/[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 335 by ph10, Sat Apr 12 14:36:14 2008 UTC revision 518 by ph10, Tue May 18 15:47:01 2010 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) 1997-2010 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 297  static const char error_texts[] = Line 397  static const char error_texts[] =
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      ;
412    
413  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
414  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 426  For convenience, we use the same bit def
426    
427  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
428    
429  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC
430    
431    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
432    UTF-8 mode. */
433    
434  static const unsigned char digitab[] =  static const unsigned char digitab[] =
435    {    {
436    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 466  static const unsigned char digitab[] =
466    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
467    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
468    
469  #else           /* This is the "abnormal" case, for EBCDIC systems */  #else
470    
471    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
472    
473  static const unsigned char digitab[] =  static const unsigned char digitab[] =
474    {    {
475    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
# Line 454  static const char * Line 566  static const char *
566  find_error_text(int n)  find_error_text(int n)
567  {  {
568  const char *s = error_texts;  const char *s = error_texts;
569  for (; n > 0; n--) while (*s++ != 0);  for (; n > 0; n--)
570      {
571      while (*s++ != 0) {};
572      if (*s == 0) return "Error text not found (please report)";
573      }
574  return s;  return s;
575  }  }
576    
# Line 502  if (c == 0) *errorcodeptr = ERR1; Line 618  if (c == 0) *errorcodeptr = ERR1;
618  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.
619  Otherwise further processing may be required. */  Otherwise further processing may be required. */
620    
621  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
622  else if (c < '0' || c > 'z') {}                           /* Not alphanumeric */  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */
623  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0) c = i;
624    
625  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
626  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 639  else
639      /* 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
640      error. */      error. */
641    
642      case 'l':      case CHAR_l:
643      case 'L':      case CHAR_L:
644      case 'N':      case CHAR_u:
645      case 'u':      case CHAR_U:
     case 'U':  
646      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
647      break;      break;
648    
649      /* \g must be followed by one of a number of specific things:      /* \g must be followed by one of a number of specific things:
650    
651      (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
652      backreference. If negative, it is a relative backreference. This is a Perl      backreference. If negative, it is a relative backreference. This is a Perl
653      5.10 feature.      5.10 feature.
654    
655      (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
656      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
657      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
658      was \k.      was \k.
659    
660      (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
661      number either in angle brackets or in single quotes. However, these are      number either in angle brackets or in single quotes. However, these are
662      (possibly recursive) subroutine calls, _not_ backreferences. Just return      (possibly recursive) subroutine calls, _not_ backreferences. Just return
663      the -ESC_g code (cf \k). */      the -ESC_g code (cf \k). */
664    
665      case 'g':      case CHAR_g:
666      if (ptr[1] == '<' || ptr[1] == '\'')      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
667        {        {
668        c = -ESC_g;        c = -ESC_g;
669        break;        break;
670        }        }
671    
672      /* Handle the Perl-compatible cases */      /* Handle the Perl-compatible cases */
673    
674      if (ptr[1] == '{')      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
675        {        {
676        const uschar *p;        const uschar *p;
677        for (p = ptr+2; *p != 0 && *p != '}'; p++)        for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
678          if (*p != '-' && (digitab[*p] & ctype_digit) == 0) break;          if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
679        if (*p != 0 && *p != '}')        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
680          {          {
681          c = -ESC_k;          c = -ESC_k;
682          break;          break;
# Line 571  else Line 686  else
686        }        }
687      else braced = FALSE;      else braced = FALSE;
688    
689      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
690        {        {
691        negated = TRUE;        negated = TRUE;
692        ptr++;        ptr++;
# Line 580  else Line 695  else
695    
696      c = 0;      c = 0;
697      while ((digitab[ptr[1]] & ctype_digit) != 0)      while ((digitab[ptr[1]] & ctype_digit) != 0)
698        c = c * 10 + *(++ptr) - '0';        c = c * 10 + *(++ptr) - CHAR_0;
699    
700      if (c < 0)   /* Integer overflow */      if (c < 0)   /* Integer overflow */
701        {        {
702        *errorcodeptr = ERR61;        *errorcodeptr = ERR61;
703        break;        break;
704        }        }
705    
706      if (braced && *(++ptr) != '}')      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
707        {        {
708        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
709        break;        break;
710        }        }
711    
712      if (c == 0)      if (c == 0)
713        {        {
714        *errorcodeptr = ERR58;        *errorcodeptr = ERR58;
715        break;        break;
716        }        }
717    
718      if (negated)      if (negated)
719        {        {
# Line 625  else Line 740  else
740      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
741      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
742    
743      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:
744      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
745    
746      if (!isclass)      if (!isclass)
747        {        {
748        oldptr = ptr;        oldptr = ptr;
749        c -= '0';        c -= CHAR_0;
750        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
751          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - CHAR_0;
752        if (c < 0)    /* Integer overflow */        if (c < 0)    /* Integer overflow */
753          {          {
754          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
# Line 651  else Line 766  else
766      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.
767      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
768    
769      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
770        {        {
771        ptr--;        ptr--;
772        c = 0;        c = 0;
# Line 664  else Line 779  else
779      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
780      than 3 octal digits. */      than 3 octal digits. */
781    
782      case '0':      case CHAR_0:
783      c -= '0';      c -= CHAR_0;
784      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
785          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
786      if (!utf8 && c > 255) *errorcodeptr = ERR51;      if (!utf8 && c > 255) *errorcodeptr = ERR51;
787      break;      break;
788    
# Line 675  else Line 790  else
790      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
791      treated as a data character. */      treated as a data character. */
792    
793      case 'x':      case CHAR_x:
794      if (ptr[1] == '{')      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
795        {        {
796        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
797        int count = 0;        int count = 0;
# Line 685  else Line 800  else
800        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
801          {          {
802          register int cc = *pt++;          register int cc = *pt++;
803          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
804          count++;          count++;
805    
806  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
807          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
808          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
809  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
810          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
811          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
812  #endif  #endif
813          }          }
814    
815        if (*pt == '}')        if (*pt == CHAR_RIGHT_CURLY_BRACKET)
816          {          {
817          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
818          ptr = pt;          ptr = pt;
# Line 713  else Line 828  else
828      c = 0;      c = 0;
829      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
830        {        {
831        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
832        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
833  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
834        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
835        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
836  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
837        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
838        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
839  #endif  #endif
840        }        }
841      break;      break;
# Line 729  else Line 844  else
844      This coding is ASCII-specific, but then the whole concept of \cx is      This coding is ASCII-specific, but then the whole concept of \cx is
845      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
846    
847      case 'c':      case CHAR_c:
848      c = *(++ptr);      c = *(++ptr);
849      if (c == 0)      if (c == 0)
850        {        {
# Line 737  else Line 852  else
852        break;        break;
853        }        }
854    
855  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
856      if (c >= 'a' && c <= 'z') c -= 32;      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
857      c ^= 0x40;      c ^= 0x40;
858  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
859      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
860      c ^= 0xC0;      c ^= 0xC0;
861  #endif  #endif
862      break;      break;
# Line 763  else Line 878  else
878      }      }
879    }    }
880    
881    /* Perl supports \N{name} for character names, as well as plain \N for "not
882    newline". PCRE does not support \N{name}. */
883    
884    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET)
885      *errorcodeptr = ERR37;
886    
887    /* If PCRE_UCP is set, we change the values for \d etc. */
888    
889    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
890      c -= (ESC_DU - ESC_D);
891    
892    /* Set the pointer to the final character before returning. */
893    
894  *ptrptr = ptr;  *ptrptr = ptr;
895  return c;  return c;
896  }  }
# Line 803  if (c == 0) goto ERROR_RETURN; Line 931  if (c == 0) goto ERROR_RETURN;
931  /* \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
932  negation. */  negation. */
933    
934  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
935    {    {
936    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
937      {      {
938      *negptr = TRUE;      *negptr = TRUE;
939      ptr++;      ptr++;
# Line 814  if (c == '{') Line 942  if (c == '{')
942      {      {
943      c = *(++ptr);      c = *(++ptr);
944      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
945      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
946      name[i] = c;      name[i] = c;
947      }      }
948    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
949    name[i] = 0;    name[i] = 0;
950    }    }
951    
# Line 882  is_counted_repeat(const uschar *p) Line 1010  is_counted_repeat(const uschar *p)
1010  {  {
1011  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
1012  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
1013  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
1014    
1015  if (*p++ != ',') return FALSE;  if (*p++ != CHAR_COMMA) return FALSE;
1016  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
1017    
1018  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
1019  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
1020    
1021  return (*p == '}');  return (*p == CHAR_RIGHT_CURLY_BRACKET);
1022  }  }
1023    
1024    
# Line 923  int max = -1; Line 1051  int max = -1;
1051  /* 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
1052  an integer overflow. */  an integer overflow. */
1053    
1054  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;
1055  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1056    {    {
1057    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 933  if (min < 0 || min > 65535) Line 1061  if (min < 0 || min > 65535)
1061  /* 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.
1062  Also, max must not be less than min. */  Also, max must not be less than min. */
1063    
1064  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1065    {    {
1066    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1067      {      {
1068      max = 0;      max = 0;
1069      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;
1070      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1071        {        {
1072        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 963  return p; Line 1091  return p;
1091    
1092    
1093  /*************************************************  /*************************************************
1094  *       Find forward referenced subpattern       *  *  Subroutine for finding forward reference      *
1095  *************************************************/  *************************************************/
1096    
1097  /* This function scans along a pattern's text looking for capturing  /* This recursive function is called only from find_parens() below. The
1098    top-level call starts at the beginning of the pattern. All other calls must
1099    start at a parenthesis. It scans along a pattern's text looking for capturing
1100  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
1101  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
1102  returns when it reaches a given numbered subpattern. This is used for forward  returns when it reaches a given numbered subpattern. We know that if (?P< is
1103  references to subpatterns. We know that if (?P< is encountered, the name will  encountered, the name will be terminated by '>' because that is checked in the
1104  be terminated by '>' because that is checked in the first pass.  first pass. Recursion is used to keep track of subpatterns that reset the
1105    capturing group numbers - the (?| feature.
1106    
1107  Arguments:  Arguments:
1108    ptr          current position in the pattern    ptrptr       address of the current character pointer (updated)
1109    count        current count of capturing parens so far encountered    cd           compile background data
1110    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1111    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1112    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1113      count        pointer to the current capturing subpattern number (updated)
1114    
1115  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1116  */  */
1117    
1118  static int  static int
1119  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,
1120    BOOL xmode)    BOOL xmode, int *count)
1121  {  {
1122  const uschar *thisname;  uschar *ptr = *ptrptr;
1123    int start_count = *count;
1124    int hwm_count = start_count;
1125    BOOL dup_parens = FALSE;
1126    
1127  for (; *ptr != 0; ptr++)  /* If the first character is a parenthesis, check on the type of group we are
1128    dealing with. The very first call may not start with a parenthesis. */
1129    
1130    if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1131    {    {
1132    int term;    if (ptr[1] == CHAR_QUESTION_MARK &&
1133          ptr[2] == CHAR_VERTICAL_LINE)
1134        {
1135        ptr += 3;
1136        dup_parens = TRUE;
1137        }
1138    
1139      /* Handle a normal, unnamed capturing parenthesis */
1140    
1141      else if (ptr[1] != CHAR_QUESTION_MARK && ptr[1] != CHAR_ASTERISK)
1142        {
1143        *count += 1;
1144        if (name == NULL && *count == lorn) return *count;
1145        ptr++;
1146        }
1147    
1148      /* Handle a condition. If it is an assertion, just carry on so that it
1149      is processed as normal. If not, skip to the closing parenthesis of the
1150      condition (there can't be any nested parens. */
1151    
1152      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1153        {
1154        ptr += 2;
1155        if (ptr[1] != CHAR_QUESTION_MARK)
1156          {
1157          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1158          if (*ptr != 0) ptr++;
1159          }
1160        }
1161    
1162      /* We have either (? or (* and not a condition */
1163    
1164      else
1165        {
1166        ptr += 2;
1167        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1168    
1169        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1170    
1171        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1172            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1173          {
1174          int term;
1175          const uschar *thisname;
1176          *count += 1;
1177          if (name == NULL && *count == lorn) return *count;
1178          term = *ptr++;
1179          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1180          thisname = ptr;
1181          while (*ptr != term) ptr++;
1182          if (name != NULL && lorn == ptr - thisname &&
1183              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1184            return *count;
1185          term++;
1186          }
1187        }
1188      }
1189    
1190    /* Past any initial parenthesis handling, scan for parentheses or vertical
1191    bars. */
1192    
1193    for (; *ptr != 0; ptr++)
1194      {
1195    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1196    
1197    if (*ptr == '\\')    if (*ptr == CHAR_BACKSLASH)
1198      {      {
1199      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1200      if (*ptr == 'Q') for (;;)      if (*ptr == CHAR_Q) for (;;)
1201        {        {
1202        while (*(++ptr) != 0 && *ptr != '\\');        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1203        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1204        if (*(++ptr) == 'E') break;        if (*(++ptr) == CHAR_E) break;
1205        }        }
1206      continue;      continue;
1207      }      }
1208    
1209    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1210      are handled for real. If the first character is '^', skip it. Also, if the
1211      first few characters (either before or after ^) are \Q\E or \E we skip them
1212      too. This makes for compatibility with Perl. Note the use of STR macros to
1213      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1214    
1215    if (*ptr == '[')    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1216      {      {
1217      while (*(++ptr) != ']')      BOOL negate_class = FALSE;
1218        for (;;)
1219          {
1220          if (ptr[1] == CHAR_BACKSLASH)
1221            {
1222            if (ptr[2] == CHAR_E)
1223              ptr+= 2;
1224            else if (strncmp((const char *)ptr+2,
1225                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1226              ptr += 4;
1227            else
1228              break;
1229            }
1230          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1231            {
1232            negate_class = TRUE;
1233            ptr++;
1234            }
1235          else break;
1236          }
1237    
1238        /* If the next character is ']', it is a data character that must be
1239        skipped, except in JavaScript compatibility mode. */
1240    
1241        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1242            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1243          ptr++;
1244    
1245        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1246        {        {
1247        if (*ptr == 0) return -1;        if (*ptr == 0) return -1;
1248        if (*ptr == '\\')        if (*ptr == CHAR_BACKSLASH)
1249          {          {
1250          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) goto FAIL_EXIT;
1251          if (*ptr == 'Q') for (;;)          if (*ptr == CHAR_Q) for (;;)
1252            {            {
1253            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1254            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1255            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1256            }            }
1257          continue;          continue;
1258          }          }
# Line 1031  for (; *ptr != 0; ptr++) Line 1262  for (; *ptr != 0; ptr++)
1262    
1263    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1264    
1265    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1266      {      {
1267      while (*(++ptr) != 0 && *ptr != '\n');      while (*(++ptr) != 0 && *ptr != CHAR_NL) {};
1268      if (*ptr == 0) return -1;      if (*ptr == 0) goto FAIL_EXIT;
1269      continue;      continue;
1270      }      }
1271    
1272    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1273    
1274    if (*ptr != '(') continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != '?' && ptr[1] != '*')  
1275      {      {
1276      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, count);
1277      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1278      continue;      if (*ptr == 0) goto FAIL_EXIT;
1279      }      }
1280    
1281    ptr += 2;    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1282    if (*ptr == 'P') ptr++;                      /* Allow optional P */      {
1283        if (dup_parens && *count < hwm_count) *count = hwm_count;
1284        *ptrptr = ptr;
1285        return -1;
1286        }
1287    
1288    /* We have to disambiguate (?<! and (?<= from (?<name> */    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1289        {
1290        if (*count > hwm_count) hwm_count = *count;
1291        *count = start_count;
1292        }
1293      }
1294    
1295    if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  FAIL_EXIT:
1296         *ptr != '\'')  *ptrptr = ptr;
1297      continue;  return -1;
1298    }
1299    
1300    
1301    
1302    
1303    /*************************************************
1304    *       Find forward referenced subpattern       *
1305    *************************************************/
1306    
1307    /* This function scans along a pattern's text looking for capturing
1308    subpatterns, and counting them. If it finds a named pattern that matches the
1309    name it is given, it returns its number. Alternatively, if the name is NULL, it
1310    returns when it reaches a given numbered subpattern. This is used for forward
1311    references to subpatterns. We used to be able to start this scan from the
1312    current compiling point, using the current count value from cd->bracount, and
1313    do it all in a single loop, but the addition of the possibility of duplicate
1314    subpattern numbers means that we have to scan from the very start, in order to
1315    take account of such duplicates, and to use a recursive function to keep track
1316    of the different types of group.
