/[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 276 by ph10, Wed Nov 21 15:39:20 2007 UTC revision 528 by ph10, Sat May 29 16:40:22 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) 1997-2007 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  static verbitem verbs[] = {    STRING_THEN;
206    { 6, OP_ACCEPT },  
207    { 6, OP_COMMIT },  static const verbitem verbs[] = {
208    { 1, OP_FAIL },    { 0, -1,        OP_MARK },
209    { 4, OP_FAIL },    { 4, -1,        OP_MARK },
210    { 5, OP_PRUNE },    { 6, OP_ACCEPT, -1 },
211    { 4, OP_SKIP  },    { 6, OP_COMMIT, -1 },
212    { 4, OP_THEN  }    { 1, OP_FAIL,   -1 },
213      { 4, OP_FAIL,   -1 },
214      { 5, OP_PRUNE,  OP_PRUNE_ARG },
215      { 4, OP_SKIP,   OP_SKIP_ARG  },
216      { 4, OP_THEN,   OP_THEN_ARG  }
217  };  };
218    
219  static int verbcount = sizeof(verbs)/sizeof(verbitem);  static const int verbcount = sizeof(verbs)/sizeof(verbitem);
220    
221    
222  /* Tables of names of POSIX character classes and their lengths. The names are  /* Tables of names of POSIX character classes and their lengths. The names are
# Line 178  length entry. The first three must be al Line 226  length entry. The first three must be al
226  for handling case independence. */  for handling case independence. */
227    
228  static const char posix_names[] =  static const char posix_names[] =
229    "alpha\0"  "lower\0"  "upper\0"  "alnum\0"  "ascii\0"  "blank\0"    STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
230    "cntrl\0"  "digit\0"  "graph\0"  "print\0"  "punct\0"  "space\0"    STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
231    "word\0"   "xdigit";    STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
232      STRING_word0  STRING_xdigit;
233    
234  static const uschar posix_name_lengths[] = {  static const uschar posix_name_lengths[] = {
235    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
# Line 212  static const int posix_class_maps[] = { Line 261  static const int posix_class_maps[] = {
261    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
262  };  };
263    
264    /* Table of substitutes for \d etc when PCRE_UCP is set. The POSIX class
265    substitutes must be in the order of the names, defined above, and there are
266    both positive and negative cases. NULL means no substitute. */
267    
268    #ifdef SUPPORT_UCP
269    static const uschar *substitutes[] = {
270      (uschar *)"\\P{Nd}",    /* \D */
271      (uschar *)"\\p{Nd}",    /* \d */
272      (uschar *)"\\P{Xsp}",   /* \S */       /* NOTE: Xsp is Perl space */
273      (uschar *)"\\p{Xsp}",   /* \s */
274      (uschar *)"\\P{Xwd}",   /* \W */
275      (uschar *)"\\p{Xwd}"    /* \w */
276    };
277    
278    static const uschar *posix_substitutes[] = {
279      (uschar *)"\\p{L}",     /* alpha */
280      (uschar *)"\\p{Ll}",    /* lower */
281      (uschar *)"\\p{Lu}",    /* upper */
282      (uschar *)"\\p{Xan}",   /* alnum */
283      NULL,                   /* ascii */
284      (uschar *)"\\h",        /* blank */
285      NULL,                   /* cntrl */
286      (uschar *)"\\p{Nd}",    /* digit */
287      NULL,                   /* graph */
288      NULL,                   /* print */
289      NULL,                   /* punct */
290      (uschar *)"\\p{Xps}",   /* space */    /* NOTE: Xps is POSIX space */
291      (uschar *)"\\p{Xwd}",   /* word */
292      NULL,                   /* xdigit */
293      /* Negated cases */
294      (uschar *)"\\P{L}",     /* ^alpha */
295      (uschar *)"\\P{Ll}",    /* ^lower */
296      (uschar *)"\\P{Lu}",    /* ^upper */
297      (uschar *)"\\P{Xan}",   /* ^alnum */
298      NULL,                   /* ^ascii */
299      (uschar *)"\\H",        /* ^blank */
300      NULL,                   /* ^cntrl */
301      (uschar *)"\\P{Nd}",    /* ^digit */
302      NULL,                   /* ^graph */
303      NULL,                   /* ^print */
304      NULL,                   /* ^punct */
305      (uschar *)"\\P{Xps}",   /* ^space */   /* NOTE: Xps is POSIX space */
306      (uschar *)"\\P{Xwd}",   /* ^word */
307      NULL                    /* ^xdigit */
308    };
309    #define POSIX_SUBSIZE (sizeof(posix_substitutes)/sizeof(uschar *))
310    #endif
311    
312  #define STRING(a)  # a  #define STRING(a)  # a
313  #define XSTRING(s) STRING(s)  #define XSTRING(s) STRING(s)
# Line 224  the number of relocations needed when a Line 320  the number of relocations needed when a
320  it is now one long string. We cannot use a table of offsets, because the  it is now one long string. We cannot use a table of offsets, because the
321  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
322  simply count through to the one we want - this isn't a performance issue  simply count through to the one we want - this isn't a performance issue
323  because these strings are used only when there is a compilation error. */  because these strings are used only when there is a compilation error.
324    
325    Each substring ends with \0 to insert a null character. This includes the final
326    substring, so that the whole string ends with \0\0, which can be detected when
327    counting through. */
328    
329  static const char error_texts[] =  static const char error_texts[] =
330    "no error\0"    "no error\0"
# Line 271  static const char error_texts[] = Line 371  static const char error_texts[] =
371    /* 35 */    /* 35 */
372    "invalid condition (?(0)\0"    "invalid condition (?(0)\0"
373    "\\C not allowed in lookbehind assertion\0"    "\\C not allowed in lookbehind assertion\0"
374    "PCRE does not support \\L, \\l, \\N, \\U, or \\u\0"    "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
375    "number after (?C is > 255\0"    "number after (?C is > 255\0"
376    "closing ) for (?C expected\0"    "closing ) for (?C expected\0"
377    /* 40 */    /* 40 */
# Line 295  static const char error_texts[] = Line 395  static const char error_texts[] =
395    /* 55 */    /* 55 */
396    "repeating a DEFINE group is not allowed\0"    "repeating a DEFINE group is not allowed\0"
397    "inconsistent NEWLINE options\0"    "inconsistent NEWLINE options\0"
398    "\\g is not followed by a braced name or an optionally braced non-zero number\0"    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
399    "(?+ or (?- or (?(+ or (?(- must be followed by a non-zero number\0"    "a numbered reference must not be zero\0"
400    "(*VERB) with an argument is not supported\0"    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
401    /* 60 */    /* 60 */
402    "(*VERB) not recognized\0"    "(*VERB) not recognized\0"
403    "number is too big\0"    "number is too big\0"
404    "subpattern name expected\0"    "subpattern name expected\0"
405    "digit expected after (?+";    "digit expected after (?+\0"
406      "] is an invalid data character in JavaScript compatibility mode\0"
407      /* 65 */
408      "different names for subpatterns of the same number are not allowed\0"
409      "(*MARK) must have an argument\0"
410      "this version of PCRE is not compiled with PCRE_UCP support\0"
411      ;
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 a number, either plain or braced. If positive, it      /* \g must be followed by one of a number of specific things:
650      is an absolute backreference. If negative, it is a relative backreference.  
651      This is a Perl 5.10 feature. Perl 5.10 also supports \g{name} as a      (1) A number, either plain or braced. If positive, it is an absolute
652      reference to a named group. This is part of Perl's movement towards a      backreference. If negative, it is a relative backreference. This is a Perl
653      unified syntax for back references. As this is synonymous with \k{name}, we      5.10 feature.
654      fudge it up by pretending it really was \k. */  
655        (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
657        this is synonymous with \k{name}, we fudge it up by pretending it really
658        was \k.
659    
660        (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
662        (possibly recursive) subroutine calls, _not_ backreferences. Just return
663        the -ESC_g code (cf \k). */
664    
665      case 'g':      case CHAR_g:
666      if (ptr[1] == '{')      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
667          {
668          c = -ESC_g;
669          break;
670          }
671    
672        /* Handle the Perl-compatible cases */
673    
674        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 554  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 563  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)      if (c < 0)   /* Integer overflow */
701        {        {
702        *errorcodeptr = ERR61;        *errorcodeptr = ERR61;
703        break;        break;
704        }        }
705    
706      if (c == 0 || (braced && *(++ptr) != '}'))      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
707        {        {
708        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
709        break;        break;
710        }        }
711    
712        if (c == 0)
713          {
714          *errorcodeptr = ERR58;
715          break;
716          }
717    
718      if (negated)      if (negated)
719        {        {
720        if (c > bracount)        if (c > bracount)
# Line 602  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)        if (c < 0)    /* Integer overflow */
753          {          {
754          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
755          break;          break;
# Line 628  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 641  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 652  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 662  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 690  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 706  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 714  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 740  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 780  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 791  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 859  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 900  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 910  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 940  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 1008  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    FAIL_EXIT:
1296    *ptrptr = ptr;
1297    return -1;
1298    }
1299    
   if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  
        *ptr != '\'')  
     continue;  
1300    
   count++;  
1301    
1302    if (name == NULL && count == lorn) return count;  
1303    term = *ptr++;  /*************************************************
1304    if (term == '<') term = '>';  *       Find forward referenced subpattern       *
1305    thisname = ptr;  *************************************************/
1306    while (*ptr != term) ptr++;  
1307    if (name != NULL && lorn == ptr - thisname &&  /* This function scans along a pattern's text looking for capturing
1308        strncmp((const char *)name, (const char *)thisname, lorn) == 0)  subpatterns, and counting them. If it finds a named pattern that matches the
1309      return count;  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    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    for (;;)
1340      {
1341      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, &count);
1342      if (rc > 0 || *ptr++ == 0) break;
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 1100  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 1116  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 1145  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 1152  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 1175  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 1188  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 1211  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 1225  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 1252  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 1306  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 1334  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 1345  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 1369  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 1400  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 1438  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 1462  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 1493  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 1508  for (;;) Line 1865  for (;;)
1865  can match the empty string or not. It is called from could_be_empty()  can match the empty string or not. It is called from could_be_empty()
1866  below and from compile_branch() when checking for an unlimited repeat of a  below and from compile_branch() when checking for an unlimited repeat of a
1867  group that can match nothing. Note that first_significant_code() skips over  group that can match nothing. Note that first_significant_code() skips over
1868  assertions. If we hit an unclosed bracket, we return "empty" - this means we've  backward and negative forward assertions when its final argument is TRUE. If we
1869  struck an inner bracket whose current branch will already have been scanned.  hit an unclosed bracket, we return "empty" - this means we've struck an inner
1870    bracket whose current branch will already have been scanned.
