/[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 295 by ph10, Mon Dec 31 17:00:24 2007 UTC revision 556 by ph10, Tue Oct 26 11:06:44 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 CHAR_g:
666        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      case 'g':      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
     if (ptr[1] == '{')  
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      utf8         TRUE if we are in UTF-8 mode
1114      count        pointer to the current capturing subpattern number (updated)
1115    
1116  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1117  */  */
1118    
1119  static int  static int
1120  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,
1121    BOOL xmode)    BOOL xmode, BOOL utf8, int *count)
1122  {  {
1123  const uschar *thisname;  uschar *ptr = *ptrptr;
1124    int start_count = *count;
1125    int hwm_count = start_count;
1126    BOOL dup_parens = FALSE;
1127    
1128  for (; *ptr != 0; ptr++)  /* If the first character is a parenthesis, check on the type of group we are
1129    dealing with. The very first call may not start with a parenthesis. */
1130    
1131    if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1132    {    {
1133    int term;    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1134    
1135      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1136    
1137      /* Handle a normal, unnamed capturing parenthesis. */
1138    
1139      else if (ptr[1] != CHAR_QUESTION_MARK)
1140        {
1141        *count += 1;
1142        if (name == NULL && *count == lorn) return *count;
1143        ptr++;
1144        }
1145    
1146      /* All cases now have (? at the start. Remember when we are in a group
1147      where the parenthesis numbers are duplicated. */
1148    
1149      else if (ptr[2] == CHAR_VERTICAL_LINE)
1150        {
1151        ptr += 3;
1152        dup_parens = TRUE;
1153        }
1154    
1155      /* Handle comments; all characters are allowed until a ket is reached. */
1156    
1157      else if (ptr[2] == CHAR_NUMBER_SIGN)
1158        {
1159        for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1160        goto FAIL_EXIT;
1161        }
1162    
1163      /* Handle a condition. If it is an assertion, just carry on so that it
1164      is processed as normal. If not, skip to the closing parenthesis of the
1165      condition (there can't be any nested parens). */
1166    
1167      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1168        {
1169        ptr += 2;
1170        if (ptr[1] != CHAR_QUESTION_MARK)
1171          {
1172          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1173          if (*ptr != 0) ptr++;
1174          }
1175        }
1176    
1177      /* Start with (? but not a condition. */
1178    
1179      else
1180        {
1181        ptr += 2;
1182        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1183    
1184        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1185    
1186        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1187            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1188          {
1189          int term;
1190          const uschar *thisname;
1191          *count += 1;
1192          if (name == NULL && *count == lorn) return *count;
1193          term = *ptr++;
1194          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1195          thisname = ptr;
1196          while (*ptr != term) ptr++;
1197          if (name != NULL && lorn == ptr - thisname &&
1198              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1199            return *count;
1200          term++;
1201          }
1202        }
1203      }
1204    
1205    /* Past any initial parenthesis handling, scan for parentheses or vertical
1206    bars. */
1207    
1208    for (; *ptr != 0; ptr++)
1209      {
1210    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1211    
1212    if (*ptr == '\\')    if (*ptr == CHAR_BACKSLASH)
1213      {      {
1214      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1215      if (*ptr == 'Q') for (;;)      if (*ptr == CHAR_Q) for (;;)
1216        {        {
1217        while (*(++ptr) != 0 && *ptr != '\\');        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1218        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1219        if (*(++ptr) == 'E') break;        if (*(++ptr) == CHAR_E) break;
1220        }        }
1221      continue;      continue;
1222      }      }
1223    
1224    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1225      are handled for real. If the first character is '^', skip it. Also, if the
1226      first few characters (either before or after ^) are \Q\E or \E we skip them
1227      too. This makes for compatibility with Perl. Note the use of STR macros to
1228      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1229    
1230    if (*ptr == '[')    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1231      {      {
1232      while (*(++ptr) != ']')      BOOL negate_class = FALSE;
1233        for (;;)
1234          {
1235          if (ptr[1] == CHAR_BACKSLASH)
1236            {
1237            if (ptr[2] == CHAR_E)
1238              ptr+= 2;
1239            else if (strncmp((const char *)ptr+2,
1240                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1241              ptr += 4;
1242            else
1243              break;
1244            }
1245          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1246            {
1247            negate_class = TRUE;
1248            ptr++;
1249            }
1250          else break;
1251          }
1252    
1253        /* If the next character is ']', it is a data character that must be
1254        skipped, except in JavaScript compatibility mode. */
1255    
1256        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1257            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1258          ptr++;
1259    
1260        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1261        {        {
1262        if (*ptr == 0) return -1;        if (*ptr == 0) return -1;
1263        if (*ptr == '\\')        if (*ptr == CHAR_BACKSLASH)
1264          {          {
1265          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) goto FAIL_EXIT;
1266          if (*ptr == 'Q') for (;;)          if (*ptr == CHAR_Q) for (;;)
1267            {            {
1268            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1269            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1270            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1271            }            }
1272          continue;          continue;
1273          }          }
# Line 1008  for (; *ptr != 0; ptr++) Line 1277  for (; *ptr != 0; ptr++)
1277    
1278    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1279    
1280    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1281      {      {
1282      while (*(++ptr) != 0 && *ptr != '\n');      ptr++;
1283      if (*ptr == 0) return -1;      while (*ptr != 0)
1284          {
1285          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1286          ptr++;
1287    #ifdef SUPPORT_UTF8
1288          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
1289    #endif
1290          }
1291        if (*ptr == 0) goto FAIL_EXIT;
1292      continue;      continue;
1293      }      }
1294    
1295    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1296    
1297    if (*ptr != '(') continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != '?' && ptr[1] != '*')  
1298      {      {
1299      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, count);
1300      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1301      continue;      if (*ptr == 0) goto FAIL_EXIT;
1302      }      }
1303    
1304    ptr += 2;    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1305    if (*ptr == 'P') ptr++;                      /* Allow optional P */      {
1306        if (dup_parens && *count < hwm_count) *count = hwm_count;
1307        goto FAIL_EXIT;
1308        }
1309    
1310    /* We have to disambiguate (?<! and (?<= from (?<name> */    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1311        {
1312        if (*count > hwm_count) hwm_count = *count;
1313        *count = start_count;
1314        }
1315      }
1316    
1317    if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  FAIL_EXIT:
1318         *ptr != '\'')  *ptrptr = ptr;
1319      continue;  return -1;
1320    }
1321    
1322    
1323    
1324    
1325    /*************************************************
1326    *       Find forward referenced subpattern       *
1327    *************************************************/
1328    
1329    /* This function scans along a pattern's text looking for capturing
1330    subpatterns, and counting them. If it finds a named pattern that matches the
1331    name it is given, it returns its number. Alternatively, if the name is NULL, it
1332    returns when it reaches a given numbered subpattern. This is used for forward
1333    references to subpatterns. We used to be able to start this scan from the
1334    current compiling point, using the current count value from cd->bracount, and
1335    do it all in a single loop, but the addition of the possibility of duplicate
1336    subpattern numbers means that we have to scan from the very start, in order to
1337    take account of such duplicates, and to use a recursive function to keep track
1338    of the different types of group.
1339    
1340    count++;  Arguments:
1341      cd           compile background data
1342      name         name to seek, or NULL if seeking a numbered subpattern
1343      lorn         name length, or subpattern number if name is NULL
1344      xmode        TRUE if we are in /x mode
1345      utf8         TRUE if we are in UTF-8 mode
1346    
1347    if (name == NULL && count == lorn) return count;  Returns:       the number of the found subpattern, or -1 if not found
1348    term = *ptr++;  */
1349    if (term == '<') term = '>';  
1350    thisname = ptr;  static int
1351    while (*ptr != term) ptr++;  find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode,
1352    if (name != NULL && lorn == ptr - thisname &&    BOOL utf8)
1353        strncmp((const char *)name, (const char *)thisname, lorn) == 0)  {
1354      return count;  uschar *ptr = (uschar *)cd->start_pattern;
1355    int count = 0;
1356    int rc;
1357    
1358    /* If the pattern does not start with an opening parenthesis, the first call
1359    to find_parens_sub() will scan right to the end (if necessary). However, if it
1360    does start with a parenthesis, find_parens_sub() will return when it hits the
1361    matching closing parens. That is why we have to have a loop. */
1362    
1363    for (;;)
1364      {
1365      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, &count);
1366      if (rc > 0 || *ptr++ == 0) break;
1367    }    }
1368    
1369  return -1;  return rc;
1370  }  }
1371    
1372    
1373    
1374    
1375  /*************************************************  /*************************************************
1376  *      Find first significant op code            *  *      Find first significant op code            *
1377  *************************************************/  *************************************************/
# Line 1100  for (;;) Line 1421  for (;;)
1421    
1422      case OP_CALLOUT:      case OP_CALLOUT:
1423      case OP_CREF:      case OP_CREF:
1424        case OP_NCREF:
1425      case OP_RREF:      case OP_RREF:
1426        case OP_NRREF:
1427      case OP_DEF:      case OP_DEF:
1428      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1429      break;      break;
# Line 1116  for (;;) Line 1439  for (;;)
1439    
1440    
1441  /*************************************************  /*************************************************
1442  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1443  *************************************************/  *************************************************/
1444    
1445  /* 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,
1446  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.
1447  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
1448    temporarily terminated with OP_END when this function is called.
1449    
1450    This function is called when a backward assertion is encountered, so that if it
1451    fails, the error message can point to the correct place in the pattern.
1452    However, we cannot do this when the assertion contains subroutine calls,
1453    because they can be forward references. We solve this by remembering this case
1454    and doing the check at the end; a flag specifies which mode we are running in.
1455    
1456  Arguments:  Arguments:
1457    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1458    options  the compiling options    options  the compiling options
1459      atend    TRUE if called when the pattern is complete
1460      cd       the "compile data" structure
1461    
1462  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1463                 or -1 if there is no fixed length,
1464               or -2 if \C was encountered               or -2 if \C was encountered
1465                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1466  */  */
1467    
1468  static int  static int
1469  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)
1470  {  {
1471  int length = -1;  int length = -1;
1472    
# Line 1145  branch, check the length against that of Line 1479  branch, check the length against that of
1479  for (;;)  for (;;)
1480    {    {
1481    int d;    int d;
1482      uschar *ce, *cs;
1483    register int op = *cc;    register int op = *cc;
1484    switch (op)    switch (op)
1485      {      {
# Line 1152  for (;;) Line 1487  for (;;)
1487      case OP_BRA:      case OP_BRA:
1488      case OP_ONCE:      case OP_ONCE:
1489      case OP_COND:      case OP_COND:
1490      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);
1491      if (d < 0) return d;      if (d < 0) return d;
1492      branchlength += d;      branchlength += d;
1493      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1175  for (;;) Line 1510  for (;;)
1510      branchlength = 0;      branchlength = 0;
1511      break;      break;
1512    
1513        /* A true recursion implies not fixed length, but a subroutine call may
1514        be OK. If the subroutine is a forward reference, we can't deal with
1515        it until the end of the pattern, so return -3. */
1516    
1517        case OP_RECURSE:
1518        if (!atend) return -3;
1519        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1520        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1521        if (cc > cs && cc < ce) return -1;                /* Recursion */
1522        d = find_fixedlength(cs + 2, options, atend, cd);
1523        if (d < 0) return d;
1524        branchlength += d;
1525        cc += 1 + LINK_SIZE;
1526        break;
1527    
1528      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1529    
1530      case OP_ASSERT:      case OP_ASSERT:
# Line 1188  for (;;) Line 1538  for (;;)
1538    
1539      case OP_REVERSE:      case OP_REVERSE:
1540      case OP_CREF:      case OP_CREF:
1541        case OP_NCREF:
1542      case OP_RREF:      case OP_RREF:
1543        case OP_NRREF:
1544      case OP_DEF:      case OP_DEF:
1545      case OP_OPT:      case OP_OPT:
1546      case OP_CALLOUT:      case OP_CALLOUT:
1547      case OP_SOD:      case OP_SOD:
1548      case OP_SOM:      case OP_SOM:
1549        case OP_SET_SOM:
1550      case OP_EOD:      case OP_EOD:
1551      case OP_EODN:      case OP_EODN:
1552      case OP_CIRC:      case OP_CIRC:
# Line 1211  for (;;) Line 1564  for (;;)
1564      branchlength++;      branchlength++;
1565      cc += 2;      cc += 2;
1566  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1567      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1568        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1569  #endif  #endif
1570      break;      break;
1571    
# Line 1225  for (;;) Line 1576  for (;;)
1576      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1577      cc += 4;      cc += 4;
1578  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1579      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1580        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1581  #endif  #endif
1582      break;      break;
1583    
# Line 1252  for (;;) Line 1601  for (;;)
1601      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1602      case OP_WORDCHAR:      case OP_WORDCHAR:
1603      case OP_ANY:      case OP_ANY:
1604        case OP_ALLANY:
1605      branchlength++;      branchlength++;
1606      cc++;      cc++;
1607      break;      break;
# Line 1306  for (;;) Line 1656  for (;;)
1656    
1657    
1658  /*************************************************  /*************************************************
1659  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1660  *************************************************/  *************************************************/
1661    
1662  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1663  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1664    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1665    so that it can be called from pcre_study() when finding the minimum matching
1666    length.
