/[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 333 by ph10, Thu Apr 10 19:55:57 2008 UTC revision 605 by ph10, Fri Jun 3 18:18:30 2011 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2008 University of Cambridge             Copyright (c) 1997-2011 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 53  supporting internal functions that are n Line 53  supporting internal functions that are n
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56  /* When DEBUG is defined, we need the pcre_printint() function, which is also  /* When PCRE_DEBUG is defined, we need the pcre_printint() function, which is
57  used by pcretest. DEBUG is not defined when building a production library. */  also used by pcretest. PCRE_DEBUG is not defined when building a production
58    library. */
59    
60  #ifdef DEBUG  #ifdef PCRE_DEBUG
61  #include "pcre_printint.src"  #include "pcre_printint.src"
62  #endif  #endif
63    
# Line 91  is 4 there is plenty of room. */ Line 92  is 4 there is plenty of room. */
92    
93  #define COMPILE_WORK_SIZE (4096)  #define COMPILE_WORK_SIZE (4096)
94    
95    /* The overrun tests check for a slightly smaller size so that they detect the
96    overrun before it actually does run off the end of the data block. */
97    
98    #define WORK_SIZE_CHECK (COMPILE_WORK_SIZE - 100)
99    
100    
101  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
102  are simple data values; negative values are for special things like \d and so  are simple data values; negative values are for special things like \d and so
103  on. Zero means further processing is needed (for things like \x), or the escape  on. Zero means further processing is needed (for things like \x), or the escape
104  is invalid. */  is invalid. */
105    
106  #ifndef EBCDIC  /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC
107    
108    /* This is the "normal" table for ASCII systems or for EBCDIC systems running
109    in UTF-8 mode. */
110    
111  static const short int escapes[] = {  static const short int escapes[] = {
112       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,                       0,
113       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,                       0,
114     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */       0,                       0,
115  -ESC_H,      0,      0, -ESC_K,      0,      0,      0,      0,   /* H - O */       0,                       0,
116  -ESC_P, -ESC_Q, -ESC_R, -ESC_S,      0,      0, -ESC_V, -ESC_W,   /* P - W */       0,                       0,
117  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */       CHAR_COLON,              CHAR_SEMICOLON,
118     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
119  -ESC_h,      0,      0, -ESC_k,      0,      0,  ESC_n,      0,   /* h - o */       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
120  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0, -ESC_v, -ESC_w,   /* p - w */       CHAR_COMMERCIAL_AT,      -ESC_A,
121       0,      0, -ESC_z                                            /* x - z */       -ESC_B,                  -ESC_C,
122         -ESC_D,                  -ESC_E,
123         0,                       -ESC_G,
124         -ESC_H,                  0,
125         0,                       -ESC_K,
126         0,                       0,
127         -ESC_N,                  0,
128         -ESC_P,                  -ESC_Q,
129         -ESC_R,                  -ESC_S,
130         0,                       0,
131         -ESC_V,                  -ESC_W,
132         -ESC_X,                  0,
133         -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,
134         CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,
135         CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,
136         CHAR_GRAVE_ACCENT,       7,
137         -ESC_b,                  0,
138         -ESC_d,                  ESC_e,
139         ESC_f,                   0,
140         -ESC_h,                  0,
141         0,                       -ESC_k,
142         0,                       0,
143         ESC_n,                   0,
144         -ESC_p,                  0,
145         ESC_r,                   -ESC_s,
146         ESC_tee,                 0,
147         -ESC_v,                  -ESC_w,
148         0,                       0,
149         -ESC_z
150  };  };
151    
152  #else           /* This is the "abnormal" table for EBCDIC systems */  #else
153    
154    /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
155    
156  static const short int escapes[] = {  static const short int escapes[] = {
157  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
158  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 130  static const short int escapes[] = { Line 171  static const short int escapes[] = {
171  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
172  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,
173  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
174  /*  D0 */   '}',     0, -ESC_K,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,-ESC_N,      0, -ESC_P,
