptr;        *endptr = ptr;
2490        return TRUE;        return TRUE;
2491        }        }
2492        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2493             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2494              ptr[1] == CHAR_EQUALS_SIGN) &&
2495            check_posix_syntax(ptr, endptr))
2496          return FALSE;
2497      }      }
2498    }    }
2499  return FALSE;  return FALSE;
# Line 1883  while ((ptr = (uschar *)find_recurse(ptr Line 2576  while ((ptr = (uschar *)find_recurse(ptr
2576    
2577    /* See if this recursion is on the forward reference list. If so, adjust the    /* See if this recursion is on the forward reference list. If so, adjust the
2578    reference. */    reference. */
2579    
2580    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
2581      {      {
2582      offset = GET(hc, 0);      offset = GET(hc, 0);
# Line 1929  auto_callout(uschar *code, const uschar Line 2622  auto_callout(uschar *code, const uschar
2622  {  {
2623  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2624  *code++ = 255;  *code++ = 255;
2625  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2626  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2627  return code + 2*LINK_SIZE;  return code + 2*LINK_SIZE;
2628  }  }
2629    
# Line 1955  Returns: nothing Line 2648  Returns: nothing
2648  static void  static void
2649  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)
2650  {  {
2651  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2652  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2653  }  }
2654    
# Line 1987  get_othercase_range(unsigned int *cptr, Line 2680  get_othercase_range(unsigned int *cptr,
2680  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2681    
2682  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2683    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2684    
2685  if (c > d) return FALSE;  if (c > d) return FALSE;
2686    
# Line 1996  next = othercase + 1; Line 2689  next = othercase + 1;
2689    
2690  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2691    {    {
2692    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2693    next++;    next++;
2694    }    }
2695    
# Line 2005  for (++c; c <= d; c++) Line 2698  for (++c; c <= d; c++)
2698    
2699  return TRUE;  return TRUE;
2700  }  }
2701    
2702    
2703    
2704    /*************************************************
2705    *        Check a character and a property        *
2706    *************************************************/
2707    
2708    /* This function is called by check_auto_possessive() when a property item
2709    is adjacent to a fixed character.
2710    
2711    Arguments:
2712      c            the character
2713      ptype        the property type
2714      pdata        the data for the type
2715      negated      TRUE if it's a negated property (\P or \p{^)
2716    
2717    Returns:       TRUE if auto-possessifying is OK
2718    */
2719    
2720    static BOOL
2721    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2722    {
2723    const ucd_record *prop = GET_UCD(c);
2724    switch(ptype)
2725      {
2726      case PT_LAMP:
2727      return (prop->chartype == ucp_Lu ||
2728              prop->chartype == ucp_Ll ||
2729              prop->chartype == ucp_Lt) == negated;
2730    
2731      case PT_GC:
2732      return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;
2733    
2734      case PT_PC:
2735      return (pdata == prop->chartype) == negated;
2736    
2737      case PT_SC:
2738      return (pdata == prop->script) == negated;
2739    
2740      /* These are specials */
2741    
2742      case PT_ALNUM:
2743      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2744              _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;
2745    
2746      case PT_SPACE:    /* Perl space */
2747      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2748              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2749              == negated;
2750    
2751      case PT_PXSPACE:  /* POSIX space */
2752      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2753              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2754              c == CHAR_FF || c == CHAR_CR)
2755              == negated;
2756    
2757      case PT_WORD:
2758      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2759              _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2760              c == CHAR_UNDERSCORE) == negated;
2761      }
2762    return FALSE;
2763    }
2764  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2765    
2766    
# Line 2018  whether the next thing could possibly ma Line 2774  whether the next thing could possibly ma
2774  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
2775    
2776  Arguments:  Arguments:
2777    op_code       the repeated op code    previous      pointer to the repeated opcode
   this          data for this item, depends on the opcode  
2778    utf8          TRUE in UTF-8 mode    utf8          TRUE in UTF-8 mode
   utf8_char     used for utf8 character bytes, NULL if not relevant  
2779    ptr           next character in pattern    ptr           next character in pattern
2780    options       options bits    options       options bits
2781    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 2030  Returns: TRUE if possessifying is Line 2784  Returns: TRUE if possessifying is
2784  */  */
2785    
2786  static BOOL  static BOOL
2787  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
2788    const uschar *ptr, int options, compile_data *cd)    int options, compile_data *cd)
2789  {  {
2790  int next;  int c, next;
2791    int op_code = *previous++;
2792    
2793  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
2794    
# Line 2042  if ((options & PCRE_EXTENDED) != 0) Line 2797  if ((options & PCRE_EXTENDED) != 0)
2797    for (;;)    for (;;)
2798      {      {
2799      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2800      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2801        {        {
2802        while (*(++ptr) != 0)        ptr++;
2803          while (*ptr != 0)
2804            {
2805          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2806            ptr++;
2807    #ifdef SUPPORT_UTF8
2808            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2809    #endif
2810            }
2811        }        }
2812      else break;      else break;
2813      }      }
# Line 2054  if ((options & PCRE_EXTENDED) != 0) Line 2816  if ((options & PCRE_EXTENDED) != 0)
2816  /* 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
2817  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2818    
2819  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2820    {    {
2821    int temperrorcode = 0;    int temperrorcode = 0;
2822    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
# Line 2079  if ((options & PCRE_EXTENDED) != 0) Line 2841  if ((options & PCRE_EXTENDED) != 0)
2841    for (;;)    for (;;)
2842      {      {
2843      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2844      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2845        {        {
2846        while (*(++ptr) != 0)        ptr++;
2847          while (*ptr != 0)
2848            {
2849          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2850            ptr++;
2851    #ifdef SUPPORT_UTF8
2852            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2853    #endif
2854            }
2855        }        }
2856      else break;      else break;
2857      }      }
# Line 2090  if ((options & PCRE_EXTENDED) != 0) Line 2859  if ((options & PCRE_EXTENDED) != 0)
2859    
2860  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
2861    
2862  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2863    return FALSE;    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2864        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. */  
   
2865    
2866  /* Handle cases when the next item is a character. */  /* Now compare the next item with the previous opcode. First, handle cases when
2867    the next item is a character. */
2868    
2869  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
2870    {    {
2871    case OP_CHAR:    case OP_CHAR:
2872  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2873    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2874    #else
2875      c = *previous;
2876  #endif  #endif
2877    return item != next;    return c != next;
2878    
2879    /* 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
2880    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
2881    high-valued characters. */    high-valued characters. */
2882    
2883    case OP_CHARNC:    case OP_CHARI:
2884  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2885    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2886    #else
2887      c = *previous;
2888  #endif  #endif
2889    if (item == next) return FALSE;    if (c == next) return FALSE;
2890  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2891    if (utf8)    if (utf8)
2892      {      {
2893      unsigned int othercase;      unsigned int othercase;