1667    
1668  Arguments:  Arguments:
1669    code        points to start of expression    code        points to start of expression
1670    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1671    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1672    
1673  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
1674  */  */
1675    
1676  static const uschar *  const uschar *
1677  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1678  {  {
1679  for (;;)  for (;;)
1680    {    {
# Line 1334  for (;;) Line 1687  for (;;)
1687    
1688    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1689    
1690      /* Handle recursion */
1691    
1692      else if (c == OP_REVERSE)
1693        {
1694        if (number < 0) return (uschar *)code;
1695        code += _pcre_OP_lengths[c];
1696        }
1697    
1698    /* Handle capturing bracket */    /* Handle capturing bracket */
1699    
1700    else if (c == OP_CBRA)    else if (c == OP_CBRA)
# Line 1345  for (;;) Line 1706  for (;;)
1706    
1707    /* 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
1708    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
1709    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1710      must add in its length. */
1711    
1712    else    else
1713      {      {
# Line 1369  for (;;) Line 1731  for (;;)
1731        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
1732        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1733        break;        break;
1734    
1735          case OP_MARK:
1736          case OP_PRUNE_ARG:
1737          case OP_SKIP_ARG:
1738          code += code[1];
1739          break;
1740    
1741          case OP_THEN_ARG:
1742          code += code[1+LINK_SIZE];
1743          break;
1744        }        }
1745    
1746      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1400  for (;;) Line 1772  for (;;)
1772        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1773        break;        break;
1774        }        }
1775    #else
1776        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1777  #endif  #endif
1778      }      }
1779    }    }
# Line 1438  for (;;) Line 1812  for (;;)
1812    
1813    /* 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
1814    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
1815    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1816      must add in its length. */
1817    
1818    else    else
1819      {      {
# Line 1462  for (;;) Line 1837  for (;;)
1837        case OP_TYPEEXACT:        case OP_TYPEEXACT:
1838        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1839        break;        break;
1840    
1841          case OP_MARK:
1842          case OP_PRUNE_ARG:
1843          case OP_SKIP_ARG:
1844          code += code[1];
1845          break;
1846    
1847          case OP_THEN_ARG:
1848          code += code[1+LINK_SIZE];
1849          break;
1850        }        }
1851    
1852      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1493  for (;;) Line 1878  for (;;)
1878        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1879        break;        break;
1880        }        }
1881    #else
1882        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1883  #endif  #endif
1884      }      }
1885    }    }
# Line 1516  Arguments: Line 1903  Arguments:
1903    code        points to start of search    code        points to start of search
1904    endcode     points to where to stop    endcode     points to where to stop
1905    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
1906      cd          contains pointers to tables etc.
1907    
1908  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
1909  */  */
1910    
1911  static BOOL  static BOOL
1912  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
1913      compile_data *cd)
1914  {  {
1915  register int c;  register int c;
1916  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 1544  for (code = first_significant_code(code Line 1933  for (code = first_significant_code(code
1933    
1934    /* Groups with zero repeats can of course be empty; skip them. */    /* Groups with zero repeats can of course be empty; skip them. */
1935    
1936    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)
1937      {      {
1938      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1939      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
# Line 1552  for (code = first_significant_code(code Line 1941  for (code = first_significant_code(code
1941      continue;      continue;
1942      }      }
1943    
1944      /* For a recursion/subroutine call, if its end has been reached, which
1945      implies a subroutine call, we can scan it. */
1946    
1947      if (c == OP_RECURSE)
1948        {
1949        BOOL empty_branch = FALSE;
1950        const uschar *scode = cd->start_code + GET(code, 1);
1951        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
1952        do
1953          {
1954          if (could_be_empty_branch(scode, endcode, utf8, cd))
1955            {
1956            empty_branch = TRUE;
1957            break;
1958            }
1959          scode += GET(scode, 1);
1960          }
1961        while (*scode == OP_ALT);
1962        if (!empty_branch) return FALSE;  /* All branches are non-empty */
1963        continue;
1964        }
1965    
1966    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
1967    
1968    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 1559  for (code = first_significant_code(code Line 1970  for (code = first_significant_code(code
1970      BOOL empty_branch;      BOOL empty_branch;
1971      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
1972    
1973      /* Scan a closed bracket */      /* If a conditional group has only one branch, there is a second, implied,
1974        empty branch, so just skip over the conditional, because it could be empty.
1975        Otherwise, scan the individual branches of the group. */
1976    
1977      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;  
1978        code += GET(code, 1);        code += GET(code, 1);
1979        else
1980          {
1981          empty_branch = FALSE;
1982          do
1983            {
1984            if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
1985              empty_branch = TRUE;
1986            code += GET(code, 1);
1987            }
1988          while (*code == OP_ALT);
1989          if (!empty_branch) return FALSE;   /* All branches are non-empty */
1990        }        }
1991      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
1992      c = *code;      c = *code;
1993      continue;      continue;
1994      }      }
# Line 1630  for (code = first_significant_code(code Line 2049  for (code = first_significant_code(code
2049      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2050      case OP_WORDCHAR:      case OP_WORDCHAR:
2051      case OP_ANY:      case OP_ANY:
2052        case OP_ALLANY:
2053      case OP_ANYBYTE:      case OP_ANYBYTE:
2054      case OP_CHAR:      case OP_CHAR:
2055      case OP_CHARNC:      case OP_CHARNC:
# Line 1686  for (code = first_significant_code(code Line 2106  for (code = first_significant_code(code
2106      case OP_QUERY:      case OP_QUERY:
2107      case OP_MINQUERY:      case OP_MINQUERY:
2108      case OP_POSQUERY:      case OP_POSQUERY:
2109        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2110        break;
2111    
2112      case OP_UPTO:      case OP_UPTO:
2113      case OP_MINUPTO:      case OP_MINUPTO:
2114      case OP_POSUPTO:      case OP_POSUPTO:
2115      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2116      break;      break;
2117  #endif  #endif
2118    
2119        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2120        string. */
2121    
2122        case OP_MARK:
2123        case OP_PRUNE_ARG:
2124        case OP_SKIP_ARG:
2125        code += code[1];
2126        break;
2127    
2128        case OP_THEN_ARG:
2129        code += code[1+LINK_SIZE];
2130        break;
2131    
2132        /* None of the remaining opcodes are required to match a character. */
2133    
2134        default:
2135        break;
2136      }      }
2137    }    }
2138    
# Line 1714  Arguments: Line 2155  Arguments:
2155    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2156    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2157    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2158      cd          pointers to tables etc
2159    
2160  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2161  */  */
2162    
2163  static BOOL  static BOOL
2164  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2165    BOOL utf8)    BOOL utf8, compile_data *cd)
2166  {  {
2167  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2168    {    {
2169    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2170        return FALSE;
2171    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2172    }    }
2173  return TRUE;  return TRUE;
# Line 1738  return TRUE; Line 2181  return TRUE;
2181    
2182  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2183  encountered in a character class. It checks whether this is followed by a  encountered in a character class. It checks whether this is followed by a
2184  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2185  reach an unescaped ']' without the special preceding character, return FALSE.  reach an unescaped ']' without the special preceding character, return FALSE.
2186    
2187  Originally, this function only recognized a sequence of letters between the  Originally, this function only recognized a sequence of letters between the
2188  terminators, but it seems that Perl recognizes any sequence of characters,  terminators, but it seems that Perl recognizes any sequence of characters,
2189  though of course unknown POSIX names are subsequently rejected. Perl gives an  though of course unknown POSIX names are subsequently rejected. Perl gives an
2190  "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE  "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2191  didn't consider this to be a POSIX class. Likewise for [:1234:].  didn't consider this to be a POSIX class. Likewise for [:1234:].
2192    
2193  The problem in trying to be exactly like Perl is in the handling of escapes. We  The problem in trying to be exactly like Perl is in the handling of escapes. We
2194  have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX  have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2195  class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code  class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2196  below handles the special case of \], but does not try to do any other escape  below handles the special case of \], but does not try to do any other escape
2197  processing. This makes it different from Perl for cases such as [:l\ower:]  processing. This makes it different from Perl for cases such as [:l\ower:]
2198  where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize  where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2199  "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,  "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2200  I think.  I think.
2201    
2202  Arguments:  Arguments:
# Line 1770  int terminator;          /* Don't combin Line 2213  int terminator;          /* Don't combin
2213  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2214  for (++ptr; *ptr != 0; ptr++)  for (++ptr; *ptr != 0; ptr++)
2215    {    {
2216    if (*ptr == '\\' && ptr[1] == ']') ptr++; else    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else
2217      {      {
2218      if (*ptr == ']') return FALSE;      if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2219      if (*ptr == terminator && ptr[1] == ']')      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2220        {        {
2221        *endptr = ptr;        *endptr = ptr;
2222        return TRUE;        return TRUE;
2223        }        }
2224      }      }
2225    }    }
2226  return FALSE;  return FALSE;
2227  }  }
2228    
# Line 1824  return -1; Line 2267  return -1;
2267  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2268  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2269  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
2270  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
2271  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
2272  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
2273  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
2274  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2275    OP_END.
2276    
2277  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2278  recursions and subroutine calls. It must also check the list of such references  recursions and subroutine calls. It must also check the list of such references
# Line 1905  auto_callout(uschar *code, const uschar Line 2349  auto_callout(uschar *code, const uschar
2349  {  {
2350  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2351  *code++ = 255;  *code++ = 255;
2352  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2353  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2354  return code + 2*LINK_SIZE;  return code + 2*LINK_SIZE;
2355  }  }
2356    
# Line 1931  Returns:             nothing Line 2375  Returns:             nothing
2375  static void  static void
2376  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)
2377  {  {
2378  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2379  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2380  }  }
2381    
# Line 1963  get_othercase_range(unsigned int *cptr, Line 2407  get_othercase_range(unsigned int *cptr,
2407  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2408    
2409  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2410    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2411    
2412  if (c > d) return FALSE;  if (c > d) return FALSE;
2413    
# Line 1972  next = othercase + 1; Line 2416  next = othercase + 1;
2416    
2417  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2418    {    {
2419    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2420    next++;    next++;
2421    }    }
2422    
# Line 1981  for (++c; c <= d; c++) Line 2425  for (++c; c <= d; c++)
2425    
2426  return TRUE;  return TRUE;
2427  }  }
2428    
2429    
2430    
2431    /*************************************************
2432    *        Check a character and a property        *
2433    *************************************************/
2434    
2435    /* This function is called by check_auto_possessive() when a property item
2436    is adjacent to a fixed character.