1317    
1318    Arguments:
1319      cd           compile background data
1320      name         name to seek, or NULL if seeking a numbered subpattern
1321      lorn         name length, or subpattern number if name is NULL
1322      xmode        TRUE if we are in /x mode
1323    
1324    Returns:       the number of the found subpattern, or -1 if not found
1325    */
1326    
1327    count++;  static int
1328    find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode)
1329    {
1330    uschar *ptr = (uschar *)cd->start_pattern;
1331    int count = 0;
1332    int rc;
1333    
1334    /* If the pattern does not start with an opening parenthesis, the first call
1335    to find_parens_sub() will scan right to the end (if necessary). However, if it
1336    does start with a parenthesis, find_parens_sub() will return when it hits the
1337    matching closing parens. That is why we have to have a loop. */
1338    
1339    if (name == NULL && count == lorn) return count;  for (;;)
1340    term = *ptr++;    {
1341    if (term == '<') term = '>';    rc = find_parens_sub(&ptr, cd, name, lorn, xmode, &count);
1342    thisname = ptr;    if (rc > 0 || *ptr++ == 0) break;
   while (*ptr != term) ptr++;  
   if (name != NULL && lorn == ptr - thisname &&  
       strncmp((const char *)name, (const char *)thisname, lorn) == 0)  
     return count;  
1343    }    }
1344    
1345  return -1;  return rc;
1346  }  }
1347    
1348    
1349    
1350    
1351  /*************************************************  /*************************************************
1352  *      Find first significant op code            *  *      Find first significant op code            *
1353  *************************************************/  *************************************************/
# Line 1123  for (;;) Line 1397  for (;;)
1397    
1398      case OP_CALLOUT:      case OP_CALLOUT:
1399      case OP_CREF:      case OP_CREF:
1400        case OP_NCREF:
1401      case OP_RREF:      case OP_RREF:
1402        case OP_NRREF:
1403      case OP_DEF:      case OP_DEF:
1404      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1405      break;      break;
# Line 1139  for (;;) Line 1415  for (;;)
1415    
1416    
1417  /*************************************************  /*************************************************
1418  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1419  *************************************************/  *************************************************/
1420    
1421  /* 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,
1422  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.
1423  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
1424    temporarily terminated with OP_END when this function is called.
1425    
1426    This function is called when a backward assertion is encountered, so that if it
1427    fails, the error message can point to the correct place in the pattern.
1428    However, we cannot do this when the assertion contains subroutine calls,
1429    because they can be forward references. We solve this by remembering this case
1430    and doing the check at the end; a flag specifies which mode we are running in.
1431    
1432  Arguments:  Arguments:
1433    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1434    options  the compiling options    options  the compiling options
1435      atend    TRUE if called when the pattern is complete
1436      cd       the "compile data" structure
1437    
1438  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1439                 or -1 if there is no fixed length,
1440               or -2 if \C was encountered               or -2 if \C was encountered
1441                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1442  */  */
1443    
1444  static int  static int
1445  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)
1446  {  {
1447  int length = -1;  int length = -1;
1448    
# Line 1168  branch, check the length against that of Line 1455  branch, check the length against that of
1455  for (;;)  for (;;)
1456    {    {
1457    int d;    int d;
1458      uschar *ce, *cs;
1459    register int op = *cc;    register int op = *cc;
1460    switch (op)    switch (op)
1461      {      {
# Line 1175  for (;;) Line 1463  for (;;)
1463      case OP_BRA:      case OP_BRA:
1464      case OP_ONCE:      case OP_ONCE:
1465      case OP_COND:      case OP_COND:
1466      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);
1467      if (d < 0) return d;      if (d < 0) return d;
1468      branchlength += d;      branchlength += d;
1469      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1198  for (;;) Line 1486  for (;;)
1486      branchlength = 0;      branchlength = 0;
1487      break;      break;
1488    
1489        /* A true recursion implies not fixed length, but a subroutine call may
1490        be OK. If the subroutine is a forward reference, we can't deal with
1491        it until the end of the pattern, so return -3. */
1492    
1493        case OP_RECURSE:
1494        if (!atend) return -3;
1495        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1496        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1497        if (cc > cs && cc < ce) return -1;                /* Recursion */
1498        d = find_fixedlength(cs + 2, options, atend, cd);
1499        if (d < 0) return d;
1500        branchlength += d;
1501        cc += 1 + LINK_SIZE;
1502        break;
1503    
1504      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1505    
1506      case OP_ASSERT:      case OP_ASSERT:
# Line 1211  for (;;) Line 1514  for (;;)
1514    
1515      case OP_REVERSE:      case OP_REVERSE:
1516      case OP_CREF:      case OP_CREF:
1517        case OP_NCREF:
1518      case OP_RREF:      case OP_RREF:
1519        case OP_NRREF:
1520      case OP_DEF:      case OP_DEF:
1521      case OP_OPT:      case OP_OPT:
1522      case OP_CALLOUT:      case OP_CALLOUT:
1523      case OP_SOD:      case OP_SOD:
1524      case OP_SOM:      case OP_SOM:
1525        case OP_SET_SOM:
1526      case OP_EOD:      case OP_EOD:
1527      case OP_EODN:      case OP_EODN:
1528      case OP_CIRC:      case OP_CIRC:
# Line 1234  for (;;) Line 1540  for (;;)
1540      branchlength++;      branchlength++;
1541      cc += 2;      cc += 2;
1542  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1543      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1544        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1545  #endif  #endif
1546      break;      break;
1547    
# Line 1248  for (;;) Line 1552  for (;;)
1552      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1553      cc += 4;      cc += 4;
1554  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1555      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1556        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1557  #endif  #endif
1558      break;      break;
1559    
# Line 1275  for (;;) Line 1577  for (;;)
1577      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1578      case OP_WORDCHAR:      case OP_WORDCHAR:
1579      case OP_ANY:      case OP_ANY:
1580        case OP_ALLANY:
1581      branchlength++;      branchlength++;
1582      cc++;      cc++;
1583      break;      break;
# Line 1329  for (;;) Line 1632  for (;;)
1632    
1633    
1634  /*************************************************  /*************************************************
1635  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1636  *************************************************/  *************************************************/
1637    
1638  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1639  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1640    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1641    so that it can be called from pcre_study() when finding the minimum matching
1642    length.
1643    
1644  Arguments:  Arguments:
1645    code        points to start of expression    code        points to start of expression
1646    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1647    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1648    
1649  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
1650  */  */
1651    
1652  static const uschar *  const uschar *
1653  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1654  {  {
1655  for (;;)  for (;;)
1656    {    {
# Line 1357  for (;;) Line 1663  for (;;)
1663    
1664    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1665    
1666      /* Handle recursion */
1667    
1668      else if (c == OP_REVERSE)
1669        {
1670        if (number < 0) return (uschar *)code;
1671        code += _pcre_OP_lengths[c];
1672        }
1673    
1674    /* Handle capturing bracket */    /* Handle capturing bracket */
1675    
1676    else if (c == OP_CBRA)    else if (c == OP_CBRA)
# Line 1368  for (;;) Line 1682  for (;;)
1682    
1683    /* 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
1684    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
1685    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1686      must add in its length. */
1687    
1688    else    else
1689      {      {
# Line 1392  for (;;) Line 1707  for (;;)
1707        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
1708        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1709        break;        break;
1710    
1711          case OP_MARK:
1712          case OP_PRUNE_ARG:
1713          case OP_SKIP_ARG:
1714          case OP_THEN_ARG:
1715          code += code[1];
1716          break;
1717        }        }
1718    
1719      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1423  for (;;) Line 1745  for (;;)
1745        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1746        break;        break;
1747        }        }
1748    #else
1749        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1750  #endif  #endif
1751      }      }
1752    }    }
# Line 1461  for (;;) Line 1785  for (;;)
1785    
1786    /* 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
1787    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
1788    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1789      must add in its length. */
1790    
1791    else    else
1792      {      {
# Line 1485  for (;;) Line 1810  for (;;)
1810        case OP_TYPEEXACT:        case OP_TYPEEXACT:
1811        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1812        break;        break;
1813    
1814          case OP_MARK:
1815          case OP_PRUNE_ARG:
1816          case OP_SKIP_ARG:
1817          case OP_THEN_ARG:
1818          code += code[1];
1819          break;
1820        }        }
1821    
1822      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1516  for (;;) Line 1848  for (;;)
1848        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1849        break;        break;
1850        }        }
1851    #else
1852        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1853  #endif  #endif
1854      }      }
1855    }    }
# Line 1539  Arguments: Line 1873  Arguments:
1873    code        points to start of search    code        points to start of search
1874    endcode     points to where to stop    endcode     points to where to stop
1875    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
1876      cd          contains pointers to tables etc.
1877    
1878  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
1879  */  */
1880    
1881  static BOOL  static BOOL
1882  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
1883      compile_data *cd)
1884  {  {
1885  register int c;  register int c;
1886  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);
# Line 1575  for (code = first_significant_code(code Line 1911  for (code = first_significant_code(code
1911      continue;      continue;
1912      }      }
1913    
1914      /* For a recursion/subroutine call, if its end has been reached, which
1915      implies a subroutine call, we can scan it. */
1916    
1917      if (c == OP_RECURSE)
1918        {
1919        BOOL empty_branch = FALSE;
1920        const uschar *scode = cd->start_code + GET(code, 1);
1921        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
1922        do
1923          {
1924          if (could_be_empty_branch(scode, endcode, utf8, cd))
1925            {
1926            empty_branch = TRUE;
1927            break;
1928            }
1929          scode += GET(scode, 1);
1930          }
1931        while (*scode == OP_ALT);
1932        if (!empty_branch) return FALSE;  /* All branches are non-empty */
1933        continue;
1934        }
1935    
1936    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
1937    
1938    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)
# Line 1582  for (code = first_significant_code(code Line 1940  for (code = first_significant_code(code
1940      BOOL empty_branch;      BOOL empty_branch;
1941      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
1942    
1943      /* Scan a closed bracket */      /* If a conditional group has only one branch, there is a second, implied,
1944        empty branch, so just skip over the conditional, because it could be empty.
1945        Otherwise, scan the individual branches of the group. */
1946    
1947      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;  
1948        code += GET(code, 1);        code += GET(code, 1);
1949        else
1950          {
1951          empty_branch = FALSE;
1952          do
1953            {
1954            if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
1955              empty_branch = TRUE;
1956            code += GET(code, 1);
1957            }
1958          while (*code == OP_ALT);
1959          if (!empty_branch) return FALSE;   /* All branches are non-empty */
1960        }        }
1961      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
1962      c = *code;      c = *code;
1963      continue;      continue;
1964      }      }
# Line 1653  for (code = first_significant_code(code Line 2019  for (code = first_significant_code(code
2019      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2020      case OP_WORDCHAR:      case OP_WORDCHAR:
2021      case OP_ANY:      case OP_ANY:
2022        case OP_ALLANY:
2023      case OP_ANYBYTE:      case OP_ANYBYTE:
2024      case OP_CHAR:      case OP_CHAR:
2025      case OP_CHARNC:      case OP_CHARNC:
# Line 1709  for (code = first_significant_code(code Line 2076  for (code = first_significant_code(code
2076      case OP_QUERY:      case OP_QUERY:
2077      case OP_MINQUERY:      case OP_MINQUERY:
2078      case OP_POSQUERY:      case OP_POSQUERY:
2079        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2080        break;
2081    
2082      case OP_UPTO:      case OP_UPTO:
2083      case OP_MINUPTO:      case OP_MINUPTO:
2084      case OP_POSUPTO:      case OP_POSUPTO:
2085      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2086      break;      break;
2087  #endif  #endif
2088    
2089        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2090        string. */
2091    
2092        case OP_MARK:
2093        case OP_PRUNE_ARG:
2094        case OP_SKIP_ARG:
2095        case OP_THEN_ARG:
2096        code += code[1];
2097        break;
2098    
2099        /* None of the remaining opcodes are required to match a character. */
2100    
2101        default:
2102        break;
2103      }      }
2104    }    }
2105    
# Line 1737  Arguments: Line 2122  Arguments:
2122    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2123    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2124    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2125      cd          pointers to tables etc
2126    
2127  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2128  */  */
2129    
2130  static BOOL  static BOOL
2131  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2132    BOOL utf8)    BOOL utf8, compile_data *cd)
2133  {  {
2134  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2135    {    {
2136    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2137        return FALSE;
2138    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2139    }    }
2140  return TRUE;  return TRUE;
# Line 1793  int terminator;          /* Don't combin Line 2180  int terminator;          /* Don't combin
2180  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2181  for (++ptr; *ptr != 0; ptr++)  for (++ptr; *ptr != 0; ptr++)
2182    {    {
2183    if (*ptr == '\\' && ptr[1] == ']') ptr++; else    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else
2184      {      {
2185      if (*ptr == ']') return FALSE;      if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2186      if (*ptr == terminator && ptr[1] == ']')      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2187        {        {
2188        *endptr = ptr;        *endptr = ptr;
2189        return TRUE;        return TRUE;
# Line 1883  while ((ptr = (uschar *)find_recurse(ptr Line 2270  while ((ptr = (uschar *)find_recurse(ptr
2270    
2271    /* 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
2272    reference. */    reference. */
2273    
2274    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
2275      {      {
2276      offset = GET(hc, 0);      offset = GET(hc, 0);
# Line 1987  get_othercase_range(unsigned int *cptr, Line 2374  get_othercase_range(unsigned int *cptr,
2374  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2375    
2376  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2377    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2378    
2379  if (c > d) return FALSE;  if (c > d) return FALSE;
2380    
# Line 1996  next = othercase + 1; Line 2383  next = othercase + 1;
2383    
2384  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2385    {    {
2386    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2387    next++;    next++;
2388    }    }
2389    
# Line 2042  if ((options & PCRE_EXTENDED) != 0) Line 2429  if ((options & PCRE_EXTENDED) != 0)
2429    for (;;)    for (;;)
2430      {      {
2431      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2432      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2433        {        {
2434        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2435          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
# Line 2054  if ((options & PCRE_EXTENDED) != 0) Line 2441  if ((options & PCRE_EXTENDED) != 0)
2441  /* 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
2442  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2443    
2444  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2445    {    {
2446    int temperrorcode = 0;    int temperrorcode = 0;
2447    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
# Line 2079  if ((options & PCRE_EXTENDED) != 0) Line 2466  if ((options & PCRE_EXTENDED) != 0)
2466    for (;;)    for (;;)
2467      {      {
2468      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2469      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2470        {        {
2471        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2472          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
# Line 2090  if ((options & PCRE_EXTENDED) != 0) Line 2477  if ((options & PCRE_EXTENDED) != 0)
2477    
2478  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
2479    
2480  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2481    return FALSE;    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2482        return FALSE;
2483    
2484  /* Now compare the next item with the previous opcode. If the previous is a  /* Now compare the next item with the previous opcode. If the previous is a
2485  positive single character match, "item" either contains the character or, if  positive single character match, "item" either contains the character or, if
# Line 2106  if (next >= 0) switch(op_code) Line 2494  if (next >= 0) switch(op_code)
2494    case OP_CHAR:    case OP_CHAR:
2495  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2496    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2497    #else
2498      (void)(utf8_char);  /* Keep compiler happy by referencing function argument */
2499  #endif  #endif
2500    return item != next;    return item != next;
2501    
# Line 2124  if (next >= 0) switch(op_code) Line 2514  if (next >= 0) switch(op_code)
2514      unsigned int othercase;      unsigned int othercase;
2515      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2516  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2517      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
2518  #else  #else
2519      othercase = NOTACHAR;      othercase = NOTACHAR;
2520  #endif  #endif
# Line 2145  if (next >= 0) switch(op_code) Line 2535  if (next >= 0) switch(op_code)
2535      unsigned int othercase;      unsigned int othercase;
2536      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2537  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2538      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
2539  #else  #else
2540      othercase = NOTACHAR;      othercase = NOTACHAR;
2541  #endif  #endif
# Line 2390  BOOL inescq = FALSE; Line 2780  BOOL inescq = FALSE;
2780  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
2781  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
2782  const uschar *tempptr;  const uschar *tempptr;
2783    const uschar *nestptr = NULL;
2784  uschar *previous = NULL;  uschar *previous = NULL;
2785  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
2786  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
# Line 2406  BOOL utf8 = FALSE; Line 2797  BOOL utf8 = FALSE;
2797  uschar *utf8_char = NULL;  uschar *utf8_char = NULL;
2798  #endif  #endif
2799    
2800  #ifdef DEBUG  #ifdef PCRE_DEBUG
2801  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2802  #endif  #endif
2803    
# Line 2459  for (;; ptr++) Line 2850  for (;; ptr++)
2850    /* Get next byte in the pattern */    /* Get next byte in the pattern */
2851    
2852    c = *ptr;    c = *ptr;
2853    
2854      /* If we are at the end of a nested substitution, revert to the outer level
2855      string. Nesting only happens one level deep. */
2856    
2857      if (c == 0 && nestptr != NULL)
2858        {
2859        ptr = nestptr;
2860        nestptr = NULL;
2861        c = *ptr;
2862        }
2863    
2864    /* 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
2865    previous cycle of this loop. */    previous cycle of this loop. */
2866    
2867    if (lengthptr != NULL)    if (lengthptr != NULL)
2868      {      {
2869  #ifdef DEBUG  #ifdef PCRE_DEBUG
2870      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
2871  #endif  #endif
2872      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
2873        {        {
2874        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
2875        goto FAILED;        goto FAILED;
# Line 2517  for (;; ptr++) Line 2918  for (;; ptr++)
2918    /* 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
2919    reference list. */    reference list. */
2920    
2921    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
2922      {      {
2923      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
2924      goto FAILED;      goto FAILED;
# Line 2527  for (;; ptr++) Line 2928  for (;; ptr++)
2928    
2929    if (inescq && c != 0)    if (inescq && c != 0)
2930      {      {
2931      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
2932        {        {
2933        inescq = FALSE;        inescq = FALSE;
2934        ptr++;        ptr++;
# Line 2553  for (;; ptr++) Line 2954  for (;; ptr++)
2954    /* 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
2955    a quantifier. */    a quantifier. */
2956    
2957    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
2958      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
2959        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
2960    
2961    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
2962         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2569  for (;; ptr++) Line 2971  for (;; ptr++)
2971    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
2972      {      {
2973      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
2974      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
2975        {        {
2976        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2977          {          {
# Line 2594  for (;; ptr++) Line 2996  for (;; ptr++)
2996      {      {
2997      /* ===================================================================*/      /* ===================================================================*/
2998      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
2999      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
3000      case ')':      case CHAR_RIGHT_PARENTHESIS:
3001      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
3002      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
3003      *codeptr = code;      *codeptr = code;
# Line 2617  for (;; ptr++) Line 3019  for (;; ptr++)
3019      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
3020      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3021    
3022      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
3023      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3024        {        {
3025        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
# Line 2626  for (;; ptr++) Line 3028  for (;; ptr++)
3028      *code++ = OP_CIRC;      *code++ = OP_CIRC;
3029      break;      break;
3030    
3031      case '$':      case CHAR_DOLLAR_SIGN:
3032      previous = NULL;      previous = NULL;
3033      *code++ = OP_DOLL;      *code++ = OP_DOLL;
3034      break;      break;
# Line 2634  for (;; ptr++) Line 3036  for (;; ptr++)
3036      /* 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
3037      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
3038    
3039      case '.':      case CHAR_DOT:
3040      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3041      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
3042      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3043      previous = code;      previous = code;
3044      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
3045      break;      break;
3046    
3047    
# Line 2654  for (;; ptr++) Line 3056  for (;; ptr++)
3056      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,
3057      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
3058      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.