2894      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2895  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2896      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
2897  #else  #else
2898      othercase = NOTACHAR;      othercase = NOTACHAR;
2899  #endif  #endif
2900      return (unsigned int)item != othercase;      return (unsigned int)c != othercase;
2901      }      }
2902    else    else
2903  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2904    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2905    
2906    /* 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
2907      opcodes are not used for multi-byte characters, because they are coded using
2908      an XCLASS instead. */
2909    
2910    case OP_NOT:    case OP_NOT:
2911    if (item == next) return TRUE;    return (c = *previous) == next;
2912    if ((options & PCRE_CASELESS) == 0) return FALSE;  
2913      case OP_NOTI:
2914      if ((c = *previous) == next) return TRUE;
2915  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2916    if (utf8)    if (utf8)
2917      {      {
2918      unsigned int othercase;      unsigned int othercase;
2919      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2920  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2921      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
2922  #else  #else
2923      othercase = NOTACHAR;      othercase = NOTACHAR;
2924  #endif  #endif
2925      return (unsigned int)item == othercase;      return (unsigned int)c == othercase;
2926      }      }
2927    else    else
2928  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2929    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2930    
2931      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2932      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
2933    
2934    case OP_DIGIT:    case OP_DIGIT:
2935    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
# Line 2196  if (next >= 0) switch(op_code) Line 2972  if (next >= 0) switch(op_code)
2972      case 0x202f:      case 0x202f:
2973      case 0x205f:      case 0x205f:
2974      case 0x3000:      case 0x3000:
2975      return op_code != OP_HSPACE;      return op_code == OP_NOT_HSPACE;
2976      default:      default:
2977      return op_code == OP_HSPACE;      return op_code != OP_NOT_HSPACE;
2978      }      }
2979    
2980      case OP_ANYNL:
2981    case OP_VSPACE:    case OP_VSPACE:
2982    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
2983    switch(next)    switch(next)
# Line 2212  if (next >= 0) switch(op_code) Line 2989  if (next >= 0) switch(op_code)
2989      case 0x85:      case 0x85:
2990      case 0x2028:      case 0x2028:
2991      case 0x2029:      case 0x2029:
2992      return op_code != OP_VSPACE;      return op_code == OP_NOT_VSPACE;
2993      default:      default:
2994      return op_code == OP_VSPACE;      return op_code != OP_NOT_VSPACE;
2995      }      }
2996    
2997    #ifdef SUPPORT_UCP
2998      case OP_PROP:
2999      return check_char_prop(next, previous[0], previous[1], FALSE);
3000    
3001      case OP_NOTPROP:
3002      return check_char_prop(next, previous[0], previous[1], TRUE);
3003    #endif
3004    
3005    default:    default:
3006    return FALSE;    return FALSE;
3007    }    }
3008    
3009    
3010  /* 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
3011    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
3012    generated only when PCRE_UCP is *not* set, that is, when only ASCII
3013    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
3014    replaced by OP_PROP codes when PCRE_UCP is set. */
3015    
3016  switch(op_code)  switch(op_code)
3017    {    {
3018    case OP_CHAR:    case OP_CHAR:
3019    case OP_CHARNC:    case OP_CHARI:
3020  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3021    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
3022    #else
3023      c = *previous;
3024  #endif  #endif
3025    switch(-next)    switch(-next)
3026      {      {
3027      case ESC_d:      case ESC_d:
3028      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
3029    
3030      case ESC_D:      case ESC_D:
3031      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
3032    
3033      case ESC_s:      case ESC_s:
3034      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
3035    
3036      case ESC_S:      case ESC_S:
3037      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
3038    
3039      case ESC_w:      case ESC_w:
3040      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
3041    
3042      case ESC_W:      case ESC_W:
3043      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
3044    
3045      case ESC_h:      case ESC_h:
3046      case ESC_H:      case ESC_H:
3047      switch(item)      switch(c)
3048        {        {
3049        case 0x09:        case 0x09:
3050        case 0x20:        case 0x20:
# Line 2281  switch(op_code) Line 3072  switch(op_code)
3072    
3073      case ESC_v:      case ESC_v:
3074      case ESC_V:      case ESC_V:
3075      switch(item)      switch(c)
3076        {        {
3077        case 0x0a:        case 0x0a:
3078        case 0x0b:        case 0x0b:
# Line 2295  switch(op_code) Line 3086  switch(op_code)
3086        return -next == ESC_v;        return -next == ESC_v;
3087        }        }
3088    
3089        /* When PCRE_UCP is set, these values get generated for \d etc. Find
3090        their substitutions and process them. The result will always be either
3091        -ESC_p or -ESC_P. Then fall through to process those values. */
3092    
3093    #ifdef SUPPORT_UCP
3094        case ESC_du:
3095        case ESC_DU:
3096        case ESC_wu:
3097        case ESC_WU:
3098        case ESC_su:
3099        case ESC_SU:
3100          {
3101          int temperrorcode = 0;
3102          ptr = substitutes[-next - ESC_DU];
3103          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
3104          if (temperrorcode != 0) return FALSE;
3105          ptr++;    /* For compatibility */
3106          }
3107        /* Fall through */
3108    
3109        case ESC_p:
3110        case ESC_P:
3111          {
3112          int ptype, pdata, errorcodeptr;
3113          BOOL negated;
3114    
3115          ptr--;      /* Make ptr point at the p or P */
3116          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
3117          if (ptype < 0) return FALSE;
3118          ptr++;      /* Point past the final curly ket */
3119    
3120          /* If the property item is optional, we have to give up. (When generated
3121          from \d etc by PCRE_UCP, this test will have been applied much earlier,
3122          to the original \d etc. At this point, ptr will point to a zero byte. */
3123    
3124          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3125            strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3126              return FALSE;
3127    
3128          /* Do the property check. */
3129    
3130          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
3131          }
3132    #endif
3133    
3134      default:      default:
3135      return FALSE;      return FALSE;
3136      }      }
3137    
3138      /* In principle, support for Unicode properties should be integrated here as
3139      well. It means re-organizing the above code so as to get hold of the property
3140      values before switching on the op-code. However, I wonder how many patterns
3141      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
3142      these op-codes are never generated.) */
3143    
3144    case OP_DIGIT:    case OP_DIGIT:
3145    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
3146           next == -ESC_h || next == -ESC_v;           next == -ESC_h || next == -ESC_v || next == -ESC_R;
3147    
3148    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
3149    return next == -ESC_d;    return next == -ESC_d;
3150    
3151    case OP_WHITESPACE:    case OP_WHITESPACE:
3152    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
3153    
3154    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
3155    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
3156    
3157    case OP_HSPACE:    case OP_HSPACE:
3158    return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
3159             next == -ESC_w || next == -ESC_v || next == -ESC_R;
3160    
3161    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
3162    return next == -ESC_h;    return next == -ESC_h;
3163    
3164    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
3165      case OP_ANYNL:
3166    case OP_VSPACE:    case OP_VSPACE:
3167    return next == -ESC_V || next == -ESC_d || next == -ESC_w;    return next == -ESC_V || next == -ESC_d || next == -ESC_w;
3168    
3169    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
3170    return next == -ESC_v;    return next == -ESC_v || next == -ESC_R;
3171    
3172    case OP_WORDCHAR:    case OP_WORDCHAR:
3173    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
3174             next == -ESC_v || next == -ESC_R;
3175    
3176    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
3177    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2358  Arguments: Line 3203  Arguments:
3203    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