1871    
1872  Arguments:  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 1531  for (code = first_significant_code(code Line 1891  for (code = first_significant_code(code
1891    
1892    c = *code;    c = *code;
1893    
1894      /* Skip over forward assertions; the other assertions are skipped by
1895      first_significant_code() with a TRUE final argument. */
1896    
1897      if (c == OP_ASSERT)
1898        {
1899        do code += GET(code, 1); while (*code == OP_ALT);
1900        c = *code;
1901        continue;
1902        }
1903    
1904    /* Groups with zero repeats can of course be empty; skip them. */    /* Groups with zero repeats can of course be empty; skip them. */
1905    
1906    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)
1907      {      {
1908      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1909      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
# Line 1541  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 1548  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 1619  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 1675  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 1703  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 1726  return TRUE; Line 2147  return TRUE;
2147  *************************************************/  *************************************************/
2148    
2149  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2150  encountered in a character class. It checks whether this is followed by an  encountered in a character class. It checks whether this is followed by a
2151  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2152  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2153    
2154    Originally, this function only recognized a sequence of letters between the
2155    terminators, but it seems that Perl recognizes any sequence of characters,
2156    though of course unknown POSIX names are subsequently rejected. Perl gives an
2157    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2158    didn't consider this to be a POSIX class. Likewise for [:1234:].
2159    
2160    The problem in trying to be exactly like Perl is in the handling of escapes. We
2161    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2162    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2163    below handles the special case of \], but does not try to do any other escape
2164    processing. This makes it different from Perl for cases such as [:l\ower:]
2165    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2166    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2167    I think.
2168    
2169  Argument:  Arguments:
2170    ptr      pointer to the initial [    ptr      pointer to the initial [
2171    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2172    
2173  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2174  */  */
2175    
2176  static BOOL  static BOOL
2177  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
2178  {  {
2179  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2180  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2181  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2182    {    {
2183    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else
2184    return TRUE;      {
2185        if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2186        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2187          {
2188          *endptr = ptr;
2189          return TRUE;
2190          }
2191        }
2192    }    }
2193  return FALSE;  return FALSE;
2194  }  }
# Line 1794  return -1; Line 2234  return -1;
2234  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2235  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2236  earlier groups that are outside the current group). However, when a group is  earlier groups that are outside the current group). However, when a group is
2237  optional (i.e. the minimum quantifier is zero), OP_BRAZERO is inserted before  optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is
2238  it, after it has been compiled. This means that any OP_RECURSE items within it  inserted before it, after it has been compiled. This means that any OP_RECURSE
2239  that refer to the group itself or any contained groups have to have their  items within it that refer to the group itself or any contained groups have to
2240  offsets adjusted. That one of the jobs of this function. Before it is called,  have their offsets adjusted. That one of the jobs of this function. Before it
2241  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2242    OP_END.
2243    
2244  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2245  recursions and subroutine calls. It must also check the list of such references  recursions and subroutine calls. It must also check the list of such references
# Line 1933  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 1942  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 1988  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 2000  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 2025  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 2036  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 2052  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 2070  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 2083  if (next >= 0) switch(op_code) Line 2527  if (next >= 0) switch(op_code)
2527    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT, "item" must be a single-byte character. */
2528    
2529    case OP_NOT:    case OP_NOT:
   if (next < 0) return FALSE;  /* Not a character */  
2530    if (item == next) return TRUE;    if (item == next) return TRUE;
2531    if ((options & PCRE_CASELESS) == 0) return FALSE;    if ((options & PCRE_CASELESS) == 0) return FALSE;
2532  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 2092  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 2101  if (next >= 0) switch(op_code) Line 2544  if (next >= 0) switch(op_code)
2544    else    else
2545  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2546    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */
2547    
2548      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2549      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
2550    
2551    case OP_DIGIT:    case OP_DIGIT:
2552    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
# Line 2143  if (next >= 0) switch(op_code) Line 2589  if (next >= 0) switch(op_code)
2589      case 0x202f:      case 0x202f:
2590      case 0x205f:      case 0x205f:
2591      case 0x3000:      case 0x3000:
2592      return op_code != OP_HSPACE;      return op_code == OP_NOT_HSPACE;
2593      default:      default:
2594      return op_code == OP_HSPACE;      return op_code != OP_NOT_HSPACE;
2595      }      }
2596    
2597      case OP_ANYNL:
2598    case OP_VSPACE:    case OP_VSPACE:
2599    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
2600    switch(next)    switch(next)
# Line 2159  if (next >= 0) switch(op_code) Line 2606  if (next >= 0) switch(op_code)
2606      case 0x85:      case 0x85:
2607      case 0x2028:      case 0x2028:
2608      case 0x2029:      case 0x2029:
2609      return op_code != OP_VSPACE;      return op_code == OP_NOT_VSPACE;
2610      default:      default:
2611      return op_code == OP_VSPACE;      return op_code != OP_NOT_VSPACE;
2612      }      }
2613    
2614    default:    default:
# Line 2169  if (next >= 0) switch(op_code) Line 2616  if (next >= 0) switch(op_code)
2616    }    }
2617    
2618    
2619  /* Handle the case when the next item is \d, \s, etc. */  /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP
2620    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
2621    generated only when PCRE_UCP is *not* set, that is, when only ASCII
2622    characteristics are recognized. */
2623    
2624  switch(op_code)  switch(op_code)
2625    {    {
# Line 2248  switch(op_code) Line 2698  switch(op_code)
2698    
2699    case OP_DIGIT:    case OP_DIGIT:
2700    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2701           next == -ESC_h || next == -ESC_v;           next == -ESC_h || next == -ESC_v || next == -ESC_R;
2702    
2703    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
2704    return next == -ESC_d;    return next == -ESC_d;
2705    
2706    case OP_WHITESPACE:    case OP_WHITESPACE:
2707    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
2708    
2709    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
2710    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2711    
2712    case OP_HSPACE:    case OP_HSPACE:
2713    return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
2714             next == -ESC_w || next == -ESC_v || next == -ESC_R;
2715    
2716    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
2717    return next == -ESC_h;    return next == -ESC_h;
2718    
2719    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
2720      case OP_ANYNL:
2721    case OP_VSPACE:    case OP_VSPACE:
2722    return next == -ESC_V || next == -ESC_d || next == -ESC_w;    return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2723    
2724    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
2725    return next == -ESC_v;    return next == -ESC_v || next == -ESC_R;
2726    
2727    case OP_WORDCHAR:    case OP_WORDCHAR:
2728    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
2729             next == -ESC_v || next == -ESC_R;
2730    
2731    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
2732    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2337  BOOL inescq = FALSE; Line 2790  BOOL inescq = FALSE;
2790  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
2791  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
2792  const uschar *tempptr;  const uschar *tempptr;
2793    const uschar *nestptr = NULL;
2794  uschar *previous = NULL;  uschar *previous = NULL;
2795  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
2796  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
# Line 2346  uschar classbits[32]; Line 2800  uschar classbits[32];
2800  BOOL class_utf8;  BOOL class_utf8;
2801  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
2802  uschar *class_utf8data;  uschar *class_utf8data;
2803    uschar *class_utf8data_base;
2804  uschar utf8_char[6];  uschar utf8_char[6];
2805  #else  #else
2806  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
2807  uschar *utf8_char = NULL;  uschar *utf8_char = NULL;
2808  #endif  #endif
2809    
2810  #ifdef DEBUG  #ifdef PCRE_DEBUG
2811  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2812  #endif  #endif
2813    
# Line 2385  req_caseopt = ((options & PCRE_CASELESS) Line 2840  req_caseopt = ((options & PCRE_CASELESS)
2840  for (;; ptr++)  for (;; ptr++)
2841    {    {
2842    BOOL negate_class;    BOOL negate_class;
2843    BOOL should_flip_negation;    BOOL should_flip_negation;
2844    BOOL possessive_quantifier;    BOOL possessive_quantifier;
2845    BOOL is_quantifier;    BOOL is_quantifier;
2846    BOOL is_recurse;    BOOL is_recurse;
# Line 2406  for (;; ptr++) Line 2861  for (;; ptr++)
2861    
2862    c = *ptr;    c = *ptr;
2863    
2864      /* If we are at the end of a nested substitution, revert to the outer level
2865      string. Nesting only happens one level deep. */
2866    
2867      if (c == 0 && nestptr != NULL)
2868        {
2869        ptr = nestptr;
2870        nestptr = NULL;
2871        c = *ptr;
2872        }
2873    
2874    /* 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
2875    previous cycle of this loop. */    previous cycle of this loop. */
2876    
2877    if (lengthptr != NULL)    if (lengthptr != NULL)
2878      {      {
2879  #ifdef DEBUG  #ifdef PCRE_DEBUG
2880      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
2881  #endif  #endif
2882      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
2883        {        {
2884        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
2885        goto FAILED;        goto FAILED;
# Line 2463  for (;; ptr++) Line 2928  for (;; ptr++)
2928    /* 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
2929    reference list. */    reference list. */
2930    
2931    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
2932      {      {
2933      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
2934      goto FAILED;      goto FAILED;
# Line 2473  for (;; ptr++) Line 2938  for (;; ptr++)
2938    
2939    if (inescq && c != 0)    if (inescq && c != 0)
2940      {      {
2941      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
2942        {        {
2943        inescq = FALSE;        inescq = FALSE;
2944        ptr++;        ptr++;
# Line 2499  for (;; ptr++) Line 2964  for (;; ptr++)
2964    /* 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
2965    a quantifier. */    a quantifier. */
2966    
2967    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
2968      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
2969        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
2970    
2971    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
2972         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2515  for (;; ptr++) Line 2981  for (;; ptr++)
2981    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
2982      {      {
2983      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
2984      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
2985        {        {
2986        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2987          {          {
# Line 2540  for (;; ptr++) Line 3006  for (;; ptr++)
3006      {      {
3007      /* ===================================================================*/      /* ===================================================================*/
3008      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
3009      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
3010      case ')':      case CHAR_RIGHT_PARENTHESIS:
3011      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
3012      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
3013      *codeptr = code;      *codeptr = code;
# Line 2563  for (;; ptr++) Line 3029  for (;; ptr++)
3029      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
3030      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3031    
3032      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
3033      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3034        {        {
3035        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
# Line 2572  for (;; ptr++) Line 3038  for (;; ptr++)
3038      *code++ = OP_CIRC;      *code++ = OP_CIRC;
3039      break;      break;
3040    
3041      case '$':      case CHAR_DOLLAR_SIGN:
3042      previous = NULL;      previous = NULL;
3043      *code++ = OP_DOLL;      *code++ = OP_DOLL;
3044      break;      break;
# Line 2580  for (;; ptr++) Line 3046  for (;; ptr++)
3046      /* 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
3047      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
3048    
3049      case '.':      case CHAR_DOT:
3050      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3051      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
3052      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3053      previous = code;      previous = code;
3054      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
3055      break;      break;
3056    
3057    
# Line 2600  for (;; ptr++) Line 3066  for (;; ptr++)
3066      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,
3067      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
3068      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.