175  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
176  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
177  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
# Line 142  static const short int escapes[] = { Line 183  static const short int escapes[] = {
183    
184  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
185  searched linearly. Put all the names into a single string, in order to reduce  searched linearly. Put all the names into a single string, in order to reduce
186  the number of relocations when a shared library is dynamically linked. */  the number of relocations when a shared library is dynamically linked. The
187    string is built from string macros so that it works in UTF-8 mode on EBCDIC
188    platforms. */
189    
190  typedef struct verbitem {  typedef struct verbitem {
191    int   len;    int   len;                 /* Length of verb name */
192    int   op;    int   op;                  /* Op when no arg, or -1 if arg mandatory */
193      int   op_arg;              /* Op when arg present, or -1 if not allowed */
194  } verbitem;  } verbitem;
195    
196  static const char verbnames[] =  static const char verbnames[] =
197    "ACCEPT\0"    "\0"                       /* Empty name is a shorthand for MARK */
198    "COMMIT\0"    STRING_MARK0
199    "F\0"    STRING_ACCEPT0
200    "FAIL\0"    STRING_COMMIT0
201    "PRUNE\0"    STRING_F0
202    "SKIP\0"    STRING_FAIL0
203    "THEN";    STRING_PRUNE0
204      STRING_SKIP0
205      STRING_THEN;
206    
207  static const verbitem verbs[] = {  static const verbitem verbs[] = {
208    { 6, OP_ACCEPT },    { 0, -1,        OP_MARK },
209    { 6, OP_COMMIT },    { 4, -1,        OP_MARK },
210    { 1, OP_FAIL },    { 6, OP_ACCEPT, -1 },
211    { 4, OP_FAIL },    { 6, OP_COMMIT, -1 },
212    { 5, OP_PRUNE },    { 1, OP_FAIL,   -1 },
213    { 4, OP_SKIP  },    { 4, OP_FAIL,   -1 },
214    { 4, OP_THEN  }    { 5, OP_PRUNE,  OP_PRUNE_ARG },
215      { 4, OP_SKIP,   OP_SKIP_ARG  },
216      { 4, OP_THEN,   OP_THEN_ARG  }
217  };  };
218    
219  static const int verbcount = sizeof(verbs)/sizeof(verbitem);  static const int verbcount = sizeof(verbs)/sizeof(verbitem);
# Line 178  length entry. The first three must be al Line 226  length entry. The first three must be al
226  for handling case independence. */  for handling case independence. */
227    
228  static const char posix_names[] =  static const char posix_names[] =
229    "alpha\0"  "lower\0"  "upper\0"  "alnum\0"  "ascii\0"  "blank\0"    STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
230    "cntrl\0"  "digit\0"  "graph\0"  "print\0"  "punct\0"  "space\0"    STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
231    "word\0"   "xdigit";    STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
232      STRING_word0  STRING_xdigit;
233    
234  static const uschar posix_name_lengths[] = {  static const uschar posix_name_lengths[] = {
235    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
# Line 212  static const int posix_class_maps[] = { Line 261  static const int posix_class_maps[] = {
261    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
262  };  };
263    
264    /* Table of substitutes for \d etc when PCRE_UCP is set. The POSIX class
265    substitutes must be in the order of the names, defined above, and there are
266    both positive and negative cases. NULL means no substitute. */
267    
268    #ifdef SUPPORT_UCP
269    static const uschar *substitutes[] = {
270      (uschar *)"\\P{Nd}",    /* \D */
271      (uschar *)"\\p{Nd}",    /* \d */
272      (uschar *)"\\P{Xsp}",   /* \S */       /* NOTE: Xsp is Perl space */
273      (uschar *)"\\p{Xsp}",   /* \s */
274      (uschar *)"\\P{Xwd}",   /* \W */
275      (uschar *)"\\p{Xwd}"    /* \w */
276    };
277    
278    static const uschar *posix_substitutes[] = {
279      (uschar *)"\\p{L}",     /* alpha */
280      (uschar *)"\\p{Ll}",    /* lower */
281      (uschar *)"\\p{Lu}",    /* upper */
282      (uschar *)"\\p{Xan}",   /* alnum */
283      NULL,                   /* ascii */
284      (uschar *)"\\h",        /* blank */
285      NULL,                   /* cntrl */
286      (uschar *)"\\p{Nd}",    /* digit */
287      NULL,                   /* graph */
288      NULL,                   /* print */
289      NULL,                   /* punct */
290      (uschar *)"\\p{Xps}",   /* space */    /* NOTE: Xps is POSIX space */
291      (uschar *)"\\p{Xwd}",   /* word */
292      NULL,                   /* xdigit */
293      /* Negated cases */
294      (uschar *)"\\P{L}",     /* ^alpha */
295      (uschar *)"\\P{Ll}",    /* ^lower */
296      (uschar *)"\\P{Lu}",    /* ^upper */
297      (uschar *)"\\P{Xan}",   /* ^alnum */
298      NULL,                   /* ^ascii */
299      (uschar *)"\\H",        /* ^blank */
300      NULL,                   /* ^cntrl */
301      (uschar *)"\\P{Nd}",    /* ^digit */
302      NULL,                   /* ^graph */
303      NULL,                   /* ^print */
304      NULL,                   /* ^punct */