3204    reqbyteptr     set to the last literal character required, else < 0    reqbyteptr     set to the last literal character required, else < 0
3205    bcptr          points to current branch chain    bcptr          points to current branch chain
3206      cond_depth     conditional nesting depth
3207    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
3208    lengthptr      NULL during the real compile phase    lengthptr      NULL during the real compile phase
3209                   points to length accumulator during pre-compile phase                   points to length accumulator during pre-compile phase
# Line 2369  Returns: TRUE on success Line 3215  Returns: TRUE on success
3215  static BOOL  static BOOL
3216  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
3217    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
3218    compile_data *cd, int *lengthptr)    int cond_depth, compile_data *cd, int *lengthptr)
3219  {  {
3220  int repeat_type, op_type;  int repeat_type, op_type;
3221  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
# Line 2378  int greedy_default, greedy_non_default; Line 3224  int greedy_default, greedy_non_default;
3224  int firstbyte, reqbyte;  int firstbyte, reqbyte;
3225  int zeroreqbyte, zerofirstbyte;  int zeroreqbyte, zerofirstbyte;
3226  int req_caseopt, reqvary, tempreqvary;  int req_caseopt, reqvary, tempreqvary;
3227  int options = *optionsptr;  int options = *optionsptr;               /* May change dynamically */
3228  int after_manual_callout = 0;  int after_manual_callout = 0;
3229  int length_prevgroup = 0;  int length_prevgroup = 0;
3230  register int c;  register int c;
# Line 2390  BOOL inescq = FALSE; Line 3236  BOOL inescq = FALSE;
3236  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
3237  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
3238  const uschar *tempptr;  const uschar *tempptr;
3239    const uschar *nestptr = NULL;
3240  uschar *previous = NULL;  uschar *previous = NULL;
3241  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
3242  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
3243  uschar classbits[32];  uschar classbits[32];
3244    
3245    /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3246    must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3247    dynamically as we process the pattern. */
3248    
3249  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3250  BOOL class_utf8;  BOOL class_utf8;
3251  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
# Line 2403  uschar *class_utf8data_base; Line 3254  uschar *class_utf8data_base;
3254  uschar utf8_char[6];  uschar utf8_char[6];
3255  #else  #else
3256  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
 uschar *utf8_char = NULL;  
3257  #endif  #endif
3258    
3259  #ifdef DEBUG  #ifdef PCRE_DEBUG
3260  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
3261  #endif  #endif
3262    
# Line 2454  for (;; ptr++) Line 3304  for (;; ptr++)
3304    int subfirstbyte;    int subfirstbyte;
3305    int terminator;    int terminator;
3306    int mclength;    int mclength;
3307      int tempbracount;
3308    uschar mcbuffer[8];    uschar mcbuffer[8];
3309    
3310    /* Get next byte in the pattern */    /* Get next byte in the pattern */
3311    
3312    c = *ptr;    c = *ptr;
3313    
3314      /* If we are at the end of a nested substitution, revert to the outer level
3315      string. Nesting only happens one level deep. */
3316    
3317      if (c == 0 && nestptr != NULL)
3318        {
3319        ptr = nestptr;
3320        nestptr = NULL;
3321        c = *ptr;
3322        }
3323    
3324    /* 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
3325    previous cycle of this loop. */    previous cycle of this loop. */
3326    
3327    if (lengthptr != NULL)    if (lengthptr != NULL)
3328      {      {
3329  #ifdef DEBUG  #ifdef PCRE_DEBUG
3330      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3331  #endif  #endif
3332      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
3333        {        {
3334        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3335        goto FAILED;        goto FAILED;
# Line 2490  for (;; ptr++) Line 3351  for (;; ptr++)
3351        goto FAILED;        goto FAILED;
3352        }        }
3353    
3354      *lengthptr += code - last_code;      *lengthptr += (int)(code - last_code);
3355      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));
3356    
3357      /* If "previous" is set and it is not at the start of the work space, move      /* If "previous" is set and it is not at the start of the work space, move
# Line 2517  for (;; ptr++) Line 3378  for (;; ptr++)
3378    /* 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
3379    reference list. */    reference list. */
3380    
3381    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
3382      {      {
3383      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3384      goto FAILED;      goto FAILED;
# Line 2527  for (;; ptr++) Line 3388  for (;; ptr++)
3388    
3389    if (inescq && c != 0)    if (inescq && c != 0)
3390      {      {
3391      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3392        {        {
3393        inescq = FALSE;        inescq = FALSE;
3394        ptr++;        ptr++;
# Line 2553  for (;; ptr++) Line 3414  for (;; ptr++)
3414    /* 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
3415    a quantifier. */    a quantifier. */
3416    
3417    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
3418      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
3419        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
3420    
3421    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
3422         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2564  for (;; ptr++) Line 3426  for (;; ptr++)
3426      previous_callout = NULL;      previous_callout = NULL;
3427      }      }
3428    
3429    /* In extended mode, skip white space and comments */    /* In extended mode, skip white space and comments. */
3430    
3431    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3432      {      {
3433      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
3434      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
3435        {        {
3436        while (*(++ptr) != 0)        ptr++;
3437          while (*ptr != 0)
3438          {          {
3439          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3440            ptr++;
3441    #ifdef SUPPORT_UTF8
3442            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
3443    #endif
3444          }          }
3445        if (*ptr != 0) continue;        if (*ptr != 0) continue;
3446    
# Line 2594  for (;; ptr++) Line 3461  for (;; ptr++)
3461      {      {
3462      /* ===================================================================*/      /* ===================================================================*/
3463      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
3464      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
3465      case ')':      case CHAR_RIGHT_PARENTHESIS:
3466      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
3467      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
3468      *codeptr = code;      *codeptr = code;
# Line 2607  for (;; ptr++) Line 3474  for (;; ptr++)
3474          *errorcodeptr = ERR20;          *errorcodeptr = ERR20;
3475          goto FAILED;          goto FAILED;
3476          }          }
3477        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += (int)(code - last_code);   /* To include callout length */
3478        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
3479        }        }
3480      return TRUE;      return TRUE;
# Line 2617  for (;; ptr++) Line 3484  for (;; ptr++)
3484      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
3485      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3486    
3487      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
3488        previous = NULL;
3489      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3490        {        {
3491        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3492          *code++ = OP_CIRCM;
3493        }        }
3494      previous = NULL;      else *code++ = OP_CIRC;
     *code++ = OP_CIRC;  
3495      break;      break;
3496    
3497      case '$':      case CHAR_DOLLAR_SIGN:
3498      previous = NULL;      previous = NULL;
3499      *code++ = OP_DOLL;      *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
3500      break;      break;
3501    
3502      /* 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
3503      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
3504    
3505      case '.':      case CHAR_DOT:
3506      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3507      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
3508      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3509      previous = code;      previous = code;
3510      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
3511      break;      break;
3512    
3513    
# Line 2654  for (;; ptr++) Line 3522  for (;; ptr++)
3522      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,
3523      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
3524      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.