     */  
3059    
3060      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
3061        default (Perl) mode, it is treated as a data character. */
3062    
3063        case CHAR_RIGHT_SQUARE_BRACKET:
3064        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3065          {
3066          *errorcodeptr = ERR64;
3067          goto FAILED;
3068          }
3069        goto NORMAL_CHAR;
3070    
3071        case CHAR_LEFT_SQUARE_BRACKET:
3072      previous = code;      previous = code;
3073    
3074      /* 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
3075      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. */
3076    
3077      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3078             ptr[1] == CHAR_EQUALS_SIGN) &&
3079          check_posix_syntax(ptr, &tempptr))          check_posix_syntax(ptr, &tempptr))
3080        {        {
3081        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
3082        goto FAILED;        goto FAILED;
3083        }        }
3084    
# Line 2677  for (;; ptr++) Line 3090  for (;; ptr++)
3090      for (;;)      for (;;)
3091        {        {
3092        c = *(++ptr);        c = *(++ptr);
3093        if (c == '\\')        if (c == CHAR_BACKSLASH)
3094          {          {
3095          if (ptr[1] == 'E') ptr++;          if (ptr[1] == CHAR_E)
3096            else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;            ptr++;
3097              else break;          else if (strncmp((const char *)ptr+1,
3098                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
3099              ptr += 3;
3100            else
3101              break;
3102          }          }
3103        else if (!negate_class && c == '^')        else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
3104          negate_class = TRUE;          negate_class = TRUE;
3105        else break;        else break;
3106        }        }
3107    
3108        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
3109        an initial ']' is taken as a data character -- the code below handles
3110        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3111        [^] must match any character, so generate OP_ALLANY. */
3112    
3113        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3114            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3115          {
3116          *code++ = negate_class? OP_ALLANY : OP_FAIL;
3117          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3118          zerofirstbyte = firstbyte;
3119          break;
3120          }
3121    
3122      /* 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
3123      negation flag at the end, so that support for characters > 255 works      negation flag at the end, so that support for characters > 255 works
3124      correctly (they are all included in the class). */      correctly (they are all included in the class). */
# Line 2727  for (;; ptr++) Line 3158  for (;; ptr++)
3158          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3159          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3160          }          }
3161    
3162        /* In the pre-compile phase, accumulate the length of any UTF-8 extra        /* In the pre-compile phase, accumulate the length of any UTF-8 extra
3163        data and reset the pointer. This is so that very large classes that        data and reset the pointer. This is so that very large classes that
3164        contain a zillion UTF-8 characters no longer overwrite the work space        contain a zillion UTF-8 characters no longer overwrite the work space
# Line 2745  for (;; ptr++) Line 3176  for (;; ptr++)
3176    
3177        if (inescq)        if (inescq)
3178          {          {
3179          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3180            {            {
3181            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
3182            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2760  for (;; ptr++) Line 3191  for (;; ptr++)
3191        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3192        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3193    
3194        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3195            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3196            check_posix_syntax(ptr, &tempptr))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3197          {          {
3198          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3199          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3200          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3201          uschar pbits[32];          uschar pbits[32];
3202    
3203          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3204            {            {
3205            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3206            goto FAILED;            goto FAILED;
3207            }            }
3208    
3209          ptr += 2;          ptr += 2;
3210          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3211            {            {
3212            local_negate = TRUE;            local_negate = TRUE;
3213            should_flip_negation = TRUE;  /* Note negative special */            should_flip_negation = TRUE;  /* Note negative special */
# Line 2796  for (;; ptr++) Line 3227  for (;; ptr++)
3227    
3228          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3229            posix_class = 0;            posix_class = 0;
3230    
3231          /* 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
3232          because we may be adding and subtracting from it, and we don't want to          different escape sequences that use Unicode properties. */
3233          subtract bits that may be in the main map already. At the end we or the  
3234          result into the bit map that is being built. */  #ifdef SUPPORT_UCP
3235            if ((options & PCRE_UCP) != 0)
3236              {
3237              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3238              if (posix_substitutes[pc] != NULL)
3239                {
3240                nestptr = tempptr + 1;
3241                ptr = posix_substitutes[pc] - 1;
3242                continue;
3243                }
3244              }
3245    #endif
3246            /* In the non-UCP case, we build the bit map for the POSIX class in a
3247            chunk of local store because we may be adding and subtracting from it,
3248            and we don't want to subtract bits that may be in the main map already.
3249            At the end we or the result into the bit map that is being built. */
3250    
3251          posix_class *= 3;          posix_class *= 3;
3252    
# Line 2844  for (;; ptr++) Line 3290  for (;; ptr++)
3290    
3291        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3292        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
3293        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
3294        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
3295        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
3296        character in them, so set class_charcount bigger than one. */        are either treated as literal characters (by default), or are faulted if
3297          PCRE_EXTRA is set. */
3298    
3299        if (c == '\\')        if (c == CHAR_BACKSLASH)
3300          {          {
3301          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3302          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3303    
3304          if (-c == ESC_b) c = '\b';       /* \b is backspace in a class */          if (-c == ESC_b) c = CHAR_BS;    /* \b is backspace in a class */
         else if (-c == ESC_X) c = 'X';   /* \X is literal X in a class */  
         else if (-c == ESC_R) c = 'R';   /* \R is literal R in a class */  
3305          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3306            {            {
3307            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3308              {              {
3309              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3310              }              }
# Line 2873  for (;; ptr++) Line 3318  for (;; ptr++)
3318            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
3319            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3320    
3321            /* Save time by not doing this in the pre-compile phase. */            switch (-c)
   
           if (lengthptr == NULL) switch (-c)  
3322              {              {
3323    #ifdef SUPPORT_UCP
3324                case ESC_du:     /* These are the values given for \d etc */
3325                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3326                case ESC_wu:     /* escape sequence with an appropriate \p */
3327                case ESC_WU:     /* or \P to test Unicode properties instead */
3328                case ESC_su:     /* of the default ASCII testing. */
3329                case ESC_SU:
3330                nestptr = ptr;
3331                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3332                class_charcount -= 2;                /* Undo! */
3333                continue;
3334    #endif
3335              case ESC_d:              case ESC_d:
3336              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3337              continue;              continue;
# Line 2906  for (;; ptr++) Line 3361  for (;; ptr++)
3361              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3362              continue;              continue;
3363    
3364              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)  
             {  
3365              SETBIT(classbits, 0x09); /* VT */              SETBIT(classbits, 0x09); /* VT */
3366              SETBIT(classbits, 0x20); /* SPACE */              SETBIT(classbits, 0x20); /* SPACE */
3367              SETBIT(classbits, 0xa0); /* NSBP */              SETBIT(classbits, 0xa0); /* NSBP */
# Line 2943  for (;; ptr++) Line 3385  for (;; ptr++)
3385                }                }
3386  #endif  #endif
3387              continue;              continue;
             }  
3388    
3389            if (-c == ESC_H)              case ESC_H:
             {  
3390              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3391                {                {
3392                int x = 0xff;                int x = 0xff;
# Line 2988  for (;; ptr++) Line 3428  for (;; ptr++)
3428                }                }
3429  #endif  #endif
3430              continue;              continue;
             }  
3431    
3432            if (-c == ESC_v)              case ESC_v:
             {  
3433              SETBIT(classbits, 0x0a); /* LF */              SETBIT(classbits, 0x0a); /* LF */
3434              SETBIT(classbits, 0x0b); /* VT */              SETBIT(classbits, 0x0b); /* VT */
3435              SETBIT(classbits, 0x0c); /* FF */              SETBIT(classbits, 0x0c); /* FF */
# Line 3007  for (;; ptr++) Line 3445  for (;; ptr++)
3445                }                }
3446  #endif  #endif
3447              continue;              continue;
             }  
3448    
3449            if (-c == ESC_V)              case ESC_V:
             {  
3450              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3451                {                {
3452                int x = 0xff;                int x = 0xff;
# Line 3040  for (;; ptr++) Line 3476  for (;; ptr++)
3476                }                }
3477  #endif  #endif
3478              continue;              continue;
             }  
   
           /* We need to deal with \P and \p in both phases. */  
3479    
3480  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3481            if (-c == ESC_p || -c == ESC_P)              case ESC_p:
3482              {              case ESC_P:
3483              BOOL negated;                {
3484              int pdata;                BOOL negated;
3485              int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                int pdata;
3486              if (ptype < 0) goto FAILED;                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3487              class_utf8 = TRUE;                if (ptype < 0) goto FAILED;
3488              *class_utf8data++ = ((-c == ESC_p) != negated)?                class_utf8 = TRUE;
3489                XCL_PROP : XCL_NOTPROP;                *class_utf8data++ = ((-c == ESC_p) != negated)?
3490              *class_utf8data++ = ptype;                  XCL_PROP : XCL_NOTPROP;
3491              *class_utf8data++ = pdata;                *class_utf8data++ = ptype;
3492              class_charcount -= 2;   /* Not a < 256 character */                *class_utf8data++ = pdata;
3493              continue;                class_charcount -= 2;   /* Not a < 256 character */
3494              }                continue;
3495                  }
3496  #endif  #endif
3497            /* Unrecognized escapes are faulted if PCRE is running in its              /* Unrecognized escapes are faulted if PCRE is running in its
3498            strict mode. By default, for compatibility with Perl, they are              strict mode. By default, for compatibility with Perl, they are
3499            treated as literals. */              treated as literals. */
3500    
3501            if ((options & PCRE_EXTRA) != 0)              default:
3502              {              if ((options & PCRE_EXTRA) != 0)
3503              *errorcodeptr = ERR7;                {
3504              goto FAILED;                *errorcodeptr = ERR7;
3505                  goto FAILED;
3506                  }
3507                class_charcount -= 2;  /* Undo the default count from above */
3508                c = *ptr;              /* Get the final character and fall through */
3509                break;
3510              }              }
   
           class_charcount -= 2;  /* Undo the default count from above */  
           c = *ptr;              /* Get the final character and fall through */  
3511            }            }
3512    
3513          /* Fall through if we have a single character (c >= 0). This may be          /* Fall through if we have a single character (c >= 0). This may be
# Line 3085  for (;; ptr++) Line 3521  for (;; ptr++)
3521        entirely. The code for handling \Q and \E is messy. */        entirely. The code for handling \Q and \E is messy. */
3522    
3523        CHECK_RANGE:        CHECK_RANGE:
3524        while (ptr[1] == '\\' && ptr[2] == 'E')        while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3525          {          {
3526          inescq = FALSE;          inescq = FALSE;
3527          ptr += 2;          ptr += 2;
# Line 3095  for (;; ptr++) Line 3531  for (;; ptr++)
3531    
3532        /* Remember \r or \n */        /* Remember \r or \n */
3533    
3534        if (c == '\r' || c == '\n') cd->external_flags |= PCRE_HASCRORLF;        if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3535    
3536        /* Check for range */        /* Check for range */
3537    
3538        if (!inescq && ptr[1] == '-')        if (!inescq && ptr[1] == CHAR_MINUS)
3539          {          {
3540          int d;          int d;
3541          ptr += 2;          ptr += 2;
3542          while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3543    
3544          /* 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
3545          mode. */          mode. */
3546    
3547          while (*ptr == '\\' && ptr[1] == 'Q')          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3548            {            {
3549            ptr += 2;            ptr += 2;
3550            if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }            if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3551                { ptr += 2; continue; }
3552            inescq = TRUE;            inescq = TRUE;
3553            break;            break;
3554            }            }
3555    
3556          if (*ptr == 0 || (!inescq && *ptr == ']'))          if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3557            {            {
3558            ptr = oldptr;            ptr = oldptr;
3559            goto LONE_SINGLE_CHARACTER;            goto LONE_SINGLE_CHARACTER;
# Line 3135  for (;; ptr++) Line 3572  for (;; ptr++)
3572          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
3573          in such circumstances. */          in such circumstances. */
3574    
3575          if (!inescq && d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3576            {            {
3577            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3578            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3579    
3580            /* \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 */  
3581    
3582            if (d < 0)            if (d < 0)
3583              {              {
3584              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  
3585                {                {
3586                ptr = oldptr;                ptr = oldptr;
3587                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 3168  for (;; ptr++) Line 3602  for (;; ptr++)
3602    
3603          /* Remember \r or \n */          /* Remember \r or \n */
3604    
3605          if (d == '\r' || d == '\n') cd->external_flags |= PCRE_HASCRORLF;          if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3606    
3607          /* 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
3608          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
# Line 3288  for (;; ptr++) Line 3722  for (;; ptr++)
3722          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3723            {            {
3724            unsigned int othercase;            unsigned int othercase;
3725            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)            if ((othercase = UCD_OTHERCASE(c)) != c)
3726              {              {
3727              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3728              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 3313  for (;; ptr++) Line 3747  for (;; ptr++)
3747          }          }
3748        }        }
3749    
3750      /* 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.