2437    
2438    Arguments:
2439      c            the character
2440      ptype        the property type
2441      pdata        the data for the type
2442      negated      TRUE if it's a negated property (\P or \p{^)
2443    
2444    Returns:       TRUE if auto-possessifying is OK
2445    */
2446    
2447    static BOOL
2448    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2449    {
2450    const ucd_record *prop = GET_UCD(c);
2451    switch(ptype)
2452      {
2453      case PT_LAMP:
2454      return (prop->chartype == ucp_Lu ||
2455              prop->chartype == ucp_Ll ||
2456              prop->chartype == ucp_Lt) == negated;
2457    
2458      case PT_GC:
2459      return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;
2460    
2461      case PT_PC:
2462      return (pdata == prop->chartype) == negated;
2463    
2464      case PT_SC:
2465      return (pdata == prop->script) == negated;
2466    
2467      /* These are specials */
2468    
2469      case PT_ALNUM:
2470      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2471              _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;
2472    
2473      case PT_SPACE:    /* Perl space */
2474      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2475              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2476              == negated;
2477    
2478      case PT_PXSPACE:  /* POSIX space */
2479      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2480              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2481              c == CHAR_FF || c == CHAR_CR)
2482              == negated;
2483    
2484      case PT_WORD:
2485      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2486              _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2487              c == CHAR_UNDERSCORE) == negated;
2488      }
2489    return FALSE;
2490    }
2491  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2492    
2493    
# Line 1994  whether the next thing could possibly ma Line 2501  whether the next thing could possibly ma
2501  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
2502    
2503  Arguments:  Arguments:
2504    op_code       the repeated op code    previous      pointer to the repeated opcode
   this          data for this item, depends on the opcode  
2505    utf8          TRUE in UTF-8 mode    utf8          TRUE in UTF-8 mode
   utf8_char     used for utf8 character bytes, NULL if not relevant  
2506    ptr           next character in pattern    ptr           next character in pattern
2507    options       options bits    options       options bits
2508    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 2006  Returns:        TRUE if possessifying is Line 2511  Returns:        TRUE if possessifying is
2511  */  */
2512    
2513  static BOOL  static BOOL
2514  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
2515    const uschar *ptr, int options, compile_data *cd)    int options, compile_data *cd)
2516  {  {
2517  int next;  int c, next;
2518    int op_code = *previous++;
2519    
2520  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
2521    
# Line 2018  if ((options & PCRE_EXTENDED) != 0) Line 2524  if ((options & PCRE_EXTENDED) != 0)
2524    for (;;)    for (;;)
2525      {      {
2526      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2527      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2528        {        {
2529        while (*(++ptr) != 0)        ptr++;
2530          while (*ptr != 0)
2531            {
2532          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2533            ptr++;
2534    #ifdef SUPPORT_UTF8
2535            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2536    #endif
2537            }
2538        }        }
2539      else break;      else break;
2540      }      }
# Line 2030  if ((options & PCRE_EXTENDED) != 0) Line 2543  if ((options & PCRE_EXTENDED) != 0)
2543  /* 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
2544  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2545    
2546  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2547    {    {
2548    int temperrorcode = 0;    int temperrorcode = 0;
2549    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
# Line 2055  if ((options & PCRE_EXTENDED) != 0) Line 2568  if ((options & PCRE_EXTENDED) != 0)
2568    for (;;)    for (;;)
2569      {      {
2570      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2571      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2572        {        {
2573        while (*(++ptr) != 0)        ptr++;
2574          while (*ptr != 0)
2575            {
2576          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2577            ptr++;
2578    #ifdef SUPPORT_UTF8
2579            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2580    #endif
2581            }
2582        }        }
2583      else break;      else break;
2584      }      }
# Line 2066  if ((options & PCRE_EXTENDED) != 0) Line 2586  if ((options & PCRE_EXTENDED) != 0)
2586    
2587  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
2588    
2589  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2590    return FALSE;    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2591        return FALSE;
 /* Now compare the next item with the previous opcode. If the previous is a  
 positive single character match, "item" either contains the character or, if  
 "item" is greater than 127 in utf8 mode, the character's bytes are in  
 utf8_char. */  
   
2592    
2593  /* Handle cases when the next item is a character. */  /* Now compare the next item with the previous opcode. First, handle cases when
2594    the next item is a character. */
2595    
2596  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
2597    {    {
2598    case OP_CHAR:    case OP_CHAR:
2599  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2600    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2601    #else
2602      c = *previous;
2603  #endif  #endif
2604    return item != next;    return c != next;
2605    
2606    /* For CHARNC (caseless character) we must check the other case. If we have    /* For CHARNC (caseless character) we must check the other case. If we have
2607    Unicode property support, we can use it to test the other case of    Unicode property support, we can use it to test the other case of
# Line 2091  if (next >= 0) switch(op_code) Line 2609  if (next >= 0) switch(op_code)
2609    
2610    case OP_CHARNC:    case OP_CHARNC:
2611  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2612    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2613    #else
2614      c = *previous;
2615  #endif  #endif
2616    if (item == next) return FALSE;    if (c == next) return FALSE;
2617  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2618    if (utf8)    if (utf8)
2619      {      {
2620      unsigned int othercase;      unsigned int othercase;
2621      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2622  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2623      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
2624  #else  #else
2625      othercase = NOTACHAR;      othercase = NOTACHAR;
2626  #endif  #endif
2627      return (unsigned int)item != othercase;      return (unsigned int)c != othercase;
2628      }      }
2629    else    else
2630  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2631    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2632    
2633    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT, its data is always a single-byte character. */
2634    
2635    case OP_NOT:    case OP_NOT:
2636    if (next < 0) return FALSE;  /* Not a character */    if ((c = *previous) == next) return TRUE;
   if (item == next) return TRUE;  
2637    if ((options & PCRE_CASELESS) == 0) return FALSE;    if ((options & PCRE_CASELESS) == 0) return FALSE;
2638  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2639    if (utf8)    if (utf8)
# Line 2122  if (next >= 0) switch(op_code) Line 2641  if (next >= 0) switch(op_code)
2641      unsigned int othercase;      unsigned int othercase;
2642      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2643  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2644      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
2645  #else  #else
2646      othercase = NOTACHAR;      othercase = NOTACHAR;
2647  #endif  #endif
2648      return (unsigned int)item == othercase;      return (unsigned int)c == othercase;
2649      }      }
2650    else    else
2651  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2652    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2653    
2654      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2655      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
2656    
2657    case OP_DIGIT:    case OP_DIGIT:
2658    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
# Line 2173  if (next >= 0) switch(op_code) Line 2695  if (next >= 0) switch(op_code)
2695      case 0x202f:      case 0x202f:
2696      case 0x205f:      case 0x205f:
2697      case 0x3000:      case 0x3000:
2698      return op_code != OP_HSPACE;      return op_code == OP_NOT_HSPACE;
2699      default:      default:
2700      return op_code == OP_HSPACE;      return op_code != OP_NOT_HSPACE;
2701      }      }
2702    
2703      case OP_ANYNL:
2704    case OP_VSPACE:    case OP_VSPACE:
2705    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
2706    switch(next)    switch(next)
# Line 2189  if (next >= 0) switch(op_code) Line 2712  if (next >= 0) switch(op_code)
2712      case 0x85:      case 0x85:
2713      case 0x2028:      case 0x2028:
2714      case 0x2029:      case 0x2029:
2715      return op_code != OP_VSPACE;      return op_code == OP_NOT_VSPACE;
2716      default:      default:
2717      return op_code == OP_VSPACE;      return op_code != OP_NOT_VSPACE;
2718      }      }
2719    
2720    #ifdef SUPPORT_UCP
2721      case OP_PROP:
2722      return check_char_prop(next, previous[0], previous[1], FALSE);
2723    
2724      case OP_NOTPROP:
2725      return check_char_prop(next, previous[0], previous[1], TRUE);
2726    #endif
2727    
2728    default:    default:
2729    return FALSE;    return FALSE;
2730    }    }
2731    
2732    
2733  /* 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
2734    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
2735    generated only when PCRE_UCP is *not* set, that is, when only ASCII
2736    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
2737    replaced by OP_PROP codes when PCRE_UCP is set. */
2738    
2739  switch(op_code)  switch(op_code)
2740    {    {
2741    case OP_CHAR:    case OP_CHAR:
2742    case OP_CHARNC:    case OP_CHARNC:
2743  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2744    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2745    #else
2746      c = *previous;
2747  #endif  #endif
2748    switch(-next)    switch(-next)
2749      {      {
2750      case ESC_d:      case ESC_d:
2751      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
2752    
2753      case ESC_D:      case ESC_D:
2754      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
2755    
2756      case ESC_s:      case ESC_s:
2757      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
2758    
2759      case ESC_S:      case ESC_S:
2760      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
2761    
2762      case ESC_w:      case ESC_w:
2763      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
2764    
2765      case ESC_W:      case ESC_W:
2766      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
2767    
2768      case ESC_h:      case ESC_h:
2769      case ESC_H:      case ESC_H:
2770      switch(item)      switch(c)
2771        {        {
2772        case 0x09:        case 0x09:
2773        case 0x20:        case 0x20:
# Line 2258  switch(op_code) Line 2795  switch(op_code)
2795    
2796      case ESC_v:      case ESC_v:
2797      case ESC_V:      case ESC_V:
2798      switch(item)      switch(c)
2799        {        {
2800        case 0x0a:        case 0x0a:
2801        case 0x0b:        case 0x0b:
# Line 2272  switch(op_code) Line 2809  switch(op_code)
2809        return -next == ESC_v;        return -next == ESC_v;
2810        }        }
2811    
2812        /* When PCRE_UCP is set, these values get generated for \d etc. Find
2813        their substitutions and process them. The result will always be either
2814        -ESC_p or -ESC_P. Then fall through to process those values. */
2815    
2816    #ifdef SUPPORT_UCP
2817        case ESC_du:
2818        case ESC_DU:
2819        case ESC_wu:
2820        case ESC_WU:
2821        case ESC_su:
2822        case ESC_SU:
2823          {
2824          int temperrorcode = 0;
2825          ptr = substitutes[-next - ESC_DU];
2826          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
2827          if (temperrorcode != 0) return FALSE;
2828          ptr++;    /* For compatibility */
2829          }
2830        /* Fall through */
2831    
2832        case ESC_p:
2833        case ESC_P:
2834          {
2835          int ptype, pdata, errorcodeptr;
2836          BOOL negated;
2837    
2838          ptr--;      /* Make ptr point at the p or P */
2839          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
2840          if (ptype < 0) return FALSE;
2841          ptr++;      /* Point past the final curly ket */
2842    
2843          /* If the property item is optional, we have to give up. (When generated
2844          from \d etc by PCRE_UCP, this test will have been applied much earlier,
2845          to the original \d etc. At this point, ptr will point to a zero byte. */
2846    
2847          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2848            strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2849              return FALSE;
2850    
2851          /* Do the property check. */
2852    
2853          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
2854          }
2855    #endif
2856    
2857      default:      default:
2858      return FALSE;      return FALSE;
2859      }      }
2860    
2861      /* In principle, support for Unicode properties should be integrated here as
2862      well. It means re-organizing the above code so as to get hold of the property
2863      values before switching on the op-code. However, I wonder how many patterns
2864      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
2865      these op-codes are never generated.) */
2866    
2867    case OP_DIGIT:    case OP_DIGIT:
2868    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2869           next == -ESC_h || next == -ESC_v;           next == -ESC_h || next == -ESC_v || next == -ESC_R;
2870    
2871    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
2872    return next == -ESC_d;    return next == -ESC_d;
2873    
2874    case OP_WHITESPACE:    case OP_WHITESPACE:
2875    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
2876    
2877    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
2878    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2879    
2880    case OP_HSPACE:    case OP_HSPACE:
2881    return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
2882             next == -ESC_w || next == -ESC_v || next == -ESC_R;
2883    
2884    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
2885    return next == -ESC_h;    return next == -ESC_h;
2886    
2887    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
2888      case OP_ANYNL:
2889    case OP_VSPACE:    case OP_VSPACE:
2890    return next == -ESC_V || next == -ESC_d || next == -ESC_w;    return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2891    
2892    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
2893    return next == -ESC_v;    return next == -ESC_v || next == -ESC_R;
2894    
2895    case OP_WORDCHAR:    case OP_WORDCHAR:
2896    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
2897             next == -ESC_v || next == -ESC_R;
2898    
2899    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
2900    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2367  BOOL inescq = FALSE; Line 2958  BOOL inescq = FALSE;
2958  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
2959  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
2960  const uschar *tempptr;  const uschar *tempptr;
2961    const uschar *nestptr = NULL;
2962  uschar *previous = NULL;  uschar *previous = NULL;
2963  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
2964  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
# Line 2376  uschar classbits[32]; Line 2968  uschar classbits[32];
2968  BOOL class_utf8;  BOOL class_utf8;
2969  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
2970  uschar *class_utf8data;  uschar *class_utf8data;
2971    uschar *class_utf8data_base;
2972  uschar utf8_char[6];  uschar utf8_char[6];
2973  #else  #else
2974  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
2975  uschar *utf8_char = NULL;  uschar *utf8_char = NULL;
2976  #endif  #endif
2977    
2978  #ifdef DEBUG  #ifdef PCRE_DEBUG
2979  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2980  #endif  #endif
2981    
# Line 2436  for (;; ptr++) Line 3029  for (;; ptr++)
3029    
3030    c = *ptr;    c = *ptr;
3031    
3032      /* If we are at the end of a nested substitution, revert to the outer level
3033      string. Nesting only happens one level deep. */
3034    
3035      if (c == 0 && nestptr != NULL)
3036        {
3037        ptr = nestptr;
3038        nestptr = NULL;
3039        c = *ptr;
3040        }
3041    
3042    /* 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
3043    previous cycle of this loop. */    previous cycle of this loop. */
3044    
3045    if (lengthptr != NULL)    if (lengthptr != NULL)
3046      {      {
3047  #ifdef DEBUG  #ifdef PCRE_DEBUG
3048      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3049  #endif  #endif
3050      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
3051        {        {
3052        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3053        goto FAILED;        goto FAILED;
# Line 2466  for (;; ptr++) Line 3069  for (;; ptr++)
3069        goto FAILED;        goto FAILED;
3070        }        }
3071    
3072      *lengthptr += code - last_code;      *lengthptr += (int)(code - last_code);
3073      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));
3074    
3075      /* If "previous" is set and it is not at the start of the work space, move      /* If "previous" is set and it is not at the start of the work space, move
# Line 2493  for (;; ptr++) Line 3096  for (;; ptr++)
3096    /* 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
3097    reference list. */    reference list. */
3098    
3099    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
3100      {      {
3101      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3102      goto FAILED;      goto FAILED;
# Line 2503  for (;; ptr++) Line 3106  for (;; ptr++)
3106    
3107    if (inescq && c != 0)    if (inescq && c != 0)
3108      {      {
3109      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3110        {        {
3111        inescq = FALSE;        inescq = FALSE;
3112        ptr++;        ptr++;
# Line 2529  for (;; ptr++) Line 3132  for (;; ptr++)
3132    /* 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
3133    a quantifier. */    a quantifier. */
3134    
3135    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
3136      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
3137        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
3138    
3139    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
3140         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2545  for (;; ptr++) Line 3149  for (;; ptr++)
3149    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3150      {      {
3151      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
3152      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
3153        {        {
3154        while (*(++ptr) != 0)        ptr++;
3155          while (*ptr != 0)
3156          {          {
3157          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3158            ptr++;
3159    #ifdef SUPPORT_UTF8
3160            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
3161    #endif
3162          }          }
3163        if (*ptr != 0) continue;        if (*ptr != 0) continue;
3164    
# Line 2570  for (;; ptr++) Line 3179  for (;; ptr++)
3179      {      {
3180      /* ===================================================================*/      /* ===================================================================*/
3181      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
3182      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
3183      case ')':      case CHAR_RIGHT_PARENTHESIS:
3184      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
3185      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
3186      *codeptr = code;      *codeptr = code;
# Line 2583  for (;; ptr++) Line 3192  for (;; ptr++)
3192          *errorcodeptr = ERR20;          *errorcodeptr = ERR20;
3193          goto FAILED;          goto FAILED;
3194          }          }
3195        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += (int)(code - last_code);   /* To include callout length */
3196        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
3197        }        }
3198      return TRUE;      return TRUE;
# Line 2593  for (;; ptr++) Line 3202  for (;; ptr++)
3202      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
3203      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3204    
3205      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
3206      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3207        {        {
3208        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
# Line 2602  for (;; ptr++) Line 3211  for (;; ptr++)
3211      *code++ = OP_CIRC;      *code++ = OP_CIRC;
3212      break;      break;
3213    
3214      case '$':      case CHAR_DOLLAR_SIGN:
3215      previous = NULL;      previous = NULL;
3216      *code++ = OP_DOLL;      *code++ = OP_DOLL;
3217      break;      break;
# Line 2610  for (;; ptr++) Line 3219  for (;; ptr++)
3219      /* 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
3220      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
3221    
3222      case '.':      case CHAR_DOT:
3223      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3224      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
3225      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3226      previous = code;      previous = code;
3227      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
3228      break;      break;
3229    
3230    
# Line 2630  for (;; ptr++) Line 3239  for (;; ptr++)
3239      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,
3240      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
3241      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.