     */  
3525    
3526      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
3527        default (Perl) mode, it is treated as a data character. */
3528    
3529        case CHAR_RIGHT_SQUARE_BRACKET:
3530        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3531          {
3532          *errorcodeptr = ERR64;
3533          goto FAILED;
3534          }
3535        goto NORMAL_CHAR;
3536    
3537        case CHAR_LEFT_SQUARE_BRACKET:
3538      previous = code;      previous = code;
3539    
3540      /* 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
3541      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. */
3542    
3543      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3544             ptr[1] == CHAR_EQUALS_SIGN) &&
3545          check_posix_syntax(ptr, &tempptr))          check_posix_syntax(ptr, &tempptr))
3546        {        {
3547        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
3548        goto FAILED;        goto FAILED;
3549        }        }
3550    
# Line 2677  for (;; ptr++) Line 3556  for (;; ptr++)
3556      for (;;)      for (;;)
3557        {        {
3558        c = *(++ptr);        c = *(++ptr);
3559        if (c == '\\')        if (c == CHAR_BACKSLASH)
3560          {          {
3561          if (ptr[1] == 'E') ptr++;          if (ptr[1] == CHAR_E)
3562            else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;            ptr++;
3563              else break;          else if (strncmp((const char *)ptr+1,
3564                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
3565              ptr += 3;
3566            else
3567              break;
3568          }          }
3569        else if (!negate_class && c == '^')        else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
3570          negate_class = TRUE;          negate_class = TRUE;
3571        else break;        else break;
3572        }        }
3573    
3574        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
3575        an initial ']' is taken as a data character -- the code below handles
3576        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3577        [^] must match any character, so generate OP_ALLANY. */
3578    
3579        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3580            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3581          {
3582          *code++ = negate_class? OP_ALLANY : OP_FAIL;
3583          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3584          zerofirstbyte = firstbyte;
3585          break;
3586          }
3587    
3588      /* 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
3589      negation flag at the end, so that support for characters > 255 works      negation flag at the end, so that support for characters > 255 works
3590      correctly (they are all included in the class). */      correctly (they are all included in the class). */
# Line 2745  for (;; ptr++) Line 3642  for (;; ptr++)
3642    
3643        if (inescq)        if (inescq)
3644          {          {
3645          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3646            {            {
3647            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
3648            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2760  for (;; ptr++) Line 3657  for (;; ptr++)
3657        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3658        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3659    
3660        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3661            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3662            check_posix_syntax(ptr, &tempptr))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3663          {          {
3664          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3665          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3666          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3667          uschar pbits[32];          uschar pbits[32];
3668    
3669          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3670            {            {
3671            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3672            goto FAILED;            goto FAILED;
3673            }            }
3674    
3675          ptr += 2;          ptr += 2;
3676          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3677            {            {
3678            local_negate = TRUE;            local_negate = TRUE;
3679            should_flip_negation = TRUE;  /* Note negative special */            should_flip_negation = TRUE;  /* Note negative special */
3680            ptr++;            ptr++;
3681            }            }
3682    
3683          posix_class = check_posix_name(ptr, tempptr - ptr);          posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3684          if (posix_class < 0)          if (posix_class < 0)
3685            {            {
3686            *errorcodeptr = ERR30;            *errorcodeptr = ERR30;
# Line 2797  for (;; ptr++) Line 3694  for (;; ptr++)
3694          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3695            posix_class = 0;            posix_class = 0;
3696    
3697          /* 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
3698          because we may be adding and subtracting from it, and we don't want to          different escape sequences that use Unicode properties. */
3699          subtract bits that may be in the main map already. At the end we or the  
3700          result into the bit map that is being built. */  #ifdef SUPPORT_UCP
3701            if ((options & PCRE_UCP) != 0)
3702              {
3703              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3704              if (posix_substitutes[pc] != NULL)
3705                {
3706                nestptr = tempptr + 1;
3707                ptr = posix_substitutes[pc] - 1;
3708                continue;
3709                }
3710              }
3711    #endif
3712            /* In the non-UCP case, we build the bit map for the POSIX class in a
3713            chunk of local store because we may be adding and subtracting from it,
3714            and we don't want to subtract bits that may be in the main map already.
