/[pcre]/code/trunk/pcre_compile.c
ViewVC logotype

Diff of /code/trunk/pcre_compile.c

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

<
revision 334 by ph10, Fri Apr 11 15:48:14 2008 UTC revision 760 by ph10, Tue Nov 22 11:23:43 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      "\\N is not supported in a class\0"
416      "too many forward references\0"
417      ;
418    
419  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
420  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 432  For convenience, we use the same bit def
432    
433  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
434    
435  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC
436    
437    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
438    UTF-8 mode. */
439    
440  static const unsigned char digitab[] =  static const unsigned char digitab[] =
441    {    {
442    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 472  static const unsigned char digitab[] =
472    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
473    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
474    
475  #else           /* This is the "abnormal" case, for EBCDIC systems */  #else
476    
477    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
478    
479  static const unsigned char digitab[] =  static const unsigned char digitab[] =
480    {    {
481    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 550  static const unsigned char ebcdic_charta
550  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
551    
552  static BOOL  static BOOL
553    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,    compile_regex(int, uschar **, const uschar **, int *, BOOL, BOOL, int, int,
554      int *, int *, branch_chain *, compile_data *, int *);      int *, int *, branch_chain *, compile_data *, int *);
555    
556    
# Line 454  static const char * Line 572  static const char *
572  find_error_text(int n)  find_error_text(int n)
573  {  {
574  const char *s = error_texts;  const char *s = error_texts;
575  for (; n > 0; n--) while (*s++ != 0);  for (; n > 0; n--)
576      {
577      while (*s++ != 0) {};
578      if (*s == 0) return "Error text not found (please report)";
579      }
580  return s;  return s;
581  }  }
582    
583    
584  /*************************************************  /*************************************************
585    *            Check for counted repeat            *
586    *************************************************/
587    
588    /* This function is called when a '{' is encountered in a place where it might
589    start a quantifier. It looks ahead to see if it really is a quantifier or not.
590    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
591    where the ddds are digits.
592    
593    Arguments:
594      p         pointer to the first char after '{'
595    
596    Returns:    TRUE or FALSE
597    */
598    
599    static BOOL
600    is_counted_repeat(const uschar *p)
601    {
602    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
603    while ((digitab[*p] & ctype_digit) != 0) p++;
604    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
605    
606    if (*p++ != CHAR_COMMA) return FALSE;
607    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
608    
609    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
610    while ((digitab[*p] & ctype_digit) != 0) p++;
611    
612    return (*p == CHAR_RIGHT_CURLY_BRACKET);
613    }
614    
615    
616    
617    /*************************************************
618  *            Handle escapes                      *  *            Handle escapes                      *
619  *************************************************/  *************************************************/
620    
# Line 502  if (c == 0) *errorcodeptr = ERR1; Line 657  if (c == 0) *errorcodeptr = ERR1;
657  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.
658  Otherwise further processing may be required. */  Otherwise further processing may be required. */
659    
660  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
661  else if (c < '0' || c > 'z') {}                           /* Not alphanumeric */  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */
662  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0) c = i;
663    
664  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
665  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 678  else
678      /* 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
679      error. */      error. */
680    
681      case 'l':      case CHAR_l:
682      case 'L':      case CHAR_L:
     case 'N':  
     case 'u':  
     case 'U':  
683      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
684      break;      break;
685    
686      /* \g must be followed by one of a number of specific things:      case CHAR_u:
687        if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
688          {
689          /* In JavaScript, \u must be followed by four hexadecimal numbers.
690          Otherwise it is a lowercase u letter. */
691          if ((digitab[ptr[1]] & ctype_xdigit) != 0 && (digitab[ptr[2]] & ctype_xdigit) != 0
692               && (digitab[ptr[3]] & ctype_xdigit) != 0 && (digitab[ptr[4]] & ctype_xdigit) != 0)
693            {
694            c = 0;
695            for (i = 0; i < 4; ++i)
696              {
697              register int cc = *(++ptr);
698    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
699              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
700              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
701    #else           /* EBCDIC coding */
702              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
703              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
704    #endif
705              }
706            }
707          }
708        else
709          *errorcodeptr = ERR37;
710        break;
711    
712        case CHAR_U:
713        /* In JavaScript, \U is an uppercase U letter. */
714        if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
715        break;
716    
717        /* In a character class, \g is just a literal "g". Outside a character
718        class, \g must be followed by one of a number of specific things:
719    
720      (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
721      backreference. If negative, it is a relative backreference. This is a Perl      backreference. If negative, it is a relative backreference. This is a Perl
722      5.10 feature.      5.10 feature.
723    
724      (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
725      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
726      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
727      was \k.      was \k.
728    
729      (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
730      number either in angle brackets or in single quotes. However, these are      number either in angle brackets or in single quotes. However, these are
731      (possibly recursive) subroutine calls, _not_ backreferences. Just return      (possibly recursive) subroutine calls, _not_ backreferences. Just return
732      the -ESC_g code (cf \k). */      the -ESC_g code (cf \k). */
733    
734      case 'g':      case CHAR_g:
735      if (ptr[1] == '<' || ptr[1] == '\'')      if (isclass) break;
736        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
737        {        {
738        c = -ESC_g;        c = -ESC_g;
739        break;        break;
740        }        }
741    
742      /* Handle the Perl-compatible cases */      /* Handle the Perl-compatible cases */
743    
744      if (ptr[1] == '{')      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
745        {        {
746        const uschar *p;        const uschar *p;
747        for (p = ptr+2; *p != 0 && *p != '}'; p++)        for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
748          if (*p != '-' && (digitab[*p] & ctype_digit) == 0) break;          if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
749        if (*p != 0 && *p != '}')        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
750          {          {
751          c = -ESC_k;          c = -ESC_k;
752          break;          break;
# Line 571  else Line 756  else
756        }        }
757      else braced = FALSE;      else braced = FALSE;
758    
759      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
760        {        {
761        negated = TRUE;        negated = TRUE;
762        ptr++;        ptr++;
# Line 580  else Line 765  else
765    
766      c = 0;      c = 0;
767      while ((digitab[ptr[1]] & ctype_digit) != 0)      while ((digitab[ptr[1]] & ctype_digit) != 0)
768        c = c * 10 + *(++ptr) - '0';        c = c * 10 + *(++ptr) - CHAR_0;
769    
770      if (c < 0)   /* Integer overflow */      if (c < 0)   /* Integer overflow */
771        {        {
772        *errorcodeptr = ERR61;        *errorcodeptr = ERR61;
773        break;        break;
774        }        }
775    
776      if (braced && *(++ptr) != '}')      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
777        {        {
778        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
779        break;        break;
780        }        }
781    
782      if (c == 0)      if (c == 0)
783        {        {
784        *errorcodeptr = ERR58;        *errorcodeptr = ERR58;
785        break;        break;
786        }        }
787    
788      if (negated)      if (negated)
789        {        {
# Line 625  else Line 810  else
810      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
811      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
812    
813      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:
814      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
815    
816      if (!isclass)      if (!isclass)
817        {        {
818        oldptr = ptr;        oldptr = ptr;
819        c -= '0';        c -= CHAR_0;
820        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
821          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - CHAR_0;
822        if (c < 0)    /* Integer overflow */        if (c < 0)    /* Integer overflow */
823          {          {
824          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
# Line 651  else Line 836  else
836      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.
837      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
838    
839      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
840        {        {
841        ptr--;        ptr--;
842        c = 0;        c = 0;
# Line 664  else Line 849  else
849      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
850      than 3 octal digits. */      than 3 octal digits. */
851    
852      case '0':      case CHAR_0:
853      c -= '0';      c -= CHAR_0;
854      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
855          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
856      if (!utf8 && c > 255) *errorcodeptr = ERR51;      if (!utf8 && c > 255) *errorcodeptr = ERR51;
857      break;      break;
858    
# Line 675  else Line 860  else
860      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
861      treated as a data character. */      treated as a data character. */
862    
863      case 'x':      case CHAR_x:
864      if (ptr[1] == '{')      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
865          {
866          /* In JavaScript, \x must be followed by two hexadecimal numbers.
867          Otherwise it is a lowercase x letter. */
868          if ((digitab[ptr[1]] & ctype_xdigit) != 0 && (digitab[ptr[2]] & ctype_xdigit) != 0)
869            {
870            c = 0;
871            for (i = 0; i < 2; ++i)
872              {
873              register int cc = *(++ptr);
874    #ifndef EBCDIC  /* ASCII/UTF-8 coding */
875              if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
876              c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
877    #else           /* EBCDIC coding */
878              if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
879              c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
880    #endif
881              }
882            }
883          break;
884          }
885    
886        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
887        {        {
888        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
889        int count = 0;        int count = 0;
# Line 685  else Line 892  else
892        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
893          {          {
894          register int cc = *pt++;          register int cc = *pt++;
895          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
896          count++;          count++;
897    
898  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
899          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
900          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
901  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
902          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
903          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
904  #endif  #endif
905          }          }
906    
907        if (*pt == '}')        if (*pt == CHAR_RIGHT_CURLY_BRACKET)
908          {          {
909          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
910          ptr = pt;          ptr = pt;
# Line 713  else Line 920  else
920      c = 0;      c = 0;
921      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
922        {        {
923        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
924        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
925  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
926        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
927        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
928  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
929        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
930        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
931  #endif  #endif
932        }        }
933      break;      break;
934    
935      /* 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.
936      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
937        coding is ASCII-specific, but then the whole concept of \cx is
938      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
939    
940      case 'c':      case CHAR_c:
941      c = *(++ptr);      c = *(++ptr);
942      if (c == 0)      if (c == 0)
943        {        {
944        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
945        break;        break;
946        }        }
947    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
948  #ifndef EBCDIC  /* ASCII coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
949      if (c >= 'a' && c <= 'z') c -= 32;        {
950          *errorcodeptr = ERR68;
951          break;
952          }
953        if (c >= CHAR_a && c <= CHAR_z) c -= 32;
954      c ^= 0x40;      c ^= 0x40;
955  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
956      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
957      c ^= 0xC0;      c ^= 0xC0;
958  #endif  #endif
959      break;      break;
# Line 763  else Line 975  else
975      }      }
976    }    }
977    
978    /* Perl supports \N{name} for character names, as well as plain \N for "not
979    newline". PCRE does not support \N{name}. However, it does support
980    quantification such as \N{2,3}. */
981    
982    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
983         !is_counted_repeat(ptr+2))
984      *errorcodeptr = ERR37;
985    
986    /* If PCRE_UCP is set, we change the values for \d etc. */
987    
988    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
989      c -= (ESC_DU - ESC_D);
990    
991    /* Set the pointer to the final character before returning. */
992    
993  *ptrptr = ptr;  *ptrptr = ptr;
994  return c;  return c;
995  }  }
# Line 803  if (c == 0) goto ERROR_RETURN; Line 1030  if (c == 0) goto ERROR_RETURN;
1030  /* \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
1031  negation. */  negation. */
1032    
1033  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
1034    {    {
1035    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1036      {      {
1037      *negptr = TRUE;      *negptr = TRUE;
1038      ptr++;      ptr++;
# Line 814  if (c == '{') Line 1041  if (c == '{')
1041      {      {
1042      c = *(++ptr);      c = *(++ptr);
1043      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
1044      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1045      name[i] = c;      name[i] = c;
1046      }      }
1047    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
1048    name[i] = 0;    name[i] = 0;
1049    }    }
1050    
# Line 863  return -1; Line 1090  return -1;
1090    
1091    
1092  /*************************************************  /*************************************************
 *            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 == '}');  
 }  
   
   
   
 /*************************************************  
1093  *         Read repeat counts                     *  *         Read repeat counts                     *
1094  *************************************************/  *************************************************/
1095    
# Line 923  int max = -1; Line 1117  int max = -1;
1117  /* 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
1118  an integer overflow. */  an integer overflow. */
1119    
1120  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;
1121  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1122    {    {
1123    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 933  if (min < 0 || min > 65535) Line 1127  if (min < 0 || min > 65535)
1127  /* 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.
1128  Also, max must not be less than min. */  Also, max must not be less than min. */
1129    
1130  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1131    {    {
1132    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1133      {      {
1134      max = 0;      max = 0;
1135      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;
1136      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1137        {        {
1138        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 963  return p; Line 1157  return p;
1157    
1158    
1159  /*************************************************  /*************************************************
1160  *       Find forward referenced subpattern       *  *  Subroutine for finding forward reference      *
1161  *************************************************/  *************************************************/
1162    
1163  /* This function scans along a pattern's text looking for capturing  /* This recursive function is called only from find_parens() below. The
1164    top-level call starts at the beginning of the pattern. All other calls must
1165    start at a parenthesis. It scans along a pattern's text looking for capturing
1166  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
1167  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
1168  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
1169  references to subpatterns. We know that if (?P< is encountered, the name will  track of subpatterns that reset the capturing group numbers - the (?| feature.
1170  be terminated by '>' because that is checked in the first pass.  