3751        If we are at the end of an internal nested string, revert to the outer
3752        string. */
3753    
3754        while (((c = *(++ptr)) != 0 ||
3755               (nestptr != NULL &&
3756                 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != 0)) &&
3757               (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
3758    
3759      while ((c = *(++ptr)) != 0 && (c != ']' || inescq));      /* Check for missing terminating ']' */
3760    
3761      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)
3762        {        {
3763        *errorcodeptr = ERR6;        *errorcodeptr = ERR6;
3764        goto FAILED;        goto FAILED;
3765        }        }
3766    
   
 /* 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  
   
   
3767      /* 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
3768      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
3769      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 3358  we set the flag only if there is a liter Line 3780  we set the flag only if there is a liter
3780      can cause firstbyte to be set. Otherwise, there can be no first char if      can cause firstbyte to be set. Otherwise, there can be no first char if
3781      this item is first, whatever repeat count may follow. In the case of      this item is first, whatever repeat count may follow. In the case of
3782      reqbyte, save the previous value for reinstating. */      reqbyte, save the previous value for reinstating. */
3783    
3784  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3785      if (class_charcount == 1 && !class_utf8 &&      if (class_charcount == 1 && !class_utf8 &&
3786        (!utf8 || !negate_class || class_lastchar < 128))        (!utf8 || !negate_class || class_lastchar < 128))
# Line 3405  we set the flag only if there is a liter Line 3827  we set the flag only if there is a liter
3827    
3828      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
3829      extended class, with its own opcode, unless there was a negated special      extended class, with its own opcode, unless there was a negated special
3830      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
3831      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
3832      (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
3833      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
3834        actual compiled code. */
3835    
3836  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3837      if (class_utf8 && !should_flip_negation)      if (class_utf8 && (!should_flip_negation || (options & PCRE_UCP) != 0))
3838        {        {
3839        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
3840        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 3437  we set the flag only if there is a liter Line 3860  we set the flag only if there is a liter
3860        }        }
3861  #endif  #endif
3862    
3863      /* 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
3864      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
3865      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
3866      map into the code vector, negating it if necessary. */      (non-UCP) in the class. Then copy the 32-byte map into the code vector,
3867        negating it if necessary. */
3868    
3869      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
3870      if (negate_class)      if (negate_class)
# Line 3460  we set the flag only if there is a liter Line 3884  we set the flag only if there is a liter
3884      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
3885      has been tested above. */      has been tested above. */
3886    
3887      case '{':      case CHAR_LEFT_CURLY_BRACKET:
3888      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
3889      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
3890      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
3891      goto REPEAT;      goto REPEAT;
3892    
3893      case '*':      case CHAR_ASTERISK:
3894      repeat_min = 0;      repeat_min = 0;
3895      repeat_max = -1;      repeat_max = -1;
3896      goto REPEAT;      goto REPEAT;
3897    
3898      case '+':      case CHAR_PLUS:
3899      repeat_min = 1;      repeat_min = 1;
3900      repeat_max = -1;      repeat_max = -1;
3901      goto REPEAT;      goto REPEAT;
3902    
3903      case '?':      case CHAR_QUESTION_MARK:
3904      repeat_min = 0;      repeat_min = 0;
3905      repeat_max = 1;      repeat_max = 1;
3906    
# Line 3511  we set the flag only if there is a liter Line 3935  we set the flag only if there is a liter
3935      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
3936      repeat type to the non-default. */      repeat type to the non-default. */
3937    
3938      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
3939        {        {
3940        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
3941        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
3942        ptr++;        ptr++;
3943        }        }
3944      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
3945        {        {
3946        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
3947        ptr++;        ptr++;
# Line 3632  we set the flag only if there is a liter Line 4056  we set the flag only if there is a liter
4056    
4057        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
4058    
4059          /*--------------------------------------------------------------------*/
4060          /* This code is obsolete from release 8.00; the restriction was finally
4061          removed: */
4062    
4063        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4064        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4065    
4066        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4067          /*--------------------------------------------------------------------*/
4068    
4069        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
4070    
# Line 3782  we set the flag only if there is a liter Line 4211  we set the flag only if there is a liter
4211          goto END_REPEAT;          goto END_REPEAT;
4212          }          }
4213    
4214          /*--------------------------------------------------------------------*/
4215          /* This code is obsolete from release 8.00; the restriction was finally
4216          removed: */
4217    
4218        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4219        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4220    
4221        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4222          /*--------------------------------------------------------------------*/
4223    
4224        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4225          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 3851  we set the flag only if there is a liter Line 4285  we set the flag only if there is a liter
4285          **   code = previous;          **   code = previous;
4286          **   goto END_REPEAT;          **   goto END_REPEAT;
4287          **   }          **   }
4288    
4289          However, that fails when a group is referenced as a subroutine from          However, that fails when a group is referenced as a subroutine from
4290          elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it          elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it
4291          so that it is skipped on execution. As we don't have a list of which          so that it is skipped on execution. As we don't have a list of which
4292          groups are referenced, we cannot do this selectively.          groups are referenced, we cannot do this selectively.
4293    
4294          If the maximum is 1 or unlimited, we just have to stick in the BRAZERO          If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4295          and do no more at this point. However, we do need to adjust any          and do no more at this point. However, we do need to adjust any
# Line 3874  we set the flag only if there is a liter Line 4308  we set the flag only if there is a liter
4308              {              {
4309              *previous++ = OP_SKIPZERO;              *previous++ = OP_SKIPZERO;
4310              goto END_REPEAT;              goto END_REPEAT;
4311              }              }
4312            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4313            }            }
4314    
# Line 3920  we set the flag only if there is a liter Line 4354  we set the flag only if there is a liter
4354            {            {
4355            /* 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
4356            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
4357            potential integer overflow. */            potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
4358              integer type when available, otherwise double. */
4359    
4360            if (lengthptr != NULL)            if (lengthptr != NULL)
4361              {              {
4362              int delta = (repeat_min - 1)*length_prevgroup;              int delta = (repeat_min - 1)*length_prevgroup;
4363              if ((double)(repeat_min - 1)*(double)length_prevgroup >              if ((INT64_OR_DOUBLE)(repeat_min - 1)*
4364                                                              (double)INT_MAX ||                    (INT64_OR_DOUBLE)length_prevgroup >
4365                        (INT64_OR_DOUBLE)INT_MAX ||
4366                  OFLOW_MAX - *lengthptr < delta)                  OFLOW_MAX - *lengthptr < delta)
4367                {                {
4368                *errorcodeptr = ERR20;                *errorcodeptr = ERR20;
# Line 3972  we set the flag only if there is a liter Line 4408  we set the flag only if there is a liter
4408          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
4409          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
4410          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
4411          paranoid checks to avoid integer overflow. */          paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
4412            a 64-bit integer type when available, otherwise double. */
4413    
4414          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4415            {            {
4416            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4417                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4418            if ((double)repeat_max *            if ((INT64_OR_DOUBLE)repeat_max *
4419                  (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)                  (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4420                    > (double)INT_MAX ||                    > (INT64_OR_DOUBLE)INT_MAX ||
4421                OFLOW_MAX - *lengthptr < delta)                OFLOW_MAX - *lengthptr < delta)
4422              {              {
4423              *errorcodeptr = ERR20;              *errorcodeptr = ERR20;
# Line 4057  we set the flag only if there is a liter Line 4494  we set the flag only if there is a liter
4494            uschar *scode = bracode;            uschar *scode = bracode;
4495            do            do
4496              {              {
4497              if (could_be_empty_branch(scode, ketcode, utf8))              if (could_be_empty_branch(scode, ketcode, utf8, cd))
4498                {                {
4499                *bracode += OP_SBRA - OP_BRA;                *bracode += OP_SBRA - OP_BRA;
4500                break;                break;
# Line 4069  we set the flag only if there is a liter Line 4506  we set the flag only if there is a liter
4506          }          }
4507        }        }
4508    
4509        /* If previous is OP_FAIL, it was generated by an empty class [] in
4510        JavaScript mode. The other ways in which OP_FAIL can be generated, that is
4511        by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
4512        error above. We can just ignore the repeat in JS case. */
4513    
4514        else if (*previous == OP_FAIL) goto END_REPEAT;
4515    
4516      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
4517    
4518      else      else
# Line 4093  we set the flag only if there is a liter Line 4537  we set the flag only if there is a liter
4537      if (possessive_quantifier)      if (possessive_quantifier)
4538        {        {
4539        int len;        int len;
4540        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||  
4541            *tempcode == OP_NOTEXACT)        if (*tempcode == OP_TYPEEXACT)
4542          tempcode += _pcre_OP_lengths[*tempcode] +          tempcode += _pcre_OP_lengths[*tempcode] +
4543            ((*tempcode == OP_TYPEEXACT &&            ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
4544               (tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP))? 2:0);  
4545          else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
4546            {
4547            tempcode += _pcre_OP_lengths[*tempcode];
4548    #ifdef SUPPORT_UTF8
4549            if (utf8 && tempcode[-1] >= 0xc0)
4550              tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];
4551    #endif
4552            }
4553    
4554        len = code - tempcode;        len = code - tempcode;
4555        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4556          {          {
# Line 4116  we set the flag only if there is a liter Line 4569  we set the flag only if there is a liter
4569          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4570          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4571    
4572            /* Because we are moving code along, we must ensure that any
4573            pending recursive references are updated. */
4574    
4575          default:          default:
4576            *code = OP_END;
4577            adjust_recurse(tempcode, 1 + LINK_SIZE, utf8, cd, save_hwm);
4578          memmove(tempcode + 1+LINK_SIZE, tempcode, len);          memmove(tempcode + 1+LINK_SIZE, tempcode, len);
4579          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
4580          len += 1 + LINK_SIZE;          len += 1 + LINK_SIZE;
# Line 4143  we set the flag only if there is a liter Line 4601  we set the flag only if there is a liter
4601      lookbehind or option setting or condition or all the other extended      lookbehind or option setting or condition or all the other extended
4602      parenthesis forms.  */      parenthesis forms.  */
4603    
4604      case '(':      case CHAR_LEFT_PARENTHESIS:
4605      newoptions = options;      newoptions = options;
4606      skipbytes = 0;      skipbytes = 0;
4607      bravalue = OP_CBRA;      bravalue = OP_CBRA;
4608      save_hwm = cd->hwm;      save_hwm = cd->hwm;
4609      reset_bracount = FALSE;      reset_bracount = FALSE;
4610    
4611      /* First deal with various "verbs" that can be introduced by '*'. */      /* First deal with various "verbs" that can be introduced by '*'. */
4612    
4613      if (*(++ptr) == '*' && (cd->ctypes[ptr[1]] & ctype_letter) != 0)      if (*(++ptr) == CHAR_ASTERISK &&
4614             ((cd->ctypes[ptr[1]] & ctype_letter) != 0 || ptr[1] == ':'))
4615        {        {
4616        int i, namelen;        int i, namelen;
4617          int arglen = 0;
4618        const char *vn = verbnames;        const char *vn = verbnames;
4619        const uschar *name = ++ptr;        const uschar *name = ptr + 1;
4620          const uschar *arg = NULL;
4621        previous = NULL;        previous = NULL;
4622        while ((cd->ctypes[*++ptr] & ctype_letter) != 0);        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
4623        if (*ptr == ':')        namelen = ptr - name;
4624    
4625          if (*ptr == CHAR_COLON)
4626          {          {
4627          *errorcodeptr = ERR59;   /* Not supported */          arg = ++ptr;
4628          goto FAILED;          while ((cd->ctypes[*ptr] & (ctype_letter|ctype_digit)) != 0
4629              || *ptr == '_') ptr++;
4630            arglen = ptr - arg;
4631          }          }
4632        if (*ptr != ')')  
4633          if (*ptr != CHAR_RIGHT_PARENTHESIS)
4634          {          {
4635          *errorcodeptr = ERR60;          *errorcodeptr = ERR60;
4636          goto FAILED;          goto FAILED;
4637          }          }
4638        namelen = ptr - name;  
4639          /* Scan the table of verb names */
4640    
4641        for (i = 0; i < verbcount; i++)        for (i = 0; i < verbcount; i++)
4642          {          {
4643          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
4644              strncmp((char *)name, vn, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
4645            {            {
4646            *code = verbs[i].op;            /* Check for open captures before ACCEPT */
4647            if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;  
4648            break;            if (verbs[i].op == OP_ACCEPT)
4649                {
4650                open_capitem *oc;
4651                cd->had_accept = TRUE;
4652                for (oc = cd->open_caps; oc != NULL; oc = oc->next)
4653                  {
4654                  *code++ = OP_CLOSE;
4655                  PUT2INC(code, 0, oc->number);
4656                  }
4657                }
4658    
4659              /* Handle the cases with/without an argument */
4660    
4661              if (arglen == 0)
4662                {
4663                if (verbs[i].op < 0)   /* Argument is mandatory */
4664                  {
4665                  *errorcodeptr = ERR66;
4666                  goto FAILED;
4667                  }
4668                *code++ = verbs[i].op;
4669                }
4670    
4671              else
4672                {
4673                if (verbs[i].op_arg < 0)   /* Argument is forbidden */
4674                  {
4675                  *errorcodeptr = ERR59;
4676                  goto FAILED;
4677                  }
4678                *code++ = verbs[i].op_arg;
4679                *code++ = arglen;
4680                memcpy(code, arg, arglen);
4681                code += arglen;
4682                *code++ = 0;
4683                }
4684    
4685              break;  /* Found verb, exit loop */
4686            }            }
4687    
4688          vn += verbs[i].len + 1;          vn += verbs[i].len + 1;
4689          }          }
4690        if (i < verbcount) continue;  
4691        *errorcodeptr = ERR60;        if (i < verbcount) continue;    /* Successfully handled a verb */
4692          *errorcodeptr = ERR60;          /* Verb not recognized */
4693        goto FAILED;        goto FAILED;
4694        }        }
4695    
4696      /* Deal with the extended parentheses; all are introduced by '?', and the      /* Deal with the extended parentheses; all are introduced by '?', and the
4697      appearance of any of them means that this is not a capturing group. */      appearance of any of them means that this is not a capturing group. */
4698    
4699      else if (*ptr == '?')      else if (*ptr == CHAR_QUESTION_MARK)
4700        {        {
4701        int i, set, unset, namelen;        int i, set, unset, namelen;
4702        int *optset;        int *optset;
# Line 4198  we set the flag only if there is a liter Line 4705  we set the flag only if there is a liter
4705    
4706        switch (*(++ptr))        switch (*(++ptr))
4707          {          {
4708          case '#':                 /* Comment; skip to ket */          case CHAR_NUMBER_SIGN:                 /* Comment; skip to ket */
4709          ptr++;          ptr++;
4710          while (*ptr != 0 && *ptr != ')') ptr++;          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
4711          if (*ptr == 0)          if (*ptr == 0)
4712            {            {
4713            *errorcodeptr = ERR18;            *errorcodeptr = ERR18;
# Line 4210  we set the flag only if there is a liter Line 4717  we set the flag only if there is a liter
4717    
4718    
4719          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4720          case '|':                 /* Reset capture count for each branch */          case CHAR_VERTICAL_LINE:  /* Reset capture count for each branch */
4721          reset_bracount = TRUE;          reset_bracount = TRUE;
4722          /* Fall through */          /* Fall through */
4723    
4724          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4725          case ':':                 /* Non-capturing bracket */          case CHAR_COLON:          /* Non-capturing bracket */
4726          bravalue = OP_BRA;          bravalue = OP_BRA;
4727          ptr++;          ptr++;
4728          break;          break;
4729    
4730    
4731          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4732          case '(':          case CHAR_LEFT_PARENTHESIS:
4733          bravalue = OP_COND;       /* Conditional group */          bravalue = OP_COND;       /* Conditional group */
4734    
4735          /* A condition can be an assertion, a number (referring to a numbered          /* A condition can be an assertion, a number (referring to a numbered
# Line 4242  we set the flag only if there is a liter Line 4749  we set the flag only if there is a liter
4749          the switch. This will take control down to where bracketed groups,          the switch. This will take control down to where bracketed groups,
4750          including assertions, are processed. */          including assertions, are processed. */
4751    
4752          if (ptr[1] == '?' && (ptr[2] == '=' || ptr[2] == '!' || ptr[2] == '<'))          if (ptr[1] == CHAR_QUESTION_MARK && (ptr[2] == CHAR_EQUALS_SIGN ||
4753                ptr[2] == CHAR_EXCLAMATION_MARK || ptr[2] == CHAR_LESS_THAN_SIGN))
4754            break;            break;
4755    
4756          /* Most other conditions use OP_CREF (a couple change to OP_RREF          /* Most other conditions use OP_CREF (a couple change to OP_RREF
# Line 4254  we set the flag only if there is a liter Line 4762  we set the flag only if there is a liter
4762    
4763          /* Check for a test for recursion in a named group. */          /* Check for a test for recursion in a named group. */
4764    
4765          if (ptr[1] == 'R' && ptr[2] == '&')          if (ptr[1] == CHAR_R && ptr[2] == CHAR_AMPERSAND)
4766            {            {
4767            terminator = -1;            terminator = -1;
4768            ptr += 2;            ptr += 2;
# Line 4264  we set the flag only if there is a liter Line 4772  we set the flag only if there is a liter
4772          /* Check for a test for a named group's having been set, using the Perl          /* Check for a test for a named group's having been set, using the Perl
4773          syntax (?(<name>) or (?('name') */          syntax (?(<name>) or (?('name') */
4774    
4775          else if (ptr[1] == '<')          else if (ptr[1] == CHAR_LESS_THAN_SIGN)
4776            {            {
4777            terminator = '>';            terminator = CHAR_GREATER_THAN_SIGN;
4778            ptr++;            ptr++;
4779            }            }
4780          else if (ptr[1] == '\'')          else if (ptr[1] == CHAR_APOSTROPHE)
4781            {            {
4782            terminator = '\'';            terminator = CHAR_APOSTROPHE;
4783            ptr++;            ptr++;
4784            }            }
4785          else          else
4786            {            {
4787            terminator = 0;            terminator = 0;
4788            if (ptr[1] == '-' || ptr[1] == '+') refsign = *(++ptr);            if (ptr[1] == CHAR_MINUS || ptr[1] == CHAR_PLUS) refsign = *(++ptr);
4789            }            }
4790    
4791          /* We now expect to read a name; any thing else is an error */          /* We now expect to read a name; any thing else is an error */
# Line 4297  we set the flag only if there is a liter Line 4805  we set the flag only if there is a liter
4805            {            {
4806            if (recno >= 0)            if (recno >= 0)
4807              recno = ((digitab[*ptr] & ctype_digit) != 0)?              recno = ((digitab[*ptr] & ctype_digit) != 0)?