     */  
3242    
3243      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
3244        default (Perl) mode, it is treated as a data character. */
3245    
3246        case CHAR_RIGHT_SQUARE_BRACKET:
3247        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3248          {
3249          *errorcodeptr = ERR64;
3250          goto FAILED;
3251          }
3252        goto NORMAL_CHAR;
3253    
3254        case CHAR_LEFT_SQUARE_BRACKET:
3255      previous = code;      previous = code;
3256    
3257      /* 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
3258      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. */
3259    
3260      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3261             ptr[1] == CHAR_EQUALS_SIGN) &&
3262          check_posix_syntax(ptr, &tempptr))          check_posix_syntax(ptr, &tempptr))
3263        {        {
3264        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
3265        goto FAILED;        goto FAILED;
3266        }        }
3267    
# Line 2653  for (;; ptr++) Line 3273  for (;; ptr++)
3273      for (;;)      for (;;)
3274        {        {
3275        c = *(++ptr);        c = *(++ptr);
3276        if (c == '\\')        if (c == CHAR_BACKSLASH)
3277          {          {
3278          if (ptr[1] == 'E') ptr++;          if (ptr[1] == CHAR_E)
3279            else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;            ptr++;
3280              else break;          else if (strncmp((const char *)ptr+1,
3281                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
3282              ptr += 3;
3283            else
3284              break;
3285          }          }
3286        else if (!negate_class && c == '^')        else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
3287          negate_class = TRUE;          negate_class = TRUE;
3288        else break;        else break;
3289        }        }
3290    
3291        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
3292        an initial ']' is taken as a data character -- the code below handles
3293        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3294        [^] must match any character, so generate OP_ALLANY. */
3295    
3296        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3297            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3298          {
3299          *code++ = negate_class? OP_ALLANY : OP_FAIL;
3300          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3301          zerofirstbyte = firstbyte;
3302          break;
3303          }
3304    
3305      /* If a class contains a negative special such as \S, we need to flip the      /* If a class contains a negative special such as \S, we need to flip the
3306      negation flag at the end, so that support for characters > 255 works      negation flag at the end, so that support for characters > 255 works
3307      correctly (they are all included in the class). */      correctly (they are all included in the class). */
# Line 2687  for (;; ptr++) Line 3325  for (;; ptr++)
3325  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3326      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
3327      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
3328        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
3329  #endif  #endif
3330    
3331      /* 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 2702  for (;; ptr++) Line 3341  for (;; ptr++)
3341          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3342          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3343          }          }
3344    
3345          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
3346          data and reset the pointer. This is so that very large classes that
3347          contain a zillion UTF-8 characters no longer overwrite the work space
3348          (which is on the stack). */
3349    
3350          if (lengthptr != NULL)
3351            {
3352            *lengthptr += class_utf8data - class_utf8data_base;
3353            class_utf8data = class_utf8data_base;
3354            }
3355    
3356  #endif  #endif
3357    
3358        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
3359    
3360        if (inescq)        if (inescq)
3361          {          {
3362          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3363            {            {
3364            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
3365            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2723  for (;; ptr++) Line 3374  for (;; ptr++)
3374        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3375        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3376    
3377        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3378            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3379            check_posix_syntax(ptr, &tempptr))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3380          {          {
3381          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3382          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3383          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3384          uschar pbits[32];          uschar pbits[32];
3385    
3386          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3387            {            {
3388            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3389            goto FAILED;            goto FAILED;
3390            }            }
3391    
3392          ptr += 2;          ptr += 2;
3393          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3394            {            {
3395            local_negate = TRUE;            local_negate = TRUE;
3396            should_flip_negation = TRUE;  /* Note negative special */            should_flip_negation = TRUE;  /* Note negative special */
3397            ptr++;            ptr++;
3398            }            }
3399    
3400          posix_class = check_posix_name(ptr, tempptr - ptr);          posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3401          if (posix_class < 0)          if (posix_class < 0)
3402            {            {
3403            *errorcodeptr = ERR30;            *errorcodeptr = ERR30;
# Line 2760  for (;; ptr++) Line 3411  for (;; ptr++)
3411          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3412            posix_class = 0;            posix_class = 0;
3413    
3414          /* 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
3415          because we may be adding and subtracting from it, and we don't want to          different escape sequences that use Unicode properties. */
3416          subtract bits that may be in the main map already. At the end we or the  
3417          result into the bit map that is being built. */  #ifdef SUPPORT_UCP
3418            if ((options & PCRE_UCP) != 0)
3419              {
3420              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3421              if (posix_substitutes[pc] != NULL)
3422                {
3423                nestptr = tempptr + 1;
3424                ptr = posix_substitutes[pc] - 1;
3425                continue;
3426                }
3427              }
3428    #endif
3429            /* In the non-UCP case, we build the bit map for the POSIX class in a
3430            chunk of local store because we may be adding and subtracting from it,
3431            and we don't want to subtract bits that may be in the main map already.
3432            At the end we or the result into the bit map that is being built. */
3433    
3434          posix_class *= 3;          posix_class *= 3;
3435    
# Line 2807  for (;; ptr++) Line 3473  for (;; ptr++)
3473    
3474        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3475        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
3476        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
3477        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
3478        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
3479        character in them, so set class_charcount bigger than one. */        are either treated as literal characters (by default), or are faulted if
3480          PCRE_EXTRA is set. */
3481    
3482        if (c == '\\')        if (c == CHAR_BACKSLASH)
3483          {          {
3484          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3485          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3486    
3487          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 */  
3488          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3489            {            {
3490            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3491              {              {
3492              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3493              }              }
# Line 2836  for (;; ptr++) Line 3501  for (;; ptr++)
3501            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
3502            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3503    
3504            /* Save time by not doing this in the pre-compile phase. */            switch (-c)
   
           if (lengthptr == NULL) switch (-c)  
3505              {              {
3506    #ifdef SUPPORT_UCP
3507                case ESC_du:     /* These are the values given for \d etc */
3508                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3509                case ESC_wu:     /* escape sequence with an appropriate \p */
3510                case ESC_WU:     /* or \P to test Unicode properties instead */
3511                case ESC_su:     /* of the default ASCII testing. */
3512                case ESC_SU:
3513                nestptr = ptr;
3514                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3515                class_charcount -= 2;                /* Undo! */
3516                continue;
3517    #endif
3518              case ESC_d:              case ESC_d:
3519              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3520              continue;              continue;
# Line 2858  for (;; ptr++) Line 3533  for (;; ptr++)
3533              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3534              continue;              continue;
3535    
3536                /* Perl 5.004 onwards omits VT from \s, but we must preserve it
3537                if it was previously set by something earlier in the character
3538                class. */
3539    
3540              case ESC_s:              case ESC_s:
3541              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];              classbits[0] |= cbits[cbit_space];
3542              classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= cbits[cbit_space+1] & ~0x08;
3543                for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3544              continue;              continue;
3545    
3546              case ESC_S:              case ESC_S:
# Line 2869  for (;; ptr++) Line 3549  for (;; ptr++)
3549              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3550              continue;              continue;
3551    
3552              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)  
             {  
3553              SETBIT(classbits, 0x09); /* VT */              SETBIT(classbits, 0x09); /* VT */
3554              SETBIT(classbits, 0x20); /* SPACE */              SETBIT(classbits, 0x20); /* SPACE */
3555              SETBIT(classbits, 0xa0); /* NSBP */              SETBIT(classbits, 0xa0); /* NSBP */
# Line 2906  for (;; ptr++) Line 3573  for (;; ptr++)
3573                }                }
3574  #endif  #endif
3575              continue;              continue;
             }  
3576    
3577            if (-c == ESC_H)              case ESC_H:
             {  
3578              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3579                {                {
3580                int x = 0xff;                int x = 0xff;
# Line 2951  for (;; ptr++) Line 3616  for (;; ptr++)
3616                }                }
3617  #endif  #endif
3618              continue;              continue;
             }  
3619    
3620            if (-c == ESC_v)              case ESC_v:
             {  
3621              SETBIT(classbits, 0x0a); /* LF */              SETBIT(classbits, 0x0a); /* LF */
3622              SETBIT(classbits, 0x0b); /* VT */              SETBIT(classbits, 0x0b); /* VT */
3623              SETBIT(classbits, 0x0c); /* FF */              SETBIT(classbits, 0x0c); /* FF */
# Line 2970  for (;; ptr++) Line 3633  for (;; ptr++)
3633                }                }
3634  #endif  #endif
3635              continue;              continue;
             }  
3636    
3637            if (-c == ESC_V)              case ESC_V:
             {  
3638              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3639                {                {
3640                int x = 0xff;                int x = 0xff;
# Line 3003  for (;; ptr++) Line 3664  for (;; ptr++)
3664                }                }
3665  #endif  #endif
3666              continue;              continue;
             }  
   
           /* We need to deal with \P and \p in both phases. */  
3667    
3668  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3669            if (-c == ESC_p || -c == ESC_P)              case ESC_p:
3670              {              case ESC_P:
3671              BOOL negated;                {
3672              int pdata;                BOOL negated;
3673              int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                int pdata;
3674              if (ptype < 0) goto FAILED;                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3675              class_utf8 = TRUE;                if (ptype < 0) goto FAILED;
3676              *class_utf8data++ = ((-c == ESC_p) != negated)?                class_utf8 = TRUE;
3677                XCL_PROP : XCL_NOTPROP;                *class_utf8data++ = ((-c == ESC_p) != negated)?
3678              *class_utf8data++ = ptype;                  XCL_PROP : XCL_NOTPROP;
3679              *class_utf8data++ = pdata;                *class_utf8data++ = ptype;
3680              class_charcount -= 2;   /* Not a < 256 character */                *class_utf8data++ = pdata;
3681              continue;                class_charcount -= 2;   /* Not a < 256 character */
3682              }                continue;
3683                  }
3684  #endif  #endif
3685            /* Unrecognized escapes are faulted if PCRE is running in its              /* Unrecognized escapes are faulted if PCRE is running in its
3686            strict mode. By default, for compatibility with Perl, they are              strict mode. By default, for compatibility with Perl, they are
3687            treated as literals. */              treated as literals. */
3688    
3689            if ((options & PCRE_EXTRA) != 0)              default:
3690              {              if ((options & PCRE_EXTRA) != 0)
3691              *errorcodeptr = ERR7;                {
3692              goto FAILED;                *errorcodeptr = ERR7;
3693                  goto FAILED;
3694                  }
3695                class_charcount -= 2;  /* Undo the default count from above */
3696                c = *ptr;              /* Get the final character and fall through */
3697                break;
3698              }              }
   
           class_charcount -= 2;  /* Undo the default count from above */  
           c = *ptr;              /* Get the final character and fall through */  
3699            }            }
3700    
3701          /* 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 3048  for (;; ptr++) Line 3709  for (;; ptr++)
3709        entirely. The code for handling \Q and \E is messy. */        entirely. The code for handling \Q and \E is messy. */
3710    
3711        CHECK_RANGE:        CHECK_RANGE:
3712        while (ptr[1] == '\\' && ptr[2] == 'E')        while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3713          {          {
3714          inescq = FALSE;          inescq = FALSE;
3715          ptr += 2;          ptr += 2;
# Line 3058  for (;; ptr++) Line 3719  for (;; ptr++)
3719    
3720        /* Remember \r or \n */        /* Remember \r or \n */
3721    
3722        if (c == '\r' || c == '\n') cd->external_flags |= PCRE_HASCRORLF;        if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3723    
3724        /* Check for range */        /* Check for range */
3725    
3726        if (!inescq && ptr[1] == '-')        if (!inescq && ptr[1] == CHAR_MINUS)
3727          {          {
3728          int d;          int d;
3729          ptr += 2;          ptr += 2;
3730          while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3731    
3732          /* 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
3733          mode. */          mode. */
3734    
3735          while (*ptr == '\\' && ptr[1] == 'Q')          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3736            {            {
3737            ptr += 2;            ptr += 2;
3738            if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }            if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3739                { ptr += 2; continue; }
3740            inescq = TRUE;            inescq = TRUE;
3741            break;            break;
3742            }            }
3743    
3744          if (*ptr == 0 || (!inescq && *ptr == ']'))          if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3745            {            {
3746            ptr = oldptr;            ptr = oldptr;
3747            goto LONE_SINGLE_CHARACTER;            goto LONE_SINGLE_CHARACTER;
# Line 3098  for (;; ptr++) Line 3760  for (;; ptr++)
3760          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
3761          in such circumstances. */          in such circumstances. */
3762    
3763          if (!inescq && d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3764            {            {
3765            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3766            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3767    
3768            /* \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 */  
3769    
3770            if (d < 0)            if (d < 0)
3771              {              {
3772              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  
3773                {                {
3774                ptr = oldptr;                ptr = oldptr;
3775                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 3131  for (;; ptr++) Line 3790  for (;; ptr++)
3790    
3791          /* Remember \r or \n */          /* Remember \r or \n */
3792    
3793          if (d == '\r' || d == '\n') cd->external_flags |= PCRE_HASCRORLF;          if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3794    
3795          /* 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
3796          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 3251  for (;; ptr++) Line 3910  for (;; ptr++)
3910          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3911            {            {
3912            unsigned int othercase;            unsigned int othercase;
3913            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)            if ((othercase = UCD_OTHERCASE(c)) != c)
3914              {              {
3915              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3916              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 3276  for (;; ptr++) Line 3935  for (;; ptr++)
3935          }          }
3936        }        }
3937    
3938      /* 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.