305      (uschar *)"\\P{Xps}",   /* ^space */   /* NOTE: Xps is POSIX space */
306      (uschar *)"\\P{Xwd}",   /* ^word */
307      NULL                    /* ^xdigit */
308    };
309    #define POSIX_SUBSIZE (sizeof(posix_substitutes)/sizeof(uschar *))
310    #endif
311    
312  #define STRING(a)  # a  #define STRING(a)  # a
313  #define XSTRING(s) STRING(s)  #define XSTRING(s) STRING(s)
# Line 224  the number of relocations needed when a Line 320  the number of relocations needed when a
320  it is now one long string. We cannot use a table of offsets, because the  it is now one long string. We cannot use a table of offsets, because the
321  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
322  simply count through to the one we want - this isn't a performance issue  simply count through to the one we want - this isn't a performance issue
323  because these strings are used only when there is a compilation error. */  because these strings are used only when there is a compilation error.
324    
325    Each substring ends with \0 to insert a null character. This includes the final
326    substring, so that the whole string ends with \0\0, which can be detected when
327    counting through. */
328    
329  static const char error_texts[] =  static const char error_texts[] =
330    "no error\0"    "no error\0"
# Line 271  static const char error_texts[] = Line 371  static const char error_texts[] =
371    /* 35 */    /* 35 */
372    "invalid condition (?(0)\0"    "invalid condition (?(0)\0"
373    "\\C not allowed in lookbehind assertion\0"    "\\C not allowed in lookbehind assertion\0"
374    "PCRE does not support \\L, \\l, \\N, \\U, or \\u\0"    "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
375    "number after (?C is > 255\0"    "number after (?C is > 255\0"
376    "closing ) for (?C expected\0"    "closing ) for (?C expected\0"
377    /* 40 */    /* 40 */
# Line 297  static const char error_texts[] = Line 397  static const char error_texts[] =
397    "inconsistent NEWLINE options\0"    "inconsistent NEWLINE options\0"
398    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
399    "a numbered reference must not be zero\0"    "a numbered reference must not be zero\0"
400    "(*VERB) with an argument is not supported\0"    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
401    /* 60 */    /* 60 */
402    "(*VERB) not recognized\0"    "(*VERB) not recognized\0"
403    "number is too big\0"    "number is too big\0"
404    "subpattern name expected\0"    "subpattern name expected\0"
405    "digit expected after (?+";    "digit expected after (?+\0"
406      "] is an invalid data character in JavaScript compatibility mode\0"
407      /* 65 */
408      "different names for subpatterns of the same number are not allowed\0"
409      "(*MARK) must have an argument\0"
410      "this version of PCRE is not compiled with PCRE_UCP support\0"
411      "\\c must be followed by an ASCII character\0"
412      ;
413    
414  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
415  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 427  For convenience, we use the same bit def
427    
428  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
429    
430  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC
431    
432    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
433    UTF-8 mode. */
434    
435  static const unsigned char digitab[] =  static const unsigned char digitab[] =
436    {    {
437    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 467  static const unsigned char digitab[] =
467    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
468    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
469    
470  #else           /* This is the "abnormal" case, for EBCDIC systems */  #else
471    
472    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
473    
474  static const unsigned char digitab[] =  static const unsigned char digitab[] =
475    {    {
476    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
# Line 432  static const unsigned char ebcdic_charta Line 545  static const unsigned char ebcdic_charta
545  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
546    
547  static BOOL  static BOOL
548    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,    compile_regex(int, uschar **, const uschar **, int *, BOOL, BOOL, int, int *,
549      int *, int *, branch_chain *, compile_data *, int *);      int *, branch_chain *, compile_data *, int *);
550    
551    
552    
# Line 454  static const char * Line 567  static const char *
567  find_error_text(int n)  find_error_text(int n)
568  {  {
569  const char *s = error_texts;  const char *s = error_texts;
570  for (; n > 0; n--) while (*s++ != 0);  for (; n > 0; n--)
571      {
572      while (*s++ != 0) {};
573      if (*s == 0) return "Error text not found (please report)";
574      }
575  return s;  return s;
576  }  }
577    
# Line 502  if (c == 0) *errorcodeptr = ERR1; Line 619  if (c == 0) *errorcodeptr = ERR1;
619  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.