4808                recno * 10 + *ptr - '0' : -1;                recno * 10 + *ptr - CHAR_0 : -1;
4809            ptr++;            ptr++;
4810            }            }
4811          namelen = ptr - name;          namelen = ptr - name;
4812    
4813          if ((terminator > 0 && *ptr++ != terminator) || *ptr++ != ')')          if ((terminator > 0 && *ptr++ != terminator) ||
4814                *ptr++ != CHAR_RIGHT_PARENTHESIS)
4815            {            {
4816            ptr--;      /* Error offset */            ptr--;      /* Error offset */
4817            *errorcodeptr = ERR26;            *errorcodeptr = ERR26;
# Line 4324  we set the flag only if there is a liter Line 4833  we set the flag only if there is a liter
4833              *errorcodeptr = ERR58;              *errorcodeptr = ERR58;
4834              goto FAILED;              goto FAILED;
4835              }              }
4836            recno = (refsign == '-')?            recno = (refsign == CHAR_MINUS)?
4837              cd->bracount - recno + 1 : recno +cd->bracount;              cd->bracount - recno + 1 : recno +cd->bracount;
4838            if (recno <= 0 || recno > cd->final_bracount)            if (recno <= 0 || recno > cd->final_bracount)
4839              {              {
# Line 4336  we set the flag only if there is a liter Line 4845  we set the flag only if there is a liter
4845            }            }
4846    
4847          /* Otherwise (did not start with "+" or "-"), start by looking for the          /* Otherwise (did not start with "+" or "-"), start by looking for the
4848          name. */          name. If we find a name, add one to the opcode to change OP_CREF or
4849            OP_RREF into OP_NCREF or OP_NRREF. These behave exactly the same,
4850            except they record that the reference was originally to a name. The
4851            information is used to check duplicate names. */
4852    
4853          slot = cd->name_table;          slot = cd->name_table;
4854          for (i = 0; i < cd->names_found; i++)          for (i = 0; i < cd->names_found; i++)
# Line 4351  we set the flag only if there is a liter Line 4863  we set the flag only if there is a liter
4863            {            {
4864            recno = GET2(slot, 0);            recno = GET2(slot, 0);
4865            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4866              code[1+LINK_SIZE]++;
4867            }            }
4868    
4869          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
4870    
4871          else if ((i = find_parens(ptr, cd->bracount, name, namelen,          else if ((i = find_parens(cd, name, namelen,
4872                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0)) > 0)
4873            {            {
4874            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
4875              code[1+LINK_SIZE]++;
4876            }            }
4877    
4878          /* If terminator == 0 it means that the name followed directly after          /* If terminator == 0 it means that the name followed directly after
# Line 4376  we set the flag only if there is a liter Line 4890  we set the flag only if there is a liter
4890          /* Check for (?(R) for recursion. Allow digits after R to specify a          /* Check for (?(R) for recursion. Allow digits after R to specify a
4891          specific group number. */          specific group number. */
4892    
4893          else if (*name == 'R')          else if (*name == CHAR_R)
4894            {            {
4895            recno = 0;            recno = 0;
4896            for (i = 1; i < namelen; i++)            for (i = 1; i < namelen; i++)
# Line 4386  we set the flag only if there is a liter Line 4900  we set the flag only if there is a liter
4900                *errorcodeptr = ERR15;                *errorcodeptr = ERR15;
4901                goto FAILED;                goto FAILED;
4902                }                }
4903              recno = recno * 10 + name[i] - '0';              recno = recno * 10 + name[i] - CHAR_0;
4904              }              }
4905            if (recno == 0) recno = RREF_ANY;            if (recno == 0) recno = RREF_ANY;
4906            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */
# Line 4396  we set the flag only if there is a liter Line 4910  we set the flag only if there is a liter
4910          /* Similarly, check for the (?(DEFINE) "condition", which is always          /* Similarly, check for the (?(DEFINE) "condition", which is always
4911          false. */          false. */
4912    
4913          else if (namelen == 6 && strncmp((char *)name, "DEFINE", 6) == 0)          else if (namelen == 6 && strncmp((char *)name, STRING_DEFINE, 6) == 0)
4914            {            {
4915            code[1+LINK_SIZE] = OP_DEF;            code[1+LINK_SIZE] = OP_DEF;
4916            skipbytes = 1;            skipbytes = 1;
# Line 4421  we set the flag only if there is a liter Line 4935  we set the flag only if there is a liter
4935    
4936    
4937          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4938          case '=':                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
4939          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
4940          ptr++;          ptr++;
4941          break;          break;
4942    
4943    
4944          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4945          case '!':                 /* Negative lookahead */          case CHAR_EXCLAMATION_MARK:            /* Negative lookahead */
4946          ptr++;          ptr++;
4947          if (*ptr == ')')          /* Optimize (?!) */          if (*ptr == CHAR_RIGHT_PARENTHESIS)    /* Optimize (?!) */
4948            {            {
4949            *code++ = OP_FAIL;            *code++ = OP_FAIL;
4950            previous = NULL;            previous = NULL;
# Line 4441  we set the flag only if there is a liter Line 4955  we set the flag only if there is a liter
4955    
4956    
4957          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4958          case '<':                 /* Lookbehind or named define */          case CHAR_LESS_THAN_SIGN:              /* Lookbehind or named define */
4959          switch (ptr[1])          switch (ptr[1])
4960            {            {
4961            case '=':               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
4962            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
4963            ptr += 2;            ptr += 2;
4964            break;            break;
4965    
4966            case '!':               /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
4967            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
4968            ptr += 2;            ptr += 2;
4969            break;            break;
# Line 4464  we set the flag only if there is a liter Line 4978  we set the flag only if there is a liter
4978    
4979    
4980          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4981          case '>':                 /* One-time brackets */          case CHAR_GREATER_THAN_SIGN:           /* One-time brackets */
4982          bravalue = OP_ONCE;          bravalue = OP_ONCE;
4983          ptr++;          ptr++;
4984          break;          break;
4985    
4986    
4987          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4988          case 'C':                 /* Callout - may be followed by digits; */          case CHAR_C:                 /* Callout - may be followed by digits; */
4989          previous_callout = code;  /* Save for later completion */          previous_callout = code;  /* Save for later completion */
4990          after_manual_callout = 1; /* Skip one item before completing */          after_manual_callout = 1; /* Skip one item before completing */
4991          *code++ = OP_CALLOUT;          *code++ = OP_CALLOUT;
4992            {            {
4993            int n = 0;            int n = 0;
4994            while ((digitab[*(++ptr)] & ctype_digit) != 0)            while ((digitab[*(++ptr)] & ctype_digit) != 0)
4995              n = n * 10 + *ptr - '0';              n = n * 10 + *ptr - CHAR_0;
4996            if (*ptr != ')')            if (*ptr != CHAR_RIGHT_PARENTHESIS)
4997              {              {
4998              *errorcodeptr = ERR39;              *errorcodeptr = ERR39;
4999              goto FAILED;              goto FAILED;
# Line 4499  we set the flag only if there is a liter Line 5013  we set the flag only if there is a liter
5013    
5014    
5015          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5016          case 'P':                 /* Python-style named subpattern handling */          case CHAR_P:              /* Python-style named subpattern handling */
5017          if (*(++ptr) == '=' || *ptr == '>')  /* Reference or recursion */          if (*(++ptr) == CHAR_EQUALS_SIGN ||
5018                *ptr == CHAR_GREATER_THAN_SIGN)  /* Reference or recursion */
5019            {            {
5020            is_recurse = *ptr == '>';            is_recurse = *ptr == CHAR_GREATER_THAN_SIGN;
5021            terminator = ')';            terminator = CHAR_RIGHT_PARENTHESIS;
5022            goto NAMED_REF_OR_RECURSE;            goto NAMED_REF_OR_RECURSE;
5023            }            }
5024          else if (*ptr != '<')    /* Test for Python-style definition */          else if (*ptr != CHAR_LESS_THAN_SIGN)  /* Test for Python-style defn */
5025            {            {
5026            *errorcodeptr = ERR41;            *errorcodeptr = ERR41;
5027            goto FAILED;            goto FAILED;
# Line 4516  we set the flag only if there is a liter Line 5031  we set the flag only if there is a liter
5031    
5032          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5033          DEFINE_NAME:    /* Come here from (?< handling */          DEFINE_NAME:    /* Come here from (?< handling */
5034          case '\'':          case CHAR_APOSTROPHE:
5035            {            {
5036            terminator = (*ptr == '<')? '>' : '\'';            terminator = (*ptr == CHAR_LESS_THAN_SIGN)?