3939        If we are at the end of an internal nested string, revert to the outer
3940        string. */
3941    
3942        while (((c = *(++ptr)) != 0 ||
3943               (nestptr != NULL &&
3944                 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != 0)) &&
3945               (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
3946    
3947      while ((c = *(++ptr)) != 0 && (c != ']' || inescq));      /* Check for missing terminating ']' */
3948    
3949      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)
3950        {        {
3951        *errorcodeptr = ERR6;        *errorcodeptr = ERR6;
3952        goto FAILED;        goto FAILED;
3953        }        }
3954    
   
 /* 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  
   
   
3955      /* 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
3956      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
3957      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 3368  we set the flag only if there is a liter Line 4015  we set the flag only if there is a liter
4015    
4016      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
4017      extended class, with its own opcode, unless there was a negated special      extended class, with its own opcode, unless there was a negated special
4018      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
4019      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
4020      (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
4021      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
4022        actual compiled code. */
4023    
4024  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4025      if (class_utf8 && !should_flip_negation)      if (class_utf8 && (!should_flip_negation || (options & PCRE_UCP) != 0))
4026        {        {
4027        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
4028        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 3400  we set the flag only if there is a liter Line 4048  we set the flag only if there is a liter
4048        }        }
4049  #endif  #endif
4050    
4051      /* 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
4052      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
4053      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
4054      map into the code vector, negating it if necessary. */      (non-UCP) in the class. Then copy the 32-byte map into the code vector,
4055        negating it if necessary. */
4056    
4057      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
4058      if (negate_class)      if (negate_class)
# Line 3423  we set the flag only if there is a liter Line 4072  we set the flag only if there is a liter
4072      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
4073      has been tested above. */      has been tested above. */
4074    
4075      case '{':      case CHAR_LEFT_CURLY_BRACKET:
4076      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
4077      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
4078      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
4079      goto REPEAT;      goto REPEAT;
4080    
4081      case '*':      case CHAR_ASTERISK:
4082      repeat_min = 0;      repeat_min = 0;
4083      repeat_max = -1;      repeat_max = -1;
4084      goto REPEAT;      goto REPEAT;
4085    
4086      case '+':      case CHAR_PLUS:
4087      repeat_min = 1;      repeat_min = 1;
4088      repeat_max = -1;      repeat_max = -1;
4089      goto REPEAT;      goto REPEAT;
4090    
4091      case '?':      case CHAR_QUESTION_MARK:
4092      repeat_min = 0;      repeat_min = 0;
4093      repeat_max = 1;      repeat_max = 1;
4094    
# Line 3474  we set the flag only if there is a liter Line 4123  we set the flag only if there is a liter
4123      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
4124      repeat type to the non-default. */      repeat type to the non-default. */
4125    
4126      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
4127        {        {
4128        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
4129        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
4130        ptr++;        ptr++;
4131        }        }
4132      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
4133        {        {
4134        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
4135        ptr++;        ptr++;
# Line 3527  we set the flag only if there is a liter Line 4176  we set the flag only if there is a liter
4176    
4177        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4178            repeat_max < 0 &&            repeat_max < 0 &&
4179            check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
             options, cd))  
4180          {          {
4181          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4182          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3549  we set the flag only if there is a liter Line 4197  we set the flag only if there is a liter
4197        c = previous[1];        c = previous[1];
4198        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4199            repeat_max < 0 &&            repeat_max < 0 &&
4200            check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4201          {          {
4202          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4203          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3573  we set the flag only if there is a liter Line 4221  we set the flag only if there is a liter
4221    
4222        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4223            repeat_max < 0 &&            repeat_max < 0 &&
4224            check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4225          {          {
4226          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4227          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3595  we set the flag only if there is a liter Line 4243  we set the flag only if there is a liter
4243    
4244        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
4245    
4246          /*--------------------------------------------------------------------*/
4247          /* This code is obsolete from release 8.00; the restriction was finally
4248          removed: */
4249    
4250        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4251        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4252    
4253        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4254          /*--------------------------------------------------------------------*/
4255    
4256        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
4257    
# Line 3745  we set the flag only if there is a liter Line 4398  we set the flag only if there is a liter
4398          goto END_REPEAT;          goto END_REPEAT;
4399          }          }
4400    
4401          /*--------------------------------------------------------------------*/
4402          /* This code is obsolete from release 8.00; the restriction was finally
4403          removed: */
4404    
4405        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4406        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4407    
4408        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4409          /*--------------------------------------------------------------------*/
4410    
4411        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4412          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 3773  we set the flag only if there is a liter Line 4431  we set the flag only if there is a liter
4431        {        {
4432        register int i;        register int i;
4433        int ketoffset = 0;        int ketoffset = 0;
4434        int len = code - previous;        int len = (int)(code - previous);
4435        uschar *bralink = NULL;        uschar *bralink = NULL;
4436    
4437        /* Repeating a DEFINE group is pointless */        /* Repeating a DEFINE group is pointless */
# Line 3794  we set the flag only if there is a liter Line 4452  we set the flag only if there is a liter
4452          {          {
4453          register uschar *ket = previous;          register uschar *ket = previous;
4454          do ket += GET(ket, 1); while (*ket != OP_KET);          do ket += GET(ket, 1); while (*ket != OP_KET);
4455          ketoffset = code - ket;          ketoffset = (int)(code - ket);
4456          }          }
4457    
4458        /* The case of a zero minimum is special because of the need to stick        /* The case of a zero minimum is special because of the need to stick
# Line 3806  we set the flag only if there is a liter Line 4464  we set the flag only if there is a liter
4464    
4465        if (repeat_min == 0)        if (repeat_min == 0)
4466          {          {
4467          /* 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
4468          altogether. */          output altogether, like this:
   
         if (repeat_max == 0)  
           {  
           code = previous;  
           goto END_REPEAT;  
           }  
4469    
4470          /* If the maximum is 1 or unlimited, we just have to stick in the          ** if (repeat_max == 0)
4471          BRAZERO and do no more at this point. However, we do need to adjust          **   {
4472          any OP_RECURSE calls inside the group that refer to the group itself or          **   code = previous;
4473          any internal or forward referenced group, because the offset is from          **   goto END_REPEAT;
4474          the start of the whole regex. Temporarily terminate the pattern while          **   }
4475          doing this. */  
4476            However, that fails when a group is referenced as a subroutine from
4477            elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it
4478            so that it is skipped on execution. As we don't have a list of which
4479            groups are referenced, we cannot do this selectively.
4480    
4481            If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4482            and do no more at this point. However, we do need to adjust any
4483            OP_RECURSE calls inside the group that refer to the group itself or any
4484            internal or forward referenced group, because the offset is from the
4485            start of the whole regex. Temporarily terminate the pattern while doing
4486            this. */
4487    
4488          if (repeat_max <= 1)          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
4489            {            {
4490            *code = OP_END;            *code = OP_END;
4491            adjust_recurse(previous, 1, utf8, cd, save_hwm);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
4492            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
4493            code++;            code++;
4494              if (repeat_max == 0)
4495                {
4496                *previous++ = OP_SKIPZERO;
4497                goto END_REPEAT;
4498                }
4499            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4500            }            }
4501    
# Line 3852  we set the flag only if there is a liter Line 4520  we set the flag only if there is a liter
4520            /* We chain together the bracket offset fields that have to be            /* We chain together the bracket offset fields that have to be
4521            filled in later when the ends of the brackets are reached. */            filled in later when the ends of the brackets are reached. */
4522    
4523            offset = (bralink == NULL)? 0 : previous - bralink;            offset = (bralink == NULL)? 0 : (int)(previous - bralink);
4524            bralink = previous;            bralink = previous;
4525            PUTINC(previous, 0, offset);            PUTINC(previous, 0, offset);
4526            }            }
# Line 3873  we set the flag only if there is a liter Line 4541  we set the flag only if there is a liter
4541            {            {
4542            /* 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
4543            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
4544            potential integer overflow. */            potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
4545              integer type when available, otherwise double. */
4546    
4547            if (lengthptr != NULL)            if (lengthptr != NULL)
4548              {              {
4549              int delta = (repeat_min - 1)*length_prevgroup;              int delta = (repeat_min - 1)*length_prevgroup;
4550              if ((double)(repeat_min - 1)*(double)length_prevgroup >              if ((INT64_OR_DOUBLE)(repeat_min - 1)*
4551                                                              (double)INT_MAX ||                    (INT64_OR_DOUBLE)length_prevgroup >
4552                        (INT64_OR_DOUBLE)INT_MAX ||
4553                  OFLOW_MAX - *lengthptr < delta)                  OFLOW_MAX - *lengthptr < delta)
4554                {                {
4555                *errorcodeptr = ERR20;                *errorcodeptr = ERR20;
# Line 3925  we set the flag only if there is a liter Line 4595  we set the flag only if there is a liter
4595          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
4596          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
4597          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
4598          paranoid checks to avoid integer overflow. */          paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
4599            a 64-bit integer type when available, otherwise double. */
4600    
4601          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4602            {            {
4603            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4604                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4605            if ((double)repeat_max *            if ((INT64_OR_DOUBLE)repeat_max *
4606                  (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)                  (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4607                    > (double)INT_MAX ||                    > (INT64_OR_DOUBLE)INT_MAX ||
4608                OFLOW_MAX - *lengthptr < delta)                OFLOW_MAX - *lengthptr < delta)
4609              {              {
4610              *errorcodeptr = ERR20;              *errorcodeptr = ERR20;
# Line 3958  we set the flag only if there is a liter Line 4629  we set the flag only if there is a liter
4629              {              {
4630              int offset;              int offset;
4631              *code++ = OP_BRA;              *code++ = OP_BRA;
4632              offset = (bralink == NULL)? 0 : code - bralink;              offset = (bralink == NULL)? 0 : (int)(code - bralink);
4633              bralink = code;              bralink = code;
4634              PUTINC(code, 0, offset);              PUTINC(code, 0, offset);