1171    This function was originally called only from the second pass, in which we know
1172    that if (?< or (?' or (?P< is encountered, the name will be correctly
1173    terminated because that is checked in the first pass. There is now one call to
1174    this function in the first pass, to check for a recursive back reference by
1175    name (so that we can make the whole group atomic). In this case, we need check
1176    only up to the current position in the pattern, and that is still OK because
1177    and previous occurrences will have been checked. To make this work, the test
1178    for "end of pattern" is a check against cd->end_pattern in the main loop,
1179    instead of looking for a binary zero. This means that the special first-pass
1180    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1181    processing items within the loop are OK, because afterwards the main loop will
1182    terminate.)
1183    
1184  Arguments:  Arguments:
1185    ptr          current position in the pattern    ptrptr       address of the current character pointer (updated)
1186    count        current count of capturing parens so far encountered    cd           compile background data
1187    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1188    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1189    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1190      utf8         TRUE if we are in UTF-8 mode
1191      count        pointer to the current capturing subpattern number (updated)
1192    
1193  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1194  */  */
1195    
1196  static int  static int
1197  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,
1198    BOOL xmode)    BOOL xmode, BOOL utf8, int *count)
1199  {  {
1200  const uschar *thisname;  uschar *ptr = *ptrptr;
1201    int start_count = *count;
1202    int hwm_count = start_count;
1203    BOOL dup_parens = FALSE;
1204    
1205  for (; *ptr != 0; ptr++)  /* If the first character is a parenthesis, check on the type of group we are
1206    dealing with. The very first call may not start with a parenthesis. */
1207    
1208    if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1209    {    {
1210    int term;    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1211    
1212      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1213    
1214      /* Handle a normal, unnamed capturing parenthesis. */
1215    
1216      else if (ptr[1] != CHAR_QUESTION_MARK)
1217        {
1218        *count += 1;
1219        if (name == NULL && *count == lorn) return *count;
1220        ptr++;
1221        }
1222    
1223      /* All cases now have (? at the start. Remember when we are in a group
1224      where the parenthesis numbers are duplicated. */
1225    
1226      else if (ptr[2] == CHAR_VERTICAL_LINE)
1227        {
1228        ptr += 3;
1229        dup_parens = TRUE;
1230        }
1231    
1232      /* Handle comments; all characters are allowed until a ket is reached. */
1233    
1234      else if (ptr[2] == CHAR_NUMBER_SIGN)
1235        {
1236        for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1237        goto FAIL_EXIT;
1238        }
1239    
1240      /* Handle a condition. If it is an assertion, just carry on so that it
1241      is processed as normal. If not, skip to the closing parenthesis of the
1242      condition (there can't be any nested parens). */
1243    
1244      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1245        {
1246        ptr += 2;
1247        if (ptr[1] != CHAR_QUESTION_MARK)
1248          {
1249          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1250          if (*ptr != 0) ptr++;
1251          }
1252        }
1253    
1254      /* Start with (? but not a condition. */
1255    
1256      else
1257        {
1258        ptr += 2;
1259        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1260    
1261        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1262    
1263        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1264            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1265          {
1266          int term;
1267          const uschar *thisname;
1268          *count += 1;
1269          if (name == NULL && *count == lorn) return *count;
1270          term = *ptr++;
1271          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1272          thisname = ptr;
1273          while (*ptr != term) ptr++;
1274          if (name != NULL && lorn == ptr - thisname &&
1275              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1276            return *count;
1277          term++;
1278          }
1279        }
1280      }
1281    
1282    /* Past any initial parenthesis handling, scan for parentheses or vertical
1283    bars. Stop if we get to cd->end_pattern. Note that this is important for the
1284    first-pass call when this value is temporarily adjusted to stop at the current
1285    position. So DO NOT change this to a test for binary zero. */
1286    
1287    for (; ptr < cd->end_pattern; ptr++)
1288      {
1289    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1290    
1291    if (*ptr == '\\')    if (*ptr == CHAR_BACKSLASH)
1292      {      {
1293      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1294      if (*ptr == 'Q') for (;;)      if (*ptr == CHAR_Q) for (;;)
1295        {        {
1296        while (*(++ptr) != 0 && *ptr != '\\');        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1297        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1298        if (*(++ptr) == 'E') break;        if (*(++ptr) == CHAR_E) break;
1299        }        }
1300      continue;      continue;
1301      }      }
1302    
1303    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1304      are handled for real. If the first character is '^', skip it. Also, if the
1305      first few characters (either before or after ^) are \Q\E or \E we skip them
1306      too. This makes for compatibility with Perl. Note the use of STR macros to
1307      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1308    
1309    if (*ptr == '[')    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1310      {      {
1311      while (*(++ptr) != ']')      BOOL negate_class = FALSE;
1312        for (;;)
1313          {
1314          if (ptr[1] == CHAR_BACKSLASH)
1315            {
1316            if (ptr[2] == CHAR_E)
1317              ptr+= 2;
1318            else if (strncmp((const char *)ptr+2,
1319                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1320              ptr += 4;
1321            else
1322              break;
1323            }
1324          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1325            {
1326            negate_class = TRUE;
1327            ptr++;
1328            }
1329          else break;
1330          }
1331    
1332        /* If the next character is ']', it is a data character that must be
1333        skipped, except in JavaScript compatibility mode. */
1334    
1335        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1336            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1337          ptr++;
1338    
1339        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1340        {        {
1341        if (*ptr == 0) return -1;        if (*ptr == 0) return -1;
1342        if (*ptr == '\\')        if (*ptr == CHAR_BACKSLASH)
1343          {          {
1344          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) goto FAIL_EXIT;
1345          if (*ptr == 'Q') for (;;)          if (*ptr == CHAR_Q) for (;;)
1346            {            {
1347            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1348            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1349            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1350            }            }
1351          continue;          continue;
1352          }          }
# Line 1031  for (; *ptr != 0; ptr++) Line 1356  for (; *ptr != 0; ptr++)
1356    
1357    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1358    
1359    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1360      {      {
1361      while (*(++ptr) != 0 && *ptr != '\n');      ptr++;
1362      if (*ptr == 0) return -1;      while (*ptr != 0)
1363          {
1364          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1365          ptr++;
1366    #ifdef SUPPORT_UTF8
1367          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
1368    #endif
1369          }
1370        if (*ptr == 0) goto FAIL_EXIT;
1371      continue;      continue;
1372      }      }
1373    
1374    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1375    
1376    if (*ptr != '(') continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != '?' && ptr[1] != '*')  
1377      {      {
1378      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, count);
1379      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1380      continue;      if (*ptr == 0) goto FAIL_EXIT;
1381      }      }
1382    
1383    ptr += 2;    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1384    if (*ptr == 'P') ptr++;                      /* Allow optional P */      {
1385        if (dup_parens && *count < hwm_count) *count = hwm_count;
1386        goto FAIL_EXIT;
1387        }
1388    
1389    /* We have to disambiguate (?<! and (?<= from (?<name> */    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1390        {
1391        if (*count > hwm_count) hwm_count = *count;
1392        *count = start_count;
1393        }
1394      }
1395    
1396    if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  FAIL_EXIT:
1397         *ptr != '\'')  *ptrptr = ptr;
1398      continue;  return -1;
1399    }
1400    
1401    
1402    
1403    
1404    /*************************************************
1405    *       Find forward referenced subpattern       *
1406    *************************************************/
1407    
1408    /* This function scans along a pattern's text looking for capturing
1409    subpatterns, and counting them. If it finds a named pattern that matches the
1410    name it is given, it returns its number. Alternatively, if the name is NULL, it
1411    returns when it reaches a given numbered subpattern. This is used for forward
1412    references to subpatterns. We used to be able to start this scan from the
1413    current compiling point, using the current count value from cd->bracount, and
1414    do it all in a single loop, but the addition of the possibility of duplicate
1415    subpattern numbers means that we have to scan from the very start, in order to
1416    take account of such duplicates, and to use a recursive function to keep track
1417    of the different types of group.
1418    
1419    Arguments:
1420      cd           compile background data
1421      name         name to seek, or NULL if seeking a numbered subpattern
1422      lorn         name length, or subpattern number if name is NULL
1423      xmode        TRUE if we are in /x mode
1424      utf8         TRUE if we are in UTF-8 mode
1425    
1426    count++;  Returns:       the number of the found subpattern, or -1 if not found
1427    */
1428    
1429    if (name == NULL && count == lorn) return count;  static int
1430    term = *ptr++;  find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode,
1431    if (term == '<') term = '>';    BOOL utf8)
1432    thisname = ptr;  {
1433    while (*ptr != term) ptr++;  uschar *ptr = (uschar *)cd->start_pattern;
1434    if (name != NULL && lorn == ptr - thisname &&  int count = 0;
1435        strncmp((const char *)name, (const char *)thisname, lorn) == 0)  int rc;
1436      return count;  
1437    /* If the pattern does not start with an opening parenthesis, the first call
1438    to find_parens_sub() will scan right to the end (if necessary). However, if it
1439    does start with a parenthesis, find_parens_sub() will return when it hits the
1440    matching closing parens. That is why we have to have a loop. */
1441    
1442    for (;;)
1443      {
1444      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, &count);
1445      if (rc > 0 || *ptr++ == 0) break;
1446    }    }
1447    
1448  return -1;  return rc;
1449  }  }
1450    
1451    
1452    
1453    
1454  /*************************************************  /*************************************************
1455  *      Find first significant op code            *  *      Find first significant op code            *
1456  *************************************************/  *************************************************/
1457    
1458  /* This is called by several functions that scan a compiled expression looking  /* This is called by several functions that scan a compiled expression looking
1459  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
1460  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
1461  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
1462  assertions, and also the \b assertion; for others it does not.  does not.
1463    
1464  Arguments:  Arguments:
1465    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  
1466    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1467    
1468  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1469  */  */
1470    
1471  static const uschar*  static const uschar*
1472  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const uschar *code, BOOL skipassert)
   BOOL skipassert)  
1473  {  {
1474  for (;;)  for (;;)
1475    {    {
1476    switch ((int)*code)    switch ((int)*code)
1477      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1478      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1479      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1480      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
# Line 1123  for (;;) Line 1490  for (;;)
1490    
1491      case OP_CALLOUT:      case OP_CALLOUT:
1492      case OP_CREF:      case OP_CREF:
1493        case OP_NCREF:
1494      case OP_RREF:      case OP_RREF:
1495        case OP_NRREF:
1496      case OP_DEF:      case OP_DEF:
1497      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1498      break;      break;
# Line 1139  for (;;) Line 1508  for (;;)
1508    
1509    
1510  /*************************************************  /*************************************************
1511  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1512  *************************************************/  *************************************************/
1513    
1514  /* 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,
1515  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.
1516  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
1517    temporarily terminated with OP_END when this function is called.
1518    
1519    This function is called when a backward assertion is encountered, so that if it
1520    fails, the error message can point to the correct place in the pattern.
1521    However, we cannot do this when the assertion contains subroutine calls,
1522    because they can be forward references. We solve this by remembering this case
1523    and doing the check at the end; a flag specifies which mode we are running in.
1524    
1525  Arguments:  Arguments:
1526    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1527    options  the compiling options    utf8     TRUE in UTF-8 mode
1528      atend    TRUE if called when the pattern is complete
1529  Returns:   the fixed length, or -1 if there is no fixed length,    cd       the "compile data" structure
1530               or -2 if \C was encountered  
1531    Returns:   the fixed length,
1532                 or -1 if there is no fixed length,
1533                 or -2 if \C was encountered (in UTF-8 mode only)
1534                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1535                 or -4 if an unknown opcode was encountered (internal error)
1536  */  */
1537    
1538  static int  static int
1539  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, BOOL utf8, BOOL atend, compile_data *cd)
1540  {  {
1541  int length = -1;  int length = -1;
1542    
# Line 1168  branch, check the length against that of Line 1549  branch, check the length against that of
1549  for (;;)  for (;;)
1550    {    {
1551    int d;    int d;
1552      uschar *ce, *cs;
1553    register int op = *cc;    register int op = *cc;
1554    switch (op)    switch (op)
1555      {      {
1556        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1557        OP_BRA (normal non-capturing bracket) because the other variants of these
1558        opcodes are all concerned with unlimited repeated groups, which of course
1559        are not of fixed length. */
1560    
1561      case OP_CBRA:      case OP_CBRA:
1562      case OP_BRA:      case OP_BRA:
1563      case OP_ONCE:      case OP_ONCE:
1564        case OP_ONCE_NC:
1565      case OP_COND:      case OP_COND:
1566      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), utf8, atend, cd);
1567      if (d < 0) return d;      if (d < 0) return d;
1568      branchlength += d;      branchlength += d;
1569      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
1570      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1571      break;      break;
1572    
1573      /* 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.