5037                CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
5038            name = ++ptr;            name = ++ptr;
5039    
5040            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
# Line 4549  we set the flag only if there is a liter Line 5065  we set the flag only if there is a liter
5065                }                }
5066              }              }
5067    
5068            /* In the real compile, create the entry in the table */            /* In the real compile, create the entry in the table, maintaining
5069              alphabetical order. Duplicate names for different numbers are
5070              permitted only if PCRE_DUPNAMES is set. Duplicate names for the same
5071              number are always OK. (An existing number can be re-used if (?|
5072              appears in the pattern.) In either event, a duplicate name results in
5073              a duplicate entry in the table, even if the number is the same. This
5074              is because the number of names, and hence the table size, is computed
5075              in the pre-compile, and it affects various numbers and pointers which
5076              would all have to be modified, and the compiled code moved down, if
5077              duplicates with the same number were omitted from the table. This
5078              doesn't seem worth the hassle. However, *different* names for the
5079              same number are not permitted. */
5080    
5081            else            else
5082              {              {
5083                BOOL dupname = FALSE;
5084              slot = cd->name_table;              slot = cd->name_table;
5085    
5086              for (i = 0; i < cd->names_found; i++)              for (i = 0; i < cd->names_found; i++)
5087                {                {
5088                int crc = memcmp(name, slot+2, namelen);                int crc = memcmp(name, slot+2, namelen);
# Line 4561  we set the flag only if there is a liter Line 5090  we set the flag only if there is a liter
5090                  {                  {
5091                  if (slot[2+namelen] == 0)                  if (slot[2+namelen] == 0)
5092                    {                    {
5093                    if ((options & PCRE_DUPNAMES) == 0)                    if (GET2(slot, 0) != cd->bracount + 1 &&
5094                          (options & PCRE_DUPNAMES) == 0)
5095                      {                      {
5096                      *errorcodeptr = ERR43;                      *errorcodeptr = ERR43;
5097                      goto FAILED;                      goto FAILED;
5098                      }                      }
5099                      else dupname = TRUE;
5100                    }                    }
5101                  else crc = -1;      /* Current name is substring */                  else crc = -1;      /* Current name is a substring */
5102                  }                  }
5103    
5104                  /* Make space in the table and break the loop for an earlier
5105                  name. For a duplicate or later name, carry on. We do this for
5106                  duplicates so that in the simple case (when ?(| is not used) they
5107                  are in order of their numbers. */
5108    
5109                if (crc < 0)                if (crc < 0)
5110                  {                  {
5111                  memmove(slot + cd->name_entry_size, slot,                  memmove(slot + cd->name_entry_size, slot,
5112                    (cd->names_found - i) * cd->name_entry_size);                    (cd->names_found - i) * cd->name_entry_size);
5113                  break;                  break;
5114                  }                  }
5115    
5116                  /* Continue the loop for a later or duplicate name */
5117    
5118                slot += cd->name_entry_size;                slot += cd->name_entry_size;
5119                }                }
5120    
5121                /* For non-duplicate names, check for a duplicate number before
5122                adding the new name. */
5123    
5124                if (!dupname)
5125                  {
5126                  uschar *cslot = cd->name_table;
5127                  for (i = 0; i < cd->names_found; i++)
5128                    {
5129                    if (cslot != slot)
5130                      {
5131                      if (GET2(cslot, 0) == cd->bracount + 1)
5132                        {
5133                        *errorcodeptr = ERR65;
5134                        goto FAILED;
5135                        }
5136                      }
5137                    else i--;
5138                    cslot += cd->name_entry_size;
5139                    }
5140                  }
5141    
5142              PUT2(slot, 0, cd->bracount + 1);              PUT2(slot, 0, cd->bracount + 1);
5143              memcpy(slot + 2, name, namelen);              memcpy(slot + 2, name, namelen);
5144              slot[2+namelen] = 0;              slot[2+namelen] = 0;
5145              }              }
5146            }            }
5147    
5148          /* In both cases, count the number of names we've encountered. */          /* In both pre-compile and compile, count the number of names we've
5149            encountered. */
5150    
         ptr++;                    /* Move past > or ' */  
5151          cd->names_found++;          cd->names_found++;
5152            ptr++;                    /* Move past > or ' */
5153          goto NUMBERED_GROUP;          goto NUMBERED_GROUP;
5154    
5155    
5156          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5157          case '&':                 /* Perl recursion/subroutine syntax */          case CHAR_AMPERSAND:            /* Perl recursion/subroutine syntax */
5158          terminator = ')';          terminator = CHAR_RIGHT_PARENTHESIS;
5159          is_recurse = TRUE;          is_recurse = TRUE;
5160          /* Fall through */          /* Fall through */
5161    
# Line 4652  we set the flag only if there is a liter Line 5214  we set the flag only if there is a liter
5214              recno = GET2(slot, 0);              recno = GET2(slot, 0);
5215              }              }
5216            else if ((recno =                /* Forward back reference */            else if ((recno =                /* Forward back reference */
5217                      find_parens(ptr, cd->bracount, name, namelen,                      find_parens(cd, name, namelen,
5218                        (options & PCRE_EXTENDED) != 0)) <= 0)                        (options & PCRE_EXTENDED) != 0)) <= 0)
5219              {              {
5220              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
# Line 4668  we set the flag only if there is a liter Line 5230  we set the flag only if there is a liter
5230    
5231    
5232          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5233          case 'R':                 /* Recursion */          case CHAR_R:              /* Recursion */
5234          ptr++;                    /* Same as (?0)      */          ptr++;                    /* Same as (?0)      */
5235          /* Fall through */          /* Fall through */
5236    
5237    
5238          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5239          case '-': case '+':          case CHAR_MINUS: case CHAR_PLUS:  /* Recursion or subroutine */
5240          case '0': case '1': case '2': case '3': case '4':   /* Recursion or */          case CHAR_0: case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4:
5241          case '5': case '6': case '7': case '8': case '9':   /* subroutine */          case CHAR_5: case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
5242            {            {
5243            const uschar *called;            const uschar *called;
5244            terminator = ')';            terminator = CHAR_RIGHT_PARENTHESIS;
5245    
5246            /* Come here from the \g<...> and \g'...' code (Oniguruma            /* Come here from the \g<...> and \g'...' code (Oniguruma
5247            compatibility). However, the syntax has been checked to ensure that            compatibility). However, the syntax has been checked to ensure that
5248            the ... are a (signed) number, so that neither ERR63 nor ERR29 will            the ... are a (signed) number, so that neither ERR63 nor ERR29 will
5249            be called on this path, nor with the jump to OTHER_CHAR_AFTER_QUERY            be called on this path, nor with the jump to OTHER_CHAR_AFTER_QUERY
5250            ever be taken. */            ever be taken. */
   
           HANDLE_NUMERICAL_RECURSION:  
5251    
5252            if ((refsign = *ptr) == '+')            HANDLE_NUMERICAL_RECURSION:
5253    
5254              if ((refsign = *ptr) == CHAR_PLUS)
5255              {              {
5256              ptr++;              ptr++;
5257              if ((digitab[*ptr] & ctype_digit) == 0)              if ((digitab[*ptr] & ctype_digit) == 0)
# Line 4698  we set the flag only if there is a liter Line 5260  we set the flag only if there is a liter
5260                goto FAILED;                goto FAILED;
5261                }                }
5262              }              }
5263            else if (refsign == '-')            else if (refsign == CHAR_MINUS)
5264              {              {
5265              if ((digitab[ptr[1]] & ctype_digit) == 0)              if ((digitab[ptr[1]] & ctype_digit) == 0)
5266                goto OTHER_CHAR_AFTER_QUERY;                goto OTHER_CHAR_AFTER_QUERY;
# Line 4707  we set the flag only if there is a liter Line 5269  we set the flag only if there is a liter
5269    
5270            recno = 0;            recno = 0;
5271            while((digitab[*ptr] & ctype_digit) != 0)            while((digitab[*ptr] & ctype_digit) != 0)
5272              recno = recno * 10 + *ptr++ - '0';              recno = recno * 10 + *ptr++ - CHAR_0;
5273    
5274            if (*ptr != terminator)            if (*ptr != terminator)
5275              {              {
# Line 4715  we set the flag only if there is a liter Line 5277  we set the flag only if there is a liter
5277              goto FAILED;              goto FAILED;
5278              }              }
5279    
5280            if (refsign == '-')            if (refsign == CHAR_MINUS)
5281              {              {
5282              if (recno == 0)              if (recno == 0)
5283                {                {
# Line 4729  we set the flag only if there is a liter Line 5291  we set the flag only if there is a liter
5291                goto FAILED;                goto FAILED;
5292                }                }
5293              }              }
5294            else if (refsign == '+')            else if (refsign == CHAR_PLUS)
5295              {              {
5296              if (recno == 0)              if (recno == 0)
5297                {                {
# Line 4756  we set the flag only if there is a liter Line 5318  we set the flag only if there is a liter
5318            if (lengthptr == NULL)            if (lengthptr == NULL)
5319              {              {
5320              *code = OP_END;              *code = OP_END;
5321              if (recno != 0) called = find_bracket(cd->start_code, utf8, recno);              if (recno != 0)
5322                  called = _pcre_find_bracket(cd->start_code, utf8, recno);
5323    
5324              /* Forward reference */              /* Forward reference */
5325    
5326              if (called == NULL)              if (called == NULL)
5327                {                {
5328                if (find_parens(ptr, cd->bracount, NULL, recno,                if (find_parens(cd, NULL, recno,
5329                     (options & PCRE_EXTENDED) != 0) < 0)                      (options & PCRE_EXTENDED) != 0) < 0)
5330                  {                  {
5331                  *errorcodeptr = ERR15;                  *errorcodeptr = ERR15;
5332                  goto FAILED;                  goto FAILED;
5333                  }                  }
5334    
5335                  /* Fudge the value of "called" so that when it is inserted as an
5336                  offset below, what it actually inserted is the reference number
5337                  of the group. */
5338    
5339                called = cd->start_code + recno;                called = cd->start_code + recno;
5340                PUTINC(cd->hwm, 0, code + 2 + LINK_SIZE - cd->start_code);                PUTINC(cd->hwm, 0, code + 2 + LINK_SIZE - cd->start_code);
5341                }                }
# Line 4777  we set the flag only if there is a liter Line 5345  we set the flag only if there is a liter
5345              recursion that could loop for ever, and diagnose that case. */              recursion that could loop for ever, and diagnose that case. */
5346    
5347              else if (GET(called, 1) == 0 &&              else if (GET(called, 1) == 0 &&
5348                       could_be_empty(called, code, bcptr, utf8))                       could_be_empty(called, code, bcptr, utf8, cd))
5349                {                {
5350                *errorcodeptr = ERR40;                *errorcodeptr = ERR40;
5351                goto FAILED;                goto FAILED;
# Line 4815  we set the flag only if there is a liter Line 5383  we set the flag only if there is a liter
5383          set = unset = 0;          set = unset = 0;
5384          optset = &set;          optset = &set;
5385    
5386          while (*ptr != ')' && *ptr != ':')          while (*ptr != CHAR_RIGHT_PARENTHESIS && *ptr != CHAR_COLON)
5387            {            {
5388            switch (*ptr++)            switch (*ptr++)
5389              {              {
5390              case '-': optset = &unset; break;              case CHAR_MINUS: optset = &unset; break;
5391    
5392              case 'J':    /* Record that it changed in the external options */              case CHAR_J:    /* Record that it changed in the external options */
5393              *optset |= PCRE_DUPNAMES;              *optset |= PCRE_DUPNAMES;
5394              cd->external_flags |= PCRE_JCHANGED;              cd->external_flags |= PCRE_JCHANGED;
5395              break;              break;
5396    
5397              case 'i': *optset |= PCRE_CASELESS; break;              case CHAR_i: *optset |= PCRE_CASELESS; break;
5398              case 'm': *optset |= PCRE_MULTILINE; break;              case CHAR_m: *optset |= PCRE_MULTILINE; break;
5399              case 's': *optset |= PCRE_DOTALL; break;              case CHAR_s: *optset |= PCRE_DOTALL; break;
5400              case 'x': *optset |= PCRE_EXTENDED; break;              case CHAR_x: *optset |= PCRE_EXTENDED; break;
5401              case 'U': *optset |= PCRE_UNGREEDY; break;              case CHAR_U: *optset |= PCRE_UNGREEDY; break;
5402              case 'X': *optset |= PCRE_EXTRA; break;              case CHAR_X: *optset |= PCRE_EXTRA; break;
5403    
5404              default:  *errorcodeptr = ERR12;              default:  *errorcodeptr = ERR12;
5405                        ptr--;    /* Correct the offset */                        ptr--;    /* Correct the offset */
# Line 4862  we set the flag only if there is a liter Line 5430  we set the flag only if there is a liter
5430          both phases.          both phases.
5431    
5432          If we are not at the pattern start, compile code to change the ims          If we are not at the pattern start, compile code to change the ims
5433          options if this setting actually changes any of them. We also pass the          options if this setting actually changes any of them, and reset the
5434          new setting back so that it can be put at the start of any following          greedy defaults and the case value for firstbyte and reqbyte. */
         branches, and when this group ends (if we are in a group), a resetting  
         item can be compiled. */  
5435    
5436          if (*ptr == ')')          if (*ptr == CHAR_RIGHT_PARENTHESIS)
5437            {            {
5438            if (code == cd->start_code + 1 + LINK_SIZE &&            if (code == cd->start_code + 1 + LINK_SIZE &&
5439                 (lengthptr == NULL || *lengthptr == 2 + 2*LINK_SIZE))                 (lengthptr == NULL || *lengthptr == 2 + 2*LINK_SIZE))
5440              {              {
5441              cd->external_options = newoptions;              cd->external_options = newoptions;
             options = newoptions;  
5442              }              }
5443           else            else
5444              {              {
5445              if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))              if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))
5446                {                {
5447                *code++ = OP_OPT;                *code++ = OP_OPT;
5448                *code++ = newoptions & PCRE_IMS;                *code++ = newoptions & PCRE_IMS;
5449                }                }
   
             /* Change options at this level, and pass them back for use  
             in subsequent branches. Reset the greedy defaults and the case  
             value for firstbyte and reqbyte. */  
   
             *optionsptr = options = newoptions;  
5450              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);
5451              greedy_non_default = greedy_default ^ 1;              greedy_non_default = greedy_default ^ 1;
5452              req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
5453              }              }
5454    
5455              /* Change options at this level, and pass them back for use
5456              in subsequent branches. When not at the start of the pattern, this
5457              information is also necessary so that a resetting item can be
5458              compiled at the end of a group (if we are in a group). */
5459    
5460              *optionsptr = options = newoptions;
5461            previous = NULL;       /* This item can't be repeated */            previous = NULL;       /* This item can't be repeated */
5462            continue;              /* It is complete */            continue;              /* It is complete */
5463            }            }
# Line 4907  we set the flag only if there is a liter Line 5472  we set the flag only if there is a liter
5472          }     /* End of switch for character following (? */          }     /* End of switch for character following (? */
5473        }       /* End of (? handling */        }       /* End of (? handling */
5474    
5475      /* Opening parenthesis not followed by '?'. If PCRE_NO_AUTO_CAPTURE is set,      /* Opening parenthesis not followed by '*' or '?'. If PCRE_NO_AUTO_CAPTURE
5476      all unadorned brackets become non-capturing and behave like (?:...)      is set, all unadorned brackets become non-capturing and behave like (?:...)
5477      brackets. */      brackets. */
5478    
5479      else if ((options & PCRE_NO_AUTO_CAPTURE) != 0)      else if ((options & PCRE_NO_AUTO_CAPTURE) != 0)
# Line 5008  we set the flag only if there is a liter Line 5573  we set the flag only if there is a liter
5573    
5574      /* Error if hit end of pattern */      /* Error if hit end of pattern */
5575    
5576      if (*ptr != ')')      if (*ptr != CHAR_RIGHT_PARENTHESIS)
5577        {        {
5578        *errorcodeptr = ERR14;        *errorcodeptr = ERR14;
5579        goto FAILED;        goto FAILED;
# Line 5100  we set the flag only if there is a liter Line 5665  we set the flag only if there is a liter
5665    
5666      /* ===================================================================*/      /* ===================================================================*/
5667      /* Handle metasequences introduced by \. For ones like \d, the ESC_ values      /* Handle metasequences introduced by \. For ones like \d, the ESC_ values
5668      are arranged to be the negation of the corresponding OP_values. For the      are arranged to be the negation of the corresponding OP_values in the
5669      back references, the values are ESC_REF plus the reference number. Only      default case when PCRE_UCP is not set. For the back references, the values
5670      back references and those types that consume a character may be repeated.      are ESC_REF plus the reference number. Only back references and those types
5671      We can test for values between ESC_b and ESC_Z for the latter; this may      that consume a character may be repeated. We can test for values between
5672      have to change if any new ones are ever created. */      ESC_b and ESC_Z for the latter; this may have to change if any new ones are
5673        ever created. */
5674      case '\\':  
5675        case CHAR_BACKSLASH:
5676      tempptr = ptr;      tempptr = ptr;
5677      c = check_escape(&ptr, errorcodeptr, cd->bracount, options, FALSE);      c = check_escape(&ptr, errorcodeptr, cd->bracount, options, FALSE);
5678      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
# Line 5115  we set the flag only if there is a liter Line 5681  we set the flag only if there is a liter
5681        {        {
5682        if (-c == ESC_Q)            /* Handle start of quoted string */        if (-c == ESC_Q)            /* Handle start of quoted string */
5683          {          {
5684          if (ptr[1] == '\\' && ptr[2] == 'E') ptr += 2; /* avoid empty string */          if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
5685            else inescq = TRUE;            ptr += 2;               /* avoid empty string */
5686                else inescq = TRUE;
5687          continue;          continue;
5688          }          }
5689    
# Line 5132  we set the flag only if there is a liter Line 5699  we set the flag only if there is a liter
5699    
5700        zerofirstbyte = firstbyte;        zerofirstbyte = firstbyte;
5701        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
5702    
5703        /* \g<name> or \g'name' is a subroutine call by name and \g<n> or \g'n'        /* \g<name> or \g'name' is a subroutine call by name and \g<n> or \g'n'
5704        is a subroutine call by number (Oniguruma syntax). In fact, the value        is a subroutine call by number (Oniguruma syntax). In fact, the value
5705        -ESC_g is returned only for these cases. So we don't need to check for <        -ESC_g is returned only for these cases. So we don't need to check for <
5706        or ' if the value is -ESC_g. For the Perl syntax \g{n} the value is        or ' if the value is -ESC_g. For the Perl syntax \g{n} the value is
5707        -ESC_REF+n, and for the Perl syntax \g{name} the result is -ESC_k (as        -ESC_REF+n, and for the Perl syntax \g{name} the result is -ESC_k (as
5708        that is a synonym for a named back reference). */        that is a synonym for a named back reference). */
5709    
5710        if (-c == ESC_g)        if (-c == ESC_g)
5711          {          {
5712          const uschar *p;          const uschar *p;
5713          save_hwm = cd->hwm;   /* Normally this is set when '(' is read */          save_hwm = cd->hwm;   /* Normally this is set when '(' is read */
5714          terminator = (*(++ptr) == '<')? '>' : '\'';          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
5715              CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
5716    
5717          /* These two statements stop the compiler for warning about possibly          /* These two statements stop the compiler for warning about possibly
5718          unset variables caused by the jump to HANDLE_NUMERICAL_RECURSION. In          unset variables caused by the jump to HANDLE_NUMERICAL_RECURSION. In
5719          fact, because we actually check for a number below, the paths that          fact, because we actually check for a number below, the paths that
5720          would actually be in error are never taken. */          would actually be in error are never taken. */
5721    
5722          skipbytes = 0;          skipbytes = 0;
5723          reset_bracount = FALSE;          reset_bracount = FALSE;
5724    
5725          /* Test for a name */          /* Test for a name */
5726    
5727          if (ptr[1] != '+' && ptr[1] != '-')          if (ptr[1] != CHAR_PLUS && ptr[1] != CHAR_MINUS)
5728            {            {
5729            BOOL isnumber = TRUE;            BOOL isnumber = TRUE;
5730            for (p = ptr + 1; *p != 0 && *p != terminator; p++)            for (p = ptr + 1; *p != 0 && *p != terminator; p++)
5731              {              {
5732              if ((cd->ctypes[*p] & ctype_digit) == 0) isnumber = FALSE;              if ((cd->ctypes[*p] & ctype_digit) == 0) isnumber = FALSE;
5733              if ((cd->ctypes[*p] & ctype_word) == 0) break;              if ((cd->ctypes[*p] & ctype_word) == 0) break;
5734              }              }
5735            if (*p != terminator)            if (*p != terminator)
5736              {              {
5737              *errorcodeptr = ERR57;              *errorcodeptr = ERR57;
5738              break;              break;
5739              }              }
5740            if (isnumber)            if (isnumber)
5741              {              {
5742              ptr++;              ptr++;
5743              goto HANDLE_NUMERICAL_RECURSION;              goto HANDLE_NUMERICAL_RECURSION;
5744              }              }
5745            is_recurse = TRUE;            is_recurse = TRUE;
5746            goto NAMED_REF_OR_RECURSE;            goto NAMED_REF_OR_RECURSE;
5747            }            }
5748    
5749          /* Test a signed number in angle brackets or quotes. */          /* Test a signed number in angle brackets or quotes. */
5750    
5751          p = ptr + 2;          p = ptr + 2;
5752          while ((digitab[*p] & ctype_digit) != 0) p++;          while ((digitab[*p] & ctype_digit) != 0) p++;
5753          if (*p != terminator)          if (*p != terminator)
# Line 5187  we set the flag only if there is a liter Line 5755  we set the flag only if there is a liter
5755            *errorcodeptr = ERR57;            *errorcodeptr = ERR57;
5756            break;            break;
5757            }            }
5758          ptr++;          ptr++;
5759          goto HANDLE_NUMERICAL_RECURSION;          goto HANDLE_NUMERICAL_RECURSION;
5760          }          }
5761    
5762        /* \k<name> or \k'name' is a back reference by name (Perl syntax).        /* \k<name> or \k'name' is a back reference by name (Perl syntax).