3715            At the end we or the result into the bit map that is being built. */
3716    
3717          posix_class *= 3;          posix_class *= 3;
3718    
# Line 2844  for (;; ptr++) Line 3756  for (;; ptr++)
3756    
3757        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3758        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
3759        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
3760        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
3761        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
3762        character in them, so set class_charcount bigger than one. */        are either treated as literal characters (by default), or are faulted if
3763          PCRE_EXTRA is set. */
3764    
3765        if (c == '\\')        if (c == CHAR_BACKSLASH)
3766          {          {
3767          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3768          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3769    
3770          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 */  
3771          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3772            {            {
3773            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3774              {              {
3775              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3776              }              }
# Line 2873  for (;; ptr++) Line 3784  for (;; ptr++)
3784            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
3785            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3786    
3787            /* Save time by not doing this in the pre-compile phase. */            switch (-c)
   
           if (lengthptr == NULL) switch (-c)  
3788              {              {
3789    #ifdef SUPPORT_UCP
3790                case ESC_du:     /* These are the values given for \d etc */
3791                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3792                case ESC_wu:     /* escape sequence with an appropriate \p */
3793                case ESC_WU:     /* or \P to test Unicode properties instead */
3794                case ESC_su:     /* of the default ASCII testing. */
3795                case ESC_SU:
3796                nestptr = ptr;
3797                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3798                class_charcount -= 2;                /* Undo! */
3799                continue;
3800    #endif
3801              case ESC_d:              case ESC_d:
3802              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3803              continue;              continue;
# Line 2895  for (;; ptr++) Line 3816  for (;; ptr++)
3816              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3817              continue;              continue;
3818    
3819                /* Perl 5.004 onwards omits VT from \s, but we must preserve it
3820                if it was previously set by something earlier in the character
3821                class. */
3822    
3823              case ESC_s:              case ESC_s:
3824              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];              classbits[0] |= cbits[cbit_space];
3825              classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= cbits[cbit_space+1] & ~0x08;
3826                for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3827              continue;              continue;
3828    
3829              case ESC_S:              case ESC_S:
# Line 2906  for (;; ptr++) Line 3832  for (;; ptr++)
3832              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3833              continue;              continue;
3834    
3835              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)  
             {  
3836              SETBIT(classbits, 0x09); /* VT */              SETBIT(classbits, 0x09); /* VT */
3837              SETBIT(classbits, 0x20); /* SPACE */              SETBIT(classbits, 0x20); /* SPACE */
3838              SETBIT(classbits, 0xa0); /* NSBP */              SETBIT(classbits, 0xa0); /* NSBP */
# Line 2943  for (;; ptr++) Line 3856  for (;; ptr++)
3856                }                }
3857  #endif  #endif
3858              continue;              continue;
             }  
3859    
3860            if (-c == ESC_H)              case ESC_H:
             {  
3861              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3862                {                {
3863                int x = 0xff;                int x = 0xff;
# Line 2988  for (;; ptr++) Line 3899  for (;; ptr++)
3899                }                }
3900  #endif  #endif
3901              continue;              continue;
             }  
3902    
3903            if (-c == ESC_v)              case ESC_v:
             {  
3904              SETBIT(classbits, 0x0a); /* LF */              SETBIT(classbits, 0x0a); /* LF */
3905              SETBIT(classbits, 0x0b); /* VT */              SETBIT(classbits, 0x0b); /* VT */
3906              SETBIT(classbits, 0x0c); /* FF */              SETBIT(classbits, 0x0c); /* FF */
# Line 3007  for (;; ptr++) Line 3916  for (;; ptr++)
3916                }                }
3917  #endif  #endif
3918              continue;              continue;
             }  
3919    
3920            if (-c == ESC_V)              case ESC_V:
             {  
3921              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3922                {                {
3923                int x = 0xff;                int x = 0xff;
# Line 3040  for (;; ptr++) Line 3947  for (;; ptr++)
3947                }                }
3948  #endif  #endif
3949              continue;              continue;
             }  
   
           /* We need to deal with \P and \p in both phases. */  
3950    
3951  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3952            if (-c == ESC_p || -c == ESC_P)              case ESC_p:
3953              {              case ESC_P:
3954              BOOL negated;                {
3955              int pdata;                BOOL negated;
3956              int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                int pdata;
3957              if (ptype < 0) goto FAILED;                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3958              class_utf8 = TRUE;                if (ptype < 0) goto FAILED;
3959              *class_utf8data++ = ((-c == ESC_p) != negated)?                class_utf8 = TRUE;
3960                XCL_PROP : XCL_NOTPROP;                *class_utf8data++ = ((-c == ESC_p) != negated)?
3961              *class_utf8data++ = ptype;                  XCL_PROP : XCL_NOTPROP;
3962              *class_utf8data++ = pdata;                *class_utf8data++ = ptype;
3963              class_charcount -= 2;   /* Not a < 256 character */                *class_utf8data++ = pdata;
3964              continue;                class_charcount -= 2;   /* Not a < 256 character */
3965              }                continue;
3966                  }
3967  #endif  #endif
3968            /* Unrecognized escapes are faulted if PCRE is running in its              /* Unrecognized escapes are faulted if PCRE is running in its
3969            strict mode. By default, for compatibility with Perl, they are              strict mode. By default, for compatibility with Perl, they are
3970            treated as literals. */              treated as literals. */
3971    
3972            if ((options & PCRE_EXTRA) != 0)              default:
3973              {              if ((options & PCRE_EXTRA) != 0)
3974              *errorcodeptr = ERR7;                {
3975              goto FAILED;                *errorcodeptr = ERR7;
3976                  goto FAILED;
3977                  }
3978                class_charcount -= 2;  /* Undo the default count from above */
3979                c = *ptr;              /* Get the final character and fall through */
3980                break;
3981              }              }
   
           class_charcount -= 2;  /* Undo the default count from above */  
           c = *ptr;              /* Get the final character and fall through */  
3982            }            }
3983    
3984          /* 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 3085  for (;; ptr++) Line 3992  for (;; ptr++)
3992        entirely. The code for handling \Q and \E is messy. */        entirely. The code for handling \Q and \E is messy. */
3993    
3994        CHECK_RANGE:        CHECK_RANGE:
3995        while (ptr[1] == '\\' && ptr[2] == 'E')        while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3996          {          {
3997          inescq = FALSE;          inescq = FALSE;
3998          ptr += 2;          ptr += 2;
# Line 3095  for (;; ptr++) Line 4002  for (;; ptr++)
4002    
4003        /* Remember \r or \n */        /* Remember \r or \n */
4004    
4005        if (c == '\r' || c == '\n') cd->external_flags |= PCRE_HASCRORLF;        if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
4006    
4007        /* Check for range */        /* Check for range */
4008    
4009        if (!inescq && ptr[1] == '-')        if (!inescq && ptr[1] == CHAR_MINUS)
4010          {          {
4011          int d;          int d;
4012          ptr += 2;          ptr += 2;
4013          while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
4014    
4015          /* 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
4016          mode. */          mode. */
4017    
4018          while (*ptr == '\\' && ptr[1] == 'Q')          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
4019            {            {
4020            ptr += 2;            ptr += 2;
4021            if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }            if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
4022                { ptr += 2; continue; }
4023            inescq = TRUE;            inescq = TRUE;
4024            break;            break;
4025            }            }
4026    
4027          if (*ptr == 0 || (!inescq && *ptr == ']'))          if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
4028            {            {
4029            ptr = oldptr;            ptr = oldptr;
4030            goto LONE_SINGLE_CHARACTER;            goto LONE_SINGLE_CHARACTER;
# Line 3135  for (;; ptr++) Line 4043  for (;; ptr++)
4043          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
4044          in such circumstances. */          in such circumstances. */
4045    
4046          if (!inescq && d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
4047            {            {
4048            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
4049            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
4050    
4051            /* \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 */  
4052    
4053            if (d < 0)            if (d < 0)
4054              {              {
4055              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  
4056                {                {
4057                ptr = oldptr;                ptr = oldptr;
4058                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 3168  for (;; ptr++) Line 4073  for (;; ptr++)
4073    
4074          /* Remember \r or \n */          /* Remember \r or \n */
4075    
4076          if (d == '\r' || d == '\n') cd->external_flags |= PCRE_HASCRORLF;          if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
4077    
4078          /* 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
4079          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 3288  for (;; ptr++) Line 4193  for (;; ptr++)
4193          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
4194            {            {
4195            unsigned int othercase;            unsigned int othercase;
4196            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)            if ((othercase = UCD_OTHERCASE(c)) != c)
4197              {              {
4198              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
4199              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 3313  for (;; ptr++) Line 4218  for (;; ptr++)
4218          }          }
4219        }        }
4220    
4221      /* 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.