1574      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
1575      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
1576        the same code. Note that we must not include the OP_KETRxxx opcodes here,
1577        because they all imply an unlimited repeat. */
1578    
1579      case OP_ALT:      case OP_ALT:
1580      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1581      case OP_END:      case OP_END:
1582        case OP_ACCEPT:
1583        case OP_ASSERT_ACCEPT:
1584      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1585        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
1586      if (*cc != OP_ALT) return length;      if (*cc != OP_ALT) return length;
# Line 1198  for (;;) Line 1588  for (;;)
1588      branchlength = 0;      branchlength = 0;
1589      break;      break;
1590    
1591        /* A true recursion implies not fixed length, but a subroutine call may
1592        be OK. If the subroutine is a forward reference, we can't deal with
1593        it until the end of the pattern, so return -3. */
1594    
1595        case OP_RECURSE:
1596        if (!atend) return -3;
1597        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1598        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1599        if (cc > cs && cc < ce) return -1;                /* Recursion */
1600        d = find_fixedlength(cs + 2, utf8, atend, cd);
1601        if (d < 0) return d;
1602        branchlength += d;
1603        cc += 1 + LINK_SIZE;
1604        break;
1605    
1606      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1607    
1608      case OP_ASSERT:      case OP_ASSERT:
# Line 1209  for (;;) Line 1614  for (;;)
1614    
1615      /* Skip over things that don't match chars */      /* Skip over things that don't match chars */
1616    
1617      case OP_REVERSE:      case OP_MARK:
1618        case OP_PRUNE_ARG:
1619        case OP_SKIP_ARG:
1620        case OP_THEN_ARG:
1621        cc += cc[1] + _pcre_OP_lengths[*cc];
1622        break;
1623    
1624        case OP_CALLOUT:
1625        case OP_CIRC:
1626        case OP_CIRCM:
1627        case OP_CLOSE:
1628        case OP_COMMIT:
1629      case OP_CREF:      case OP_CREF:
     case OP_RREF:  
1630      case OP_DEF:      case OP_DEF:
1631      case OP_OPT:      case OP_DOLL:
1632      case OP_CALLOUT:      case OP_DOLLM:
     case OP_SOD:  
     case OP_SOM:  
1633      case OP_EOD:      case OP_EOD:
1634      case OP_EODN:      case OP_EODN:
1635      case OP_CIRC:      case OP_FAIL:
1636      case OP_DOLL:      case OP_NCREF:
1637        case OP_NRREF:
1638      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1639        case OP_PRUNE:
1640        case OP_REVERSE:
1641        case OP_RREF:
1642        case OP_SET_SOM:
1643        case OP_SKIP:
1644        case OP_SOD:
1645        case OP_SOM:
1646        case OP_THEN:
1647      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1648      cc += _pcre_OP_lengths[*cc];      cc += _pcre_OP_lengths[*cc];
1649      break;      break;
# Line 1229  for (;;) Line 1651  for (;;)
1651      /* Handle literal characters */      /* Handle literal characters */
1652    
1653      case OP_CHAR:      case OP_CHAR:
1654      case OP_CHARNC:      case OP_CHARI:
1655      case OP_NOT:      case OP_NOT:
1656        case OP_NOTI:
1657      branchlength++;      branchlength++;
1658      cc += 2;      cc += 2;
1659  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1660      if ((options & PCRE_UTF8) != 0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       {  
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1661  #endif  #endif
1662      break;      break;
1663    
# Line 1245  for (;;) Line 1665  for (;;)
1665      need to skip over a multibyte character in UTF8 mode.  */      need to skip over a multibyte character in UTF8 mode.  */
1666    
1667      case OP_EXACT:      case OP_EXACT:
1668        case OP_EXACTI:
1669        case OP_NOTEXACT:
1670        case OP_NOTEXACTI:
1671      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1672      cc += 4;      cc += 4;
1673  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1674      if ((options & PCRE_UTF8) != 0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       {  
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1675  #endif  #endif
1676      break;      break;
1677    
# Line 1268  for (;;) Line 1688  for (;;)
1688      cc += 2;      cc += 2;
1689      /* Fall through */      /* Fall through */
1690    
1691        case OP_HSPACE:
1692        case OP_VSPACE:
1693        case OP_NOT_HSPACE:
1694        case OP_NOT_VSPACE:
1695      case OP_NOT_DIGIT:      case OP_NOT_DIGIT:
1696      case OP_DIGIT:      case OP_DIGIT:
1697      case OP_NOT_WHITESPACE:      case OP_NOT_WHITESPACE:
# Line 1275  for (;;) Line 1699  for (;;)
1699      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1700      case OP_WORDCHAR:      case OP_WORDCHAR:
1701      case OP_ANY:      case OP_ANY:
1702        case OP_ALLANY:
1703      branchlength++;      branchlength++;
1704      cc++;      cc++;
1705      break;      break;
1706    
1707      /* The single-byte matcher isn't allowed */      /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1708        otherwise \C is coded as OP_ALLANY. */
1709    
1710      case OP_ANYBYTE:      case OP_ANYBYTE:
1711      return -2;      return -2;
# Line 1298  for (;;) Line 1724  for (;;)
1724    
1725      switch (*cc)      switch (*cc)
1726        {        {
1727          case OP_CRPLUS:
1728          case OP_CRMINPLUS:
1729        case OP_CRSTAR:        case OP_CRSTAR:
1730        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1731        case OP_CRQUERY:        case OP_CRQUERY:
# Line 1316  for (;;) Line 1744  for (;;)
1744        }        }
1745      break;      break;
1746    
1747      /* Anything else is variable length */      /* Anything else is variable length */
1748    
1749        case OP_ANYNL:
1750        case OP_BRAMINZERO:
1751        case OP_BRAPOS:
1752        case OP_BRAPOSZERO:
1753        case OP_BRAZERO:
1754        case OP_CBRAPOS:
1755        case OP_EXTUNI:
1756        case OP_KETRMAX:
1757        case OP_KETRMIN:
1758        case OP_KETRPOS:
1759        case OP_MINPLUS:
1760        case OP_MINPLUSI:
1761        case OP_MINQUERY:
1762        case OP_MINQUERYI:
1763        case OP_MINSTAR:
1764        case OP_MINSTARI:
1765        case OP_MINUPTO:
1766        case OP_MINUPTOI:
1767        case OP_NOTMINPLUS:
1768        case OP_NOTMINPLUSI:
1769        case OP_NOTMINQUERY:
1770        case OP_NOTMINQUERYI:
1771        case OP_NOTMINSTAR:
1772        case OP_NOTMINSTARI:
1773        case OP_NOTMINUPTO:
1774        case OP_NOTMINUPTOI:
1775        case OP_NOTPLUS:
1776        case OP_NOTPLUSI:
1777        case OP_NOTPOSPLUS:
1778        case OP_NOTPOSPLUSI:
1779        case OP_NOTPOSQUERY:
1780        case OP_NOTPOSQUERYI:
1781        case OP_NOTPOSSTAR:
1782        case OP_NOTPOSSTARI:
1783        case OP_NOTPOSUPTO:
1784        case OP_NOTPOSUPTOI:
1785        case OP_NOTQUERY:
1786        case OP_NOTQUERYI:
1787        case OP_NOTSTAR:
1788        case OP_NOTSTARI:
1789        case OP_NOTUPTO:
1790        case OP_NOTUPTOI:
1791        case OP_PLUS:
1792        case OP_PLUSI:
1793        case OP_POSPLUS:
1794        case OP_POSPLUSI:
1795        case OP_POSQUERY:
1796        case OP_POSQUERYI:
1797        case OP_POSSTAR:
1798        case OP_POSSTARI:
1799        case OP_POSUPTO:
1800        case OP_POSUPTOI:
1801        case OP_QUERY:
1802        case OP_QUERYI:
1803        case OP_REF:
1804        case OP_REFI:
1805        case OP_SBRA:
1806        case OP_SBRAPOS:
1807        case OP_SCBRA:
1808        case OP_SCBRAPOS:
1809        case OP_SCOND:
1810        case OP_SKIPZERO:
1811        case OP_STAR:
1812        case OP_STARI:
1813        case OP_TYPEMINPLUS:
1814        case OP_TYPEMINQUERY:
1815        case OP_TYPEMINSTAR:
1816        case OP_TYPEMINUPTO:
1817        case OP_TYPEPLUS:
1818        case OP_TYPEPOSPLUS:
1819        case OP_TYPEPOSQUERY:
1820        case OP_TYPEPOSSTAR:
1821        case OP_TYPEPOSUPTO:
1822        case OP_TYPEQUERY:
1823        case OP_TYPESTAR:
1824        case OP_TYPEUPTO:
1825        case OP_UPTO:
1826        case OP_UPTOI:
1827        return -1;
1828    
1829        /* Catch unrecognized opcodes so that when new ones are added they
1830        are not forgotten, as has happened in the past. */
1831    
1832      default:      default:
1833      return -1;      return -4;
1834      }      }
1835    }    }
1836  /* Control never gets here */  /* Control never gets here */
# Line 1329  for (;;) Line 1840  for (;;)
1840    
1841    
1842  /*************************************************  /*************************************************
1843  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1844  *************************************************/  *************************************************/
1845    
1846  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1847  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1848    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1849    so that it can be called from pcre_study() when finding the minimum matching
1850    length.
1851    
1852  Arguments:  Arguments:
1853    code        points to start of expression    code        points to start of expression
1854    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1855    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1856    
1857  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
1858  */  */
1859    
1860  static const uschar *  const uschar *
1861  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1862  {  {
1863  for (;;)  for (;;)
1864    {    {
1865    register int c = *code;    register int c = *code;
1866    
1867    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1868    
1869    /* 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 1872  for (;;)
1872    
1873    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1874    
1875      /* Handle recursion */
1876    
1877      else if (c == OP_REVERSE)
1878        {
1879        if (number < 0) return (uschar *)code;
1880        code += _pcre_OP_lengths[c];
1881        }
1882    
1883    /* Handle capturing bracket */    /* Handle capturing bracket */
1884    
1885    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
1886               c == OP_CBRAPOS || c == OP_SCBRAPOS)
1887      {      {
1888      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
1889      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
# Line 1368  for (;;) Line 1892  for (;;)
1892    
1893    /* 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
1894    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
1895    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1896      must add in its length. */
1897    
1898    else    else
1899      {      {
# Line 1392  for (;;) Line 1917  for (;;)
1917        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
1918        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1919        break;        break;
1920    
1921          case OP_MARK:
1922          case OP_PRUNE_ARG:
1923          case OP_SKIP_ARG:
1924          code += code[1];
1925          break;
1926    
1927          case OP_THEN_ARG:
1928          code += code[1];
1929          break;
1930        }        }
1931    
1932      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1406  for (;;) Line 1941  for (;;)
1941      if (utf8) switch(c)      if (utf8) switch(c)
1942        {        {
1943        case OP_CHAR:        case OP_CHAR:
1944        case OP_CHARNC:        case OP_CHARI:
1945        case OP_EXACT:        case OP_EXACT:
1946          case OP_EXACTI:
1947        case OP_UPTO:        case OP_UPTO:
1948          case OP_UPTOI:
1949        case OP_MINUPTO:        case OP_MINUPTO:
1950          case OP_MINUPTOI:
1951        case OP_POSUPTO:        case OP_POSUPTO:
1952          case OP_POSUPTOI:
1953        case OP_STAR:        case OP_STAR:
1954          case OP_STARI:
1955        case OP_MINSTAR:        case OP_MINSTAR:
1956          case OP_MINSTARI:
1957        case OP_POSSTAR:        case OP_POSSTAR:
1958          case OP_POSSTARI:
1959        case OP_PLUS:        case OP_PLUS:
1960          case OP_PLUSI:
1961        case OP_MINPLUS:        case OP_MINPLUS:
1962          case OP_MINPLUSI:
1963        case OP_POSPLUS:        case OP_POSPLUS:
1964          case OP_POSPLUSI:
1965        case OP_QUERY:        case OP_QUERY:
1966          case OP_QUERYI:
1967        case OP_MINQUERY:        case OP_MINQUERY:
1968          case OP_MINQUERYI:
1969        case OP_POSQUERY:        case OP_POSQUERY:
1970          case OP_POSQUERYI:
1971        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1972        break;        break;
1973        }        }
1974    #else
1975        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1976  #endif  #endif
1977      }      }
1978    }    }
# Line 1461  for (;;) Line 2011  for (;;)
2011    
2012    /* 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
2013    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
2014    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2015      must add in its length. */
2016    
2017    else    else
2018      {      {
# Line 1485  for (;;) Line 2036  for (;;)
2036        case OP_TYPEEXACT:        case OP_TYPEEXACT:
2037        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
2038        break;        break;
2039    
2040          case OP_MARK:
2041          case OP_PRUNE_ARG:
2042          case OP_SKIP_ARG:
2043          code += code[1];
2044          break;
2045    
2046          case OP_THEN_ARG:
2047          code += code[1];
2048          break;
2049        }        }
2050    
2051      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1499  for (;;) Line 2060  for (;;)
2060      if (utf8) switch(c)      if (utf8) switch(c)
2061        {        {
2062        case OP_CHAR:        case OP_CHAR:
2063        case OP_CHARNC:        case OP_CHARI:
2064        case OP_EXACT:        case OP_EXACT:
2065          case OP_EXACTI:
2066        case OP_UPTO:        case OP_UPTO:
2067          case OP_UPTOI:
2068        case OP_MINUPTO:        case OP_MINUPTO:
2069          case OP_MINUPTOI:
2070        case OP_POSUPTO:        case OP_POSUPTO:
2071          case OP_POSUPTOI:
2072        case OP_STAR:        case OP_STAR:
2073          case OP_STARI:
2074        case OP_MINSTAR:        case OP_MINSTAR:
2075          case OP_MINSTARI:
2076        case OP_POSSTAR:        case OP_POSSTAR:
2077          case OP_POSSTARI:
2078        case OP_PLUS:        case OP_PLUS:
2079          case OP_PLUSI:
2080        case OP_MINPLUS:        case OP_MINPLUS:
2081          case OP_MINPLUSI:
2082        case OP_POSPLUS:        case OP_POSPLUS:
2083          case OP_POSPLUSI:
2084        case OP_QUERY:        case OP_QUERY:
2085          case OP_QUERYI:
2086        case OP_MINQUERY:        case OP_MINQUERY:
2087          case OP_MINQUERYI:
2088        case OP_POSQUERY:        case OP_POSQUERY:
2089          case OP_POSQUERYI:
2090        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
2091        break;        break;
2092        }        }
2093    #else
2094        (void)(utf8);  /* Keep compiler happy by referencing function argument */
2095  #endif  #endif
2096      }      }
2097    }    }
# Line 1539  Arguments: Line 2115  Arguments:
2115    code        points to start of search    code        points to start of search
2116    endcode     points to where to stop    endcode     points to where to stop
2117    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
2118      cd          contains pointers to tables etc.