4635              }              }
# Line 3979  we set the flag only if there is a liter Line 4650  we set the flag only if there is a liter
4650          while (bralink != NULL)          while (bralink != NULL)
4651            {            {
4652            int oldlinkoffset;            int oldlinkoffset;
4653            int offset = code - bralink + 1;            int offset = (int)(code - bralink + 1);
4654            uschar *bra = code - offset;            uschar *bra = code - offset;
4655            oldlinkoffset = GET(bra, 1);            oldlinkoffset = GET(bra, 1);
4656            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;
# Line 4010  we set the flag only if there is a liter Line 4681  we set the flag only if there is a liter
4681            uschar *scode = bracode;            uschar *scode = bracode;
4682            do            do
4683              {              {
4684              if (could_be_empty_branch(scode, ketcode, utf8))              if (could_be_empty_branch(scode, ketcode, utf8, cd))
4685                {                {
4686                *bracode += OP_SBRA - OP_BRA;                *bracode += OP_SBRA - OP_BRA;
4687                break;                break;
# Line 4022  we set the flag only if there is a liter Line 4693  we set the flag only if there is a liter
4693          }          }
4694        }        }
4695    
4696        /* If previous is OP_FAIL, it was generated by an empty class [] in
4697        JavaScript mode. The other ways in which OP_FAIL can be generated, that is
4698        by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
4699        error above. We can just ignore the repeat in JS case. */
4700    
4701        else if (*previous == OP_FAIL) goto END_REPEAT;
4702    
4703      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
4704    
4705      else      else
# Line 4046  we set the flag only if there is a liter Line 4724  we set the flag only if there is a liter
4724      if (possessive_quantifier)      if (possessive_quantifier)
4725        {        {
4726        int len;        int len;
4727        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||  
4728            *tempcode == OP_NOTEXACT)        if (*tempcode == OP_TYPEEXACT)
4729          tempcode += _pcre_OP_lengths[*tempcode] +          tempcode += _pcre_OP_lengths[*tempcode] +
4730            ((*tempcode == OP_TYPEEXACT &&            ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
4731               (tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP))? 2:0);  
4732        len = code - tempcode;        else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
4733            {
4734            tempcode += _pcre_OP_lengths[*tempcode];
4735    #ifdef SUPPORT_UTF8
4736            if (utf8 && tempcode[-1] >= 0xc0)
4737              tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];
4738    #endif
4739            }
4740    
4741          len = (int)(code - tempcode);
4742        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4743          {          {
4744          case OP_STAR:  *tempcode = OP_POSSTAR; break;          case OP_STAR:  *tempcode = OP_POSSTAR; break;
# Line 4069  we set the flag only if there is a liter Line 4756  we set the flag only if there is a liter
4756          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4757          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4758    
4759            /* Because we are moving code along, we must ensure that any
4760            pending recursive references are updated. */
4761    
4762          default:          default:
4763            *code = OP_END;
4764            adjust_recurse(tempcode, 1 + LINK_SIZE, utf8, cd, save_hwm);
4765          memmove(tempcode + 1+LINK_SIZE, tempcode, len);          memmove(tempcode + 1+LINK_SIZE, tempcode, len);
4766          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
4767          len += 1 + LINK_SIZE;          len += 1 + LINK_SIZE;
# Line 4096  we set the flag only if there is a liter Line 4788  we set the flag only if there is a liter
4788      lookbehind or option setting or condition or all the other extended      lookbehind or option setting or condition or all the other extended
4789      parenthesis forms.  */      parenthesis forms.  */
4790    
4791      case '(':      case CHAR_LEFT_PARENTHESIS:
4792      newoptions = options;      newoptions = options;
4793      skipbytes = 0;      skipbytes = 0;
4794      bravalue = OP_CBRA;      bravalue = OP_CBRA;
# Line 4105  we set the flag only if there is a liter Line 4797  we set the flag only if there is a liter
4797    
4798      /* First deal with various "verbs" that can be introduced by '*'. */      /* First deal with various "verbs" that can be introduced by '*'. */
4799    
4800      if (*(++ptr) == '*' && (cd->ctypes[ptr[1]] & ctype_letter) != 0)      if (*(++ptr) == CHAR_ASTERISK &&
4801             ((cd->ctypes[ptr[1]] & ctype_letter) != 0 || ptr[1] == ':'))
4802        {        {
4803        int i, namelen;        int i, namelen;
4804          int arglen = 0;
4805        const char *vn = verbnames;        const char *vn = verbnames;
4806        const uschar *name = ++ptr;        const uschar *name = ptr + 1;
4807          const uschar *arg = NULL;
4808        previous = NULL;        previous = NULL;
4809        while ((cd->ctypes[*++ptr] & ctype_letter) != 0);        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
4810        if (*ptr == ':')        namelen = (int)(ptr - name);
4811    
4812          if (*ptr == CHAR_COLON)
4813          {          {
4814          *errorcodeptr = ERR59;   /* Not supported */          arg = ++ptr;
4815          goto FAILED;          while ((cd->ctypes[*ptr] & (ctype_letter|ctype_digit)) != 0
4816              || *ptr == '_') ptr++;
4817            arglen = (int)(ptr - arg);
4818          }          }
4819        if (*ptr != ')')  
4820          if (*ptr != CHAR_RIGHT_PARENTHESIS)
4821          {          {
4822          *errorcodeptr = ERR60;          *errorcodeptr = ERR60;
4823          goto FAILED;          goto FAILED;
4824          }          }
4825        namelen = ptr - name;  
4826          /* Scan the table of verb names */
4827    
4828        for (i = 0; i < verbcount; i++)        for (i = 0; i < verbcount; i++)
4829          {          {
4830          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
4831              strncmp((char *)name, vn, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
4832            {            {
4833            *code = verbs[i].op;            /* Check for open captures before ACCEPT */
4834            if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;  
4835            break;            if (verbs[i].op == OP_ACCEPT)
4836                {
4837                open_capitem *oc;
4838                cd->had_accept = TRUE;
4839                for (oc = cd->open_caps; oc != NULL; oc = oc->next)
4840                  {
4841                  *code++ = OP_CLOSE;
4842                  PUT2INC(code, 0, oc->number);
4843                  }
4844                }
4845    
4846              /* Handle the cases with/without an argument */
4847    
4848              if (arglen == 0)
4849                {
4850                if (verbs[i].op < 0)   /* Argument is mandatory */
4851                  {
4852                  *errorcodeptr = ERR66;
4853                  goto FAILED;
4854                  }
4855                *code = verbs[i].op;
4856                if (*code++ == OP_THEN)
4857                  {
4858                  PUT(code, 0, code - bcptr->current_branch - 1);
4859                  code += LINK_SIZE;
4860                  }
4861                }
4862    
4863              else
4864                {
4865                if (verbs[i].op_arg < 0)   /* Argument is forbidden */
4866                  {
4867                  *errorcodeptr = ERR59;
4868                  goto FAILED;
4869                  }
4870                *code = verbs[i].op_arg;
4871                if (*code++ == OP_THEN_ARG)
4872                  {
4873                  PUT(code, 0, code - bcptr->current_branch - 1);
4874                  code += LINK_SIZE;
4875                  }
4876                *code++ = arglen;
4877                memcpy(code, arg, arglen);
4878                code += arglen;
4879                *code++ = 0;
4880                }
4881    
4882              break;  /* Found verb, exit loop */
4883            }            }
4884    
4885          vn += verbs[i].len + 1;          vn += verbs[i].len + 1;
4886          }          }
4887        if (i < verbcount) continue;  
4888        *errorcodeptr = ERR60;        if (i < verbcount) continue;    /* Successfully handled a verb */
4889          *errorcodeptr = ERR60;          /* Verb not recognized */
4890        goto FAILED;        goto FAILED;
4891        }        }
4892    
4893      /* Deal with the extended parentheses; all are introduced by '?', and the      /* Deal with the extended parentheses; all are introduced by '?', and the
4894      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. */
4895    
4896      else if (*ptr == '?')      else if (*ptr == CHAR_QUESTION_MARK)
4897        {        {
4898        int i, set, unset, namelen;        int i, set, unset, namelen;
4899        int *optset;        int *optset;
# Line 4151  we set the flag only if there is a liter Line 4902  we set the flag only if there is a liter
4902    
4903        switch (*(++ptr))        switch (*(++ptr))
4904          {          {
4905          case '#':                 /* Comment; skip to ket */          case CHAR_NUMBER_SIGN:                 /* Comment; skip to ket */
4906          ptr++;          ptr++;
4907          while (*ptr != 0 && *ptr != ')') ptr++;          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
4908          if (*ptr == 0)          if (*ptr == 0)
4909            {            {
4910            *errorcodeptr = ERR18;            *errorcodeptr = ERR18;
# Line 4163  we set the flag only if there is a liter Line 4914  we set the flag only if there is a liter
4914    
4915    
4916          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4917          case '|':                 /* Reset capture count for each branch */          case CHAR_VERTICAL_LINE:  /* Reset capture count for each branch */
4918          reset_bracount = TRUE;          reset_bracount = TRUE;
4919          /* Fall through */          /* Fall through */
4920    
4921          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4922          case ':':                 /* Non-capturing bracket */          case CHAR_COLON:          /* Non-capturing bracket */
4923          bravalue = OP_BRA;          bravalue = OP_BRA;
4924          ptr++;          ptr++;
4925          break;          break;
4926    
4927    
4928          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4929          case '(':          case CHAR_LEFT_PARENTHESIS:
4930          bravalue = OP_COND;       /* Conditional group */          bravalue = OP_COND;       /* Conditional group */
4931    
4932          /* 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 4195  we set the flag only if there is a liter Line 4946  we set the flag only if there is a liter
4946          the switch. This will take control down to where bracketed groups,          the switch. This will take control down to where bracketed groups,
4947          including assertions, are processed. */          including assertions, are processed. */
4948    
4949          if (ptr[1] == '?' && (ptr[2] == '=' || ptr[2] == '!' || ptr[2] == '<'))          if (ptr[1] == CHAR_QUESTION_MARK && (ptr[2] == CHAR_EQUALS_SIGN ||
4950                ptr[2] == CHAR_EXCLAMATION_MARK || ptr[2] == CHAR_LESS_THAN_SIGN))
4951            break;            break;
4952    
4953          /* 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 4207  we set the flag only if there is a liter Line 4959  we set the flag only if there is a liter
4959    
4960          /* Check for a test for recursion in a named group. */          /* Check for a test for recursion in a named group. */
4961    
4962          if (ptr[1] == 'R' && ptr[2] == '&')          if (ptr[1] == CHAR_R && ptr[2] == CHAR_AMPERSAND)
4963            {            {
4964            terminator = -1;            terminator = -1;
4965            ptr += 2;            ptr += 2;
# Line 4217  we set the flag only if there is a liter Line 4969  we set the flag only if there is a liter
4969          /* 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
4970          syntax (?(<name>) or (?('name') */          syntax (?(<name>) or (?('name') */
4971    
4972          else if (ptr[1] == '<')          else if (ptr[1] == CHAR_LESS_THAN_SIGN)
4973            {            {
4974            terminator = '>';            terminator = CHAR_GREATER_THAN_SIGN;
4975            ptr++;            ptr++;
4976            }            }
4977          else if (ptr[1] == '\'')          else if (ptr[1] == CHAR_APOSTROPHE)
4978            {            {
4979            terminator = '\'';            terminator = CHAR_APOSTROPHE;
4980            ptr++;            ptr++;
4981            }            }
4982          else          else
4983            {            {
4984            terminator = 0;            terminator = 0;
4985            if (ptr[1] == '-' || ptr[1] == '+') refsign = *(++ptr);            if (ptr[1] == CHAR_MINUS || ptr[1] == CHAR_PLUS) refsign = *(++ptr);
4986            }            }
4987    
4988          /* 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 4250  we set the flag only if there is a liter Line 5002  we set the flag only if there is a liter
5002            {            {
5003            if (recno >= 0)            if (recno >= 0)
5004              recno = ((digitab[*ptr] & ctype_digit) != 0)?              recno = ((digitab[*ptr] & ctype_digit) != 0)?
5005                recno * 10 + *ptr - '0' : -1;                recno * 10 + *ptr - CHAR_0 : -1;
5006            ptr++;            ptr++;
5007            }            }
5008          namelen = ptr - name;          namelen = (int)(ptr - name);
5009    
5010          if ((terminator > 0 && *ptr++ != terminator) || *ptr++ != ')')          if ((terminator > 0 && *ptr++ != terminator) ||
5011                *ptr++ != CHAR_RIGHT_PARENTHESIS)
5012            {            {
5013            ptr--;      /* Error offset */            ptr--;      /* Error offset */
5014            *errorcodeptr = ERR26;            *errorcodeptr = ERR26;
# Line 4277  we set the flag only if there is a liter Line 5030  we set the flag only if there is a liter
5030              *errorcodeptr = ERR58;              *errorcodeptr = ERR58;
5031              goto FAILED;              goto FAILED;
5032              }              }
5033            recno = (refsign == '-')?            recno = (refsign == CHAR_MINUS)?