     */  
3069    
3070      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
3071        default (Perl) mode, it is treated as a data character. */
3072    
3073        case CHAR_RIGHT_SQUARE_BRACKET:
3074        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3075          {
3076          *errorcodeptr = ERR64;
3077          goto FAILED;
3078          }
3079        goto NORMAL_CHAR;
3080    
3081        case CHAR_LEFT_SQUARE_BRACKET:
3082      previous = code;      previous = code;
3083    
3084      /* 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
3085      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. */
3086    
3087      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3088          check_posix_syntax(ptr, &tempptr, cd))           ptr[1] == CHAR_EQUALS_SIGN) &&
3089            check_posix_syntax(ptr, &tempptr))
3090        {        {
3091        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
3092        goto FAILED;        goto FAILED;
3093        }        }
3094    
# Line 2623  for (;; ptr++) Line 3100  for (;; ptr++)
3100      for (;;)      for (;;)
3101        {        {
3102        c = *(++ptr);        c = *(++ptr);
3103        if (c == '\\')        if (c == CHAR_BACKSLASH)
3104          {          {
3105          if (ptr[1] == 'E') ptr++;          if (ptr[1] == CHAR_E)
3106            else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;            ptr++;
3107              else break;          else if (strncmp((const char *)ptr+1,
3108                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
3109              ptr += 3;
3110            else
3111              break;
3112          }          }
3113        else if (!negate_class && c == '^')        else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
3114          negate_class = TRUE;          negate_class = TRUE;
3115        else break;        else break;
3116        }        }
3117    
3118      /* If a class contains a negative special such as \S, we need to flip the      /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
3119      negation flag at the end, so that support for characters > 255 works      an initial ']' is taken as a data character -- the code below handles
3120        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3121        [^] must match any character, so generate OP_ALLANY. */
3122    
3123        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3124            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3125          {
3126          *code++ = negate_class? OP_ALLANY : OP_FAIL;
3127          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3128          zerofirstbyte = firstbyte;
3129          break;
3130          }
3131    
3132        /* If a class contains a negative special such as \S, we need to flip the
3133        negation flag at the end, so that support for characters > 255 works
3134      correctly (they are all included in the class). */      correctly (they are all included in the class). */
3135    
3136      should_flip_negation = FALSE;      should_flip_negation = FALSE;
# Line 2657  for (;; ptr++) Line 3152  for (;; ptr++)
3152  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3153      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
3154      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
3155        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
3156  #endif  #endif
3157    
3158      /* Process characters until ] is reached. By writing this as a "do" it      /* Process characters until ] is reached. By writing this as a "do" it
# Line 2672  for (;; ptr++) Line 3168  for (;; ptr++)
3168          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3169          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3170          }          }
3171    
3172          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
3173          data and reset the pointer. This is so that very large classes that
3174          contain a zillion UTF-8 characters no longer overwrite the work space
3175          (which is on the stack). */
3176    
3177          if (lengthptr != NULL)
3178            {
3179            *lengthptr += class_utf8data - class_utf8data_base;
3180            class_utf8data = class_utf8data_base;
3181            }
3182    
3183  #endif  #endif
3184    
3185        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
3186    
3187        if (inescq)        if (inescq)
3188          {          {
3189          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3190            {            {
3191            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
3192            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2693  for (;; ptr++) Line 3201  for (;; ptr++)
3201        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3202        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3203    
3204        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3205            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3206            check_posix_syntax(ptr, &tempptr, cd))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3207          {          {
3208          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3209          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3210          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3211          uschar pbits[32];          uschar pbits[32];
3212    
3213          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3214            {            {
3215            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3216            goto FAILED;            goto FAILED;
3217            }            }
3218    
3219          ptr += 2;          ptr += 2;
3220          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3221            {            {
3222            local_negate = TRUE;            local_negate = TRUE;
3223            should_flip_negation = TRUE;  /* Note negative special */            should_flip_negation = TRUE;  /* Note negative special */
3224            ptr++;            ptr++;
3225            }            }
3226    
# Line 2729  for (;; ptr++) Line 3237  for (;; ptr++)
3237    
3238          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3239            posix_class = 0;            posix_class = 0;
3240    
3241          /* 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
3242          because we may be adding and subtracting from it, and we don't want to          different escape sequences that use Unicode properties. */
3243          subtract bits that may be in the main map already. At the end we or the  
3244          result into the bit map that is being built. */  #ifdef SUPPORT_UCP
3245            if ((options & PCRE_UCP) != 0)
3246              {
3247              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3248              if (posix_substitutes[pc] != NULL)
3249                {
3250                nestptr = tempptr + 1;
3251                ptr = posix_substitutes[pc] - 1;
3252                continue;
3253                }
3254              }
3255    #endif
3256            /* In the non-UCP case, we build the bit map for the POSIX class in a
3257            chunk of local store because we may be adding and subtracting from it,
3258            and we don't want to subtract bits that may be in the main map already.
3259            At the end we or the result into the bit map that is being built. */
3260    
3261          posix_class *= 3;          posix_class *= 3;
3262    
# Line 2777  for (;; ptr++) Line 3300  for (;; ptr++)
3300    
3301        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3302        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
3303        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
3304        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
3305        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
3306        character in them, so set class_charcount bigger than one. */        are either treated as literal characters (by default), or are faulted if
3307          PCRE_EXTRA is set. */
3308    
3309        if (c == '\\')        if (c == CHAR_BACKSLASH)
3310          {          {
3311          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3312          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3313    
3314          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 */  
3315          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3316            {            {
3317            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3318              {              {
3319              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3320              }              }
# Line 2806  for (;; ptr++) Line 3328  for (;; ptr++)
3328            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
3329            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3330    
3331            /* Save time by not doing this in the pre-compile phase. */            switch (-c)
   
           if (lengthptr == NULL) switch (-c)  
3332              {              {
3333    #ifdef SUPPORT_UCP
3334                case ESC_du:     /* These are the values given for \d etc */
3335                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3336                case ESC_wu:     /* escape sequence with an appropriate \p */
3337                case ESC_WU:     /* or \P to test Unicode properties instead */
3338                case ESC_su:     /* of the default ASCII testing. */
3339                case ESC_SU:
3340                nestptr = ptr;
3341                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3342                class_charcount -= 2;                /* Undo! */
3343                continue;
3344    #endif
3345              case ESC_d:              case ESC_d:
3346              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3347              continue;              continue;
3348    
3349              case ESC_D:              case ESC_D:
3350              should_flip_negation = TRUE;              should_flip_negation = TRUE;
3351              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
3352              continue;              continue;
3353    
# Line 2824  for (;; ptr++) Line 3356  for (;; ptr++)
3356              continue;              continue;
3357    
3358              case ESC_W:              case ESC_W:
3359              should_flip_negation = TRUE;              should_flip_negation = TRUE;
3360              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3361              continue;              continue;
3362    
# Line 2834  for (;; ptr++) Line 3366  for (;; ptr++)
3366              continue;              continue;
3367    
3368              case ESC_S:              case ESC_S:
3369              should_flip_negation = TRUE;              should_flip_negation = TRUE;
3370              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
3371              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3372              continue;              continue;
3373    
3374              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)  
             {  
3375              SETBIT(classbits, 0x09); /* VT */              SETBIT(classbits, 0x09); /* VT */
3376              SETBIT(classbits, 0x20); /* SPACE */              SETBIT(classbits, 0x20); /* SPACE */
3377              SETBIT(classbits, 0xa0); /* NSBP */              SETBIT(classbits, 0xa0); /* NSBP */
# Line 2876  for (;; ptr++) Line 3395  for (;; ptr++)
3395                }                }
3396  #endif  #endif
3397              continue;              continue;
             }  
3398    
3399            if (-c == ESC_H)              case ESC_H:
             {  
3400              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3401                {                {
3402                int x = 0xff;                int x = 0xff;
# Line 2921  for (;; ptr++) Line 3438  for (;; ptr++)
3438                }                }
3439  #endif  #endif
3440              continue;              continue;
             }  
3441    
3442            if (-c == ESC_v)              case ESC_v:
             {  
3443              SETBIT(classbits, 0x0a); /* LF */              SETBIT(classbits, 0x0a); /* LF */
3444              SETBIT(classbits, 0x0b); /* VT */              SETBIT(classbits, 0x0b); /* VT */
3445              SETBIT(classbits, 0x0c); /* FF */              SETBIT(classbits, 0x0c); /* FF */
# Line 2940  for (;; ptr++) Line 3455  for (;; ptr++)
3455                }                }
3456  #endif  #endif
3457              continue;              continue;
             }  
3458    
3459            if (-c == ESC_V)              case ESC_V:
             {  
3460              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3461                {                {
3462                int x = 0xff;                int x = 0xff;
# Line 2973  for (;; ptr++) Line 3486  for (;; ptr++)
3486                }                }
3487  #endif  #endif
3488              continue;              continue;
             }  
   
           /* We need to deal with \P and \p in both phases. */  
3489    
3490  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3491            if (-c == ESC_p || -c == ESC_P)              case ESC_p:
3492              {              case ESC_P:
3493              BOOL negated;                {
3494              int pdata;                BOOL negated;
3495              int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                int pdata;
3496              if (ptype < 0) goto FAILED;                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3497              class_utf8 = TRUE;                if (ptype < 0) goto FAILED;
3498              *class_utf8data++ = ((-c == ESC_p) != negated)?                class_utf8 = TRUE;
3499                XCL_PROP : XCL_NOTPROP;                *class_utf8data++ = ((-c == ESC_p) != negated)?
3500              *class_utf8data++ = ptype;                  XCL_PROP : XCL_NOTPROP;
3501              *class_utf8data++ = pdata;                *class_utf8data++ = ptype;
3502              class_charcount -= 2;   /* Not a < 256 character */                *class_utf8data++ = pdata;
3503              continue;                class_charcount -= 2;   /* Not a < 256 character */
3504              }                continue;
3505                  }
3506  #endif  #endif
3507            /* Unrecognized escapes are faulted if PCRE is running in its              /* Unrecognized escapes are faulted if PCRE is running in its
3508            strict mode. By default, for compatibility with Perl, they are              strict mode. By default, for compatibility with Perl, they are
3509            treated as literals. */              treated as literals. */
3510    
3511            if ((options & PCRE_EXTRA) != 0)              default:
3512              {              if ((options & PCRE_EXTRA) != 0)
3513              *errorcodeptr = ERR7;                {
3514              goto FAILED;                *errorcodeptr = ERR7;
3515                  goto FAILED;
3516                  }
3517                class_charcount -= 2;  /* Undo the default count from above */
3518                c = *ptr;              /* Get the final character and fall through */
3519                break;
3520              }              }
   
           class_charcount -= 2;  /* Undo the default count from above */  
           c = *ptr;              /* Get the final character and fall through */  
3521            }            }
3522    
3523          /* 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 3018  for (;; ptr++) Line 3531  for (;; ptr++)
3531        entirely. The code for handling \Q and \E is messy. */        entirely. The code for handling \Q and \E is messy. */
3532    
3533        CHECK_RANGE:        CHECK_RANGE:
3534        while (ptr[1] == '\\' && ptr[2] == 'E')        while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3535          {          {
3536          inescq = FALSE;          inescq = FALSE;
3537          ptr += 2;          ptr += 2;
# Line 3028  for (;; ptr++) Line 3541  for (;; ptr++)
3541    
3542        /* Remember \r or \n */        /* Remember \r or \n */
3543    
3544        if (c == '\r' || c == '\n') cd->external_flags |= PCRE_HASCRORLF;        if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3545    
3546        /* Check for range */        /* Check for range */
3547    
3548        if (!inescq && ptr[1] == '-')        if (!inescq && ptr[1] == CHAR_MINUS)
3549          {          {
3550          int d;          int d;
3551          ptr += 2;          ptr += 2;
3552          while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3553    
3554          /* 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
3555          mode. */          mode. */
3556    
3557          while (*ptr == '\\' && ptr[1] == 'Q')          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3558            {            {
3559            ptr += 2;            ptr += 2;
3560            if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }            if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3561                { ptr += 2; continue; }
3562            inescq = TRUE;            inescq = TRUE;
3563            break;            break;
3564            }            }
3565    
3566          if (*ptr == 0 || (!inescq && *ptr == ']'))          if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3567            {            {
3568            ptr = oldptr;            ptr = oldptr;
3569            goto LONE_SINGLE_CHARACTER;            goto LONE_SINGLE_CHARACTER;
# Line 3068  for (;; ptr++) Line 3582  for (;; ptr++)
3582          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
3583          in such circumstances. */          in such circumstances. */
3584    
3585          if (!inescq && d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3586            {            {
3587            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3588            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3589    
3590            /* \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 */  
3591    
3592            if (d < 0)            if (d < 0)
3593              {              {
3594              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  
3595                {                {
3596                ptr = oldptr;                ptr = oldptr;
3597                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 3101  for (;; ptr++) Line 3612  for (;; ptr++)
3612    
3613          /* Remember \r or \n */          /* Remember \r or \n */
3614    
3615          if (d == '\r' || d == '\n') cd->external_flags |= PCRE_HASCRORLF;          if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3616    
3617          /* 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
3618          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 3221  for (;; ptr++) Line 3732  for (;; ptr++)
3732          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3733            {            {
3734            unsigned int othercase;            unsigned int othercase;
3735            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)            if ((othercase = UCD_OTHERCASE(c)) != c)
3736              {              {
3737              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3738              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 3246  for (;; ptr++) Line 3757  for (;; ptr++)
3757          }          }
3758        }        }
3759    
3760      /* 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.