2119    
2120  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2121  */  */
2122    
2123  static BOOL  static BOOL
2124  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
2125      compile_data *cd)
2126  {  {
2127  register int c;  register int c;
2128  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], TRUE);
2129       code < endcode;       code < endcode;
2130       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], TRUE))
2131    {    {
2132    const uschar *ccode;    const uschar *ccode;
2133    
# Line 1565  for (code = first_significant_code(code Line 2143  for (code = first_significant_code(code
2143      continue;      continue;
2144      }      }
2145    
2146      /* For a recursion/subroutine call, if its end has been reached, which
2147      implies a backward reference subroutine call, we can scan it. If it's a
2148      forward reference subroutine call, we can't. To detect forward reference
2149      we have to scan up the list that is kept in the workspace. This function is
2150      called only when doing the real compile, not during the pre-compile that
2151      measures the size of the compiled pattern. */
2152    
2153      if (c == OP_RECURSE)
2154        {
2155        const uschar *scode;
2156        BOOL empty_branch;
2157    
2158        /* Test for forward reference */
2159    
2160        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
2161          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
2162    
2163        /* Not a forward reference, test for completed backward reference */
2164    
2165        empty_branch = FALSE;
2166        scode = cd->start_code + GET(code, 1);
2167        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2168    
2169        /* Completed backwards reference */
2170    
2171        do
2172          {
2173          if (could_be_empty_branch(scode, endcode, utf8, cd))
2174            {
2175            empty_branch = TRUE;
2176            break;
2177            }
2178          scode += GET(scode, 1);
2179          }
2180        while (*scode == OP_ALT);
2181    
2182        if (!empty_branch) return FALSE;  /* All branches are non-empty */
2183        continue;
2184        }
2185    
2186    /* Groups with zero repeats can of course be empty; skip them. */    /* Groups with zero repeats can of course be empty; skip them. */
2187    
2188    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2189          c == OP_BRAPOSZERO)
2190      {      {
2191      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
2192      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 2194  for (code = first_significant_code(code
2194      continue;      continue;
2195      }      }
2196    
2197      /* A nested group that is already marked as "could be empty" can just be
2198      skipped. */
2199    
2200      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2201          c == OP_SCBRA || c == OP_SCBRAPOS)
2202        {
2203        do code += GET(code, 1); while (*code == OP_ALT);
2204        c = *code;
2205        continue;
2206        }
2207    
2208    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
2209    
2210    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_BRA  || c == OP_BRAPOS ||
2211          c == OP_CBRA || c == OP_CBRAPOS ||
2212          c == OP_ONCE || c == OP_ONCE_NC ||
2213          c == OP_COND)
2214      {      {
2215      BOOL empty_branch;      BOOL empty_branch;
2216      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
2217    
2218      /* Scan a closed bracket */      /* If a conditional group has only one branch, there is a second, implied,
2219        empty branch, so just skip over the conditional, because it could be empty.
2220        Otherwise, scan the individual branches of the group. */
2221    
2222      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;  
2223        code += GET(code, 1);        code += GET(code, 1);
2224        else
2225          {
2226          empty_branch = FALSE;
2227          do
2228            {
2229            if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
2230              empty_branch = TRUE;
2231            code += GET(code, 1);
2232            }
2233          while (*code == OP_ALT);
2234          if (!empty_branch) return FALSE;   /* All branches are non-empty */
2235        }        }
2236      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
2237      c = *code;      c = *code;
2238      continue;      continue;
2239      }      }
# Line 1653  for (code = first_significant_code(code Line 2294  for (code = first_significant_code(code
2294      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2295      case OP_WORDCHAR:      case OP_WORDCHAR:
2296      case OP_ANY:      case OP_ANY:
2297        case OP_ALLANY:
2298      case OP_ANYBYTE:      case OP_ANYBYTE:
2299      case OP_CHAR:      case OP_CHAR:
2300      case OP_CHARNC:      case OP_CHARI:
2301      case OP_NOT:      case OP_NOT:
2302        case OP_NOTI:
2303      case OP_PLUS:      case OP_PLUS:
2304      case OP_MINPLUS:      case OP_MINPLUS:
2305      case OP_POSPLUS:      case OP_POSPLUS:
# Line 1696  for (code = first_significant_code(code Line 2339  for (code = first_significant_code(code
2339      case OP_KET:      case OP_KET:
2340      case OP_KETRMAX:      case OP_KETRMAX:
2341      case OP_KETRMIN:      case OP_KETRMIN:
2342        case OP_KETRPOS:
2343      case OP_ALT:      case OP_ALT:
2344      return TRUE;      return TRUE;
2345    
# Line 1704  for (code = first_significant_code(code Line 2348  for (code = first_significant_code(code
2348    
2349  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2350      case OP_STAR:      case OP_STAR:
2351        case OP_STARI:
2352      case OP_MINSTAR:      case OP_MINSTAR:
2353        case OP_MINSTARI:
2354      case OP_POSSTAR:      case OP_POSSTAR:
2355        case OP_POSSTARI:
2356      case OP_QUERY:      case OP_QUERY:
2357        case OP_QUERYI:
2358      case OP_MINQUERY:      case OP_MINQUERY:
2359        case OP_MINQUERYI:
2360      case OP_POSQUERY:      case OP_POSQUERY:
2361        case OP_POSQUERYI:
2362        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2363        break;
2364    
2365      case OP_UPTO:      case OP_UPTO:
2366        case OP_UPTOI:
2367      case OP_MINUPTO:      case OP_MINUPTO:
2368        case OP_MINUPTOI:
2369      case OP_POSUPTO:      case OP_POSUPTO:
2370      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_POSUPTOI:
2371        if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2372      break;      break;
2373  #endif  #endif
2374    
2375        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2376        string. */
2377    
2378        case OP_MARK:
2379        case OP_PRUNE_ARG:
2380        case OP_SKIP_ARG:
2381        code += code[1];
2382        break;
2383    
2384        case OP_THEN_ARG:
2385        code += code[1];
2386        break;
2387    
2388        /* None of the remaining opcodes are required to match a character. */
2389    
2390        default:
2391        break;
2392      }      }
2393    }    }
2394    
# Line 1731  return TRUE; Line 2405  return TRUE;
2405  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
2406  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,
2407  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.
2408    This function is called only during the real compile, not during the
2409    pre-compile.
2410    
2411  Arguments:  Arguments:
2412    code        points to start of the recursion    code        points to start of the recursion
2413    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2414    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2415    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2416      cd          pointers to tables etc
2417    
2418  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2419  */  */
2420    
2421  static BOOL  static BOOL
2422  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2423    BOOL utf8)    BOOL utf8, compile_data *cd)
2424  {  {
2425  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2426    {    {
2427    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2428        return FALSE;
2429    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2430    }    }
2431  return TRUE;  return TRUE;
# Line 1779  where Perl recognizes it as the POSIX cl Line 2457  where Perl recognizes it as the POSIX cl
2457  "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,
2458  I think.  I think.
2459    
2460    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2461    It seems that the appearance of a nested POSIX class supersedes an apparent
2462    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2463    a digit.
2464    
2465    In Perl, unescaped square brackets may also appear as part of class names. For
2466    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2467    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2468    seem right at all. PCRE does not allow closing square brackets in POSIX class
2469    names.
2470    
2471  Arguments:  Arguments:
2472    ptr      pointer to the initial [    ptr      pointer to the initial [
2473    endptr   where to return the end pointer    endptr   where to return the end pointer
# Line 1793  int terminator;          /* Don't combin Line 2482  int terminator;          /* Don't combin
2482  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2483  for (++ptr; *ptr != 0; ptr++)  for (++ptr; *ptr != 0; ptr++)
2484    {    {
2485    if (*ptr == '\\' && ptr[1] == ']') ptr++; else    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2486        ptr++;
2487      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2488      else
2489      {      {
2490      if (*ptr == ']') return FALSE;      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
     if (*ptr == terminator && ptr[1] == ']')  
2491        {        {
2492        *endptr = ptr;        *endptr = ptr;
2493        return TRUE;        return TRUE;
2494        }        }
2495        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2496             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2497              ptr[1] == CHAR_EQUALS_SIGN) &&
2498            check_posix_syntax(ptr, endptr))
2499          return FALSE;
2500      }      }
2501    }    }
2502  return FALSE;  return FALSE;
# Line 1847  return -1; Line 2543  return -1;
2543  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2544  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2545  earlier groups that are outside the current group). However, when a group is  earlier groups that are outside the current group). However, when a group is
2546  optional (i.e. the minimum quantifier is zero), OP_BRAZERO is inserted before  optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is
2547  it, after it has been compiled. This means that any OP_RECURSE items within it  inserted before it, after it has been compiled. This means that any OP_RECURSE
2548  that refer to the group itself or any contained groups have to have their  items within it that refer to the group itself or any contained groups have to
2549  offsets adjusted. That one of the jobs of this function. Before it is called,  have their offsets adjusted. That one of the jobs of this function. Before it
2550  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2551    OP_END.
2552    
2553  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2554  recursions and subroutine calls. It must also check the list of such references  recursions and subroutine calls. It must also check the list of such references
# Line 1882  while ((ptr = (uschar *)find_recurse(ptr Line 2579  while ((ptr = (uschar *)find_recurse(ptr
2579    
2580    /* 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
2581    reference. */    reference. */
2582    
2583    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
2584      {      {
2585      offset = GET(hc, 0);      offset = GET(hc, 0);
# Line 1928  auto_callout(uschar *code, const uschar Line 2625  auto_callout(uschar *code, const uschar
2625  {  {
2626  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2627  *code++ = 255;  *code++ = 255;
2628  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2629  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2630  return code + 2*LINK_SIZE;  return code + 2*LINK_SIZE;
2631  }  }
2632    
# Line 1954  Returns:             nothing Line 2651  Returns:             nothing
2651  static void  static void
2652  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)
2653  {  {
2654  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2655  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2656  }  }
2657    
# Line 1986  get_othercase_range(unsigned int *cptr, Line 2683  get_othercase_range(unsigned int *cptr,
2683  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2684    
2685  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2686    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2687    
2688  if (c > d) return FALSE;  if (c > d) return FALSE;
2689    
# Line 1995  next = othercase + 1; Line 2692  next = othercase + 1;
2692    
2693  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2694    {    {
2695    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2696    next++;    next++;
2697    }    }
2698    
# Line 2004  for (++c; c <= d; c++) Line 2701  for (++c; c <= d; c++)
2701    
2702  return TRUE;  return TRUE;
2703  }  }
2704    
2705    
2706    
2707    /*************************************************
2708    *        Check a character and a property        *
2709    *************************************************/
2710    
2711    /* This function is called by check_auto_possessive() when a property item
2712    is adjacent to a fixed character.