5034              cd->bracount - recno + 1 : recno +cd->bracount;              cd->bracount - recno + 1 : recno +cd->bracount;
5035            if (recno <= 0 || recno > cd->final_bracount)            if (recno <= 0 || recno > cd->final_bracount)
5036              {              {
# Line 4289  we set the flag only if there is a liter Line 5042  we set the flag only if there is a liter
5042            }            }
5043    
5044          /* Otherwise (did not start with "+" or "-"), start by looking for the          /* Otherwise (did not start with "+" or "-"), start by looking for the
5045          name. */          name. If we find a name, add one to the opcode to change OP_CREF or
5046            OP_RREF into OP_NCREF or OP_NRREF. These behave exactly the same,
5047            except they record that the reference was originally to a name. The
5048            information is used to check duplicate names. */
5049    
5050          slot = cd->name_table;          slot = cd->name_table;
5051          for (i = 0; i < cd->names_found; i++)          for (i = 0; i < cd->names_found; i++)
# Line 4304  we set the flag only if there is a liter Line 5060  we set the flag only if there is a liter
5060            {            {
5061            recno = GET2(slot, 0);            recno = GET2(slot, 0);
5062            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
5063              code[1+LINK_SIZE]++;
5064            }            }
5065    
5066          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
5067    
5068          else if ((i = find_parens(ptr, cd->bracount, name, namelen,          else if ((i = find_parens(cd, name, namelen,
5069                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0, utf8)) > 0)
5070            {            {
5071            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
5072              code[1+LINK_SIZE]++;
5073            }            }
5074    
5075          /* If terminator == 0 it means that the name followed directly after          /* If terminator == 0 it means that the name followed directly after
# Line 4329  we set the flag only if there is a liter Line 5087  we set the flag only if there is a liter
5087          /* Check for (?(R) for recursion. Allow digits after R to specify a          /* Check for (?(R) for recursion. Allow digits after R to specify a
5088          specific group number. */          specific group number. */
5089    
5090          else if (*name == 'R')          else if (*name == CHAR_R)
5091            {            {
5092            recno = 0;            recno = 0;
5093            for (i = 1; i < namelen; i++)            for (i = 1; i < namelen; i++)
# Line 4339  we set the flag only if there is a liter Line 5097  we set the flag only if there is a liter
5097                *errorcodeptr = ERR15;                *errorcodeptr = ERR15;
5098                goto FAILED;                goto FAILED;
5099                }                }
5100              recno = recno * 10 + name[i] - '0';              recno = recno * 10 + name[i] - CHAR_0;
5101              }              }
5102            if (recno == 0) recno = RREF_ANY;            if (recno == 0) recno = RREF_ANY;
5103            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */
# Line 4349  we set the flag only if there is a liter Line 5107  we set the flag only if there is a liter
5107          /* Similarly, check for the (?(DEFINE) "condition", which is always          /* Similarly, check for the (?(DEFINE) "condition", which is always
5108          false. */          false. */
5109    
5110          else if (namelen == 6 && strncmp((char *)name, "DEFINE", 6) == 0)          else if (namelen == 6 && strncmp((char *)name, STRING_DEFINE, 6) == 0)
5111            {            {
5112            code[1+LINK_SIZE] = OP_DEF;            code[1+LINK_SIZE] = OP_DEF;
5113            skipbytes = 1;            skipbytes = 1;
# Line 4374  we set the flag only if there is a liter Line 5132  we set the flag only if there is a liter
5132    
5133    
5134          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5135          case '=':                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
5136          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
5137          ptr++;          ptr++;
5138          break;          break;
5139    
5140    
5141          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5142          case '!':                 /* Negative lookahead */          case CHAR_EXCLAMATION_MARK:            /* Negative lookahead */
5143          ptr++;          ptr++;
5144          if (*ptr == ')')          /* Optimize (?!) */          if (*ptr == CHAR_RIGHT_PARENTHESIS)    /* Optimize (?!) */
5145            {            {
5146            *code++ = OP_FAIL;            *code++ = OP_FAIL;
5147            previous = NULL;            previous = NULL;
# Line 4394  we set the flag only if there is a liter Line 5152  we set the flag only if there is a liter
5152    
5153    
5154          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5155          case '<':                 /* Lookbehind or named define */          case CHAR_LESS_THAN_SIGN:              /* Lookbehind or named define */
5156          switch (ptr[1])          switch (ptr[1])
5157            {            {
5158            case '=':               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
5159            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
5160            ptr += 2;            ptr += 2;
5161            break;            break;
5162    
5163            case '!':               /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
5164            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
5165            ptr += 2;            ptr += 2;
5166            break;            break;
# Line 4417  we set the flag only if there is a liter Line 5175  we set the flag only if there is a liter
5175    
5176    
5177          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5178          case '>':                 /* One-time brackets */          case CHAR_GREATER_THAN_SIGN:           /* One-time brackets */
5179          bravalue = OP_ONCE;          bravalue = OP_ONCE;
5180          ptr++;          ptr++;
5181          break;          break;
5182    
5183    
5184          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5185          case 'C':                 /* Callout - may be followed by digits; */          case CHAR_C:                 /* Callout - may be followed by digits; */
5186          previous_callout = code;  /* Save for later completion */          previous_callout = code;  /* Save for later completion */
5187          after_manual_callout = 1; /* Skip one item before completing */          after_manual_callout = 1; /* Skip one item before completing */
5188          *code++ = OP_CALLOUT;          *code++ = OP_CALLOUT;
5189            {            {
5190            int n = 0;            int n = 0;
5191            while ((digitab[*(++ptr)] & ctype_digit) != 0)            while ((digitab[*(++ptr)] & ctype_digit) != 0)
5192              n = n * 10 + *ptr - '0';              n = n * 10 + *ptr - CHAR_0;
5193            if (*ptr != ')')            if (*ptr != CHAR_RIGHT_PARENTHESIS)
5194              {              {
5195              *errorcodeptr = ERR39;              *errorcodeptr = ERR39;
5196              goto FAILED;              goto FAILED;
# Line 4443  we set the flag only if there is a liter Line 5201  we set the flag only if there is a liter
5201              goto FAILED;              goto FAILED;
5202              }              }
5203            *code++ = n;            *code++ = n;
5204            PUT(code, 0, ptr - cd->start_pattern + 1);  /* Pattern offset */            PUT(code, 0, (int)(ptr - cd->start_pattern + 1)); /* Pattern offset */
5205            PUT(code, LINK_SIZE, 0);                    /* Default length */            PUT(code, LINK_SIZE, 0);                          /* Default length */
5206            code += 2 * LINK_SIZE;            code += 2 * LINK_SIZE;
5207            }            }
5208          previous = NULL;          previous = NULL;
# Line 4452  we set the flag only if there is a liter Line 5210  we set the flag only if there is a liter
5210    
5211    
5212          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5213          case 'P':                 /* Python-style named subpattern handling */          case CHAR_P:              /* Python-style named subpattern handling */
5214          if (*(++ptr) == '=' || *ptr == '>')  /* Reference or recursion */          if (*(++ptr) == CHAR_EQUALS_SIGN ||
5215                *ptr == CHAR_GREATER_THAN_SIGN)  /* Reference or recursion */
5216            {            {
5217            is_recurse = *ptr == '>';            is_recurse = *ptr == CHAR_GREATER_THAN_SIGN;
5218            terminator = ')';            terminator = CHAR_RIGHT_PARENTHESIS;
5219            goto NAMED_REF_OR_RECURSE;            goto NAMED_REF_OR_RECURSE;
5220            }            }
5221          else if (*ptr != '<')    /* Test for Python-style definition */          else if (*ptr != CHAR_LESS_THAN_SIGN)  /* Test for Python-style defn */
5222            {            {
5223            *errorcodeptr = ERR41;            *errorcodeptr = ERR41;
5224            goto FAILED;            goto FAILED;
# Line 4469  we set the flag only if there is a liter Line 5228  we set the flag only if there is a liter
5228    
5229          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5230          DEFINE_NAME:    /* Come here from (?< handling */          DEFINE_NAME:    /* Come here from (?< handling */
5231          case '\'':          case CHAR_APOSTROPHE:
5232            {            {
5233            terminator = (*ptr == '<')? '>' : '\'';            terminator = (*ptr == CHAR_LESS_THAN_SIGN)?
5234                CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
5235            name = ++ptr;            name = ++ptr;
5236    
5237            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
5238            namelen = ptr - name;            namelen = (int)(ptr - name);
5239    
5240            /* In the pre-compile phase, just do a syntax check. */            /* In the pre-compile phase, just do a syntax check. */
5241    
# Line 4502  we set the flag only if there is a liter Line 5262  we set the flag only if there is a liter
5262                }                }
5263              }              }
5264    
5265            /* In the real compile, create the entry in the table */            /* In the real compile, create the entry in the table, maintaining
5266              alphabetical order. Duplicate names for different numbers are
5267              permitted only if PCRE_DUPNAMES is set. Duplicate names for the same
5268              number are always OK. (An existing number can be re-used if (?|
5269              appears in the pattern.) In either event, a duplicate name results in
5270              a duplicate entry in the table, even if the number is the same. This
5271              is because the number of names, and hence the table size, is computed
5272              in the pre-compile, and it affects various numbers and pointers which
5273              would all have to be modified, and the compiled code moved down, if
5274              duplicates with the same number were omitted from the table. This
5275              doesn't seem worth the hassle. However, *different* names for the
5276              same number are not permitted. */
5277    
5278            else            else
5279              {              {
5280                BOOL dupname = FALSE;
5281              slot = cd->name_table;              slot = cd->name_table;
5282    
5283              for (i = 0; i < cd->names_found; i++)              for (i = 0; i < cd->names_found; i++)
5284                {                {
5285                int crc = memcmp(name, slot+2, namelen);                int crc = memcmp(name, slot+2, namelen);
# Line 4514  we set the flag only if there is a liter Line 5287  we set the flag only if there is a liter
5287                  {                  {
5288                  if (slot[2+namelen] == 0)                  if (slot[2+namelen] == 0)
5289                    {                    {
5290                    if ((options & PCRE_DUPNAMES) == 0)                    if (GET2(slot, 0) != cd->bracount + 1 &&
5291                          (options & PCRE_DUPNAMES) == 0)
5292                      {                      {
5293                      *errorcodeptr = ERR43;                      *errorcodeptr = ERR43;
5294                      goto FAILED;                      goto FAILED;
5295                      }                      }
5296                      else dupname = TRUE;
5297                    }                    }
5298                  else crc = -1;      /* Current name is substring */                  else crc = -1;      /* Current name is a substring */
5299                  }                  }
5300    
5301                  /* Make space in the table and break the loop for an earlier
5302                  name. For a duplicate or later name, carry on. We do this for
5303                  duplicates so that in the simple case (when ?(| is not used) they
5304                  are in order of their numbers. */
5305    
5306                if (crc < 0)                if (crc < 0)
5307                  {                  {
5308                  memmove(slot + cd->name_entry_size, slot,                  memmove(slot + cd->name_entry_size, slot,
5309                    (cd->names_found - i) * cd->name_entry_size);                    (cd->names_found - i) * cd->name_entry_size);
5310                  break;                  break;
5311                  }                  }
5312    
5313                  /* Continue the loop for a later or duplicate name */
5314    
5315                slot += cd->name_entry_size;                slot += cd->name_entry_size;
5316                }                }
5317    
5318                /* For non-duplicate names, check for a duplicate number before
5319                adding the new name. */
5320    
5321                if (!dupname)
5322                  {
5323                  uschar *cslot = cd->name_table;
5324                  for (i = 0; i < cd->names_found; i++)
5325                    {
5326                    if (cslot != slot)
5327                      {
5328                      if (GET2(cslot, 0) == cd->bracount + 1)
5329                        {
5330                        *errorcodeptr = ERR65;
5331                        goto FAILED;
5332                        }
5333                      }
5334                    else i--;
5335                    cslot += cd->name_entry_size;
5336                    }
5337                  }
5338    
5339              PUT2(slot, 0, cd->bracount + 1);              PUT2(slot, 0, cd->bracount + 1);
5340              memcpy(slot + 2, name, namelen);              memcpy(slot + 2, name, namelen);
5341              slot[2+namelen] = 0;              slot[2+namelen] = 0;
5342              }              }
5343            }            }
5344    
5345          /* In both cases, count the number of names we've encountered. */          /* In both pre-compile and compile, count the number of names we've
5346            encountered. */
5347    
         ptr++;                    /* Move past > or ' */  
5348          cd->names_found++;          cd->names_found++;
5349            ptr++;                    /* Move past > or ' */
5350          goto NUMBERED_GROUP;          goto NUMBERED_GROUP;
5351    
5352    
5353          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5354          case '&':                 /* Perl recursion/subroutine syntax */          case CHAR_AMPERSAND:            /* Perl recursion/subroutine syntax */
5355          terminator = ')';          terminator = CHAR_RIGHT_PARENTHESIS;
5356          is_recurse = TRUE;          is_recurse = TRUE;
5357          /* Fall through */          /* Fall through */
5358    
# Line 4554  we set the flag only if there is a liter Line 5360  we set the flag only if there is a liter
5360          references (?P=name) and recursion (?P>name), as well as falling          references (?P=name) and recursion (?P>name), as well as falling
5361          through from the Perl recursion syntax (?&name). We also come here from          through from the Perl recursion syntax (?&name). We also come here from
5362          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}
5363          .NET syntax. */          .NET syntax, and the Oniguruma \g<...> and \g'...' subroutine syntax. */
5364    
5365          NAMED_REF_OR_RECURSE:          NAMED_REF_OR_RECURSE:
5366          name = ++ptr;          name = ++ptr;
5367          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
5368          namelen = ptr - name;          namelen = (int)(ptr - name);
5369    
5370          /* In the pre-compile phase, do a syntax check and set a dummy          /* In the pre-compile phase, do a syntax check and set a dummy
5371          reference number. */          reference number. */
# Line 4605  we set the flag only if there is a liter Line 5411  we set the flag only if there is a liter
5411              recno = GET2(slot, 0);              recno = GET2(slot, 0);
5412              }              }
5413            else if ((recno =                /* Forward back reference */            else if ((recno =                /* Forward back reference */
5414                      find_parens(ptr, cd->bracount, name, namelen,                      find_parens(cd, name, namelen,
5415                        (options & PCRE_EXTENDED) != 0)) <= 0)                        (options & PCRE_EXTENDED) != 0, utf8)) <= 0)
5416              {              {
5417              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
5418              goto FAILED;              goto FAILED;