620  Otherwise further processing may be required. */  Otherwise further processing may be required. */
621    
622  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
623  else if (c < '0' || c > 'z') {}                           /* Not alphanumeric */  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */
624  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0) c = i;
625    
626  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
627  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 640  else
640      /* 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
641      error. */      error. */
642    
643      case 'l':      case CHAR_l:
644      case 'L':      case CHAR_L:
645      case 'N':      case CHAR_u:
646      case 'u':      case CHAR_U:
     case 'U':  
647      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
648      break;      break;
649    
650      /* \g must be followed by one of a number of specific things:      /* \g must be followed by one of a number of specific things:
651    
652      (1) A number, either plain or braced. If positive, it is an absolute      (1) A number, either plain or braced. If positive, it is an absolute
653      backreference. If negative, it is a relative backreference. This is a Perl      backreference. If negative, it is a relative backreference. This is a Perl
654      5.10 feature.      5.10 feature.
655    
656      (2) Perl 5.10 also supports \g{name} as a reference to a named group. This      (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
657      is part of Perl's movement towards a unified syntax for back references. As      is part of Perl's movement towards a unified syntax for back references. As
658      this is synonymous with \k{name}, we fudge it up by pretending it really      this is synonymous with \k{name}, we fudge it up by pretending it really
659      was \k.      was \k.
660    
661      (3) For Oniguruma compatibility we also support \g followed by a name or a      (3) For Oniguruma compatibility we also support \g followed by a name or a
662      number either in angle brackets or in single quotes. However, these are      number either in angle brackets or in single quotes. However, these are
663      (possibly recursive) subroutine calls, _not_ backreferences. Just return      (possibly recursive) subroutine calls, _not_ backreferences. Just return
664      the -ESC_g code (cf \k). */      the -ESC_g code (cf \k). */
665    
666      case 'g':      case CHAR_g:
667      if (ptr[1] == '<' || ptr[1] == '\'')      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
668        {        {
669        c = -ESC_g;        c = -ESC_g;
670        break;        break;
671        }        }
672    
673      /* Handle the Perl-compatible cases */      /* Handle the Perl-compatible cases */
674    
675      if (ptr[1] == '{')      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
676        {        {
677        const uschar *p;        const uschar *p;
678        for (p = ptr+2; *p != 0 && *p != '}'; p++)        for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
679          if (*p != '-' && (digitab[*p] & ctype_digit) == 0) break;          if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
680        if (*p != 0 && *p != '}')        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
681          {          {
682          c = -ESC_k;          c = -ESC_k;
683          break;          break;
# Line 571  else Line 687  else
687        }        }
688      else braced = FALSE;      else braced = FALSE;
689    
690      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
691        {        {
692        negated = TRUE;        negated = TRUE;
693        ptr++;        ptr++;
# Line 580  else Line 696  else
696    
697      c = 0;      c = 0;
698      while ((digitab[ptr[1]] & ctype_digit) != 0)      while ((digitab[ptr[1]] & ctype_digit) != 0)
699        c = c * 10 + *(++ptr) - '0';        c = c * 10 + *(++ptr) - CHAR_0;
700    
701      if (c < 0)   /* Integer overflow */      if (c < 0)   /* Integer overflow */
702        {        {
703        *errorcodeptr = ERR61;        *errorcodeptr = ERR61;
704        break;        break;
705        }        }
706    
707      if (braced && *(++ptr) != '}')      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
708        {        {
709        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
710        break;        break;
711        }        }
712    
713      if (c == 0)      if (c == 0)
714        {        {
715        *errorcodeptr = ERR58;        *errorcodeptr = ERR58;
716        break;        break;
717        }        }
718    
719      if (negated)      if (negated)
720        {        {
# Line 625  else Line 741  else
741      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
742      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
743    
744      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:
745      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
746    
747      if (!isclass)      if (!isclass)
748        {        {
749        oldptr = ptr;        oldptr = ptr;
750        c -= '0';        c -= CHAR_0;
751        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
752          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - CHAR_0;
753        if (c < 0)    /* Integer overflow */        if (c < 0)    /* Integer overflow */
754          {          {
755          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
# Line 651  else Line 767  else
767      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.