3761        If we are at the end of an internal nested string, revert to the outer
3762        string. */
3763    
3764        while (((c = *(++ptr)) != 0 ||
3765               (nestptr != NULL &&
3766                 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != 0)) &&
3767               (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
3768    
3769      while ((c = *(++ptr)) != 0 && (c != ']' || inescq));      /* Check for missing terminating ']' */
3770    
3771      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)
3772        {        {
3773        *errorcodeptr = ERR6;        *errorcodeptr = ERR6;
3774        goto FAILED;        goto FAILED;
3775        }        }
3776    
   
 /* 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  
   
   
3777      /* 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
3778      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
3779      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 3291  we set the flag only if there is a liter Line 3790  we set the flag only if there is a liter
3790      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
3791      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
3792      reqbyte, save the previous value for reinstating. */      reqbyte, save the previous value for reinstating. */
3793    
3794  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3795      if (class_charcount == 1 && !class_utf8 &&      if (class_charcount == 1 && !class_utf8 &&
3796        (!utf8 || !negate_class || class_lastchar < 128))        (!utf8 || !negate_class || class_lastchar < 128))
# Line 3337  we set the flag only if there is a liter Line 3836  we set the flag only if there is a liter
3836      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3837    
3838      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
3839      extended class, with its own opcode, unless there was a negated special      extended class, with its own opcode, unless there was a negated special
3840      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
3841      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
3842      (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
3843      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
3844        actual compiled code. */
3845    
3846  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3847      if (class_utf8 && !should_flip_negation)      if (class_utf8 && (!should_flip_negation || (options & PCRE_UCP) != 0))
3848        {        {
3849        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
3850        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 3370  we set the flag only if there is a liter Line 3870  we set the flag only if there is a liter
3870        }        }
3871  #endif  #endif
3872    
3873      /* 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
3874      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
3875      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
3876      map into the code vector, negating it if necessary. */      (non-UCP) in the class. Then copy the 32-byte map into the code vector,
3877        negating it if necessary. */
3878    
3879      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
3880      if (negate_class)      if (negate_class)
3881        {        {
# Line 3393  we set the flag only if there is a liter Line 3894  we set the flag only if there is a liter
3894      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
3895      has been tested above. */      has been tested above. */
3896    
3897      case '{':      case CHAR_LEFT_CURLY_BRACKET:
3898      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
3899      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
3900      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
3901      goto REPEAT;      goto REPEAT;
3902    
3903      case '*':      case CHAR_ASTERISK:
3904      repeat_min = 0;      repeat_min = 0;
3905      repeat_max = -1;      repeat_max = -1;
3906      goto REPEAT;      goto REPEAT;
3907    
3908      case '+':      case CHAR_PLUS:
3909      repeat_min = 1;      repeat_min = 1;
3910      repeat_max = -1;      repeat_max = -1;
3911      goto REPEAT;      goto REPEAT;
3912    
3913      case '?':      case CHAR_QUESTION_MARK:
3914      repeat_min = 0;      repeat_min = 0;
3915      repeat_max = 1;      repeat_max = 1;
3916    
# Line 3444  we set the flag only if there is a liter Line 3945  we set the flag only if there is a liter
3945      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
3946      repeat type to the non-default. */      repeat type to the non-default. */
3947    
3948      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
3949        {        {
3950        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
3951        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
3952        ptr++;        ptr++;
3953        }        }
3954      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
3955        {        {
3956        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
3957        ptr++;        ptr++;
# Line 3565  we set the flag only if there is a liter Line 4066  we set the flag only if there is a liter
4066    
4067        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
4068    
4069          /*--------------------------------------------------------------------*/
4070          /* This code is obsolete from release 8.00; the restriction was finally
4071          removed: */
4072    
4073        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4074        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4075    
4076        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4077          /*--------------------------------------------------------------------*/
4078    
4079        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
4080    
# Line 3715  we set the flag only if there is a liter Line 4221  we set the flag only if there is a liter
4221          goto END_REPEAT;          goto END_REPEAT;
4222          }          }
4223    
4224          /*--------------------------------------------------------------------*/
4225          /* This code is obsolete from release 8.00; the restriction was finally
4226          removed: */
4227    
4228        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4229        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4230    
4231        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4232          /*--------------------------------------------------------------------*/
4233    
4234        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4235          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 3776  we set the flag only if there is a liter Line 4287  we set the flag only if there is a liter
4287    
4288        if (repeat_min == 0)        if (repeat_min == 0)
4289          {          {
4290          /* If the maximum is also zero, we just omit the group from the output          /* If the maximum is also zero, we used to just omit the group from the
4291          altogether. */          output altogether, like this:
   
         if (repeat_max == 0)  
           {  
           code = previous;  
           goto END_REPEAT;  
           }  
4292    
4293          /* If the maximum is 1 or unlimited, we just have to stick in the          ** if (repeat_max == 0)
4294          BRAZERO and do no more at this point. However, we do need to adjust          **   {
4295          any OP_RECURSE calls inside the group that refer to the group itself or          **   code = previous;
4296          any internal or forward referenced group, because the offset is from          **   goto END_REPEAT;
4297          the start of the whole regex. Temporarily terminate the pattern while          **   }
4298          doing this. */  
4299            However, that fails when a group is referenced as a subroutine from
4300            elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it
4301            so that it is skipped on execution. As we don't have a list of which
4302            groups are referenced, we cannot do this selectively.
4303    
4304            If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4305            and do no more at this point. However, we do need to adjust any
4306            OP_RECURSE calls inside the group that refer to the group itself or any
4307            internal or forward referenced group, because the offset is from the
4308            start of the whole regex. Temporarily terminate the pattern while doing
4309            this. */
4310    
4311          if (repeat_max <= 1)          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
4312            {            {
4313            *code = OP_END;            *code = OP_END;
4314            adjust_recurse(previous, 1, utf8, cd, save_hwm);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
4315            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
4316            code++;            code++;
4317              if (repeat_max == 0)
4318                {
4319                *previous++ = OP_SKIPZERO;
4320                goto END_REPEAT;
4321                }
4322            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4323            }            }
4324    
# Line 3843  we set the flag only if there is a liter Line 4364  we set the flag only if there is a liter
4364            {            {
4365            /* 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
4366            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
4367            potential integer overflow. */            potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
4368              integer type when available, otherwise double. */
4369    
4370            if (lengthptr != NULL)            if (lengthptr != NULL)
4371              {              {
4372              int delta = (repeat_min - 1)*length_prevgroup;              int delta = (repeat_min - 1)*length_prevgroup;
4373              if ((double)(repeat_min - 1)*(double)length_prevgroup >              if ((INT64_OR_DOUBLE)(repeat_min - 1)*
4374                                                              (double)INT_MAX ||                    (INT64_OR_DOUBLE)length_prevgroup >
4375                        (INT64_OR_DOUBLE)INT_MAX ||
4376                  OFLOW_MAX - *lengthptr < delta)                  OFLOW_MAX - *lengthptr < delta)
4377                {                {
4378                *errorcodeptr = ERR20;                *errorcodeptr = ERR20;
# Line 3895  we set the flag only if there is a liter Line 4418  we set the flag only if there is a liter
4418          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
4419          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
4420          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
4421          paranoid checks to avoid integer overflow. */          paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
4422            a 64-bit integer type when available, otherwise double. */
4423    
4424          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4425            {            {
4426            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4427                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4428            if ((double)repeat_max *            if ((INT64_OR_DOUBLE)repeat_max *
4429                  (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)                  (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4430                    > (double)INT_MAX ||                    > (INT64_OR_DOUBLE)INT_MAX ||
4431                OFLOW_MAX - *lengthptr < delta)                OFLOW_MAX - *lengthptr < delta)
4432              {              {
4433              *errorcodeptr = ERR20;              *errorcodeptr = ERR20;
# Line 3980  we set the flag only if there is a liter Line 4504  we set the flag only if there is a liter
4504            uschar *scode = bracode;            uschar *scode = bracode;
4505            do            do
4506              {              {
4507              if (could_be_empty_branch(scode, ketcode, utf8))              if (could_be_empty_branch(scode, ketcode, utf8, cd))
4508                {                {
4509                *bracode += OP_SBRA - OP_BRA;                *bracode += OP_SBRA - OP_BRA;
4510                break;                break;
# Line 3992  we set the flag only if there is a liter Line 4516  we set the flag only if there is a liter
4516          }          }
4517        }        }
4518    
4519        /* If previous is OP_FAIL, it was generated by an empty class [] in