2713    
2714    Arguments:
2715      c            the character
2716      ptype        the property type
2717      pdata        the data for the type
2718      negated      TRUE if it's a negated property (\P or \p{^)
2719    
2720    Returns:       TRUE if auto-possessifying is OK
2721    */
2722    
2723    static BOOL
2724    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2725    {
2726    const ucd_record *prop = GET_UCD(c);
2727    switch(ptype)
2728      {
2729      case PT_LAMP:
2730      return (prop->chartype == ucp_Lu ||
2731              prop->chartype == ucp_Ll ||
2732              prop->chartype == ucp_Lt) == negated;
2733    
2734      case PT_GC:
2735      return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;
2736    
2737      case PT_PC:
2738      return (pdata == prop->chartype) == negated;
2739    
2740      case PT_SC:
2741      return (pdata == prop->script) == negated;
2742    
2743      /* These are specials */
2744    
2745      case PT_ALNUM:
2746      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2747              _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;
2748    
2749      case PT_SPACE:    /* Perl space */
2750      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2751              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2752              == negated;
2753    
2754      case PT_PXSPACE:  /* POSIX space */
2755      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2756              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2757              c == CHAR_FF || c == CHAR_CR)
2758              == negated;
2759    
2760      case PT_WORD:
2761      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2762              _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2763              c == CHAR_UNDERSCORE) == negated;
2764      }
2765    return FALSE;
2766    }
2767  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2768    
2769    
# Line 2017  whether the next thing could possibly ma Line 2777  whether the next thing could possibly ma
2777  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
2778    
2779  Arguments:  Arguments:
2780    op_code       the repeated op code    previous      pointer to the repeated opcode
   this          data for this item, depends on the opcode  
2781    utf8          TRUE in UTF-8 mode    utf8          TRUE in UTF-8 mode
   utf8_char     used for utf8 character bytes, NULL if not relevant  
2782    ptr           next character in pattern    ptr           next character in pattern
2783    options       options bits    options       options bits
2784    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 2029  Returns:        TRUE if possessifying is Line 2787  Returns:        TRUE if possessifying is
2787  */  */
2788    
2789  static BOOL  static BOOL
2790  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
2791    const uschar *ptr, int options, compile_data *cd)    int options, compile_data *cd)
2792  {  {
2793  int next;  int c, next;
2794    int op_code = *previous++;
2795    
2796  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
2797    
# Line 2041  if ((options & PCRE_EXTENDED) != 0) Line 2800  if ((options & PCRE_EXTENDED) != 0)
2800    for (;;)    for (;;)
2801      {      {
2802      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2803      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2804        {        {
2805        while (*(++ptr) != 0)        ptr++;
2806          while (*ptr != 0)
2807            {
2808          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2809            ptr++;
2810    #ifdef SUPPORT_UTF8
2811            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2812    #endif
2813            }
2814        }        }
2815      else break;      else break;
2816      }      }
# Line 2053  if ((options & PCRE_EXTENDED) != 0) Line 2819  if ((options & PCRE_EXTENDED) != 0)
2819  /* 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
2820  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2821    
2822  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2823    {    {
2824    int temperrorcode = 0;    int temperrorcode = 0;
2825    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
# Line 2078  if ((options & PCRE_EXTENDED) != 0) Line 2844  if ((options & PCRE_EXTENDED) != 0)
2844    for (;;)    for (;;)
2845      {      {
2846      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2847      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2848        {        {
2849        while (*(++ptr) != 0)        ptr++;
2850          while (*ptr != 0)
2851            {
2852          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2853            ptr++;
2854    #ifdef SUPPORT_UTF8
2855            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2856    #endif
2857            }
2858        }        }
2859      else break;      else break;
2860      }      }
# Line 2089  if ((options & PCRE_EXTENDED) != 0) Line 2862  if ((options & PCRE_EXTENDED) != 0)
2862    
2863  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
2864    
2865  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2866    return FALSE;    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2867        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. */  
   
2868    
2869  /* Handle cases when the next item is a character. */  /* Now compare the next item with the previous opcode. First, handle cases when
2870    the next item is a character. */
2871    
2872  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
2873    {    {
2874    case OP_CHAR:    case OP_CHAR:
2875  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2876    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2877    #else
2878      c = *previous;
2879  #endif  #endif
2880    return item != next;    return c != next;
2881    
2882    /* 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
2883    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
2884    high-valued characters. */    high-valued characters. */
2885    
2886    case OP_CHARNC:    case OP_CHARI:
2887  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2888    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2889    #else
2890      c = *previous;
2891  #endif  #endif
2892    if (item == next) return FALSE;    if (c == next) return FALSE;
2893  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2894    if (utf8)    if (utf8)
2895      {      {
2896      unsigned int othercase;      unsigned int othercase;
2897      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2898  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2899      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
2900  #else  #else
2901      othercase = NOTACHAR;      othercase = NOTACHAR;
2902  #endif  #endif
2903      return (unsigned int)item != othercase;      return (unsigned int)c != othercase;
2904      }      }
2905    else    else
2906  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2907    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2908    
2909    /* 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
2910      opcodes are not used for multi-byte characters, because they are coded using
2911      an XCLASS instead. */
2912    
2913    case OP_NOT:    case OP_NOT:
2914    if (item == next) return TRUE;    return (c = *previous) == next;
2915    if ((options & PCRE_CASELESS) == 0) return FALSE;  
2916      case OP_NOTI:
2917      if ((c = *previous) == next) return TRUE;
2918  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2919    if (utf8)    if (utf8)
2920      {      {
2921      unsigned int othercase;      unsigned int othercase;
2922      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2923  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2924      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
2925  #else  #else
2926      othercase = NOTACHAR;      othercase = NOTACHAR;
2927  #endif  #endif
2928      return (unsigned int)item == othercase;      return (unsigned int)c == othercase;
2929      }      }
2930    else    else
2931  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2932    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2933    
2934      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2935      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
2936    
2937    case OP_DIGIT:    case OP_DIGIT:
2938    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
# Line 2195  if (next >= 0) switch(op_code) Line 2975  if (next >= 0) switch(op_code)
2975      case 0x202f:      case 0x202f:
2976      case 0x205f:      case 0x205f:
2977      case 0x3000:      case 0x3000:
2978      return op_code != OP_HSPACE;      return op_code == OP_NOT_HSPACE;
2979      default:      default:
2980      return op_code == OP_HSPACE;      return op_code != OP_NOT_HSPACE;
2981      }      }
2982    
2983      case OP_ANYNL:
2984    case OP_VSPACE:    case OP_VSPACE:
2985    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
2986    switch(next)    switch(next)
# Line 2211  if (next >= 0) switch(op_code) Line 2992  if (next >= 0) switch(op_code)
2992      case 0x85:      case 0x85:
2993      case 0x2028:      case 0x2028:
2994      case 0x2029:      case 0x2029:
2995      return op_code != OP_VSPACE;      return op_code == OP_NOT_VSPACE;
2996      default:      default:
2997      return op_code == OP_VSPACE;      return op_code != OP_NOT_VSPACE;
2998      }      }
2999    
3000    #ifdef SUPPORT_UCP
3001      case OP_PROP:
3002      return check_char_prop(next, previous[0], previous[1], FALSE);
3003    
3004      case OP_NOTPROP:
3005      return check_char_prop(next, previous[0], previous[1], TRUE);
3006    #endif
3007    
3008    default:    default:
3009    return FALSE;    return FALSE;
3010    }    }
3011    
3012    
3013  /* 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
3014    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
3015    generated only when PCRE_UCP is *not* set, that is, when only ASCII
3016    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
3017    replaced by OP_PROP codes when PCRE_UCP is set. */
3018    
3019  switch(op_code)  switch(op_code)
3020    {    {
3021    case OP_CHAR:    case OP_CHAR:
3022    case OP_CHARNC:    case OP_CHARI:
3023  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3024    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
3025    #else
3026      c = *previous;
3027  #endif  #endif
3028    switch(-next)    switch(-next)
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_D:      case ESC_D:
3034      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 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_S:      case ESC_S:
3040      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_space) != 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_W:      case ESC_W:
3046      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
3047    
3048      case ESC_h:      case ESC_h:
3049      case ESC_H:      case ESC_H:
3050      switch(item)      switch(c)
3051        {        {
3052        case 0x09:        case 0x09:
3053        case 0x20:        case 0x20:
# Line 2280  switch(op_code) Line 3075  switch(op_code)
3075    
3076      case ESC_v:      case ESC_v:
3077      case ESC_V:      case ESC_V:
3078      switch(item)      switch(c)
3079        {        {
3080        case 0x0a:        case 0x0a:
3081        case 0x0b:        case 0x0b:
# Line 2294  switch(op_code) Line 3089  switch(op_code)
3089        return -next == ESC_v;        return -next == ESC_v;
3090        }        }
3091    
3092        /* When PCRE_UCP is set, these values get generated for \d etc. Find
3093        their substitutions and process them. The result will always be either
3094        -ESC_p or -ESC_P. Then fall through to process those values. */
3095    
3096    #ifdef SUPPORT_UCP
3097        case ESC_du:
3098        case ESC_DU:
3099        case ESC_wu:
3100        case ESC_WU:
3101        case ESC_su:
3102        case ESC_SU:
3103          {
3104          int temperrorcode = 0;
3105          ptr = substitutes[-next - ESC_DU];
3106          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
3107          if (temperrorcode != 0) return FALSE;
3108          ptr++;    /* For compatibility */
3109          }
3110        /* Fall through */
3111    
3112        case ESC_p:
3113        case ESC_P:
3114          {
3115          int ptype, pdata, errorcodeptr;
3116          BOOL negated;
3117    
3118          ptr--;      /* Make ptr point at the p or P */
3119          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
3120          if (ptype < 0) return FALSE;
3121          ptr++;      /* Point past the final curly ket */
3122    
3123          /* If the property item is optional, we have to give up. (When generated
3124          from \d etc by PCRE_UCP, this test will have been applied much earlier,
3125          to the original \d etc. At this point, ptr will point to a zero byte. */
3126    
3127          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
3128            strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
3129              return FALSE;
3130    
3131          /* Do the property check. */
3132    
3133          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
3134          }
3135    #endif
3136    
3137      default:      default:
3138      return FALSE;      return FALSE;
3139      }      }
3140    
3141      /* In principle, support for Unicode properties should be integrated here as
3142      well. It means re-organizing the above code so as to get hold of the property
3143      values before switching on the op-code. However, I wonder how many patterns
3144      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
3145      these op-codes are never generated.) */
3146    
3147    case OP_DIGIT:    case OP_DIGIT:
3148    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
3149           next == -ESC_h || next == -ESC_v;           next == -ESC_h || next == -ESC_v || next == -ESC_R;
3150    
3151    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
3152    return next == -ESC_d;    return next == -ESC_d;
3153    
3154    case OP_WHITESPACE:    case OP_WHITESPACE:
3155    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
3156    
3157    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
3158    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
3159    
3160    case OP_HSPACE:    case OP_HSPACE:
3161    return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
3162             next == -ESC_w || next == -ESC_v || next == -ESC_R;
3163    
3164    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
3165    return next == -ESC_h;    return next == -ESC_h;
3166    
3167    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
3168      case OP_ANYNL:
3169    case OP_VSPACE:    case OP_VSPACE:
3170    return next == -ESC_V || next == -ESC_d || next == -ESC_w;    return next == -ESC_V || next == -ESC_d || next == -ESC_w;
3171    
3172    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
3173    return next == -ESC_v;    return next == -ESC_v || next == -ESC_R;
3174    
3175    case OP_WORDCHAR:    case OP_WORDCHAR:
3176    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
3177             next == -ESC_v || next == -ESC_R;
3178    
3179    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
3180    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2357  Arguments: Line 3206  Arguments:
3206    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
3207    reqbyteptr     set to the last literal character required, else < 0    reqbyteptr     set to the last literal character required, else < 0
3208    bcptr          points to current branch chain    bcptr          points to current branch chain
3209      cond_depth     conditional nesting depth
3210    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
3211    lengthptr      NULL during the real compile phase    lengthptr      NULL during the real compile phase
3212                   points to length accumulator during pre-compile phase                   points to length accumulator during pre-compile phase
# Line 2368  Returns:         TRUE on success Line 3218  Returns:         TRUE on success
3218  static BOOL  static BOOL
3219  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
3220    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
3221    compile_data *cd, int *lengthptr)    int cond_depth, compile_data *cd, int *lengthptr)
3222  {  {
3223  int repeat_type, op_type;  int repeat_type, op_type;
3224  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
# Line 2377  int greedy_default, greedy_non_default; Line 3227  int greedy_default, greedy_non_default;
3227  int firstbyte, reqbyte;  int firstbyte, reqbyte;
3228  int zeroreqbyte, zerofirstbyte;  int zeroreqbyte, zerofirstbyte;
3229  int req_caseopt, reqvary, tempreqvary;  int req_caseopt, reqvary, tempreqvary;
3230  int options = *optionsptr;  int options = *optionsptr;               /* May change dynamically */
3231  int after_manual_callout = 0;  int after_manual_callout = 0;
3232  int length_prevgroup = 0;  int length_prevgroup = 0;
3233  register int c;  register int c;
# Line 2389  BOOL inescq = FALSE; Line 3239  BOOL inescq = FALSE;
3239  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
3240  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
3241  const uschar *tempptr;  const uschar *tempptr;
3242    const uschar *nestptr = NULL;
3243  uschar *previous = NULL;  uschar *previous = NULL;
3244  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
3245  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
3246  uschar classbits[32];  uschar classbits[32];
3247    
3248    /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3249    must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3250    dynamically as we process the pattern. */
3251    
3252  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3253  BOOL class_utf8;  BOOL class_utf8;
3254  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
# Line 2402  uschar *class_utf8data_base; Line 3257  uschar *class_utf8data_base;
3257  uschar utf8_char[6];  uschar utf8_char[6];
3258  #else  #else
3259  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
 uschar *utf8_char = NULL;  
3260  #endif  #endif
3261    
3262  #ifdef DEBUG  #ifdef PCRE_DEBUG
3263  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
3264  #endif  #endif
3265    
# Line 2453  for (;; ptr++) Line 3307  for (;; ptr++)
3307    int subfirstbyte;    int subfirstbyte;
3308    int terminator;    int terminator;
3309    int mclength;    int mclength;
3310      int tempbracount;
3311    uschar mcbuffer[8];    uschar mcbuffer[8];
3312    
3313    /* Get next byte in the pattern */    /* Get next byte in the pattern */
3314    
3315    c = *ptr;    c = *ptr;
3316    
3317      /* If we are at the end of a nested substitution, revert to the outer level
3318      string. Nesting only happens one level deep. */