4222        If we are at the end of an internal nested string, revert to the outer
4223        string. */
4224    
4225        while (((c = *(++ptr)) != 0 ||
4226               (nestptr != NULL &&
4227                 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != 0)) &&
4228               (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
4229    
4230      while ((c = *(++ptr)) != 0 && (c != ']' || inescq));      /* Check for missing terminating ']' */
4231    
4232      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)
4233        {        {
4234        *errorcodeptr = ERR6;        *errorcodeptr = ERR6;
4235        goto FAILED;        goto FAILED;
4236        }        }
4237    
   
 /* 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  
   
   
4238      /* 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
4239      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
4240      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 3349  we set the flag only if there is a liter Line 4242  we set the flag only if there is a liter
4242    
4243      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
4244      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
4245      operate on single-bytes only. This is an historical hangover. Maybe one day      operate on single-bytes characters only. This is an historical hangover.
4246      we can tidy these opcodes to handle multi-byte characters.      Maybe one day we can tidy these opcodes to handle multi-byte characters.
4247    
4248      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
4249      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.
4250      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
4251      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
4252      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
4253      reqbyte, save the previous value for reinstating. */      of reqbyte, save the previous value for reinstating. */
4254    
4255  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4256      if (class_charcount == 1 && !class_utf8 &&      if (class_charcount == 1 && !class_utf8 &&
# Line 3368  we set the flag only if there is a liter Line 4261  we set the flag only if there is a liter
4261        {        {
4262        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
4263    
4264        /* The OP_NOT opcode works on one-byte characters only. */        /* The OP_NOT[I] opcodes work on one-byte characters only. */
4265    
4266        if (negate_class)        if (negate_class)
4267          {          {
4268          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
4269          zerofirstbyte = firstbyte;          zerofirstbyte = firstbyte;
4270          *code++ = OP_NOT;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
4271          *code++ = class_lastchar;          *code++ = class_lastchar;
4272          break;          break;
4273          }          }
# Line 3405  we set the flag only if there is a liter Line 4298  we set the flag only if there is a liter
4298    
4299      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
4300      extended class, with its own opcode, unless there was a negated special      extended class, with its own opcode, unless there was a negated special
4301      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
4302      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
4303      (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
4304      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
4305        actual compiled code. */
4306    
4307  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4308      if (class_utf8 && !should_flip_negation)      if (class_utf8 && (!should_flip_negation || (options & PCRE_UCP) != 0))
4309        {        {
4310        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
4311        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 3437  we set the flag only if there is a liter Line 4331  we set the flag only if there is a liter
4331        }        }
4332  #endif  #endif
4333    
4334      /* 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
4335      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
4336      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
4337      map into the code vector, negating it if necessary. */      (non-UCP) in the class. Then copy the 32-byte map into the code vector,
4338        negating it if necessary. */
4339    
4340      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
4341      if (negate_class)      if (negate_class)
# Line 3460  we set the flag only if there is a liter Line 4355  we set the flag only if there is a liter
4355      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
4356      has been tested above. */      has been tested above. */
4357    
4358      case '{':      case CHAR_LEFT_CURLY_BRACKET:
4359      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
4360      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
4361      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
4362      goto REPEAT;      goto REPEAT;
4363    
4364      case '*':      case CHAR_ASTERISK:
4365      repeat_min = 0;      repeat_min = 0;
4366      repeat_max = -1;      repeat_max = -1;
4367      goto REPEAT;      goto REPEAT;
4368    
4369      case '+':      case CHAR_PLUS:
4370      repeat_min = 1;      repeat_min = 1;
4371      repeat_max = -1;      repeat_max = -1;
4372      goto REPEAT;      goto REPEAT;
4373    
4374      case '?':      case CHAR_QUESTION_MARK:
4375      repeat_min = 0;      repeat_min = 0;
4376      repeat_max = 1;      repeat_max = 1;
4377    
# Line 3500  we set the flag only if there is a liter Line 4395  we set the flag only if there is a liter
4395      op_type = 0;                    /* Default single-char op codes */      op_type = 0;                    /* Default single-char op codes */
4396      possessive_quantifier = FALSE;  /* Default not possessive quantifier */      possessive_quantifier = FALSE;  /* Default not possessive quantifier */
4397    
4398      /* 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
4399      for an inserted OP_ONCE for the additional '+' extension. */      insert something before it. */
4400    
4401      tempcode = previous;      tempcode = previous;
4402    
# Line 3511  we set the flag only if there is a liter Line 4406  we set the flag only if there is a liter
4406      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
4407      repeat type to the non-default. */      repeat type to the non-default. */
4408    
4409      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
4410        {        {
4411        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
4412        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
4413        ptr++;        ptr++;
4414        }        }
4415      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
4416        {        {
4417        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
4418        ptr++;        ptr++;
4419        }        }
4420      else repeat_type = greedy_default;      else repeat_type = greedy_default;
4421    
4422        /* If previous was a recursion call, wrap it in atomic brackets so that
4423        previous becomes the atomic group. All recursions were so wrapped in the
4424        past, but it no longer happens for non-repeated recursions. In fact, the
4425        repeated ones could be re-implemented independently so as not to need this,
4426        but for the moment we rely on the code for repeating groups. */
4427    
4428        if (*previous == OP_RECURSE)
4429          {
4430          memmove(previous + 1 + LINK_SIZE, previous, 1 + LINK_SIZE);
4431          *previous = OP_ONCE;
4432          PUT(previous, 1, 2 + 2*LINK_SIZE);
4433          previous[2 + 2*LINK_SIZE] = OP_KET;
4434          PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
4435          code += 2 + 2 * LINK_SIZE;
4436          length_prevgroup = 3 + 3*LINK_SIZE;
4437    
4438          /* When actually compiling, we need to check whether this was a forward
4439          reference, and if so, adjust the offset. */
4440    
4441          if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
4442            {
4443            int offset = GET(cd->hwm, -LINK_SIZE);
4444            if (offset == previous + 1 - cd->start_code)
4445              PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
4446            }
4447          }
4448    
4449        /* Now handle repetition for the different types of item. */
4450    