4520        JavaScript mode. The other ways in which OP_FAIL can be generated, that is
4521        by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
4522        error above. We can just ignore the repeat in JS case. */
4523    
4524        else if (*previous == OP_FAIL) goto END_REPEAT;
4525    
4526      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
4527    
4528      else      else
# Line 4016  we set the flag only if there is a liter Line 4547  we set the flag only if there is a liter
4547      if (possessive_quantifier)      if (possessive_quantifier)
4548        {        {
4549        int len;        int len;
4550        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||  
4551            *tempcode == OP_NOTEXACT)        if (*tempcode == OP_TYPEEXACT)
4552            tempcode += _pcre_OP_lengths[*tempcode] +
4553              ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
4554    
4555          else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
4556            {
4557          tempcode += _pcre_OP_lengths[*tempcode];          tempcode += _pcre_OP_lengths[*tempcode];
4558    #ifdef SUPPORT_UTF8
4559            if (utf8 && tempcode[-1] >= 0xc0)
4560              tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];
4561    #endif
4562            }
4563    
4564        len = code - tempcode;        len = code - tempcode;
4565        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4566          {          {
# Line 4037  we set the flag only if there is a liter Line 4579  we set the flag only if there is a liter
4579          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4580          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4581    
4582            /* Because we are moving code along, we must ensure that any
4583            pending recursive references are updated. */
4584    
4585          default:          default:
4586            *code = OP_END;
4587            adjust_recurse(tempcode, 1 + LINK_SIZE, utf8, cd, save_hwm);
4588          memmove(tempcode + 1+LINK_SIZE, tempcode, len);          memmove(tempcode + 1+LINK_SIZE, tempcode, len);
4589          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
4590          len += 1 + LINK_SIZE;          len += 1 + LINK_SIZE;
# Line 4064  we set the flag only if there is a liter Line 4611  we set the flag only if there is a liter
4611      lookbehind or option setting or condition or all the other extended      lookbehind or option setting or condition or all the other extended
4612      parenthesis forms.  */      parenthesis forms.  */
4613    
4614      case '(':      case CHAR_LEFT_PARENTHESIS:
4615      newoptions = options;      newoptions = options;
4616      skipbytes = 0;      skipbytes = 0;
4617      bravalue = OP_CBRA;      bravalue = OP_CBRA;
# Line 4073  we set the flag only if there is a liter Line 4620  we set the flag only if there is a liter
4620    
4621      /* First deal with various "verbs" that can be introduced by '*'. */      /* First deal with various "verbs" that can be introduced by '*'. */
4622    
4623      if (*(++ptr) == '*' && (cd->ctypes[ptr[1]] & ctype_letter) != 0)      if (*(++ptr) == CHAR_ASTERISK &&
4624             ((cd->ctypes[ptr[1]] & ctype_letter) != 0 || ptr[1] == ':'))
4625        {        {
4626        int i, namelen;        int i, namelen;
4627          int arglen = 0;
4628        const char *vn = verbnames;        const char *vn = verbnames;
4629        const uschar *name = ++ptr;        const uschar *name = ptr + 1;
4630          const uschar *arg = NULL;
4631        previous = NULL;        previous = NULL;
4632        while ((cd->ctypes[*++ptr] & ctype_letter) != 0);        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
4633        if (*ptr == ':')        namelen = ptr - name;
4634    
4635          if (*ptr == CHAR_COLON)
4636          {          {
4637          *errorcodeptr = ERR59;   /* Not supported */          arg = ++ptr;
4638          goto FAILED;          while ((cd->ctypes[*ptr] & (ctype_letter|ctype_digit)) != 0
4639              || *ptr == '_') ptr++;
4640            arglen = ptr - arg;
4641          }          }
4642        if (*ptr != ')')  
4643          if (*ptr != CHAR_RIGHT_PARENTHESIS)
4644          {          {
4645          *errorcodeptr = ERR60;          *errorcodeptr = ERR60;
4646          goto FAILED;          goto FAILED;
4647          }          }
4648        namelen = ptr - name;  
4649          /* Scan the table of verb names */
4650    
4651        for (i = 0; i < verbcount; i++)        for (i = 0; i < verbcount; i++)
4652          {          {
4653          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
4654              strncmp((char *)name, vn, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
4655            {            {
4656            *code = verbs[i].op;            /* Check for open captures before ACCEPT */
4657            if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;  
4658            break;            if (verbs[i].op == OP_ACCEPT)
4659                {
4660                open_capitem *oc;
4661                cd->had_accept = TRUE;
4662                for (oc = cd->open_caps; oc != NULL; oc = oc->next)
4663                  {
4664                  *code++ = OP_CLOSE;
4665                  PUT2INC(code, 0, oc->number);
4666                  }
4667                }
4668    
4669              /* Handle the cases with/without an argument */
4670    
4671              if (arglen == 0)
4672                {
4673                if (verbs[i].op < 0)   /* Argument is mandatory */
4674                  {
4675                  *errorcodeptr = ERR66;
4676                  goto FAILED;
4677                  }
4678                *code++ = verbs[i].op;
4679                }
4680    
4681              else
4682                {
4683                if (verbs[i].op_arg < 0)   /* Argument is forbidden */
4684                  {
4685                  *errorcodeptr = ERR59;
4686                  goto FAILED;
4687                  }
4688                *code++ = verbs[i].op_arg;
4689                *code++ = arglen;
4690                memcpy(code, arg, arglen);
4691                code += arglen;
4692                *code++ = 0;
4693                }
4694    
4695              break;  /* Found verb, exit loop */
4696            }            }
4697    
4698          vn += verbs[i].len + 1;          vn += verbs[i].len + 1;
4699          }          }
4700        if (i < verbcount) continue;  
4701        *errorcodeptr = ERR60;        if (i < verbcount) continue;    /* Successfully handled a verb */
4702          *errorcodeptr = ERR60;          /* Verb not recognized */
4703        goto FAILED;        goto FAILED;
4704        }        }
4705    
4706      /* Deal with the extended parentheses; all are introduced by '?', and the      /* Deal with the extended parentheses; all are introduced by '?', and the
4707      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. */
4708    
4709      else if (*ptr == '?')      else if (*ptr == CHAR_QUESTION_MARK)
4710        {        {
4711        int i, set, unset, namelen;        int i, set, unset, namelen;
4712        int *optset;        int *optset;
# Line 4119  we set the flag only if there is a liter Line 4715  we set the flag only if there is a liter
4715    
4716        switch (*(++ptr))        switch (*(++ptr))
4717          {          {
4718          case '#':                 /* Comment; skip to ket */          case CHAR_NUMBER_SIGN:                 /* Comment; skip to ket */
4719          ptr++;          ptr++;
4720          while (*ptr != 0 && *ptr != ')') ptr++;          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
4721          if (*ptr == 0)          if (*ptr == 0)
4722            {            {
4723            *errorcodeptr = ERR18;            *errorcodeptr = ERR18;
# Line 4131  we set the flag only if there is a liter Line 4727  we set the flag only if there is a liter
4727    
4728    
4729          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4730          case '|':                 /* Reset capture count for each branch */          case CHAR_VERTICAL_LINE:  /* Reset capture count for each branch */
4731          reset_bracount = TRUE;          reset_bracount = TRUE;
4732          /* Fall through */          /* Fall through */
4733    
4734          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4735          case ':':                 /* Non-capturing bracket */          case CHAR_COLON:          /* Non-capturing bracket */
4736          bravalue = OP_BRA;          bravalue = OP_BRA;
4737          ptr++;          ptr++;
4738          break;          break;
4739    
4740    
4741          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4742          case '(':          case CHAR_LEFT_PARENTHESIS:
4743          bravalue = OP_COND;       /* Conditional group */          bravalue = OP_COND;       /* Conditional group */
4744    
4745          /* 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 4163  we set the flag only if there is a liter Line 4759  we set the flag only if there is a liter
4759          the switch. This will take control down to where bracketed groups,          the switch. This will take control down to where bracketed groups,
4760          including assertions, are processed. */          including assertions, are processed. */
4761    
4762          if (ptr[1] == '?' && (ptr[2] == '=' || ptr[2] == '!' || ptr[2] == '<'))          if (ptr[1] == CHAR_QUESTION_MARK && (ptr[2] == CHAR_EQUALS_SIGN ||
4763                ptr[2] == CHAR_EXCLAMATION_MARK || ptr[2] == CHAR_LESS_THAN_SIGN))
4764            break;            break;
4765    
4766          /* 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 4175  we set the flag only if there is a liter Line 4772  we set the flag only if there is a liter
4772    
4773          /* Check for a test for recursion in a named group. */          /* Check for a test for recursion in a named group. */
4774    
4775          if (ptr[1] == 'R' && ptr[2] == '&')          if (ptr[1] == CHAR_R && ptr[2] == CHAR_AMPERSAND)
4776            {            {
4777            terminator = -1;            terminator = -1;
4778            ptr += 2;            ptr += 2;
# Line 4185  we set the flag only if there is a liter Line 4782  we set the flag only if there is a liter
4782          /* 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
4783          syntax (?(<name>) or (?('name') */          syntax (?(<name>) or (?('name') */
4784    
4785          else if (ptr[1] == '<')          else if (ptr[1] == CHAR_LESS_THAN_SIGN)
4786            {            {
4787            terminator = '>';            terminator = CHAR_GREATER_THAN_SIGN;
4788            ptr++;            ptr++;
4789            }            }
4790          else if (ptr[1] == '\'')          else if (ptr[1] == CHAR_APOSTROPHE)
4791            {            {
4792            terminator = '\'';            terminator = CHAR_APOSTROPHE;
4793            ptr++;            ptr++;
4794            }            }
4795          else          else
4796            {            {
4797            terminator = 0;            terminator = 0;
4798            if (ptr[1] == '-' || ptr[1] == '+') refsign = *(++ptr);            if (ptr[1] == CHAR_MINUS || ptr[1] == CHAR_PLUS) refsign = *(++ptr);
4799            }            }
4800    
4801          /* 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 4218  we set the flag only if there is a liter Line 4815  we set the flag only if there is a liter
4815            {            {
4816            if (recno >= 0)            if (recno >= 0)
4817              recno = ((digitab[*ptr] & ctype_digit) != 0)?              recno = ((digitab[*ptr] & ctype_digit) != 0)?
4818                recno * 10 + *ptr - '0' : -1;                recno * 10 + *ptr - CHAR_0 : -1;
4819            ptr++;            ptr++;
4820            }            }
4821          namelen = ptr - name;          namelen = ptr - name;
4822    
4823          if ((terminator > 0 && *ptr++ != terminator) || *ptr++ != ')')          if ((terminator > 0 && *ptr++ != terminator) ||
4824                *ptr++ != CHAR_RIGHT_PARENTHESIS)
4825            {            {
4826            ptr--;      /* Error offset */            ptr--;      /* Error offset */
4827            *errorcodeptr = ERR26;            *errorcodeptr = ERR26;
# Line 4245  we set the flag only if there is a liter Line 4843  we set the flag only if there is a liter
4843              *errorcodeptr = ERR58;              *errorcodeptr = ERR58;
4844              goto FAILED;              goto FAILED;
4845              }              }
4846            recno = (refsign == '-')?            recno = (refsign == CHAR_MINUS)?