3319    
3320      if (c == 0 && nestptr != NULL)
3321        {
3322        ptr = nestptr;
3323        nestptr = NULL;
3324        c = *ptr;
3325        }
3326    
3327    /* 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
3328    previous cycle of this loop. */    previous cycle of this loop. */
3329    
3330    if (lengthptr != NULL)    if (lengthptr != NULL)
3331      {      {
3332  #ifdef DEBUG  #ifdef PCRE_DEBUG
3333      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3334  #endif  #endif
3335      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
3336        {        {
3337        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3338        goto FAILED;        goto FAILED;
# Line 2489  for (;; ptr++) Line 3354  for (;; ptr++)
3354        goto FAILED;        goto FAILED;
3355        }        }
3356    
3357      *lengthptr += code - last_code;      *lengthptr += (int)(code - last_code);
3358      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, (int)(code - last_code),
3359          c));
3360    
3361      /* 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
3362      it back to there, in order to avoid filling up the work space. Otherwise,      it back to there, in order to avoid filling up the work space. Otherwise,
# Line 2516  for (;; ptr++) Line 3382  for (;; ptr++)
3382    /* 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
3383    reference list. */    reference list. */
3384    
3385    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
3386      {      {
3387      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3388      goto FAILED;      goto FAILED;
# Line 2526  for (;; ptr++) Line 3392  for (;; ptr++)
3392    
3393    if (inescq && c != 0)    if (inescq && c != 0)
3394      {      {
3395      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3396        {        {
3397        inescq = FALSE;        inescq = FALSE;
3398        ptr++;        ptr++;
# Line 2552  for (;; ptr++) Line 3418  for (;; ptr++)
3418    /* 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
3419    a quantifier. */    a quantifier. */
3420    
3421    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
3422      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
3423        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
3424    
3425    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
3426         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2563  for (;; ptr++) Line 3430  for (;; ptr++)
3430      previous_callout = NULL;      previous_callout = NULL;
3431      }      }
3432    
3433    /* In extended mode, skip white space and comments */    /* In extended mode, skip white space and comments. */
3434    
3435    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3436      {      {
3437      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
3438      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
3439        {        {
3440        while (*(++ptr) != 0)        ptr++;
3441          while (*ptr != 0)
3442          {          {
3443          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3444            ptr++;
3445    #ifdef SUPPORT_UTF8
3446            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
3447    #endif
3448          }          }
3449        if (*ptr != 0) continue;        if (*ptr != 0) continue;
3450    
# Line 2593  for (;; ptr++) Line 3465  for (;; ptr++)
3465      {      {
3466      /* ===================================================================*/      /* ===================================================================*/
3467      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
3468      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
3469      case ')':      case CHAR_RIGHT_PARENTHESIS:
3470      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
3471      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
3472      *codeptr = code;      *codeptr = code;
# Line 2606  for (;; ptr++) Line 3478  for (;; ptr++)
3478          *errorcodeptr = ERR20;          *errorcodeptr = ERR20;
3479          goto FAILED;          goto FAILED;
3480          }          }
3481        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += (int)(code - last_code);   /* To include callout length */
3482        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
3483        }        }
3484      return TRUE;      return TRUE;
# Line 2616  for (;; ptr++) Line 3488  for (;; ptr++)
3488      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
3489      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3490    
3491      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
3492        previous = NULL;
3493      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3494        {        {
3495        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3496          *code++ = OP_CIRCM;
3497        }        }
3498      previous = NULL;      else *code++ = OP_CIRC;
     *code++ = OP_CIRC;  
3499      break;      break;
3500    
3501      case '$':      case CHAR_DOLLAR_SIGN:
3502      previous = NULL;      previous = NULL;
3503      *code++ = OP_DOLL;      *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
3504      break;      break;
3505    
3506      /* 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
3507      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
3508    
3509      case '.':      case CHAR_DOT:
3510      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3511      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
3512      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3513      previous = code;      previous = code;
3514      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
3515      break;      break;
3516    
3517    
# Line 2653  for (;; ptr++) Line 3526  for (;; ptr++)
3526      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,
3527      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
3528      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.
     */  
3529    
3530      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
3531        default (Perl) mode, it is treated as a data character. */
3532    
3533        case CHAR_RIGHT_SQUARE_BRACKET:
3534        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3535          {
3536          *errorcodeptr = ERR64;
3537          goto FAILED;
3538          }
3539        goto NORMAL_CHAR;
3540    
3541        case CHAR_LEFT_SQUARE_BRACKET:
3542      previous = code;      previous = code;
3543    
3544      /* 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
3545      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. */
3546    
3547      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3548             ptr[1] == CHAR_EQUALS_SIGN) &&
3549          check_posix_syntax(ptr, &tempptr))          check_posix_syntax(ptr, &tempptr))
3550        {        {
3551        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
3552        goto FAILED;        goto FAILED;
3553        }        }
3554    
# Line 2676  for (;; ptr++) Line 3560  for (;; ptr++)
3560      for (;;)      for (;;)
3561        {        {
3562        c = *(++ptr);        c = *(++ptr);
3563        if (c == '\\')        if (c == CHAR_BACKSLASH)
3564          {          {
3565          if (ptr[1] == 'E') ptr++;          if (ptr[1] == CHAR_E)
3566            else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;            ptr++;
3567              else break;          else if (strncmp((const char *)ptr+1,
3568                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
3569              ptr += 3;
3570            else
3571              break;
3572          }          }
3573        else if (!negate_class && c == '^')        else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
3574          negate_class = TRUE;          negate_class = TRUE;
3575        else break;        else break;
3576        }        }
3577    
3578        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
3579        an initial ']' is taken as a data character -- the code below handles
3580        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3581        [^] must match any character, so generate OP_ALLANY. */
3582    
3583        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3584            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3585          {
3586          *code++ = negate_class? OP_ALLANY : OP_FAIL;
3587          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3588          zerofirstbyte = firstbyte;
3589          break;
3590          }
3591    
3592      /* 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
3593      negation flag at the end, so that support for characters > 255 works      negation flag at the end, so that support for characters > 255 works
3594      correctly (they are all included in the class). */      correctly (they are all included in the class). */
# Line 2744  for (;; ptr++) Line 3646  for (;; ptr++)
3646    
3647        if (inescq)        if (inescq)
3648          {          {
3649          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3650            {            {
3651            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
3652            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2759  for (;; ptr++) Line 3661  for (;; ptr++)
3661        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3662        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3663    
3664        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3665            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3666            check_posix_syntax(ptr, &tempptr))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3667          {          {
3668          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3669          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3670          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3671          uschar pbits[32];          uschar pbits[32];
3672    
3673          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3674            {            {
3675            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3676            goto FAILED;            goto FAILED;
3677            }            }
3678    
3679          ptr += 2;          ptr += 2;
3680          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3681            {            {
3682            local_negate = TRUE;            local_negate = TRUE;
3683            should_flip_negation = TRUE;  /* Note negative special */            should_flip_negation = TRUE;  /* Note negative special */
3684            ptr++;            ptr++;
3685            }            }
3686    
3687          posix_class = check_posix_name(ptr, tempptr - ptr);          posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3688          if (posix_class < 0)          if (posix_class < 0)
3689            {            {
3690            *errorcodeptr = ERR30;            *errorcodeptr = ERR30;
# Line 2796  for (;; ptr++) Line 3698  for (;; ptr++)
3698          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3699            posix_class = 0;            posix_class = 0;
3700    
3701          /* 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
3702          because we may be adding and subtracting from it, and we don't want to          different escape sequences that use Unicode properties. */
3703          subtract bits that may be in the main map already. At the end we or the  
3704          result into the bit map that is being built. */  #ifdef SUPPORT_UCP
3705            if ((options & PCRE_UCP) != 0)
3706              {
3707              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3708              if (posix_substitutes[pc] != NULL)
3709                {
3710                nestptr = tempptr + 1;
3711                ptr = posix_substitutes[pc] - 1;
3712                continue;
3713                }
3714              }
3715    #endif
3716            /* In the non-UCP case, we build the bit map for the POSIX class in a
3717            chunk of local store because we may be adding and subtracting from it,
3718            and we don't want to subtract bits that may be in the main map already.
3719            At the end we or the result into the bit map that is being built. */
3720    
3721          posix_class *= 3;          posix_class *= 3;
3722    
# Line 2843  for (;; ptr++) Line 3760  for (;; ptr++)
3760    
3761        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3762        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
3763        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
3764        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
3765        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
3766        character in them, so set class_charcount bigger than one. */        are either treated as literal characters (by default), or are faulted if
3767          PCRE_EXTRA is set. */
3768    
3769        if (c == '\\')        if (c == CHAR_BACKSLASH)
3770          {          {
3771          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3772          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3773    
3774          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 */
3775          else if (-c == ESC_X) c = 'X';   /* \X is literal X in a class */          else if (-c == ESC_N)            /* \N is not supported in a class */
3776          else if (-c == ESC_R) c = 'R';   /* \R is literal R in a class */            {
3777              *errorcodeptr = ERR71;
3778              goto FAILED;
3779              }
3780          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3781            {            {
3782            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3783              {              {
3784              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3785              }              }
# Line 2872  for (;; ptr++) Line 3793  for (;; ptr++)
3793            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
3794            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3795    
3796            /* Save time by not doing this in the pre-compile phase. */            switch (-c)
   
           if (lengthptr == NULL) switch (-c)  
3797              {              {
3798    #ifdef SUPPORT_UCP
3799                case ESC_du:     /* These are the values given for \d etc */
3800                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3801                case ESC_wu:     /* escape sequence with an appropriate \p */
3802                case ESC_WU:     /* or \P to test Unicode properties instead */
3803                case ESC_su:     /* of the default ASCII testing. */
3804                case ESC_SU:
3805                nestptr = ptr;
3806                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3807                class_charcount -= 2;                /* Undo! */
3808                continue;
3809    #endif
3810              case ESC_d:              case ESC_d:
3811              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3812              continue;              continue;
# Line 2894  for (;; ptr++) Line 3825  for (;; ptr++)
3825              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3826              continue;              continue;
3827    
3828                /* Perl 5.004 onwards omits VT from \s, but we must preserve it
3829                if it was previously set by something earlier in the character
3830                class. */
3831    
3832              case ESC_s:              case ESC_s:
3833              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];              classbits[0] |= cbits[cbit_space];
3834              classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= cbits[cbit_space+1] & ~0x08;
3835                for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3836              continue;              continue;
3837    
3838              case ESC_S:              case ESC_S:
# Line 2905  for (;; ptr++) Line 3841  for (;; ptr++)
3841              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3842              continue;              continue;
3843    
3844              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)  
             {  
3845              SETBIT(classbits, 0x09); /* VT */              SETBIT(classbits, 0x09); /* VT */
3846              SETBIT(classbits, 0x20); /* SPACE */              SETBIT(classbits, 0x20); /* SPACE */
3847              SETBIT(classbits, 0xa0); /* NSBP */              SETBIT(classbits, 0xa0); /* NSBP */
# Line 2942  for (;; ptr++) Line 3865  for (;; ptr++)
3865                }                }
3866  #endif  #endif
3867              continue;              continue;
             }  
3868    
3869            if (-c == ESC_H)              case ESC_H:
             {  
3870              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3871                {                {
3872                int x = 0xff;                int x = 0xff;
# Line 2987  for (;; ptr++) Line 3908  for (;; ptr++)
3908                }                }
3909  #endif  #endif
3910              continue;              continue;
             }  
3911    
3912            if (-c == ESC_v)              case ESC_v:
             {  
3913              SETBIT(classbits, 0x0a); /* LF */              SETBIT(classbits, 0x0a); /* LF */
3914              SETBIT(classbits, 0x0b); /* VT */              SETBIT(classbits, 0x0b); /* VT */
3915              SETBIT(classbits, 0x0c); /* FF */              SETBIT(classbits, 0x0c); /* FF */
# Line 3006  for (;; ptr++) Line 3925  for (;; ptr++)
3925                }                }
3926  #endif  #endif
3927              continue;              continue;
             }  
3928    
3929            if (-c == ESC_V)              case ESC_V:
             {  
3930              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3931                {                {
3932                int x = 0xff;                int x = 0xff;
# Line 3039  for (;; ptr++) Line 3956  for (;; ptr++)
3956                }                }
3957  #endif  #endif
3958              continue;              continue;
             }  
   
           /* We need to deal with \P and \p in both phases. */  
3959    
3960  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3961            if (-c == ESC_p || -c == ESC_P)              case ESC_p:
3962              {              case ESC_P:
3963              BOOL negated;                {
3964              int pdata;                BOOL negated;
3965              int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                int pdata;
3966              if (ptype < 0) goto FAILED;                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3967              class_utf8 = TRUE;                if (ptype < 0) goto FAILED;
3968              *class_utf8data++ = ((-c == ESC_p) != negated)?                class_utf8 = TRUE;