768      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
769    
770      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
771        {        {
772        ptr--;        ptr--;
773        c = 0;        c = 0;
# Line 664  else Line 780  else
780      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
781      than 3 octal digits. */      than 3 octal digits. */
782    
783      case '0':      case CHAR_0:
784      c -= '0';      c -= CHAR_0;
785      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
786          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
787      if (!utf8 && c > 255) *errorcodeptr = ERR51;      if (!utf8 && c > 255) *errorcodeptr = ERR51;
788      break;      break;
789    
# Line 675  else Line 791  else
791      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
792      treated as a data character. */      treated as a data character. */
793    
794      case 'x':      case CHAR_x:
795      if (ptr[1] == '{')      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
796        {        {
797        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
798        int count = 0;        int count = 0;
# Line 685  else Line 801  else
801        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
802          {          {
803          register int cc = *pt++;          register int cc = *pt++;
804          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
805          count++;          count++;
806    
807  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
808          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
809          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
810  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
811          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
812          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
813  #endif  #endif
814          }          }
815    
816        if (*pt == '}')        if (*pt == CHAR_RIGHT_CURLY_BRACKET)
817          {          {
818          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
819          ptr = pt;          ptr = pt;
# Line 713  else Line 829  else
829      c = 0;      c = 0;
830      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
831        {        {
832        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
833        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
834  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
835        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
836        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
837  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
838        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
839        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
840  #endif  #endif
841        }        }
842      break;      break;
843    
844      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
845      This coding is ASCII-specific, but then the whole concept of \cx is      An error is given if the byte following \c is not an ASCII character. This
846        coding is ASCII-specific, but then the whole concept of \cx is
847      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
848    
849      case 'c':      case CHAR_c:
850      c = *(++ptr);      c = *(++ptr);
851      if (c == 0)      if (c == 0)
852        {        {
853        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
854        break;        break;
855        }        }
856    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
857  #ifndef EBCDIC  /* ASCII coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
858      if (c >= 'a' && c <= 'z') c -= 32;        {
859          *errorcodeptr = ERR68;
860          break;
861          }
862        if (c >= CHAR_a && c <= CHAR_z) c -= 32;
863      c ^= 0x40;      c ^= 0x40;
864  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
865      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
866      c ^= 0xC0;      c ^= 0xC0;
867  #endif  #endif
868      break;      break;
# Line 763  else Line 884  else
884      }      }
885    }    }
886    
887    /* Perl supports \N{name} for character names, as well as plain \N for "not
888    newline". PCRE does not support \N{name}. */
889    
890    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET)
891      *errorcodeptr = ERR37;
892    
893    /* If PCRE_UCP is set, we change the values for \d etc. */
894    
895    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
896      c -= (ESC_DU - ESC_D);
897    
898    /* Set the pointer to the final character before returning. */
899    
900  *ptrptr = ptr;  *ptrptr = ptr;
901  return c;  return c;
902  }  }
# Line 803  if (c == 0) goto ERROR_RETURN; Line 937  if (c == 0) goto ERROR_RETURN;
937  /* \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
938  negation. */  negation. */
939    
940  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
941    {    {
942    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
943      {      {
944      *negptr = TRUE;      *negptr = TRUE;
945      ptr++;      ptr++;
# Line 814  if (c == '{') Line 948  if (c == '{')
948      {      {
949      c = *(++ptr);      c = *(++ptr);
950      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
951      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
952      name[i] = c;      name[i] = c;
953      }      }
954    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
955    name[i] = 0;    name[i] = 0;
956    }    }
957    
# Line 882  is_counted_repeat(const uschar *p) Line 1016  is_counted_repeat(const uschar *p)
1016  {  {
1017  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
1018  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
1019  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
1020    
1021  if (*p++ != ',') return FALSE;  if (*p++ != CHAR_COMMA) return FALSE;
1022  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
1023    
1024  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
1025  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
1026    
1027  return (*p == '}');  return (*p == CHAR_RIGHT_CURLY_BRACKET);
1028  }  }
1029    
1030    
# Line 923  int max = -1; Line 1057  int max = -1;
1057  /* 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
1058  an integer overflow. */  an integer overflow. */
1059    
1060  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;
1061  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1062    {    {
1063    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 933  if (min < 0 || min > 65535) Line 1067  if (min < 0 || min > 65535)
1067  /* 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.