4451      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
4452      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
4453      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
4454      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
4455      instead.  */      instead.  */
4456    
4457      if (*previous == OP_CHAR || *previous == OP_CHARNC)      if (*previous == OP_CHAR || *previous == OP_CHARI)
4458        {        {
4459          op_type = (*previous == OP_CHAR)? 0 : OP_STARI - OP_STAR;
4460    
4461        /* 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
4462        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
4463        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 3564  we set the flag only if there is a liter Line 4490  we set the flag only if there is a liter
4490    
4491        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4492            repeat_max < 0 &&            repeat_max < 0 &&
4493            check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
             options, cd))  
4494          {          {
4495          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4496          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3577  we set the flag only if there is a liter Line 4502  we set the flag only if there is a liter
4502      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
4503      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-
4504      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
4505      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
4506      currently used only for single-byte chars. */      are currently used only for single-byte chars. */
4507    
4508      else if (*previous == OP_NOT)      else if (*previous == OP_NOT || *previous == OP_NOTI)
4509        {        {
4510        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = ((*previous == OP_NOT)? OP_NOTSTAR : OP_NOTSTARI) - OP_STAR;
4511        c = previous[1];        c = previous[1];
4512        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4513            repeat_max < 0 &&            repeat_max < 0 &&
4514            check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4515          {          {
4516          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4517          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3610  we set the flag only if there is a liter Line 4535  we set the flag only if there is a liter
4535    
4536        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4537            repeat_max < 0 &&            repeat_max < 0 &&
4538            check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4539          {          {
4540          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4541          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3632  we set the flag only if there is a liter Line 4557  we set the flag only if there is a liter
4557    
4558        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
4559    
4560          /*--------------------------------------------------------------------*/
4561          /* This code is obsolete from release 8.00; the restriction was finally
4562          removed: */
4563    
4564        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4565        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4566    
4567        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4568          /*--------------------------------------------------------------------*/
4569    
4570        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
4571    
# Line 3774  we set the flag only if there is a liter Line 4704  we set the flag only if there is a liter
4704  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4705               *previous == OP_XCLASS ||               *previous == OP_XCLASS ||
4706  #endif  #endif
4707               *previous == OP_REF)               *previous == OP_REF ||
4708                 *previous == OP_REFI)
4709        {        {
4710        if (repeat_max == 0)        if (repeat_max == 0)
4711          {          {
# Line 3782  we set the flag only if there is a liter Line 4713  we set the flag only if there is a liter
4713          goto END_REPEAT;          goto END_REPEAT;
4714          }          }
4715    
4716          /*--------------------------------------------------------------------*/
4717          /* This code is obsolete from release 8.00; the restriction was finally
4718          removed: */
4719    
4720        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4721        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4722    
4723        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4724          /*--------------------------------------------------------------------*/
4725    
4726        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4727          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 3803  we set the flag only if there is a liter Line 4739  we set the flag only if there is a liter
4739        }        }
4740    
4741      /* 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
4742      cases. */      cases. Note that at this point we can encounter only the "basic" bracket
4743        opcodes such as BRA and CBRA, as this is the place where they get converted
4744        into the more special varieties such as BRAPOS and SBRA. A test for >=
4745        OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
4746        ASSERTBACK_NOT, ONCE, BRA, CBRA, and COND. Originally, PCRE did not allow
4747        repetition of assertions, but now it does, for Perl compatibility. */
4748    
4749      else if (*previous == OP_BRA  || *previous == OP_CBRA ||      else if (*previous >= OP_ASSERT && *previous <= OP_COND)
              *previous == OP_ONCE || *previous == OP_COND)  
4750        {        {
4751        register int i;        register int i;
4752        int ketoffset = 0;        int len = (int)(code - previous);
       int len = code - previous;  
4753        uschar *bralink = NULL;        uschar *bralink = NULL;
4754          uschar *brazeroptr = NULL;
4755    
4756        /* Repeating a DEFINE group is pointless */        /* Repeating a DEFINE group is pointless, but Perl allows the syntax, so
4757          we just ignore the repeat. */
4758    
4759        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
4760          {          goto END_REPEAT;
4761          *errorcodeptr = ERR55;  
4762          goto FAILED;        /* There is no sense in actually repeating assertions. The only potential
4763          }        use of repetition is in cases when the assertion is optional. Therefore,
4764          if the minimum is greater than zero, just ignore the repeat. If the
4765          maximum is not not zero or one, set it to 1. */
4766    
4767        /* If the maximum repeat count is unlimited, find the end of the bracket        if (*previous < OP_ONCE)    /* Assertion */
4768        by scanning through from the start, and compute the offset back to it          {
4769        from the current code pointer. There may be an OP_OPT setting following          if (repeat_min > 0) goto END_REPEAT;
4770        the final KET, so we can't find the end just by going back from the code          if (repeat_max < 0 || repeat_max > 1) repeat_max = 1;
       pointer. */  
   
       if (repeat_max == -1)  
         {  
         register uschar *ket = previous;  
         do ket += GET(ket, 1); while (*ket != OP_KET);  
         ketoffset = code - ket;  
4771          }          }
4772    
4773        /* 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 3851  we set the flag only if there is a liter Line 4787  we set the flag only if there is a liter
4787          **   code = previous;          **   code = previous;
4788          **   goto END_REPEAT;          **   goto END_REPEAT;
4789          **   }          **   }
4790    
4791          However, that fails when a group is referenced as a subroutine from          However, that fails when a group or a subgroup within it is referenced
4792          elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it          as a subroutine from elsewhere in the pattern, so now we stick in
4793          so that it is skipped on execution. As we don't have a list of which          OP_SKIPZERO in front of it so that it is skipped on execution. As we
4794          groups are referenced, we cannot do this selectively.          don't have a list of which groups are referenced, we cannot do this
4795            selectively.