4847              cd->bracount - recno + 1 : recno +cd->bracount;              cd->bracount - recno + 1 : recno +cd->bracount;
4848            if (recno <= 0 || recno > cd->final_bracount)            if (recno <= 0 || recno > cd->final_bracount)
4849              {              {
# Line 4257  we set the flag only if there is a liter Line 4855  we set the flag only if there is a liter
4855            }            }
4856    
4857          /* Otherwise (did not start with "+" or "-"), start by looking for the          /* Otherwise (did not start with "+" or "-"), start by looking for the
4858          name. */          name. If we find a name, add one to the opcode to change OP_CREF or
4859            OP_RREF into OP_NCREF or OP_NRREF. These behave exactly the same,
4860            except they record that the reference was originally to a name. The
4861            information is used to check duplicate names. */
4862    
4863          slot = cd->name_table;          slot = cd->name_table;
4864          for (i = 0; i < cd->names_found; i++)          for (i = 0; i < cd->names_found; i++)
# Line 4272  we set the flag only if there is a liter Line 4873  we set the flag only if there is a liter
4873            {            {
4874            recno = GET2(slot, 0);            recno = GET2(slot, 0);
4875            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4876              code[1+LINK_SIZE]++;
4877            }            }
4878    
4879          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
4880    
4881          else if ((i = find_parens(ptr, cd->bracount, name, namelen,          else if ((i = find_parens(cd, name, namelen,
4882                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0)) > 0)
4883            {            {
4884            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
4885              code[1+LINK_SIZE]++;
4886            }            }
4887    
4888          /* If terminator == 0 it means that the name followed directly after          /* If terminator == 0 it means that the name followed directly after
# Line 4297  we set the flag only if there is a liter Line 4900  we set the flag only if there is a liter
4900          /* Check for (?(R) for recursion. Allow digits after R to specify a          /* Check for (?(R) for recursion. Allow digits after R to specify a
4901          specific group number. */          specific group number. */
4902    
4903          else if (*name == 'R')          else if (*name == CHAR_R)
4904            {            {
4905            recno = 0;            recno = 0;
4906            for (i = 1; i < namelen; i++)            for (i = 1; i < namelen; i++)
# Line 4307  we set the flag only if there is a liter Line 4910  we set the flag only if there is a liter
4910                *errorcodeptr = ERR15;                *errorcodeptr = ERR15;
4911                goto FAILED;                goto FAILED;
4912                }                }
4913              recno = recno * 10 + name[i] - '0';              recno = recno * 10 + name[i] - CHAR_0;
4914              }              }
4915            if (recno == 0) recno = RREF_ANY;            if (recno == 0) recno = RREF_ANY;
4916            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */
# Line 4317  we set the flag only if there is a liter Line 4920  we set the flag only if there is a liter
4920          /* Similarly, check for the (?(DEFINE) "condition", which is always          /* Similarly, check for the (?(DEFINE) "condition", which is always
4921          false. */          false. */
4922    
4923          else if (namelen == 6 && strncmp((char *)name, "DEFINE", 6) == 0)          else if (namelen == 6 && strncmp((char *)name, STRING_DEFINE, 6) == 0)
4924            {            {
4925            code[1+LINK_SIZE] = OP_DEF;            code[1+LINK_SIZE] = OP_DEF;
4926            skipbytes = 1;            skipbytes = 1;
4927            }            }
4928    
4929          /* Check for the "name" actually being a subpattern number. We are          /* Check for the "name" actually being a subpattern number. We are
4930          in the second pass here, so final_bracount is set. */          in the second pass here, so final_bracount is set. */
4931    
4932          else if (recno > 0 && recno <= cd->final_bracount)          else if (recno > 0 && recno <= cd->final_bracount)
4933            {            {
# Line 4342  we set the flag only if there is a liter Line 4945  we set the flag only if there is a liter
4945    
4946    
4947          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4948          case '=':                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
4949          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
4950          ptr++;          ptr++;
4951          break;          break;
4952    
4953    
4954          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4955          case '!':                 /* Negative lookahead */          case CHAR_EXCLAMATION_MARK:            /* Negative lookahead */
4956          ptr++;          ptr++;
4957          if (*ptr == ')')          /* Optimize (?!) */          if (*ptr == CHAR_RIGHT_PARENTHESIS)    /* Optimize (?!) */
4958            {            {
4959            *code++ = OP_FAIL;            *code++ = OP_FAIL;
4960            previous = NULL;            previous = NULL;
# Line 4362  we set the flag only if there is a liter Line 4965  we set the flag only if there is a liter
4965    
4966    
4967          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4968          case '<':                 /* Lookbehind or named define */          case CHAR_LESS_THAN_SIGN:              /* Lookbehind or named define */
4969          switch (ptr[1])          switch (ptr[1])
4970            {            {
4971            case '=':               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
4972            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
4973            ptr += 2;            ptr += 2;
4974            break;            break;
4975    
4976            case '!':               /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
4977            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
4978            ptr += 2;            ptr += 2;
4979            break;            break;
# Line 4385  we set the flag only if there is a liter Line 4988  we set the flag only if there is a liter
4988    
4989    
4990          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4991          case '>':                 /* One-time brackets */          case CHAR_GREATER_THAN_SIGN:           /* One-time brackets */
4992          bravalue = OP_ONCE;          bravalue = OP_ONCE;
4993          ptr++;          ptr++;
4994          break;          break;
4995    
4996    
4997          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4998          case 'C':                 /* Callout - may be followed by digits; */          case CHAR_C:                 /* Callout - may be followed by digits; */
4999          previous_callout = code;  /* Save for later completion */          previous_callout = code;  /* Save for later completion */
5000          after_manual_callout = 1; /* Skip one item before completing */          after_manual_callout = 1; /* Skip one item before completing */
5001          *code++ = OP_CALLOUT;          *code++ = OP_CALLOUT;
5002            {            {
5003            int n = 0;            int n = 0;
5004            while ((digitab[*(++ptr)] & ctype_digit) != 0)            while ((digitab[*(++ptr)] & ctype_digit) != 0)
5005              n = n * 10 + *ptr - '0';              n = n * 10 + *ptr - CHAR_0;
5006            if (*ptr != ')')            if (*ptr != CHAR_RIGHT_PARENTHESIS)
5007              {              {
5008              *errorcodeptr = ERR39;              *errorcodeptr = ERR39;
5009              goto FAILED;              goto FAILED;
# Line 4420  we set the flag only if there is a liter Line 5023  we set the flag only if there is a liter
5023    
5024    
5025          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5026          case 'P':                 /* Python-style named subpattern handling */          case CHAR_P:              /* Python-style named subpattern handling */
5027          if (*(++ptr) == '=' || *ptr == '>')  /* Reference or recursion */          if (*(++ptr) == CHAR_EQUALS_SIGN ||
5028                *ptr == CHAR_GREATER_THAN_SIGN)  /* Reference or recursion */
5029            {            {
5030            is_recurse = *ptr == '>';            is_recurse = *ptr == CHAR_GREATER_THAN_SIGN;
5031            terminator = ')';            terminator = CHAR_RIGHT_PARENTHESIS;
5032            goto NAMED_REF_OR_RECURSE;            goto NAMED_REF_OR_RECURSE;
5033            }            }
5034          else if (*ptr != '<')    /* Test for Python-style definition */          else if (*ptr != CHAR_LESS_THAN_SIGN)  /* Test for Python-style defn */
5035            {            {
5036            *errorcodeptr = ERR41;            *errorcodeptr = ERR41;
5037            goto FAILED;            goto FAILED;
# Line 4437  we set the flag only if there is a liter Line 5041  we set the flag only if there is a liter
5041    
5042          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5043          DEFINE_NAME:    /* Come here from (?< handling */          DEFINE_NAME:    /* Come here from (?< handling */
5044          case '\'':          case CHAR_APOSTROPHE:
5045            {            {
5046            terminator = (*ptr == '<')? '>' : '\'';            terminator = (*ptr == CHAR_LESS_THAN_SIGN)?
5047                CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
5048            name = ++ptr;            name = ++ptr;
5049    
5050            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
# Line 4470  we set the flag only if there is a liter Line 5075  we set the flag only if there is a liter
5075                }                }
5076              }              }
5077    
5078            /* In the real compile, create the entry in the table */            /* In the real compile, create the entry in the table, maintaining
5079              alphabetical order. Duplicate names for different numbers are
5080              permitted only if PCRE_DUPNAMES is set. Duplicate names for the same
5081              number are always OK. (An existing number can be re-used if (?|
5082              appears in the pattern.) In either event, a duplicate name results in
5083              a duplicate entry in the table, even if the number is the same. This
5084              is because the number of names, and hence the table size, is computed
5085              in the pre-compile, and it affects various numbers and pointers which
5086              would all have to be modified, and the compiled code moved down, if
5087              duplicates with the same number were omitted from the table. This
5088              doesn't seem worth the hassle. However, *different* names for the
5089              same number are not permitted. */
5090    
5091            else            else
5092              {              {
5093                BOOL dupname = FALSE;
5094              slot = cd->name_table;              slot = cd->name_table;
5095    
5096              for (i = 0; i < cd->names_found; i++)              for (i = 0; i < cd->names_found; i++)
5097                {                {
5098                int crc = memcmp(name, slot+2, namelen);                int crc = memcmp(name, slot+2, namelen);
# Line 4482  we set the flag only if there is a liter Line 5100  we set the flag only if there is a liter
5100                  {                  {
5101                  if (slot[2+namelen] == 0)                  if (slot[2+namelen] == 0)
5102                    {                    {
5103                    if ((options & PCRE_DUPNAMES) == 0)                    if (GET2(slot, 0) != cd->bracount + 1 &&
5104                          (options & PCRE_DUPNAMES) == 0)
5105                      {                      {
5106                      *errorcodeptr = ERR43;                      *errorcodeptr = ERR43;
5107                      goto FAILED;                      goto FAILED;
5108                      }                      }
5109                      else dupname = TRUE;
5110                    }                    }
5111                  else crc = -1;      /* Current name is substring */                  else crc = -1;      /* Current name is a substring */
5112                  }                  }
5113    
5114                  /* Make space in the table and break the loop for an earlier
5115                  name. For a duplicate or later name, carry on. We do this for
5116                  duplicates so that in the simple case (when ?(| is not used) they
5117                  are in order of their numbers. */
5118    
5119                if (crc < 0)                if (crc < 0)
5120                  {                  {
5121                  memmove(slot + cd->name_entry_size, slot,                  memmove(slot + cd->name_entry_size, slot,
5122                    (cd->names_found - i) * cd->name_entry_size);                    (cd->names_found - i) * cd->name_entry_size);
5123                  break;                  break;
5124                  }                  }
5125    
5126                  /* Continue the loop for a later or duplicate name */
5127    
5128                slot += cd->name_entry_size;                slot += cd->name_entry_size;
5129                }                }
5130    
5131                /* For non-duplicate names, check for a duplicate number before
5132                adding the new name. */
5133    
5134                if (!dupname)
5135                  {
5136                  uschar *cslot = cd->name_table;
5137                  for (i = 0; i < cd->names_found; i++)
5138                    {
5139                    if (cslot != slot)
5140                      {
5141                      if (GET2(cslot, 0) == cd->bracount + 1)
5142                        {
5143                        *errorcodeptr = ERR65;
5144                        goto FAILED;
5145                        }
5146                      }
5147                    else i--;
5148                    cslot += cd->name_entry_size;
5149                    }
5150                  }
5151    
5152              PUT2(slot, 0, cd->bracount + 1);              PUT2(slot, 0, cd->bracount + 1);
5153              memcpy(slot + 2, name, namelen);              memcpy(slot + 2, name, namelen);
5154              slot[2+namelen] = 0;              slot[2+namelen] = 0;
5155              }              }
5156            }            }
5157    
5158          /* In both cases, count the number of names we've encountered. */          /* In both pre-compile and compile, count the number of names we've
5159            encountered. */
5160    
         ptr++;                    /* Move past > or ' */  
5161          cd->names_found++;          cd->names_found++;
5162            ptr++;                    /* Move past > or ' */