1068  Also, max must not be less than min. */  Also, max must not be less than min. */
1069    
1070  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1071    {    {
1072    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1073      {      {
1074      max = 0;      max = 0;
1075      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;
1076      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1077        {        {
1078        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 963  return p; Line 1097  return p;
1097    
1098    
1099  /*************************************************  /*************************************************
1100  *       Find forward referenced subpattern       *  *  Subroutine for finding forward reference      *
1101  *************************************************/  *************************************************/
1102    
1103  /* This function scans along a pattern's text looking for capturing  /* This recursive function is called only from find_parens() below. The
1104    top-level call starts at the beginning of the pattern. All other calls must
1105    start at a parenthesis. It scans along a pattern's text looking for capturing
1106  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
1107  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
1108  returns when it reaches a given numbered subpattern. This is used for forward  returns when it reaches a given numbered subpattern. Recursion is used to keep
1109  references to subpatterns. We know that if (?P< is encountered, the name will  track of subpatterns that reset the capturing group numbers - the (?| feature.
1110  be terminated by '>' because that is checked in the first pass.  
1111    This function was originally called only from the second pass, in which we know
1112    that if (?< or (?' or (?P< is encountered, the name will be correctly
1113    terminated because that is checked in the first pass. There is now one call to
1114    this function in the first pass, to check for a recursive back reference by
1115    name (so that we can make the whole group atomic). In this case, we need check
1116    only up to the current position in the pattern, and that is still OK because
1117    and previous occurrences will have been checked. To make this work, the test
1118    for "end of pattern" is a check against cd->end_pattern in the main loop,
1119    instead of looking for a binary zero. This means that the special first-pass
1120    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1121    processing items within the loop are OK, because afterwards the main loop will
1122    terminate.)
1123    
1124  Arguments:  Arguments:
1125    ptr          current position in the pattern    ptrptr       address of the current character pointer (updated)
1126    count        current count of capturing parens so far encountered    cd           compile background data
1127    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1128    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1129    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1130      utf8         TRUE if we are in UTF-8 mode
1131      count        pointer to the current capturing subpattern number (updated)
1132    
1133  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1134  */  */
1135    
1136  static int  static int
1137  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,
1138    BOOL xmode)    BOOL xmode, BOOL utf8, int *count)
1139  {  {
1140  const uschar *thisname;  uschar *ptr = *ptrptr;
1141    int start_count = *count;
1142    int hwm_count = start_count;
1143    BOOL dup_parens = FALSE;
1144    
1145    /* If the first character is a parenthesis, check on the type of group we are
1146    dealing with. The very first call may not start with a parenthesis. */
1147    
1148  for (; *ptr != 0; ptr++)  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1149    {    {
1150    int term;    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1151    
1152      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1153    
1154      /* Handle a normal, unnamed capturing parenthesis. */
1155    
1156      else if (ptr[1] != CHAR_QUESTION_MARK)
1157        {
1158        *count += 1;
1159        if (name == NULL && *count == lorn) return *count;
1160        ptr++;
1161        }
1162    
1163      /* All cases now have (? at the start. Remember when we are in a group
1164      where the parenthesis numbers are duplicated. */
1165    
1166      else if (ptr[2] == CHAR_VERTICAL_LINE)
1167        {
1168        ptr += 3;
1169        dup_parens = TRUE;
1170        }
1171    
1172      /* Handle comments; all characters are allowed until a ket is reached. */
1173    
1174      else if (ptr[2] == CHAR_NUMBER_SIGN)
1175        {
1176        for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1177        goto FAIL_EXIT;
1178        }
1179    
1180      /* Handle a condition. If it is an assertion, just carry on so that it
1181      is processed as normal. If not, skip to the closing parenthesis of the
1182      condition (there can't be any nested parens). */
1183    
1184      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1185        {