5763        We also support \k{name} (.NET syntax) */        We also support \k{name} (.NET syntax) */
5764    
5765        if (-c == ESC_k && (ptr[1] == '<' || ptr[1] == '\'' || ptr[1] == '{'))        if (-c == ESC_k && (ptr[1] == CHAR_LESS_THAN_SIGN ||
5766              ptr[1] == CHAR_APOSTROPHE || ptr[1] == CHAR_LEFT_CURLY_BRACKET))
5767          {          {
5768          is_recurse = FALSE;          is_recurse = FALSE;
5769          terminator = (*(++ptr) == '<')? '>' : (*ptr == '\'')? '\'' : '}';          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
5770              CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?
5771              CHAR_APOSTROPHE : CHAR_RIGHT_CURLY_BRACKET;
5772          goto NAMED_REF_OR_RECURSE;          goto NAMED_REF_OR_RECURSE;
5773          }          }
5774    
# Line 5207  we set the flag only if there is a liter Line 5778  we set the flag only if there is a liter
5778    
5779        if (-c >= ESC_REF)        if (-c >= ESC_REF)
5780          {          {
5781            open_capitem *oc;
5782          recno = -c - ESC_REF;          recno = -c - ESC_REF;
5783    
5784          HANDLE_REFERENCE:    /* Come here from named backref handling */          HANDLE_REFERENCE:    /* Come here from named backref handling */
# Line 5216  we set the flag only if there is a liter Line 5788  we set the flag only if there is a liter
5788          PUT2INC(code, 0, recno);          PUT2INC(code, 0, recno);
5789          cd->backref_map |= (recno < 32)? (1 << recno) : 1;          cd->backref_map |= (recno < 32)? (1 << recno) : 1;
5790          if (recno > cd->top_backref) cd->top_backref = recno;          if (recno > cd->top_backref) cd->top_backref = recno;
5791    
5792            /* Check to see if this back reference is recursive, that it, it
5793            is inside the group that it references. A flag is set so that the
5794            group can be made atomic. */
5795    
5796            for (oc = cd->open_caps; oc != NULL; oc = oc->next)
5797              {
5798              if (oc->number == recno)
5799                {
5800                oc->flag = TRUE;
5801                break;
5802                }
5803              }
5804          }          }
5805    
5806        /* So are Unicode property matches, if supported. */        /* So are Unicode property matches, if supported. */
# Line 5245  we set the flag only if there is a liter Line 5830  we set the flag only if there is a liter
5830  #endif  #endif
5831    
5832        /* For the rest (including \X when Unicode properties are supported), we        /* For the rest (including \X when Unicode properties are supported), we
5833        can obtain the OP value by negating the escape value. */        can obtain the OP value by negating the escape value in the default
5834          situation when PCRE_UCP is not set. When it *is* set, we substitute
5835          Unicode property tests. */
5836    
5837        else        else
5838          {          {
5839          previous = (-c > ESC_b && -c < ESC_Z)? code : NULL;  #ifdef SUPPORT_UCP
5840          *code++ = -c;          if (-c >= ESC_DU && -c <= ESC_wu)
5841              {
5842              nestptr = ptr + 1;                   /* Where to resume */
5843              ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
5844              }
5845            else
5846    #endif
5847              {
5848              previous = (-c > ESC_b && -c < ESC_Z)? code : NULL;
5849              *code++ = -c;
5850              }
5851          }          }
5852        continue;        continue;
5853        }        }
# Line 5300  we set the flag only if there is a liter Line 5897  we set the flag only if there is a liter
5897    
5898      /* Remember if \r or \n were seen */      /* Remember if \r or \n were seen */
5899    
5900      if (mcbuffer[0] == '\r' || mcbuffer[0] == '\n')      if (mcbuffer[0] == CHAR_CR || mcbuffer[0] == CHAR_NL)
5901        cd->external_flags |= PCRE_HASCRORLF;        cd->external_flags |= PCRE_HASCRORLF;
5902    
5903      /* Set the first and required bytes appropriately. If no previous first      /* Set the first and required bytes appropriately. If no previous first
# Line 5398  uschar *code = *codeptr; Line 5995  uschar *code = *codeptr;
5995  uschar *last_branch = code;  uschar *last_branch = code;
5996  uschar *start_bracket = code;  uschar *start_bracket = code;
5997  uschar *reverse_count = NULL;  uschar *reverse_count = NULL;
5998    open_capitem capitem;
5999    int capnumber = 0;
6000  int firstbyte, reqbyte;  int firstbyte, reqbyte;
6001  int branchfirstbyte, branchreqbyte;  int branchfirstbyte, branchreqbyte;
6002  int length;  int length;
6003  int orig_bracount;  int orig_bracount;
6004  int max_bracount;  int max_bracount;
6005    int old_external_options = cd->external_options;
6006  branch_chain bc;  branch_chain bc;
6007    
6008  bc.outer = bcptr;  bc.outer = bcptr;
6009  bc.current = code;  bc.current_branch = code;
6010    
6011  firstbyte = reqbyte = REQ_UNSET;  firstbyte = reqbyte = REQ_UNSET;
6012    
# Line 5424  the code that abstracts option settings Line 6024  the code that abstracts option settings
6024  them global. It tests the value of length for (2 + 2*LINK_SIZE) in the  them global. It tests the value of length for (2 + 2*LINK_SIZE) in the
6025  pre-compile phase to find out whether anything has yet been compiled or not. */  pre-compile phase to find out whether anything has yet been compiled or not. */
6026    
6027    /* If this is a capturing subpattern, add to the chain of open capturing items
6028    so that we can detect them if (*ACCEPT) is encountered. This is also used to
6029    detect groups that contain recursive back references to themselves. */
6030    
6031    if (*code == OP_CBRA)
6032      {
6033      capnumber = GET2(code, 1 + LINK_SIZE);
6034      capitem.number = capnumber;
6035      capitem.next = cd->open_caps;
6036      capitem.flag = FALSE;
6037      cd->open_caps = &capitem;
6038      }
6039    
6040  /* Offset is set zero to mark that this bracket is still open */  /* Offset is set zero to mark that this bracket is still open */
6041    
6042  PUT(code, 1, 0);  PUT(code, 1, 0);
# Line 5468  for (;;) Line 6081  for (;;)
6081      return FALSE;      return FALSE;
6082      }      }
6083    
6084      /* If the external options have changed during this branch, it means that we
6085      are at the top level, and a leading option setting has been encountered. We
6086      need to re-set the original option values to take account of this so that,
6087      during the pre-compile phase, we know to allow for a re-set at the start of
6088      subsequent branches. */
6089    
6090      if (old_external_options != cd->external_options)
6091        oldims = cd->external_options & PCRE_IMS;
6092    
6093    /* Keep the highest bracket count in case (?| was used and some branch    /* Keep the highest bracket count in case (?| was used and some branch
6094    has fewer than the rest. */    has fewer than the rest. */
6095    
# Line 5518  for (;;) Line 6140  for (;;)
6140    
6141      /* If lookbehind, check that this branch matches a fixed-length string, and      /* If lookbehind, check that this branch matches a fixed-length string, and
6142      put the length into the OP_REVERSE item. Temporarily mark the end of the      put the length into the OP_REVERSE item. Temporarily mark the end of the
6143      branch with OP_END. */      branch with OP_END. If the branch contains OP_RECURSE, the result is -3
6144        because there may be forward references that we can't check here. Set a
6145        flag to cause another lookbehind check at the end. Why not do it all at the
6146        end? Because common, erroneous checks are picked up here and the offset of
6147        the problem can be shown. */
6148    
6149      if (lookbehind)      if (lookbehind)
6150        {        {
6151        int fixed_length;        int fixed_length;
6152        *code = OP_END;        *code = OP_END;
6153        fixed_length = find_fixedlength(last_branch, options);        fixed_length = find_fixedlength(last_branch, options, FALSE, cd);
6154        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
6155        if (fixed_length < 0)        if (fixed_length == -3)
6156            {
6157            cd->check_lookbehind = TRUE;
6158            }
6159          else if (fixed_length < 0)
6160          {          {
6161          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;
6162          *ptrptr = ptr;          *ptrptr = ptr;
6163          return FALSE;          return FALSE;
6164          }          }
6165        PUT(reverse_count, 0, fixed_length);        else { PUT(reverse_count, 0, fixed_length); }
6166        }        }
6167      }      }
6168    
# Line 5545  for (;;) Line 6175  for (;;)
6175    compile a resetting op-code following, except at the very end of the pattern.    compile a resetting op-code following, except at the very end of the pattern.
6176    Return leaving the pointer at the terminating char. */    Return leaving the pointer at the terminating char. */
6177    
6178    if (*ptr != '|')    if (*ptr != CHAR_VERTICAL_LINE)
6179      {      {
6180      if (lengthptr == NULL)      if (lengthptr == NULL)
6181        {        {
# Line 5566  for (;;) Line 6196  for (;;)
6196      PUT(code, 1, code - start_bracket);      PUT(code, 1, code - start_bracket);
6197      code += 1 + LINK_SIZE;      code += 1 + LINK_SIZE;
6198    
6199      /* Resetting option if needed */      /* If it was a capturing subpattern, check to see if it contained any
6200        recursive back references. If so, we must wrap it in atomic brackets.