5163          goto NUMBERED_GROUP;          goto NUMBERED_GROUP;
5164    
5165    
5166          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5167          case '&':                 /* Perl recursion/subroutine syntax */          case CHAR_AMPERSAND:            /* Perl recursion/subroutine syntax */
5168          terminator = ')';          terminator = CHAR_RIGHT_PARENTHESIS;
5169          is_recurse = TRUE;          is_recurse = TRUE;
5170          /* Fall through */          /* Fall through */
5171    
# Line 4522  we set the flag only if there is a liter Line 5173  we set the flag only if there is a liter
5173          references (?P=name) and recursion (?P>name), as well as falling          references (?P=name) and recursion (?P>name), as well as falling
5174          through from the Perl recursion syntax (?&name). We also come here from          through from the Perl recursion syntax (?&name). We also come here from
5175          the Perl \k<name> or \k'name' back reference syntax and the \k{name}          the Perl \k<name> or \k'name' back reference syntax and the \k{name}
5176          .NET syntax. */          .NET syntax, and the Oniguruma \g<...> and \g'...' subroutine syntax. */
5177    
5178          NAMED_REF_OR_RECURSE:          NAMED_REF_OR_RECURSE:
5179          name = ++ptr;          name = ++ptr;
# Line 4538  we set the flag only if there is a liter Line 5189  we set the flag only if there is a liter
5189              {              {
5190              *errorcodeptr = ERR62;              *errorcodeptr = ERR62;
5191              goto FAILED;              goto FAILED;
5192              }              }
5193            if (*ptr != terminator)            if (*ptr != terminator)
5194              {              {
5195              *errorcodeptr = ERR42;              *errorcodeptr = ERR42;
# Line 4552  we set the flag only if there is a liter Line 5203  we set the flag only if there is a liter
5203            recno = 0;            recno = 0;
5204            }            }
5205    
5206          /* In the real compile, seek the name in the table. We check the name          /* In the real compile, seek the name in the table. We check the name
5207          first, and then check that we have reached the end of the name in the          first, and then check that we have reached the end of the name in the
5208          table. That way, if the name that is longer than any in the table,          table. That way, if the name that is longer than any in the table,
5209          the comparison will fail without reading beyond the table entry. */          the comparison will fail without reading beyond the table entry. */
5210    
# Line 4563  we set the flag only if there is a liter Line 5214  we set the flag only if there is a liter
5214            for (i = 0; i < cd->names_found; i++)            for (i = 0; i < cd->names_found; i++)
5215              {              {
5216              if (strncmp((char *)name, (char *)slot+2, namelen) == 0 &&              if (strncmp((char *)name, (char *)slot+2, namelen) == 0 &&
5217                  slot[2+namelen] == 0)                  slot[2+namelen] == 0)
5218                break;                break;
5219              slot += cd->name_entry_size;              slot += cd->name_entry_size;
5220              }              }
# Line 4573  we set the flag only if there is a liter Line 5224  we set the flag only if there is a liter
5224              recno = GET2(slot, 0);              recno = GET2(slot, 0);
5225              }              }
5226            else if ((recno =                /* Forward back reference */            else if ((recno =                /* Forward back reference */
5227                      find_parens(ptr, cd->bracount, name, namelen,                      find_parens(cd, name, namelen,
5228                        (options & PCRE_EXTENDED) != 0)) <= 0)                        (options & PCRE_EXTENDED) != 0)) <= 0)
5229              {              {
5230              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
# Line 4589  we set the flag only if there is a liter Line 5240  we set the flag only if there is a liter
5240    
5241    
5242          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5243          case 'R':                 /* Recursion */          case CHAR_R:              /* Recursion */
5244          ptr++;                    /* Same as (?0)      */          ptr++;                    /* Same as (?0)      */
5245          /* Fall through */          /* Fall through */
5246    
5247    
5248          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5249          case '-': case '+':          case CHAR_MINUS: case CHAR_PLUS:  /* Recursion or subroutine */
5250          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:
5251          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:
5252            {            {
5253            const uschar *called;            const uschar *called;
5254              terminator = CHAR_RIGHT_PARENTHESIS;
5255    
5256              /* Come here from the \g<...> and \g'...' code (Oniguruma
5257              compatibility). However, the syntax has been checked to ensure that
5258              the ... are a (signed) number, so that neither ERR63 nor ERR29 will
5259              be called on this path, nor with the jump to OTHER_CHAR_AFTER_QUERY
5260              ever be taken. */
5261    
5262            if ((refsign = *ptr) == '+')            HANDLE_NUMERICAL_RECURSION:
5263    
5264              if ((refsign = *ptr) == CHAR_PLUS)
5265              {              {
5266              ptr++;              ptr++;
5267              if ((digitab[*ptr] & ctype_digit) == 0)              if ((digitab[*ptr] & ctype_digit) == 0)
5268                {                {
5269                *errorcodeptr = ERR63;                *errorcodeptr = ERR63;
5270                goto FAILED;                goto FAILED;
5271                }                }
5272              }              }
5273            else if (refsign == '-')            else if (refsign == CHAR_MINUS)
5274              {              {
5275              if ((digitab[ptr[1]] & ctype_digit) == 0)              if ((digitab[ptr[1]] & ctype_digit) == 0)
5276                goto OTHER_CHAR_AFTER_QUERY;                goto OTHER_CHAR_AFTER_QUERY;
# Line 4619  we set the flag only if there is a liter Line 5279  we set the flag only if there is a liter
5279    
5280            recno = 0;            recno = 0;
5281            while((digitab[*ptr] & ctype_digit) != 0)            while((digitab[*ptr] & ctype_digit) != 0)
5282              recno = recno * 10 + *ptr++ - '0';              recno = recno * 10 + *ptr++ - CHAR_0;
5283    
5284            if (*ptr != ')')            if (*ptr != terminator)
5285              {              {
5286              *errorcodeptr = ERR29;              *errorcodeptr = ERR29;
5287              goto FAILED;              goto FAILED;
5288              }              }
5289    
5290            if (refsign == '-')            if (refsign == CHAR_MINUS)
5291              {              {
5292              if (recno == 0)              if (recno == 0)
5293                {                {
# Line 4641  we set the flag only if there is a liter Line 5301  we set the flag only if there is a liter
5301                goto FAILED;                goto FAILED;
5302                }                }
5303              }              }
5304            else if (refsign == '+')            else if (refsign == CHAR_PLUS)
5305              {              {
5306              if (recno == 0)              if (recno == 0)
5307                {                {
# Line 4668  we set the flag only if there is a liter Line 5328  we set the flag only if there is a liter
5328            if (lengthptr == NULL)            if (lengthptr == NULL)
5329              {              {
5330              *code = OP_END;              *code = OP_END;
5331              if (recno != 0) called = find_bracket(cd->start_code, utf8, recno);              if (recno != 0)
5332                  called = _pcre_find_bracket(cd->start_code, utf8, recno);
5333    
5334              /* Forward reference */              /* Forward reference */
5335    
5336              if (called == NULL)              if (called == NULL)
5337                {                {
5338                if (find_parens(ptr, cd->bracount, NULL, recno,                if (find_parens(cd, NULL, recno,
5339                     (options & PCRE_EXTENDED) != 0) < 0)                      (options & PCRE_EXTENDED) != 0) < 0)
5340                  {                  {
5341                  *errorcodeptr = ERR15;                  *errorcodeptr = ERR15;
5342                  goto FAILED;                  goto FAILED;
5343                  }                  }
5344    
5345                  /* Fudge the value of "called" so that when it is inserted as an
5346                  offset below, what it actually inserted is the reference number
5347                  of the group. */
5348    
5349                called = cd->start_code + recno;                called = cd->start_code + recno;
5350                PUTINC(cd->hwm, 0, code + 2 + LINK_SIZE - cd->start_code);                PUTINC(cd->hwm, 0, code + 2 + LINK_SIZE - cd->start_code);
5351                }                }
# Line 4689  we set the flag only if there is a liter Line 5355  we set the flag only if there is a liter
5355              recursion that could loop for ever, and diagnose that case. */              recursion that could loop for ever, and diagnose that case. */
5356    
5357              else if (GET(called, 1) == 0 &&              else if (GET(called, 1) == 0 &&
5358                       could_be_empty(called, code, bcptr, utf8))                       could_be_empty(called, code, bcptr, utf8, cd))
5359                {                {
5360                *errorcodeptr = ERR40;                *errorcodeptr = ERR40;
5361                goto FAILED;                goto FAILED;
# Line 4727  we set the flag only if there is a liter Line 5393  we set the flag only if there is a liter
5393          set = unset = 0;          set = unset = 0;
5394          optset = &set;          optset = &set;
5395    
5396          while (*ptr != ')' && *ptr != ':')          while (*ptr != CHAR_RIGHT_PARENTHESIS && *ptr != CHAR_COLON)
5397            {            {
5398            switch (*ptr++)            switch (*ptr++)
5399              {              {
5400              case '-': optset = &unset; break;              case CHAR_MINUS: optset = &unset; break;
5401    
5402              case 'J':    /* Record that it changed in the external options */              case CHAR_J:    /* Record that it changed in the external options */
5403              *optset |= PCRE_DUPNAMES;              *optset |= PCRE_DUPNAMES;
5404              cd->external_flags |= PCRE_JCHANGED;              cd->external_flags |= PCRE_JCHANGED;
5405              break;              break;
5406    
5407              case 'i': *optset |= PCRE_CASELESS; break;              case CHAR_i: *optset |= PCRE_CASELESS; break;
5408              case 'm': *optset |= PCRE_MULTILINE; break;              case CHAR_m: *optset |= PCRE_MULTILINE; break;
5409              case 's': *optset |= PCRE_DOTALL; break;              case CHAR_s: *optset |= PCRE_DOTALL; break;
5410              case 'x': *optset |= PCRE_EXTENDED; break;              case CHAR_x: *optset |= PCRE_EXTENDED; break;
5411              case 'U': *optset |= PCRE_UNGREEDY; break;              case CHAR_U: *optset |= PCRE_UNGREEDY; break;
5412              case 'X': *optset |= PCRE_EXTRA; break;              case CHAR_X: *optset |= PCRE_EXTRA; break;
5413    
5414              default:  *errorcodeptr = ERR12;              default:  *errorcodeptr = ERR12;
5415                        ptr--;    /* Correct the offset */                        ptr--;    /* Correct the offset */
# Line 4774  we set the flag only if there is a liter Line 5440  we set the flag only if there is a liter
5440          both phases.          both phases.
5441    
5442          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
5443          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
5444          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. */  
5445    
5446          if (*ptr == ')')          if (*ptr == CHAR_RIGHT_PARENTHESIS)
5447            {            {
5448            if (code == cd->start_code + 1 + LINK_SIZE &&            if (code == cd->start_code + 1 + LINK_SIZE &&
5449                 (lengthptr == NULL || *lengthptr == 2 + 2*LINK_SIZE))                 (lengthptr == NULL || *lengthptr == 2 + 2*LINK_SIZE))
5450              {              {
5451              cd->external_options = newoptions;              cd->external_options = newoptions;
             options = newoptions;  
5452              }              }
5453           else            else
5454              {              {
5455              if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))              if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))
5456                {                {
5457                *code++ = OP_OPT;                *code++ = OP_OPT;
5458                *code++ = newoptions & PCRE_IMS;                *code++ = newoptions & PCRE_IMS;
5459                }                }
   
             /* 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;  
5460              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);
5461              greedy_non_default = greedy_default ^ 1;              greedy_non_default = greedy_default ^ 1;
5462              req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
5463              }              }
5464    
5465              /* Change options at this level, and pass them back for use
5466              in subsequent branches. When not at the start of the pattern, this
5467              information is also necessary so that a resetting item can be
5468              compiled at the end of a group (if we are in a group). */
5469    
5470              *optionsptr = options = newoptions;
5471            previous = NULL;       /* This item can't be repeated */            previous = NULL;       /* This item can't be repeated */
5472            continue;              /* It is complete */            continue;              /* It is complete */
5473            }            }
# Line 4819  we set the flag only if there is a liter Line 5482  we set the flag only if there is a liter
5482          }     /* End of switch for character following (? */          }     /* End of switch for character following (? */
5483        }       /* End of (? handling */        }       /* End of (? handling */
5484    
5485      /* Opening parenthesis not followed by '?'. If PCRE_NO_AUTO_CAPTURE is set,      /* Opening parenthesis not followed by '*' or '?'. If PCRE_NO_AUTO_CAPTURE
5486      all unadorned brackets become non-capturing and behave like (?:...)      is set, all unadorned brackets become non-capturing and behave like (?:...)