1186        ptr += 2;
1187        if (ptr[1] != CHAR_QUESTION_MARK)
1188          {
1189          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1190          if (*ptr != 0) ptr++;
1191          }
1192        }
1193    
1194      /* Start with (? but not a condition. */
1195    
1196      else
1197        {
1198        ptr += 2;
1199        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1200    
1201        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1202    
1203        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1204            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1205          {
1206          int term;
1207          const uschar *thisname;
1208          *count += 1;
1209          if (name == NULL && *count == lorn) return *count;
1210          term = *ptr++;
1211          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1212          thisname = ptr;
1213          while (*ptr != term) ptr++;
1214          if (name != NULL && lorn == ptr - thisname &&
1215              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1216            return *count;
1217          term++;
1218          }
1219        }
1220      }
1221    
1222    /* Past any initial parenthesis handling, scan for parentheses or vertical
1223    bars. Stop if we get to cd->end_pattern. Note that this is important for the
1224    first-pass call when this value is temporarily adjusted to stop at the current
1225    position. So DO NOT change this to a test for binary zero. */
1226    
1227    for (; ptr < cd->end_pattern; ptr++)
1228      {
1229    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1230    
1231    if (*ptr == '\\')    if (*ptr == CHAR_BACKSLASH)
1232      {      {
1233      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1234      if (*ptr == 'Q') for (;;)      if (*ptr == CHAR_Q) for (;;)
1235        {        {
1236        while (*(++ptr) != 0 && *ptr != '\\');        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1237        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1238        if (*(++ptr) == 'E') break;        if (*(++ptr) == CHAR_E) break;
1239        }        }
1240      continue;      continue;
1241      }      }
1242    
1243    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1244      are handled for real. If the first character is '^', skip it. Also, if the
1245      first few characters (either before or after ^) are \Q\E or \E we skip them
1246      too. This makes for compatibility with Perl. Note the use of STR macros to
1247      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1248    
1249    if (*ptr == '[')    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1250      {      {
1251      while (*(++ptr) != ']')      BOOL negate_class = FALSE;
1252        for (;;)
1253          {
1254          if (ptr[1] == CHAR_BACKSLASH)
1255            {
1256            if (ptr[2] == CHAR_E)
1257              ptr+= 2;
1258            else if (strncmp((const char *)ptr+2,
1259                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1260              ptr += 4;
1261            else
1262              break;
1263            }
1264          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1265            {
1266            negate_class = TRUE;
1267            ptr++;
1268            }
1269          else break;
1270          }
1271    
1272        /* If the next character is ']', it is a data character that must be
1273        skipped, except in JavaScript compatibility mode. */
1274    
1275        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1276            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1277          ptr++;
1278    
1279        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1280        {        {
1281        if (*ptr == 0) return -1;        if (*ptr == 0) return -1;
1282        if (*ptr == '\\')        if (*ptr == CHAR_BACKSLASH)
1283          {          {
1284          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) goto FAIL_EXIT;
1285          if (*ptr == 'Q') for (;;)          if (*ptr == CHAR_Q) for (;;)
1286            {            {
1287            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1288            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1289            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1290            }            }
1291          continue;          continue;
1292          }          }
# Line 1031  for (; *ptr != 0; ptr++) Line 1296  for (; *ptr != 0; ptr++)
1296    
1297    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1298    
1299    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1300      {      {
1301      while (*(++ptr) != 0 && *ptr != '\n');      ptr++;
1302      if (*ptr == 0) return -1;      while (*ptr != 0)
1303          {
1304          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1305          ptr++;
1306    #ifdef SUPPORT_UTF8
1307          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
1308    #endif
1309          }
1310        if (*ptr == 0) goto FAIL_EXIT;
1311      continue;      continue;
1312      }      }
1313    
1314    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1315    