4796    
4797          If the maximum is 1 or unlimited, we just have to stick in the BRAZERO          If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4798          and do no more at this point. However, we do need to adjust any          and do no more at this point. However, we do need to adjust any
# Line 3874  we set the flag only if there is a liter Line 4811  we set the flag only if there is a liter
4811              {              {
4812              *previous++ = OP_SKIPZERO;              *previous++ = OP_SKIPZERO;
4813              goto END_REPEAT;              goto END_REPEAT;
4814              }              }
4815              brazeroptr = previous;    /* Save for possessive optimizing */
4816            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4817            }            }
4818    
# Line 3899  we set the flag only if there is a liter Line 4837  we set the flag only if there is a liter
4837            /* We chain together the bracket offset fields that have to be            /* We chain together the bracket offset fields that have to be
4838            filled in later when the ends of the brackets are reached. */            filled in later when the ends of the brackets are reached. */
4839    
4840            offset = (bralink == NULL)? 0 : previous - bralink;            offset = (bralink == NULL)? 0 : (int)(previous - bralink);
4841            bralink = previous;            bralink = previous;
4842            PUTINC(previous, 0, offset);            PUTINC(previous, 0, offset);
4843            }            }
# Line 3920  we set the flag only if there is a liter Line 4858  we set the flag only if there is a liter
4858            {            {
4859            /* 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
4860            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
4861            potential integer overflow. */            potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
4862              integer type when available, otherwise double. */
4863    
4864            if (lengthptr != NULL)            if (lengthptr != NULL)
4865              {              {
4866              int delta = (repeat_min - 1)*length_prevgroup;              int delta = (repeat_min - 1)*length_prevgroup;
4867              if ((double)(repeat_min - 1)*(double)length_prevgroup >              if ((INT64_OR_DOUBLE)(repeat_min - 1)*
4868                                                              (double)INT_MAX ||                    (INT64_OR_DOUBLE)length_prevgroup >
4869                        (INT64_OR_DOUBLE)INT_MAX ||
4870                  OFLOW_MAX - *lengthptr < delta)                  OFLOW_MAX - *lengthptr < delta)
4871                {                {
4872                *errorcodeptr = ERR20;                *errorcodeptr = ERR20;
# Line 3972  we set the flag only if there is a liter Line 4912  we set the flag only if there is a liter
4912          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
4913          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
4914          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
4915          paranoid checks to avoid integer overflow. */          paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
4916            a 64-bit integer type when available, otherwise double. */
4917    
4918          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4919            {            {
4920            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4921                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4922            if ((double)repeat_max *            if ((INT64_OR_DOUBLE)repeat_max *
4923                  (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)                  (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4924                    > (double)INT_MAX ||                    > (INT64_OR_DOUBLE)INT_MAX ||
4925                OFLOW_MAX - *lengthptr < delta)                OFLOW_MAX - *lengthptr < delta)
4926              {              {
4927              *errorcodeptr = ERR20;              *errorcodeptr = ERR20;
# Line 4005  we set the flag only if there is a liter Line 4946  we set the flag only if there is a liter
4946              {              {
4947              int offset;              int offset;
4948              *code++ = OP_BRA;              *code++ = OP_BRA;
4949              offset = (bralink == NULL)? 0 : code - bralink;              offset = (bralink == NULL)? 0 : (int)(code - bralink);
4950              bralink = code;              bralink = code;
4951              PUTINC(code, 0, offset);              PUTINC(code, 0, offset);
4952              }              }
# Line 4026  we set the flag only if there is a liter Line 4967  we set the flag only if there is a liter
4967          while (bralink != NULL)          while (bralink != NULL)
4968            {            {
4969            int oldlinkoffset;            int oldlinkoffset;
4970            int offset = code - bralink + 1;            int offset = (int)(code - bralink + 1);
4971            uschar *bra = code - offset;            uschar *bra = code - offset;
4972            oldlinkoffset = GET(bra, 1);            oldlinkoffset = GET(bra, 1);
4973            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;
# Line 4036  we set the flag only if there is a liter Line 4977  we set the flag only if there is a liter
4977            }            }
4978          }          }
4979    
4980        /* 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
4981        can't just offset backwards from the current code point, because we        ONCE brackets, that's all we need to do. However, possessively repeated
4982        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
4983        correct offset was computed above.        behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to
4984          deal with possessive ONCEs specially.
4985        Then, when we are doing the actual compile phase, check to see whether  
4986        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
4987          whether this group is one that could match an empty string. If so,
4988        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
4989        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
4990        atomic groups at runtime, but in a different way.] */        groups at runtime, but in a different way.]
4991    
4992          Then, if the quantifier was possessive and the bracket is not a
4993          conditional, we convert the BRA code to the POS form, and the KET code to
4994          KETRPOS. (It turns out to be convenient at runtime to detect this kind of
4995          subpattern at both the start and at the end.) The use of special opcodes
4996          makes it possible to reduce greatly the stack usage in pcre_exec(). If
4997          the group is preceded by OP_BRAZERO, convert this to OP_BRAPOSZERO. Then
4998          cancel the possessive flag so that the default action below, of wrapping
4999          everything inside atomic brackets, does not happen. */
5000    
5001        else        else
5002          {          {
5003          uschar *ketcode = code - ketoffset;          uschar *ketcode = code - 1 - LINK_SIZE;
5004          uschar *bracode = ketcode - GET(ketcode, 1);          uschar *bracode = ketcode - GET(ketcode, 1);
5005          *ketcode = OP_KETRMAX + repeat_type;  
5006          if (lengthptr == NULL && *bracode != OP_ONCE)          /* Convert possessive ONCE brackets to non-capturing */
5007    
5008            if ((*bracode == OP_ONCE || *bracode == OP_ONCE_NC) &&
5009                possessive_quantifier) *bracode = OP_BRA;
5010    
5011            /* For non-possessive ONCE brackets, all we need to do is to
5012            set the KET. */
5013    
5014            if (*bracode == OP_ONCE || *bracode == OP_ONCE_NC)
5015              *ketcode = OP_KETRMAX + repeat_type;
5016    
5017            /* Handle non-ONCE brackets and possessive ONCEs (which have been
5018            converted to non-capturing above). */
5019    
5020            else
5021            {            {
5022            uschar *scode = bracode;            /* In the compile phase, check for empty string matching. */
5023            do  
5024              if (lengthptr == NULL)
5025              {              {
5026              if (could_be_empty_branch(scode, ketcode, utf8))              uschar *scode = bracode;
5027                do
5028                {                {
5029                *bracode += OP_SBRA - OP_BRA;                if (could_be_empty_branch(scode, ketcode, utf8, cd))
5030                break;                  {
5031                    *bracode += OP_SBRA - OP_BRA;
5032                    break;
5033                    }
5034                  scode += GET(scode, 1);
5035                  }
5036                while (*scode == OP_ALT);
5037                }
5038    
5039              /* Handle possessive quantifiers. */
5040    
5041              if (possessive_quantifier)
5042                {
5043                /* For COND brackets, we wrap the whole thing in a possessively
5044                repeated non-capturing bracket, because we have not invented POS
5045                versions of the COND opcodes. Because we are moving code along, we
5046                must ensure that any pending recursive references are updated. */
5047    
5048                if (*bracode == OP_COND || *bracode == OP_SCOND)
5049                  {
5050                  int nlen = (int)(code - bracode);
5051                  *code = OP_END;
5052                  adjust_recurse(bracode, 1 + LINK_SIZE, utf8, cd, save_hwm);
5053                  memmove(bracode + 1+LINK_SIZE, bracode, nlen);
5054                  code += 1 + LINK_SIZE;
5055                  nlen += 1 + LINK_SIZE;
5056                  *bracode = OP_BRAPOS;
5057                  *code++ = OP_KETRPOS;
5058                  PUTINC(code, 0, nlen);
5059                  PUT(bracode, 1, nlen);
5060                  }
5061    
5062                /* For non-COND brackets, we modify the BRA code and use KETRPOS. */
5063    
5064                else
5065                  {
5066                  *bracode += 1;              /* Switch to xxxPOS opcodes */
5067                  *ketcode = OP_KETRPOS;
5068                }                }
5069              scode += GET(scode, 1);  
5070                /* If the minimum is zero, mark it as possessive, then unset the
5071                possessive flag. */
5072    
5073                if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
5074                possessive_quantifier = FALSE;