3969                XCL_PROP : XCL_NOTPROP;                *class_utf8data++ = ((-c == ESC_p) != negated)?
3970              *class_utf8data++ = ptype;                  XCL_PROP : XCL_NOTPROP;
3971              *class_utf8data++ = pdata;                *class_utf8data++ = ptype;
3972              class_charcount -= 2;   /* Not a < 256 character */                *class_utf8data++ = pdata;
3973              continue;                class_charcount -= 2;   /* Not a < 256 character */
3974              }                continue;
3975                  }
3976  #endif  #endif
3977            /* Unrecognized escapes are faulted if PCRE is running in its              /* Unrecognized escapes are faulted if PCRE is running in its
3978            strict mode. By default, for compatibility with Perl, they are              strict mode. By default, for compatibility with Perl, they are
3979            treated as literals. */              treated as literals. */
3980    
3981            if ((options & PCRE_EXTRA) != 0)              default:
3982              {              if ((options & PCRE_EXTRA) != 0)
3983              *errorcodeptr = ERR7;                {
3984              goto FAILED;                *errorcodeptr = ERR7;
3985                  goto FAILED;
3986                  }
3987                class_charcount -= 2;  /* Undo the default count from above */
3988                c = *ptr;              /* Get the final character and fall through */
3989                break;
3990              }              }
   
           class_charcount -= 2;  /* Undo the default count from above */  
           c = *ptr;              /* Get the final character and fall through */  
3991            }            }
3992    
3993          /* Fall through if we have a single character (c >= 0). This may be          /* Fall through if we have a single character (c >= 0). This may be
# Line 3084  for (;; ptr++) Line 4001  for (;; ptr++)
4001        entirely. The code for handling \Q and \E is messy. */        entirely. The code for handling \Q and \E is messy. */
4002    
4003        CHECK_RANGE:        CHECK_RANGE:
4004        while (ptr[1] == '\\' && ptr[2] == 'E')        while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
4005          {          {
4006          inescq = FALSE;          inescq = FALSE;
4007          ptr += 2;          ptr += 2;
# Line 3094  for (;; ptr++) Line 4011  for (;; ptr++)
4011    
4012        /* Remember \r or \n */        /* Remember \r or \n */
4013    
4014        if (c == '\r' || c == '\n') cd->external_flags |= PCRE_HASCRORLF;        if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
4015    
4016        /* Check for range */        /* Check for range */
4017    
4018        if (!inescq && ptr[1] == '-')        if (!inescq && ptr[1] == CHAR_MINUS)
4019          {          {
4020          int d;          int d;
4021          ptr += 2;          ptr += 2;
4022          while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
4023    
4024          /* 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
4025          mode. */          mode. */
4026    
4027          while (*ptr == '\\' && ptr[1] == 'Q')          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
4028            {            {
4029            ptr += 2;            ptr += 2;
4030            if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }            if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
4031                { ptr += 2; continue; }
4032            inescq = TRUE;            inescq = TRUE;
4033            break;            break;
4034            }            }
4035    
4036          if (*ptr == 0 || (!inescq && *ptr == ']'))          if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
4037            {            {
4038            ptr = oldptr;            ptr = oldptr;
4039            goto LONE_SINGLE_CHARACTER;            goto LONE_SINGLE_CHARACTER;
# Line 3134  for (;; ptr++) Line 4052  for (;; ptr++)
4052          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
4053          in such circumstances. */          in such circumstances. */
4054    
4055          if (!inescq && d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
4056            {            {
4057            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
4058            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
4059    
4060            /* \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 */  
4061    
4062            if (d < 0)            if (d < 0)
4063              {              {
4064              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  
4065                {                {
4066                ptr = oldptr;                ptr = oldptr;
4067                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 3167  for (;; ptr++) Line 4082  for (;; ptr++)
4082    
4083          /* Remember \r or \n */          /* Remember \r or \n */
4084    
4085          if (d == '\r' || d == '\n') cd->external_flags |= PCRE_HASCRORLF;          if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
4086    
4087          /* 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
4088          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
# Line 3287  for (;; ptr++) Line 4202  for (;; ptr++)
4202          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
4203            {            {
4204            unsigned int othercase;            unsigned int othercase;
4205            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)            if ((othercase = UCD_OTHERCASE(c)) != c)
4206              {              {
4207              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
4208              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 3312  for (;; ptr++) Line 4227  for (;; ptr++)
4227          }          }
4228        }        }
4229    
4230      /* 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.
4231        If we are at the end of an internal nested string, revert to the outer
4232        string. */
4233    
4234        while (((c = *(++ptr)) != 0 ||
4235               (nestptr != NULL &&
4236                 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != 0)) &&
4237               (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
4238    
4239      while ((c = *(++ptr)) != 0 && (c != ']' || inescq));      /* Check for missing terminating ']' */
4240    
4241      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)
4242        {        {
4243        *errorcodeptr = ERR6;        *errorcodeptr = ERR6;
4244        goto FAILED;        goto FAILED;
4245        }        }
4246    
   
 /* 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  
   
   
4247      /* 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
4248      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
4249      use of \p or \P, in other words, no use of any XCLASS features, we can      use of \p or \P, in other words, no use of any XCLASS features, we can
# Line 3348  we set the flag only if there is a liter Line 4251  we set the flag only if there is a liter
4251    
4252      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
4253      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
4254      operate on single-bytes only. This is an historical hangover. Maybe one day      operate on single-bytes characters only. This is an historical hangover.
4255      we can tidy these opcodes to handle multi-byte characters.      Maybe one day we can tidy these opcodes to handle multi-byte characters.
4256    
4257      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
4258      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.
4259      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
4260      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
4261      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
4262      reqbyte, save the previous value for reinstating. */      of reqbyte, save the previous value for reinstating. */
4263    
4264  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4265      if (class_charcount == 1 && !class_utf8 &&      if (class_charcount == 1 && !class_utf8 &&
# Line 3367  we set the flag only if there is a liter Line 4270  we set the flag only if there is a liter
4270        {        {
4271        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
4272    
4273        /* The OP_NOT opcode works on one-byte characters only. */        /* The OP_NOT[I] opcodes work on one-byte characters only. */
4274    
4275        if (negate_class)        if (negate_class)
4276          {          {
4277          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
4278          zerofirstbyte = firstbyte;          zerofirstbyte = firstbyte;
4279          *code++ = OP_NOT;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
4280          *code++ = class_lastchar;          *code++ = class_lastchar;
4281          break;          break;
4282          }          }
# Line 3404  we set the flag only if there is a liter Line 4307  we set the flag only if there is a liter
4307    
4308      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
4309      extended class, with its own opcode, unless there was a negated special      extended class, with its own opcode, unless there was a negated special
4310      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
4311      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
4312      (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
4313      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
4314        actual compiled code. */
4315    
4316  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4317      if (class_utf8 && !should_flip_negation)      if (class_utf8 && (!should_flip_negation || (options & PCRE_UCP) != 0))
4318        {        {
4319        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
4320        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 3436  we set the flag only if there is a liter Line 4340  we set the flag only if there is a liter
4340        }        }
4341  #endif  #endif
4342    
4343      /* 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
4344      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
4345      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
4346      map into the code vector, negating it if necessary. */      (non-UCP) in the class. Then copy the 32-byte map into the code vector,
4347        negating it if necessary. */
4348    
4349      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
4350      if (negate_class)      if (negate_class)
# Line 3459  we set the flag only if there is a liter Line 4364  we set the flag only if there is a liter
4364      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
4365      has been tested above. */      has been tested above. */
4366    
4367      case '{':      case CHAR_LEFT_CURLY_BRACKET:
4368      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
4369      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
4370      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
4371      goto REPEAT;      goto REPEAT;
4372    
4373      case '*':      case CHAR_ASTERISK:
4374      repeat_min = 0;      repeat_min = 0;
4375      repeat_max = -1;      repeat_max = -1;
4376      goto REPEAT;      goto REPEAT;
4377    
4378      case '+':      case CHAR_PLUS:
4379      repeat_min = 1;      repeat_min = 1;
4380      repeat_max = -1;      repeat_max = -1;
4381      goto REPEAT;      goto REPEAT;
4382    
4383      case '?':      case CHAR_QUESTION_MARK:
4384      repeat_min = 0;      repeat_min = 0;
4385      repeat_max = 1;      repeat_max = 1;
4386    
# Line 3499  we set the flag only if there is a liter Line 4404  we set the flag only if there is a liter
4404      op_type = 0;                    /* Default single-char op codes */      op_type = 0;                    /* Default single-char op codes */
4405      possessive_quantifier = FALSE;  /* Default not possessive quantifier */      possessive_quantifier = FALSE;  /* Default not possessive quantifier */
4406    
4407      /* 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
4408      for an inserted OP_ONCE for the additional '+' extension. */      insert something before it. */
4409    
4410      tempcode = previous;      tempcode = previous;
4411    
# Line 3510  we set the flag only if there is a liter Line 4415  we set the flag only if there is a liter
4415      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
4416      repeat type to the non-default. */      repeat type to the non-default. */
4417    
4418      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
4419        {        {
4420        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
4421        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
4422        ptr++;        ptr++;
4423        }        }
4424      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
4425        {        {
4426        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
4427        ptr++;        ptr++;
4428        }        }
4429      else repeat_type = greedy_default;      else repeat_type = greedy_default;
4430    
4431        /* If previous was a recursion call, wrap it in atomic brackets so that
4432        previous becomes the atomic group. All recursions were so wrapped in the
4433        past, but it no longer happens for non-repeated recursions. In fact, the
4434        repeated ones could be re-implemented independently so as not to need this,
4435        but for the moment we rely on the code for repeating groups. */
4436    
4437        if (*previous == OP_RECURSE)
4438          {
4439          memmove(previous + 1 + LINK_SIZE, previous, 1 + LINK_SIZE);
4440          *previous = OP_ONCE;
4441          PUT(previous, 1, 2 + 2*LINK_SIZE);
4442          previous[2 + 2*LINK_SIZE] = OP_KET;
4443          PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
4444          code += 2 + 2 * LINK_SIZE;
4445          length_prevgroup = 3 + 3*LINK_SIZE;
4446    
4447          /* When actually compiling, we need to check whether this was a forward
4448          reference, and if so, adjust the offset. */
4449    
4450          if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
4451            {
4452            int offset = GET(cd->hwm, -LINK_SIZE);
4453            if (offset == previous + 1 - cd->start_code)
4454              PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
4455            }
4456          }
4457    
4458        /* Now handle repetition for the different types of item. */
4459    
4460      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
4461      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
4462      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
4463      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
4464      instead.  */      instead.  */
4465    
4466      if (*previous == OP_CHAR || *previous == OP_CHARNC)      if (*previous == OP_CHAR || *previous == OP_CHARI)
4467        {        {
4468          op_type = (*previous == OP_CHAR)? 0 : OP_STARI - OP_STAR;
4469    
4470        /* 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
4471        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
4472        hold the length of the character in bytes, plus 0x80 to flag that it's a        hold the length of the character in bytes, plus 0x80 to flag that it's a
# Line 3563  we set the flag only if there is a liter Line 4499  we set the flag only if there is a liter
4499    
4500        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4501            repeat_max < 0 &&            repeat_max < 0 &&
4502            check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
             options, cd))  
4503          {          {
4504          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4505          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3576  we set the flag only if there is a liter Line 4511  we set the flag only if there is a liter
4511      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
4512      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-
4513      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
4514      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
4515      currently used only for single-byte chars. */      are currently used only for single-byte chars. */
4516    
4517      else if (*previous == OP_NOT)      else if (*previous == OP_NOT || *previous == OP_NOTI)
4518        {        {
4519        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = ((*previous == OP_NOT)? OP_NOTSTAR : OP_NOTSTARI) - OP_STAR;
4520        c = previous[1];        c = previous[1];
4521        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4522            repeat_max < 0 &&            repeat_max < 0 &&
4523            check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4524          {          {
4525          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4526          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3609  we set the flag only if there is a liter Line 4544  we set the flag only if there is a liter
4544    
4545        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4546            repeat_max < 0 &&            repeat_max < 0 &&