# Line 4621  we set the flag only if there is a liter Line 5427  we set the flag only if there is a liter
5427    
5428    
5429          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5430          case 'R':                 /* Recursion */          case CHAR_R:              /* Recursion */
5431          ptr++;                    /* Same as (?0)      */          ptr++;                    /* Same as (?0)      */
5432          /* Fall through */          /* Fall through */
5433    
5434    
5435          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5436          case '-': case '+':          case CHAR_MINUS: case CHAR_PLUS:  /* Recursion or subroutine */
5437          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:
5438          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:
5439            {            {
5440            const uschar *called;            const uschar *called;
5441              terminator = CHAR_RIGHT_PARENTHESIS;
5442    
5443              /* Come here from the \g<...> and \g'...' code (Oniguruma
5444              compatibility). However, the syntax has been checked to ensure that
5445              the ... are a (signed) number, so that neither ERR63 nor ERR29 will
5446              be called on this path, nor with the jump to OTHER_CHAR_AFTER_QUERY
5447              ever be taken. */
5448    
5449            if ((refsign = *ptr) == '+')            HANDLE_NUMERICAL_RECURSION:
5450    
5451              if ((refsign = *ptr) == CHAR_PLUS)
5452              {              {
5453              ptr++;              ptr++;
5454              if ((digitab[*ptr] & ctype_digit) == 0)              if ((digitab[*ptr] & ctype_digit) == 0)
# Line 4642  we set the flag only if there is a liter Line 5457  we set the flag only if there is a liter
5457                goto FAILED;                goto FAILED;
5458                }                }
5459              }              }
5460            else if (refsign == '-')            else if (refsign == CHAR_MINUS)
5461              {              {
5462              if ((digitab[ptr[1]] & ctype_digit) == 0)              if ((digitab[ptr[1]] & ctype_digit) == 0)
5463                goto OTHER_CHAR_AFTER_QUERY;                goto OTHER_CHAR_AFTER_QUERY;
# Line 4651  we set the flag only if there is a liter Line 5466  we set the flag only if there is a liter
5466    
5467            recno = 0;            recno = 0;
5468            while((digitab[*ptr] & ctype_digit) != 0)            while((digitab[*ptr] & ctype_digit) != 0)
5469              recno = recno * 10 + *ptr++ - '0';              recno = recno * 10 + *ptr++ - CHAR_0;
5470    
5471            if (*ptr != ')')            if (*ptr != terminator)
5472              {              {
5473              *errorcodeptr = ERR29;              *errorcodeptr = ERR29;
5474              goto FAILED;              goto FAILED;
5475              }              }
5476    
5477            if (refsign == '-')            if (refsign == CHAR_MINUS)
5478              {              {
5479              if (recno == 0)              if (recno == 0)
5480                {                {
# Line 4673  we set the flag only if there is a liter Line 5488  we set the flag only if there is a liter
5488                goto FAILED;                goto FAILED;
5489                }                }
5490              }              }
5491            else if (refsign == '+')            else if (refsign == CHAR_PLUS)
5492              {              {
5493              if (recno == 0)              if (recno == 0)
5494                {                {
# Line 4700  we set the flag only if there is a liter Line 5515  we set the flag only if there is a liter
5515            if (lengthptr == NULL)            if (lengthptr == NULL)
5516              {              {
5517              *code = OP_END;              *code = OP_END;
5518              if (recno != 0) called = find_bracket(cd->start_code, utf8, recno);              if (recno != 0)
5519                  called = _pcre_find_bracket(cd->start_code, utf8, recno);
5520    
5521              /* Forward reference */              /* Forward reference */
5522    
5523              if (called == NULL)              if (called == NULL)
5524                {                {
5525                if (find_parens(ptr, cd->bracount, NULL, recno,                if (find_parens(cd, NULL, recno,
5526                     (options & PCRE_EXTENDED) != 0) < 0)                      (options & PCRE_EXTENDED) != 0, utf8) < 0)
5527                  {                  {
5528                  *errorcodeptr = ERR15;                  *errorcodeptr = ERR15;
5529                  goto FAILED;                  goto FAILED;
5530                  }                  }
5531    
5532                  /* Fudge the value of "called" so that when it is inserted as an
5533                  offset below, what it actually inserted is the reference number
5534                  of the group. */
5535    
5536                called = cd->start_code + recno;                called = cd->start_code + recno;
5537                PUTINC(cd->hwm, 0, code + 2 + LINK_SIZE - cd->start_code);                PUTINC(cd->hwm, 0, (int)(code + 2 + LINK_SIZE - cd->start_code));
5538                }                }
5539    
5540              /* If not a forward reference, and the subpattern is still open,              /* If not a forward reference, and the subpattern is still open,
# Line 4721  we set the flag only if there is a liter Line 5542  we set the flag only if there is a liter
5542              recursion that could loop for ever, and diagnose that case. */              recursion that could loop for ever, and diagnose that case. */
5543    
5544              else if (GET(called, 1) == 0 &&              else if (GET(called, 1) == 0 &&
5545                       could_be_empty(called, code, bcptr, utf8))                       could_be_empty(called, code, bcptr, utf8, cd))
5546                {                {
5547                *errorcodeptr = ERR40;                *errorcodeptr = ERR40;
5548                goto FAILED;                goto FAILED;
# Line 4737  we set the flag only if there is a liter Line 5558  we set the flag only if there is a liter
5558            code += 1 + LINK_SIZE;            code += 1 + LINK_SIZE;
5559    
5560            *code = OP_RECURSE;            *code = OP_RECURSE;
5561            PUT(code, 1, called - cd->start_code);            PUT(code, 1, (int)(called - cd->start_code));
5562            code += 1 + LINK_SIZE;            code += 1 + LINK_SIZE;
5563    
5564            *code = OP_KET;            *code = OP_KET;
# Line 4759  we set the flag only if there is a liter Line 5580  we set the flag only if there is a liter
5580          set = unset = 0;          set = unset = 0;
5581          optset = &set;          optset = &set;
5582    
5583          while (*ptr != ')' && *ptr != ':')          while (*ptr != CHAR_RIGHT_PARENTHESIS && *ptr != CHAR_COLON)
5584            {            {
5585            switch (*ptr++)            switch (*ptr++)
5586              {              {
5587              case '-': optset = &unset; break;              case CHAR_MINUS: optset = &unset; break;
5588    
5589              case 'J':    /* Record that it changed in the external options */              case CHAR_J:    /* Record that it changed in the external options */
5590              *optset |= PCRE_DUPNAMES;              *optset |= PCRE_DUPNAMES;
5591              cd->external_flags |= PCRE_JCHANGED;              cd->external_flags |= PCRE_JCHANGED;
5592              break;              break;
5593    
5594              case 'i': *optset |= PCRE_CASELESS; break;              case CHAR_i: *optset |= PCRE_CASELESS; break;
5595              case 'm': *optset |= PCRE_MULTILINE; break;              case CHAR_m: *optset |= PCRE_MULTILINE; break;
5596              case 's': *optset |= PCRE_DOTALL; break;              case CHAR_s: *optset |= PCRE_DOTALL; break;
5597              case 'x': *optset |= PCRE_EXTENDED; break;              case CHAR_x: *optset |= PCRE_EXTENDED; break;
5598              case 'U': *optset |= PCRE_UNGREEDY; break;              case CHAR_U: *optset |= PCRE_UNGREEDY; break;
5599              case 'X': *optset |= PCRE_EXTRA; break;              case CHAR_X: *optset |= PCRE_EXTRA; break;
5600    
5601              default:  *errorcodeptr = ERR12;              default:  *errorcodeptr = ERR12;
5602                        ptr--;    /* Correct the offset */                        ptr--;    /* Correct the offset */
# Line 4806  we set the flag only if there is a liter Line 5627  we set the flag only if there is a liter
5627          both phases.          both phases.
5628    
5629          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
5630          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
5631          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. */  
5632    
5633          if (*ptr == ')')          if (*ptr == CHAR_RIGHT_PARENTHESIS)
5634            {            {
5635            if (code == cd->start_code + 1 + LINK_SIZE &&            if (code == cd->start_code + 1 + LINK_SIZE &&
5636                 (lengthptr == NULL || *lengthptr == 2 + 2*LINK_SIZE))                 (lengthptr == NULL || *lengthptr == 2 + 2*LINK_SIZE))
5637              {              {
5638              cd->external_options = newoptions;              cd->external_options = newoptions;
             options = newoptions;  
5639              }              }
5640           else            else
5641              {              {
5642              if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))              if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))
5643                {                {
5644                *code++ = OP_OPT;                *code++ = OP_OPT;
5645                *code++ = newoptions & PCRE_IMS;                *code++ = newoptions & PCRE_IMS;
5646                }                }
   
             /* 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;  
5647              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);
5648              greedy_non_default = greedy_default ^ 1;              greedy_non_default = greedy_default ^ 1;
5649              req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
5650              }              }
5651    
5652              /* Change options at this level, and pass them back for use
5653              in subsequent branches. When not at the start of the pattern, this
5654              information is also necessary so that a resetting item can be
5655              compiled at the end of a group (if we are in a group). */
5656    
5657              *optionsptr = options = newoptions;
5658            previous = NULL;       /* This item can't be repeated */            previous = NULL;       /* This item can't be repeated */
5659            continue;              /* It is complete */            continue;              /* It is complete */
5660            }            }
# Line 4851  we set the flag only if there is a liter Line 5669  we set the flag only if there is a liter
5669          }     /* End of switch for character following (? */          }     /* End of switch for character following (? */
5670        }       /* End of (? handling */        }       /* End of (? handling */
5671    
5672      /* Opening parenthesis not followed by '?'. If PCRE_NO_AUTO_CAPTURE is set,      /* Opening parenthesis not followed by '*' or '?'. If PCRE_NO_AUTO_CAPTURE
5673      all unadorned brackets become non-capturing and behave like (?:...)      is set, all unadorned brackets become non-capturing and behave like (?:...)
5674      brackets. */      brackets. */
5675    
5676      else if ((options & PCRE_NO_AUTO_CAPTURE) != 0)      else if ((options & PCRE_NO_AUTO_CAPTURE) != 0)
# Line 4952  we set the flag only if there is a liter Line 5770  we set the flag only if there is a liter
5770    
5771      /* Error if hit end of pattern */      /* Error if hit end of pattern */
5772    
5773      if (*ptr != ')')      if (*ptr != CHAR_RIGHT_PARENTHESIS)
5774        {        {
5775        *errorcodeptr = ERR14;        *errorcodeptr = ERR14;
5776        goto FAILED;        goto FAILED;
# Line 5044  we set the flag only if there is a liter Line 5862  we set the flag only if there is a liter
5862    
5863      /* ===================================================================*/      /* ===================================================================*/
5864      /* Handle metasequences introduced by \. For ones like \d, the ESC_ values      /* Handle metasequences introduced by \. For ones like \d, the ESC_ values
5865      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
5866      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
5867      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
5868      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
5869      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
5870        ever created. */
5871    
5872      case '\\':      case CHAR_BACKSLASH:
5873      tempptr = ptr;      tempptr = ptr;
5874      c = check_escape(&ptr, errorcodeptr, cd->bracount, options, FALSE);      c = check_escape(&ptr, errorcodeptr, cd->bracount, options, FALSE);
5875      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
# Line 5059  we set the flag only if there is a liter Line 5878  we set the flag only if there is a liter
5878        {        {
5879        if (-c == ESC_Q)            /* Handle start of quoted string */        if (-c == ESC_Q)            /* Handle start of quoted string */
5880          {          {
5881          if (ptr[1] == '\\' && ptr[2] == 'E') ptr += 2; /* avoid empty string */          if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
5882            else inescq = TRUE;            ptr += 2;               /* avoid empty string */
5883                else inescq = TRUE;
5884          continue;          continue;
5885          }          }
5886    
# Line 5077  we set the flag only if there is a liter Line 5897  we set the flag only if there is a liter
5897        zerofirstbyte = firstbyte;        zerofirstbyte = firstbyte;
5898        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
5899    
5900          /* \g<name> or \g'name' is a subroutine call by name and \g<n> or \g'n'
5901          is a subroutine call by number (Oniguruma syntax). In fact, the value
5902          -ESC_g is returned only for these cases. So we don't need to check for <
5903          or ' if the value is -ESC_g. For the Perl syntax \g{n} the value is
5904          -ESC_REF+n, and for the Perl syntax \g{name} the result is -ESC_k (as
5905          that is a synonym for a named back reference). */
5906    
5907          if (-c == ESC_g)
5908            {
5909            const uschar *p;
5910            save_hwm = cd->hwm;   /* Normally this is set when '(' is read */
5911            terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
5912              CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
5913    
5914            /* These two statements stop the compiler for warning about possibly
5915            unset variables caused by the jump to HANDLE_NUMERICAL_RECURSION. In
5916            fact, because we actually check for a number below, the paths that
5917            would actually be in error are never taken. */
5918    
5919            skipbytes = 0;
5920            reset_bracount = FALSE;
5921    
5922            /* Test for a name */
5923    
5924            if (ptr[1] != CHAR_PLUS && ptr[1] != CHAR_MINUS)
5925              {
5926              BOOL isnumber = TRUE;
5927              for (p = ptr + 1; *p != 0 && *p != terminator; p++)
5928                {
5929                if ((cd->ctypes[*p] & ctype_digit) == 0) isnumber = FALSE;
5930                if ((cd->ctypes[*p] & ctype_word) == 0) break;
5931                }
5932              if (*p != terminator)
5933                {
5934                *errorcodeptr = ERR57;
5935                break;
5936                }
5937              if (isnumber)
5938                {
5939                ptr++;
5940                goto HANDLE_NUMERICAL_RECURSION;
5941                }
5942              is_recurse = TRUE;
5943              goto NAMED_REF_OR_RECURSE;
5944              }
5945    
5946            /* Test a signed number in angle brackets or quotes. */
5947    
5948            p = ptr + 2;
5949            while ((digitab[*p] & ctype_digit) != 0) p++;
5950            if (*p != terminator)
5951              {
5952              *errorcodeptr = ERR57;
5953              break;
5954              }
5955            ptr++;
5956            goto HANDLE_NUMERICAL_RECURSION;
5957            }
5958    
5959        /* \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).
5960        We also support \k{name} (.NET syntax) */        We also support \k{name} (.NET syntax) */
5961    
5962        if (-c == ESC_k && (ptr[1] == '<' || ptr[1] == '\'' || ptr[1] == '{'))        if (-c == ESC_k && (ptr[1] == CHAR_LESS_THAN_SIGN ||
5963              ptr[1] == CHAR_APOSTROPHE || ptr[1] == CHAR_LEFT_CURLY_BRACKET))
5964          {          {
5965          is_recurse = FALSE;          is_recurse = FALSE;
5966          terminator = (*(++ptr) == '<')? '>' : (*ptr == '\'')? '\'' : '}';          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
5967              CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?
5968              CHAR_APOSTROPHE : CHAR_RIGHT_CURLY_BRACKET;
5969          goto NAMED_REF_OR_RECURSE;          goto NAMED_REF_OR_RECURSE;
5970          }          }
5971    
# Line 5093  we set the flag only if there is a liter Line 5975  we set the flag only if there is a liter
5975    
5976        if (-c >= ESC_REF)        if (-c >= ESC_REF)
5977          {          {
5978            open_capitem *oc;
5979          recno = -c - ESC_REF;          recno = -c - ESC_REF;