5487      brackets. */      brackets. */
5488    
5489      else if ((options & PCRE_NO_AUTO_CAPTURE) != 0)      else if ((options & PCRE_NO_AUTO_CAPTURE) != 0)
# Line 4920  we set the flag only if there is a liter Line 5583  we set the flag only if there is a liter
5583    
5584      /* Error if hit end of pattern */      /* Error if hit end of pattern */
5585    
5586      if (*ptr != ')')      if (*ptr != CHAR_RIGHT_PARENTHESIS)
5587        {        {
5588        *errorcodeptr = ERR14;        *errorcodeptr = ERR14;
5589        goto FAILED;        goto FAILED;
# Line 5012  we set the flag only if there is a liter Line 5675  we set the flag only if there is a liter
5675    
5676      /* ===================================================================*/      /* ===================================================================*/
5677      /* Handle metasequences introduced by \. For ones like \d, the ESC_ values      /* Handle metasequences introduced by \. For ones like \d, the ESC_ values
5678      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
5679      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
5680      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
5681      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
5682      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
5683        ever created. */
5684    
5685      case '\\':      case CHAR_BACKSLASH:
5686      tempptr = ptr;      tempptr = ptr;
5687      c = check_escape(&ptr, errorcodeptr, cd->bracount, options, FALSE);      c = check_escape(&ptr, errorcodeptr, cd->bracount, options, FALSE);
5688      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
# Line 5027  we set the flag only if there is a liter Line 5691  we set the flag only if there is a liter
5691        {        {
5692        if (-c == ESC_Q)            /* Handle start of quoted string */        if (-c == ESC_Q)            /* Handle start of quoted string */
5693          {          {
5694          if (ptr[1] == '\\' && ptr[2] == 'E') ptr += 2; /* avoid empty string */          if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
5695            else inescq = TRUE;            ptr += 2;               /* avoid empty string */
5696                else inescq = TRUE;
5697          continue;          continue;
5698          }          }
5699    
# Line 5045  we set the flag only if there is a liter Line 5710  we set the flag only if there is a liter
5710        zerofirstbyte = firstbyte;        zerofirstbyte = firstbyte;
5711        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
5712    
5713          /* \g<name> or \g'name' is a subroutine call by name and \g<n> or \g'n'
5714          is a subroutine call by number (Oniguruma syntax). In fact, the value
5715          -ESC_g is returned only for these cases. So we don't need to check for <
5716          or ' if the value is -ESC_g. For the Perl syntax \g{n} the value is
5717          -ESC_REF+n, and for the Perl syntax \g{name} the result is -ESC_k (as
5718          that is a synonym for a named back reference). */
5719    
5720          if (-c == ESC_g)
5721            {
5722            const uschar *p;
5723            save_hwm = cd->hwm;   /* Normally this is set when '(' is read */
5724            terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
5725              CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
5726    
5727            /* These two statements stop the compiler for warning about possibly
5728            unset variables caused by the jump to HANDLE_NUMERICAL_RECURSION. In
5729            fact, because we actually check for a number below, the paths that
5730            would actually be in error are never taken. */
5731    
5732            skipbytes = 0;
5733            reset_bracount = FALSE;
5734    
5735            /* Test for a name */
5736    
5737            if (ptr[1] != CHAR_PLUS && ptr[1] != CHAR_MINUS)
5738              {
5739              BOOL isnumber = TRUE;
5740              for (p = ptr + 1; *p != 0 && *p != terminator; p++)
5741                {
5742                if ((cd->ctypes[*p] & ctype_digit) == 0) isnumber = FALSE;
5743                if ((cd->ctypes[*p] & ctype_word) == 0) break;
5744                }
5745              if (*p != terminator)
5746                {
5747                *errorcodeptr = ERR57;
5748                break;
5749                }
5750              if (isnumber)
5751                {
5752                ptr++;
5753                goto HANDLE_NUMERICAL_RECURSION;
5754                }
5755              is_recurse = TRUE;
5756              goto NAMED_REF_OR_RECURSE;
5757              }
5758    
5759            /* Test a signed number in angle brackets or quotes. */
5760    
5761            p = ptr + 2;
5762            while ((digitab[*p] & ctype_digit) != 0) p++;
5763            if (*p != terminator)
5764              {
5765              *errorcodeptr = ERR57;
5766              break;
5767              }
5768            ptr++;
5769            goto HANDLE_NUMERICAL_RECURSION;
5770            }
5771    
5772        /* \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).
5773        We also support \k{name} (.NET syntax) */        We also support \k{name} (.NET syntax) */
5774    
5775        if (-c == ESC_k && (ptr[1] == '<' || ptr[1] == '\'' || ptr[1] == '{'))        if (-c == ESC_k && (ptr[1] == CHAR_LESS_THAN_SIGN ||
5776              ptr[1] == CHAR_APOSTROPHE || ptr[1] == CHAR_LEFT_CURLY_BRACKET))
5777          {          {
5778          is_recurse = FALSE;          is_recurse = FALSE;
5779          terminator = (*(++ptr) == '<')? '>' : (*ptr == '\'')? '\'' : '}';          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
5780              CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?
5781              CHAR_APOSTROPHE : CHAR_RIGHT_CURLY_BRACKET;
5782          goto NAMED_REF_OR_RECURSE;          goto NAMED_REF_OR_RECURSE;
5783          }          }
5784    
# Line 5061  we set the flag only if there is a liter Line 5788  we set the flag only if there is a liter
5788    
5789        if (-c >= ESC_REF)        if (-c >= ESC_REF)
5790          {          {
5791            open_capitem *oc;
5792          recno = -c - ESC_REF;          recno = -c - ESC_REF;
5793    
5794          HANDLE_REFERENCE:    /* Come here from named backref handling */          HANDLE_REFERENCE:    /* Come here from named backref handling */
# Line 5070  we set the flag only if there is a liter Line 5798  we set the flag only if there is a liter
5798          PUT2INC(code, 0, recno);          PUT2INC(code, 0, recno);
5799          cd->backref_map |= (recno < 32)? (1 << recno) : 1;          cd->backref_map |= (recno < 32)? (1 << recno) : 1;
5800          if (recno > cd->top_backref) cd->top_backref = recno;          if (recno > cd->top_backref) cd->top_backref = recno;
5801    
5802            /* Check to see if this back reference is recursive, that it, it
5803            is inside the group that it references. A flag is set so that the
5804            group can be made atomic. */
5805    
5806            for (oc = cd->open_caps; oc != NULL; oc = oc->next)
5807              {
5808              if (oc->number == recno)
5809                {
5810                oc->flag = TRUE;
5811                break;
5812                }
5813              }
5814          }          }
5815    
5816        /* So are Unicode property matches, if supported. */        /* So are Unicode property matches, if supported. */
# Line 5099  we set the flag only if there is a liter Line 5840  we set the flag only if there is a liter
5840  #endif  #endif
5841    
5842        /* For the rest (including \X when Unicode properties are supported), we        /* For the rest (including \X when Unicode properties are supported), we
5843        can obtain the OP value by negating the escape value. */        can obtain the OP value by negating the escape value in the default
5844          situation when PCRE_UCP is not set. When it *is* set, we substitute
5845          Unicode property tests. */
5846    
5847        else        else
5848          {          {
5849          previous = (-c > ESC_b && -c < ESC_Z)? code : NULL;  #ifdef SUPPORT_UCP
5850          *code++ = -c;          if (-c >= ESC_DU && -c <= ESC_wu)
5851              {
5852              nestptr = ptr + 1;                   /* Where to resume */
5853              ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
5854              }
5855            else
5856    #endif
5857              {
5858              previous = (-c > ESC_b && -c < ESC_Z)? code : NULL;
5859              *code++ = -c;
5860              }
5861          }          }
5862        continue;        continue;
5863        }        }
# Line 5154  we set the flag only if there is a liter Line 5907  we set the flag only if there is a liter
5907    
5908      /* Remember if \r or \n were seen */      /* Remember if \r or \n were seen */
5909    
5910      if (mcbuffer[0] == '\r' || mcbuffer[0] == '\n')      if (mcbuffer[0] == CHAR_CR || mcbuffer[0] == CHAR_NL)
5911        cd->external_flags |= PCRE_HASCRORLF;        cd->external_flags |= PCRE_HASCRORLF;
5912    
5913      /* Set the first and required bytes appropriately. If no previous first      /* Set the first and required bytes appropriately. If no previous first
# Line 5252  uschar *code = *codeptr; Line 6005  uschar *code = *codeptr;
6005  uschar *last_branch = code;  uschar *last_branch = code;
6006  uschar *start_bracket = code;  uschar *start_bracket = code;
6007  uschar *reverse_count = NULL;  uschar *reverse_count = NULL;
6008    open_capitem capitem;
6009    int capnumber = 0;
6010  int firstbyte, reqbyte;  int firstbyte, reqbyte;
6011  int branchfirstbyte, branchreqbyte;  int branchfirstbyte, branchreqbyte;
6012  int length;  int length;
6013  int orig_bracount;  int orig_bracount;
6014  int max_bracount;  int max_bracount;
6015    int old_external_options = cd->external_options;
6016  branch_chain bc;  branch_chain bc;
6017    
6018  bc.outer = bcptr;  bc.outer = bcptr;
6019  bc.current = code;  bc.current_branch = code;
6020    
6021  firstbyte = reqbyte = REQ_UNSET;  firstbyte = reqbyte = REQ_UNSET;
6022    
# Line 5278  the code that abstracts option settings Line 6034  the code that abstracts option settings
6034  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
6035  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. */
6036    
6037    /* If this is a capturing subpattern, add to the chain of open capturing items
6038    so that we can detect them if (*ACCEPT) is encountered. This is also used to
6039    detect groups that contain recursive back references to themselves. */
6040    
6041    if (*code == OP_CBRA)
6042      {
6043      capnumber = GET2(code, 1 + LINK_SIZE);
6044      capitem.number = capnumber;
6045      capitem.next = cd->open_caps;
6046      capitem.flag = FALSE;
6047      cd->open_caps = &capitem;
6048      }
6049    
6050  /* Offset is set zero to mark that this bracket is still open */  /* Offset is set zero to mark that this bracket is still open */
6051    
6052  PUT(code, 1, 0);  PUT(code, 1, 0);
# Line 5322  for (;;) Line 6091  for (;;)
6091      return FALSE;      return FALSE;
6092      }      }
6093    
6094      /* If the external options have changed during this branch, it means that we
6095      are at the top level, and a leading option setting has been encountered. We
6096      need to re-set the original option values to take account of this so that,
6097      during the pre-compile phase, we know to allow for a re-set at the start of
6098      subsequent branches. */
6099    
6100      if (old_external_options != cd->external_options)
6101        oldims = cd->external_options & PCRE_IMS;