4547            check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4548          {          {
4549          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4550          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3631  we set the flag only if there is a liter Line 4566  we set the flag only if there is a liter
4566    
4567        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
4568    
4569          /*--------------------------------------------------------------------*/
4570          /* This code is obsolete from release 8.00; the restriction was finally
4571          removed: */
4572    
4573        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4574        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4575    
4576        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4577          /*--------------------------------------------------------------------*/
4578    
4579        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
4580    
# Line 3773  we set the flag only if there is a liter Line 4713  we set the flag only if there is a liter
4713  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4714               *previous == OP_XCLASS ||               *previous == OP_XCLASS ||
4715  #endif  #endif
4716               *previous == OP_REF)               *previous == OP_REF ||
4717                 *previous == OP_REFI)
4718        {        {
4719        if (repeat_max == 0)        if (repeat_max == 0)
4720          {          {
# Line 3781  we set the flag only if there is a liter Line 4722  we set the flag only if there is a liter
4722          goto END_REPEAT;          goto END_REPEAT;
4723          }          }
4724    
4725          /*--------------------------------------------------------------------*/
4726          /* This code is obsolete from release 8.00; the restriction was finally
4727          removed: */
4728    
4729        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4730        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4731    
4732        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4733          /*--------------------------------------------------------------------*/
4734    
4735        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4736          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 3802  we set the flag only if there is a liter Line 4748  we set the flag only if there is a liter
4748        }        }
4749    
4750      /* 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
4751      cases. */      cases. Note that at this point we can encounter only the "basic" bracket
4752        opcodes such as BRA and CBRA, as this is the place where they get converted
4753        into the more special varieties such as BRAPOS and SBRA. A test for >=
4754        OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
4755        ASSERTBACK_NOT, ONCE, BRA, CBRA, and COND. Originally, PCRE did not allow
4756        repetition of assertions, but now it does, for Perl compatibility. */
4757    
4758      else if (*previous == OP_BRA  || *previous == OP_CBRA ||      else if (*previous >= OP_ASSERT && *previous <= OP_COND)
              *previous == OP_ONCE || *previous == OP_COND)  
4759        {        {
4760        register int i;        register int i;
4761        int ketoffset = 0;        int len = (int)(code - previous);
       int len = code - previous;  
4762        uschar *bralink = NULL;        uschar *bralink = NULL;
4763          uschar *brazeroptr = NULL;
4764    
4765        /* Repeating a DEFINE group is pointless */        /* Repeating a DEFINE group is pointless, but Perl allows the syntax, so
4766          we just ignore the repeat. */
4767    
4768        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
4769          {          goto END_REPEAT;
4770          *errorcodeptr = ERR55;  
4771          goto FAILED;        /* There is no sense in actually repeating assertions. The only potential
4772          }        use of repetition is in cases when the assertion is optional. Therefore,
4773          if the minimum is greater than zero, just ignore the repeat. If the
4774          maximum is not not zero or one, set it to 1. */
4775    
4776        /* If the maximum repeat count is unlimited, find the end of the bracket        if (*previous < OP_ONCE)    /* Assertion */
4777        by scanning through from the start, and compute the offset back to it          {
4778        from the current code pointer. There may be an OP_OPT setting following          if (repeat_min > 0) goto END_REPEAT;
4779        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;  
4780          }          }
4781    
4782        /* The case of a zero minimum is special because of the need to stick        /* The case of a zero minimum is special because of the need to stick
# Line 3842  we set the flag only if there is a liter Line 4788  we set the flag only if there is a liter
4788    
4789        if (repeat_min == 0)        if (repeat_min == 0)
4790          {          {
4791          /* If the maximum is also zero, we just omit the group from the output          /* If the maximum is also zero, we used to just omit the group from the
4792          altogether. */          output altogether, like this:
   
         if (repeat_max == 0)  
           {  
           code = previous;  
           goto END_REPEAT;  
           }  
4793    
4794          /* If the maximum is 1 or unlimited, we just have to stick in the          ** if (repeat_max == 0)
4795          BRAZERO and do no more at this point. However, we do need to adjust          **   {
4796          any OP_RECURSE calls inside the group that refer to the group itself or          **   code = previous;
4797          any internal or forward referenced group, because the offset is from          **   goto END_REPEAT;
4798          the start of the whole regex. Temporarily terminate the pattern while          **   }
4799          doing this. */  
4800            However, that fails when a group or a subgroup within it is referenced
4801            as a subroutine from elsewhere in the pattern, so now we stick in
4802            OP_SKIPZERO in front of it so that it is skipped on execution. As we
4803            don't have a list of which groups are referenced, we cannot do this
4804            selectively.
4805    
4806            If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4807            and do no more at this point. However, we do need to adjust any
4808            OP_RECURSE calls inside the group that refer to the group itself or any
4809            internal or forward referenced group, because the offset is from the
4810            start of the whole regex. Temporarily terminate the pattern while doing
4811            this. */
4812    
4813          if (repeat_max <= 1)          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
4814            {            {
4815            *code = OP_END;            *code = OP_END;
4816            adjust_recurse(previous, 1, utf8, cd, save_hwm);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
4817            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
4818            code++;            code++;
4819              if (repeat_max == 0)
4820                {
4821                *previous++ = OP_SKIPZERO;
4822                goto END_REPEAT;
4823                }
4824              brazeroptr = previous;    /* Save for possessive optimizing */
4825            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4826            }            }
4827    
# Line 3888  we set the flag only if there is a liter Line 4846  we set the flag only if there is a liter
4846            /* We chain together the bracket offset fields that have to be            /* We chain together the bracket offset fields that have to be
4847            filled in later when the ends of the brackets are reached. */            filled in later when the ends of the brackets are reached. */
4848    
4849            offset = (bralink == NULL)? 0 : previous - bralink;            offset = (bralink == NULL)? 0 : (int)(previous - bralink);
4850            bralink = previous;            bralink = previous;
4851            PUTINC(previous, 0, offset);            PUTINC(previous, 0, offset);
4852            }            }
# Line 3909  we set the flag only if there is a liter Line 4867  we set the flag only if there is a liter
4867            {            {
4868            /* 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
4869            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
4870            potential integer overflow. */            potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
4871              integer type when available, otherwise double. */
4872    
4873            if (lengthptr != NULL)            if (lengthptr != NULL)
4874              {              {
4875              int delta = (repeat_min - 1)*length_prevgroup;              int delta = (repeat_min - 1)*length_prevgroup;
4876              if ((double)(repeat_min - 1)*(double)length_prevgroup >              if ((INT64_OR_DOUBLE)(repeat_min - 1)*
4877                                                              (double)INT_MAX ||                    (INT64_OR_DOUBLE)length_prevgroup >
4878                        (INT64_OR_DOUBLE)INT_MAX ||
4879                  OFLOW_MAX - *lengthptr < delta)                  OFLOW_MAX - *lengthptr < delta)
4880                {                {
4881                *errorcodeptr = ERR20;                *errorcodeptr = ERR20;
# Line 3936  we set the flag only if there is a liter Line 4896  we set the flag only if there is a liter
4896                memcpy(code, previous, len);                memcpy(code, previous, len);
4897                for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)                for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
4898                  {                  {
4899                    if (cd->hwm >= cd->start_workspace + WORK_SIZE_CHECK)
4900                      {
4901                      *errorcodeptr = ERR72;
4902                      goto FAILED;
4903                      }
4904                  PUT(cd->hwm, 0, GET(hc, 0) + len);                  PUT(cd->hwm, 0, GET(hc, 0) + len);
4905                  cd->hwm += LINK_SIZE;                  cd->hwm += LINK_SIZE;
4906                  }                  }
# Line 3961  we set the flag only if there is a liter Line 4926  we set the flag only if there is a liter
4926          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
4927          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
4928          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
4929          paranoid checks to avoid integer overflow. */          paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
4930            a 64-bit integer type when available, otherwise double. */
4931    
4932          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4933            {            {
4934            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4935                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4936            if ((double)repeat_max *            if ((INT64_OR_DOUBLE)repeat_max *
4937                  (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)                  (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4938                    > (double)INT_MAX ||                    > (INT64_OR_DOUBLE)INT_MAX ||
4939                OFLOW_MAX - *lengthptr < delta)                OFLOW_MAX - *lengthptr < delta)
4940              {              {
4941              *errorcodeptr = ERR20;              *errorcodeptr = ERR20;
# Line 3994  we set the flag only if there is a liter Line 4960  we set the flag only if there is a liter
4960              {              {
4961              int offset;              int offset;
4962              *code++ = OP_BRA;              *code++ = OP_BRA;
4963              offset = (bralink == NULL)? 0 : code - bralink;              offset = (bralink == NULL)? 0 : (int)(code - bralink);
4964              bralink = code;              bralink = code;
4965              PUTINC(code, 0, offset);              PUTINC(code, 0, offset);
4966              }              }
# Line 4002  we set the flag only if there is a liter Line 4968  we set the flag only if there is a liter
4968            memcpy(code, previous, len);            memcpy(code, previous, len);
4969            for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)            for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
4970              {              {
4971                if (cd->hwm >= cd->start_workspace + WORK_SIZE_CHECK)
4972                  {
4973                  *errorcodeptr = ERR72;
4974                  goto FAILED;
4975                  }
4976              PUT(cd->hwm, 0, GET(hc, 0) + len + ((i != 0)? 2+LINK_SIZE : 1));              PUT(cd->hwm, 0, GET(hc, 0) + len + ((i != 0)? 2+LINK_SIZE : 1));
4977              cd->hwm += LINK_SIZE;              cd->hwm += LINK_SIZE;
4978              }              }
# Line 4015  we set the flag only if there is a liter Line 4986  we set the flag only if there is a liter
4986          while (bralink != NULL)          while (bralink != NULL)
4987            {            {
4988            int oldlinkoffset;            int oldlinkoffset;
4989            int offset = code - bralink + 1;            int offset = (int)(code - bralink + 1);
4990            uschar *bra = code - offset;            uschar *bra = code - offset;
4991            oldlinkoffset = GET(bra, 1);            oldlinkoffset = GET(bra, 1);
4992            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;
# Line 4025  we set the flag only if there is a liter Line 4996  we set the flag only if there is a liter
4996            }            }
4997          }          }
4998    
4999        /* 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
5000        can't just offset backwards from the current code point, because we        ONCE brackets, that's all we need to do. However, possessively repeated
5001        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
5002        correct offset was computed above.        behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to
5003          deal with possessive ONCEs specially.
5004        Then, when we are doing the actual compile phase, check to see whether  
5005        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
5006          whether this group is one that could match an empty string. If so,
5007        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
5008        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
5009        atomic groups at runtime, but in a different way.] */        groups at runtime, but in a different way.]
5010    
5011          Then, if the quantifier was possessive and the bracket is not a
5012          conditional, we convert the BRA code to the POS form, and the KET code to
5013          KETRPOS. (It turns out to be convenient at runtime to detect this kind of
5014          subpattern at both the start and at the end.) The use of special opcodes
5015          makes it possible to reduce greatly the stack usage in pcre_exec(). If
5016          the group is preceded by OP_BRAZERO, convert this to OP_BRAPOSZERO.
5017    
5018          Then, if the minimum number of matches is 1 or 0, cancel the possessive
5019          flag so that the default action below, of wrapping everything inside
5020          atomic brackets, does not happen. When the minimum is greater than 1,
5021          there will be earlier copies of the group, and so we still have to wrap
5022          the whole thing. */
5023    
5024        else        else
5025          {          {
5026          uschar *ketcode = code - ketoffset;          uschar *ketcode = code - 1 - LINK_SIZE;
5027          uschar *bracode = ketcode - GET(ketcode, 1);          uschar *bracode = ketcode - GET(ketcode, 1);
5028          *ketcode = OP_KETRMAX + repeat_type;  
5029          if (lengthptr == NULL && *bracode != OP_ONCE)          /* Convert possessive ONCE brackets to non-capturing */
5030    
5031            if ((*bracode == OP_ONCE || *bracode == OP_ONCE_NC) &&
5032                possessive_quantifier) *bracode = OP_BRA;
5033    
5034            /* For non-possessive ONCE brackets, all we need to do is to
5035            set the KET. */
5036    
5037            if (*bracode == OP_ONCE || *bracode == OP_ONCE_NC)
5038              *ketcode = OP_KETRMAX + repeat_type;
5039    
5040            /* Handle non-ONCE brackets and possessive ONCEs (which have been
5041            converted to non-capturing above). */
5042    
5043            else
5044            {            {
5045            uschar *scode = bracode;            /* In the compile phase, check for empty string matching. */
5046            do  
5047              if (lengthptr == NULL)
5048              {              {
5049              if (could_be_empty_branch(scode, ketcode, utf8))              uschar *scode = bracode;
5050                do
5051                {                {
5052                *bracode += OP_SBRA - OP_BRA;                if (could_be_empty_branch(scode, ketcode, utf8, cd))
5053                break;                  {
5054                    *bracode += OP_SBRA - OP_BRA;
5055                    break;