6201        In any event, remove the block from the chain. */
6202    
6203      if ((options & PCRE_IMS) != oldims && *ptr == ')')      if (capnumber > 0)
6204          {
6205          if (cd->open_caps->flag)
6206            {
6207            memmove(start_bracket + 1 + LINK_SIZE, start_bracket,
6208              code - start_bracket);
6209            *start_bracket = OP_ONCE;
6210            code += 1 + LINK_SIZE;
6211            PUT(start_bracket, 1, code - start_bracket);
6212            *code = OP_KET;
6213            PUT(code, 1, code - start_bracket);
6214            code += 1 + LINK_SIZE;
6215            length += 2 + 2*LINK_SIZE;
6216            }
6217          cd->open_caps = cd->open_caps->next;
6218          }
6219    
6220        /* Reset options if needed. */
6221    
6222        if ((options & PCRE_IMS) != oldims && *ptr == CHAR_RIGHT_PARENTHESIS)
6223        {        {
6224        *code++ = OP_OPT;        *code++ = OP_OPT;
6225        *code++ = oldims;        *code++ = oldims;
# Line 5615  for (;;) Line 6266  for (;;)
6266      {      {
6267      *code = OP_ALT;      *code = OP_ALT;
6268      PUT(code, 1, code - last_branch);      PUT(code, 1, code - last_branch);
6269      bc.current = last_branch = code;      bc.current_branch = last_branch = code;
6270      code += 1 + LINK_SIZE;      code += 1 + LINK_SIZE;
6271      }      }
6272    
# Line 5697  do { Line 6348  do {
6348       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;
6349       }       }
6350    
6351     /* .* is not anchored unless DOTALL is set and it isn't in brackets that     /* .* is not anchored unless DOTALL is set (which generates OP_ALLANY) and
6352     are or may be referenced. */     it isn't in brackets that are or may be referenced. */
6353    
6354     else if ((op == OP_TYPESTAR || op == OP_TYPEMINSTAR ||     else if ((op == OP_TYPESTAR || op == OP_TYPEMINSTAR ||
6355               op == OP_TYPEPOSSTAR) &&               op == OP_TYPEPOSSTAR))
             (*options & PCRE_DOTALL) != 0)  
6356       {       {
6357       if (scode[1] != OP_ANY || (bracket_map & backref_map) != 0) return FALSE;       if (scode[1] != OP_ALLANY || (bracket_map & backref_map) != 0)
6358           return FALSE;
6359       }       }
6360    
6361     /* Check for explicit anchoring */     /* Check for explicit anchoring */
# Line 5750  do { Line 6401  do {
6401       NULL, 0, FALSE);       NULL, 0, FALSE);
6402     register int op = *scode;     register int op = *scode;
6403    
6404       /* If we are at the start of a conditional assertion group, *both* the
6405       conditional assertion *and* what follows the condition must satisfy the test
6406       for start of line. Other kinds of condition fail. Note that there may be an
6407       auto-callout at the start of a condition. */
6408    
6409       if (op == OP_COND)
6410         {
6411         scode += 1 + LINK_SIZE;
6412         if (*scode == OP_CALLOUT) scode += _pcre_OP_lengths[OP_CALLOUT];
6413         switch (*scode)
6414           {
6415           case OP_CREF:
6416           case OP_NCREF:
6417           case OP_RREF:
6418           case OP_NRREF:
6419           case OP_DEF:
6420           return FALSE;
6421    
6422           default:     /* Assertion */
6423           if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
6424           do scode += GET(scode, 1); while (*scode == OP_ALT);
6425           scode += 1 + LINK_SIZE;
6426           break;
6427           }
6428         scode = first_significant_code(scode, NULL, 0, FALSE);
6429         op = *scode;
6430         }
6431    
6432     /* Non-capturing brackets */     /* Non-capturing brackets */
6433    
6434     if (op == OP_BRA)     if (op == OP_BRA)
# Line 5768  do { Line 6447  do {
6447    
6448     /* Other brackets */     /* Other brackets */
6449    
6450     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_COND)     else if (op == OP_ASSERT || op == OP_ONCE)
6451       { if (!is_startline(scode, bracket_map, backref_map)) return FALSE; }       {
6452         if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
6453         }
6454    
6455     /* .* means "start at start or after \n" if it isn't in brackets that     /* .* means "start at start or after \n" if it isn't in brackets that
6456     may be referenced. */     may be referenced. */
# Line 5886  Returns:        pointer to compiled data Line 6567  Returns:        pointer to compiled data
6567                  with errorptr and erroroffset set                  with errorptr and erroroffset set
6568  */  */
6569    
6570  PCRE_EXP_DEFN pcre *  PCRE_EXP_DEFN pcre * PCRE_CALL_CONVENTION
6571  pcre_compile(const char *pattern, int options, const char **errorptr,  pcre_compile(const char *pattern, int options, const char **errorptr,
6572    int *erroroffset, const unsigned char *tables)    int *erroroffset, const unsigned char *tables)
6573  {  {
# Line 5894  return pcre_compile2(pattern, options, N Line 6575  return pcre_compile2(pattern, options, N
6575  }  }
6576    
6577    
6578  PCRE_EXP_DEFN pcre *  PCRE_EXP_DEFN pcre * PCRE_CALL_CONVENTION
6579  pcre_compile2(const char *pattern, int options, int *errorcodeptr,  pcre_compile2(const char *pattern, int options, int *errorcodeptr,
6580    const char **errorptr, int *erroroffset, const unsigned char *tables)    const char **errorptr, int *erroroffset, const unsigned char *tables)
6581  {  {
# Line 5903  int length = 1;  /* For final END opcode Line 6584  int length = 1;  /* For final END opcode
6584  int firstbyte, reqbyte, newline;  int firstbyte, reqbyte, newline;
6585  int errorcode = 0;  int errorcode = 0;
6586  int skipatstart = 0;  int skipatstart = 0;
 #ifdef SUPPORT_UTF8  
6587  BOOL utf8;  BOOL utf8;
 #endif  
6588  size_t size;  size_t size;
6589  uschar *code;  uschar *code;
6590  const uschar *codestart;  const uschar *codestart;
# Line 5948  if (erroroffset == NULL) Line 6627  if (erroroffset == NULL)
6627    
6628  *erroroffset = 0;  *erroroffset = 0;
6629    
 /* Can't support UTF8 unless PCRE has been compiled to include the code. */  
   
 #ifdef SUPPORT_UTF8  
 utf8 = (options & PCRE_UTF8) != 0;  
 if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&  
      (*erroroffset = _pcre_valid_utf8((uschar *)pattern, -1)) >= 0)  
   {  
   errorcode = ERR44;  
   goto PCRE_EARLY_ERROR_RETURN2;  
   }  
 #else  
 if ((options & PCRE_UTF8) != 0)  
   {  
   errorcode = ERR32;  
   goto PCRE_EARLY_ERROR_RETURN;  
   }  
 #endif  
   
 if ((options & ~PUBLIC_OPTIONS) != 0)  
   {  
   errorcode = ERR17;  
   goto PCRE_EARLY_ERROR_RETURN;  
   }  
   
6630  /* Set up pointers to the individual character tables */  /* Set up pointers to the individual character tables */
6631    
6632  if (tables == NULL) tables = _pcre_default_tables;  if (tables == NULL) tables = _pcre_default_tables;
# Line 5980  cd->fcc = tables + fcc_offset; Line 6635  cd->fcc = tables + fcc_offset;
6635  cd->cbits = tables + cbits_offset;  cd->cbits = tables + cbits_offset;
6636  cd->ctypes = tables + ctypes_offset;  cd->ctypes = tables + ctypes_offset;
6637    
6638    /* Check that all undefined public option bits are zero */
6639    
6640    if ((options & ~PUBLIC_COMPILE_OPTIONS) != 0)
6641      {
6642      errorcode = ERR17;
6643      goto PCRE_EARLY_ERROR_RETURN;
6644      }
6645    
6646  /* Check for global one-time settings at the start of the pattern, and remember  /* Check for global one-time settings at the start of the pattern, and remember
6647  the offset for later. */  the offset for later. */
6648    
6649  while (ptr[skipatstart] == '(' && ptr[skipatstart+1] == '*')  while (ptr[skipatstart] == CHAR_LEFT_PARENTHESIS &&
6650           ptr[skipatstart+1] == CHAR_ASTERISK)
6651    {    {
6652    int newnl = 0;    int newnl = 0;
6653    int newbsr = 0;    int newbsr = 0;
6654    
6655    if (strncmp((char *)(ptr+skipatstart+2), "CR)", 3) == 0)    if (strncmp((char *)(ptr+skipatstart+2), STRING_UTF8_RIGHTPAR, 5) == 0)
6656        { skipatstart += 7; options |= PCRE_UTF8; continue; }
6657      else if (strncmp((char *)(ptr+skipatstart+2), STRING_UCP_RIGHTPAR, 4) == 0)
6658        { skipatstart += 6; options |= PCRE_UCP; continue; }
6659    
6660      if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)
6661      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }
6662    else if (strncmp((char *)(ptr+skipatstart+2), "LF)", 3)  == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_LF_RIGHTPAR, 3)  == 0)
6663      { skipatstart += 5; newnl = PCRE_NEWLINE_LF; }      { skipatstart += 5; newnl = PCRE_NEWLINE_LF; }
6664    else if (strncmp((char *)(ptr+skipatstart+2), "CRLF)", 5)  == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_CRLF_RIGHTPAR, 5)  == 0)
6665      { skipatstart += 7; newnl = PCRE_NEWLINE_CR + PCRE_NEWLINE_LF; }      { skipatstart += 7; newnl = PCRE_NEWLINE_CR + PCRE_NEWLINE_LF; }
6666    else if (strncmp((char *)(ptr+skipatstart+2), "ANY)", 4) == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_ANY_RIGHTPAR, 4) == 0)
6667      { skipatstart += 6; newnl = PCRE_NEWLINE_ANY; }      { skipatstart += 6; newnl = PCRE_NEWLINE_ANY; }
6668    else if (strncmp((char *)(ptr+skipatstart+2), "ANYCRLF)", 8)  == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_ANYCRLF_RIGHTPAR, 8) == 0)
6669      { skipatstart += 10; newnl = PCRE_NEWLINE_ANYCRLF; }      { skipatstart += 10; newnl = PCRE_NEWLINE_ANYCRLF; }
6670    
6671    else if (strncmp((char *)(ptr+skipatstart+2), "BSR_ANYCRLF)", 12) == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_BSR_ANYCRLF_RIGHTPAR, 12) == 0)
6672      { skipatstart += 14; newbsr = PCRE_BSR_ANYCRLF; }      { skipatstart += 14; newbsr = PCRE_BSR_ANYCRLF; }
6673    else if (strncmp((char *)(ptr+skipatstart+2), "BSR_UNICODE)", 12) == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_BSR_UNICODE_RIGHTPAR, 12) == 0)
6674      { skipatstart += 14; newbsr = PCRE_BSR_UNICODE; }      { skipatstart += 14; newbsr = PCRE_BSR_UNICODE; }
6675    
6676    if (newnl != 0)    if (newnl != 0)
# Line 6010  while (ptr[skipatstart] == '(' && ptr[sk Line 6679  while (ptr[skipatstart] == '(' && ptr[sk
6679      options = (options & ~(PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) | newbsr;      options = (options & ~(PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) | newbsr;
6680    else break;    else break;
6681    }    }
6682    
6683    utf8 = (options & PCRE_UTF8) != 0;
6684    
6685    /* Can't support UTF8 unless PCRE has been compiled to include the code. */
6686    
6687    #ifdef SUPPORT_UTF8
6688    if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&
6689         (*erroroffset = _pcre_valid_utf8((USPTR)pattern, -1)) >= 0)
6690      {
6691      errorcode = ERR44;
6692      goto PCRE_EARLY_ERROR_RETURN2;
6693      }
6694    #else
6695    if (utf8)
6696      {
6697      errorcode = ERR32;
6698      goto PCRE_EARLY_ERROR_RETURN;
6699      }
6700    #endif
6701    
6702    /* Can't support UCP unless PCRE has been compiled to include the code. */
6703    
6704    #ifndef SUPPORT_UCP
6705    if ((options & PCRE_UCP) != 0)
6706      {
6707      errorcode = ERR67;
6708      goto PCRE_EARLY_ERROR_RETURN;
6709      }
6710    #endif
6711    
6712  /* Check validity of \R options. */  /* Check validity of \R options. */
6713    
# Line 6029  current code allows for fixed one- or tw Line 6727  current code allows for fixed one- or tw
6727  switch (options & PCRE_NEWLINE_BITS)  switch (options & PCRE_NEWLINE_BITS)
6728    {    {
6729    case 0: newline = NEWLINE; break;   /* Build-time default */    case 0: newline = NEWLINE; break;   /* Build-time default */
6730    case PCRE_NEWLINE_CR: newline = '\r'; break;    case PCRE_NEWLINE_CR: newline = CHAR_CR; break;
6731    case PCRE_NEWLINE_LF: newline = '\n'; break;    case PCRE_NEWLINE_LF: newline = CHAR_NL; break;
6732    case PCRE_NEWLINE_CR+    case PCRE_NEWLINE_CR+
6733         PCRE_NEWLINE_LF: newline = ('\r' << 8) | '\n'; break;         PCRE_NEWLINE_LF: newline = (CHAR_CR << 8) | CHAR_NL; break;
6734    case PCRE_NEWLINE_ANY: newline = -1; break;    case PCRE_NEWLINE_ANY: newline = -1; break;
6735    case PCRE_NEWLINE_ANYCRLF: newline = -2; break;    case PCRE_NEWLINE_ANYCRLF: newline = -2; break;
6736    default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;    default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;
# Line 6093  cd->end_pattern = (const uschar *)(patte Line 6791  cd->end_pattern = (const uschar *)(patte
6791  cd->req_varyopt = 0;  cd->req_varyopt = 0;
6792  cd->external_options = options;  cd->external_options = options;
6793  cd->external_flags = 0;  cd->external_flags = 0;
6794    cd->open_caps = NULL;
6795    
6796  /* Now do the pre-compile. On error, errorcode will be set non-zero, so we  /* Now do the pre-compile. On error, errorcode will be set non-zero, so we
6797  don't need to look at the result of the function here. The initial options have  don't need to look at the result of the function here. The initial options have
# Line 6167  cd->start_code = codestart; Line 6866  cd->start_code = codestart;
6866  cd->hwm = cworkspace;  cd->hwm = cworkspace;
6867  cd->req_varyopt = 0;  cd->req_varyopt = 0;
6868  cd->had_accept = FALSE;  cd->had_accept = FALSE;
6869    cd->check_lookbehind = FALSE;
6870    cd->open_caps = NULL;
6871    
6872  /* Set up a starting, non-extracting bracket, then compile the expression. On  /* Set up a starting, non-extracting bracket, then compile the expression. On
6873  error, errorcode will be set non-zero, so we don't need to look at the result  error, errorcode will be set non-zero, so we don't need to look at the result
# Line 6192  if debugging, leave the test till after Line 6893  if debugging, leave the test till after
6893    
6894  *code++ = OP_END;  *code++ = OP_END;
6895    
6896  #ifndef DEBUG  #ifndef PCRE_DEBUG
6897  if (code - codestart > length) errorcode = ERR23;  if (code - codestart > length) errorcode = ERR23;
6898  #endif  #endif
6899    
# Line 6205  while (errorcode == 0 && cd->hwm > cwork Line 6906  while (errorcode == 0 && cd->hwm > cwork
6906    cd->hwm -= LINK_SIZE;    cd->hwm -= LINK_SIZE;
6907    offset = GET(cd->hwm, 0);    offset = GET(cd->hwm, 0);
6908    recno = GET(codestart, offset);    recno = GET(codestart, offset);
6909    groupptr = find_bracket(codestart, (re->options & PCRE_UTF8) != 0, recno);    groupptr = _pcre_find_bracket(codestart, utf8, recno);
6910    if (groupptr == NULL) errorcode = ERR53;    if (groupptr == NULL) errorcode = ERR53;
6911      else PUT(((uschar *)codestart), offset, groupptr - codestart);      else PUT(((uschar *)codestart), offset, groupptr - codestart);
6912    }    }
6913    
6914  /* Give an error if there's back reference to a non-existent capturing  /* Give an error if there's back reference to a non-existent capturing
6915  subpattern. */  subpattern. */
6916    
6917  if (errorcode == 0 && re->top_backref > re->top_bracket) errorcode = ERR15;  if (errorcode == 0 && re->top_backref > re->top_bracket) errorcode = ERR15;
6918    
6919    /* If there were any lookbehind assertions that contained OP_RECURSE
6920    (recursions or subroutine calls), a flag is set for them to be checked here,
6921    because they may contain forward references. Actual recursions can't be fixed
6922    length, but subroutine calls can. It is done like this so that those without
6923    OP_RECURSE that are not fixed length get a diagnosic with a useful offset. The
6924    exceptional ones forgo this. We scan the pattern to check that they are fixed
6925    length, and set their lengths. */
6926    
6927    if (cd->check_lookbehind)
6928      {
6929      uschar *cc = (uschar *)codestart;
6930    
6931      /* Loop, searching for OP_REVERSE items, and process those that do not have
6932      their length set. (Actually, it will also re-process any that have a length
6933      of zero, but that is a pathological case, and it does no harm.) When we find
6934      one, we temporarily terminate the branch it is in while we scan it. */
6935    
6936      for (cc = (uschar *)_pcre_find_bracket(codestart, utf8, -1);
6937           cc != NULL;
6938           cc = (uschar *)_pcre_find_bracket(cc, utf8, -1))
6939        {
6940        if (GET(cc, 1) == 0)
6941          {
6942          int fixed_length;
6943          uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);
6944          int end_op = *be;
6945          *be = OP_END;
6946          fixed_length = find_fixedlength(cc, re->options, TRUE, cd);
6947          *be = end_op;
6948          DPRINTF(("fixed length = %d\n", fixed_length));
6949          if (fixed_length < 0)
6950            {
6951            errorcode = (fixed_length == -2)? ERR36 : ERR25;
6952            break;
6953            }
6954          PUT(cc, 1, fixed_length);
6955          }
6956        cc += 1 + LINK_SIZE;
6957        }
6958      }
6959    
6960  /* Failed to compile, or error while post-processing */  /* Failed to compile, or error while post-processing */
6961    
6962  if (errorcode != 0)  if (errorcode != 0)
# Line 6275  if (reqbyte >= 0 && Line 7017  if (reqbyte >= 0 &&
7017  /* Print out the compiled data if debugging is enabled. This is never the  /* Print out the compiled data if debugging is enabled. This is never the
7018  case when building a production library. */  case when building a production library. */
7019    
7020  #ifdef DEBUG  #ifdef PCRE_DEBUG
   
7021  printf("Length = %d top_bracket = %d top_backref = %d\n",  printf("Length = %d top_bracket = %d top_backref = %d\n",
7022    length, re->top_bracket, re->top_backref);    length, re->top_bracket, re->top_backref);
7023    
# Line 6313  if (code - codestart > length) Line 7054  if (code - codestart > length)
7054    if (errorcodeptr != NULL) *errorcodeptr = ERR23;    if (errorcodeptr != NULL) *errorcodeptr = ERR23;
7055    return NULL;    return NULL;
7056    }    }
7057  #endif   /* DEBUG */  #endif   /* PCRE_DEBUG */
7058    
7059  return (pcre *)re;  return (pcre *)re;
7060  }  }

Legend:
Removed from v.335  
changed lines
  Added in v.518

  ViewVC Help
Powered by ViewVC 1.1.5