1316    if (*ptr != '(') continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != '?' && ptr[1] != '*')  
1317      {      {
1318      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, count);
1319      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1320      continue;      if (*ptr == 0) goto FAIL_EXIT;
1321      }      }
1322    
1323    ptr += 2;    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1324    if (*ptr == 'P') ptr++;                      /* Allow optional P */      {
1325        if (dup_parens && *count < hwm_count) *count = hwm_count;
1326        goto FAIL_EXIT;
1327        }
1328    
1329    /* We have to disambiguate (?<! and (?<= from (?<name> */    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1330        {
1331        if (*count > hwm_count) hwm_count = *count;
1332        *count = start_count;
1333        }
1334      }
1335    
1336    if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  FAIL_EXIT:
1337         *ptr != '\'')  *ptrptr = ptr;
1338      continue;  return -1;
1339    }
1340    
1341    
1342    
1343    
1344    /*************************************************
1345    *       Find forward referenced subpattern       *
1346    *************************************************/
1347    
1348    /* This function scans along a pattern's text looking for capturing
1349    subpatterns, and counting them. If it finds a named pattern that matches the
1350    name it is given, it returns its number. Alternatively, if the name is NULL, it
1351    returns when it reaches a given numbered subpattern. This is used for forward
1352    references to subpatterns. We used to be able to start this scan from the
1353    current compiling point, using the current count value from cd->bracount, and
1354    do it all in a single loop, but the addition of the possibility of duplicate
1355    subpattern numbers means that we have to scan from the very start, in order to
1356    take account of such duplicates, and to use a recursive function to keep track
1357    of the different types of group.
1358    
1359    Arguments:
1360      cd           compile background data
1361      name         name to seek, or NULL if seeking a numbered subpattern
1362      lorn         name length, or subpattern number if name is NULL
1363      xmode        TRUE if we are in /x mode
1364      utf8         TRUE if we are in UTF-8 mode
1365    
1366    count++;  Returns:       the number of the found subpattern, or -1 if not found
1367    */
1368    
1369    static int
1370    find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode,
1371      BOOL utf8)
1372    {
1373    uschar *ptr = (uschar *)cd->start_pattern;
1374    int count = 0;
1375    int rc;
1376    
1377    /* If the pattern does not start with an opening parenthesis, the first call
1378    to find_parens_sub() will scan right to the end (if necessary). However, if it
1379    does start with a parenthesis, find_parens_sub() will return when it hits the
1380    matching closing parens. That is why we have to have a loop. */
1381    
1382    if (name == NULL && count == lorn) return count;  for (;;)
1383    term = *ptr++;    {
1384    if (term == '<') term = '>';    rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, &count);
1385    thisname = ptr;    if (rc > 0 || *ptr++ == 0) break;
   while (*ptr != term) ptr++;  
   if (name != NULL && lorn == ptr - thisname &&  
       strncmp((const char *)name, (const char *)thisname, lorn) == 0)  
     return count;  
1386    }    }
1387    
1388  return -1;  return rc;
1389  }  }
1390    
1391    
1392    
1393    
1394  /*************************************************  /*************************************************
1395  *      Find first significant op code            *  *      Find first significant op code            *
1396  *************************************************/  *************************************************/
1397    
1398  /* This is called by several functions that scan a compiled expression looking  /* This is called by several functions that scan a compiled expression looking
1399  for a fixed first character, or an anchoring op code etc. It skips over things  for a fixed first character, or an anchoring op code etc. It skips over things
1400  that do not influence this. For some calls, a change of option is important.  that do not influence this. For some calls, it makes sense to skip negative
1401  For some calls, it makes sense to skip negative forward and all backward  forward and all backward assertions, and also the \b assertion; for others it
1402  assertions, and also the \b assertion; for others it does not.  does not.
1403    
1404  Arguments:  Arguments:
1405    code         pointer to the start of the group    code         pointer to the start of the group
   options      pointer to external options  
   optbit       the option bit whose changing is significant, or  
                  zero if none are  
1406    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1407    
1408  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1409  */  */
1410    
1411  static const uschar*  static const uschar*
1412  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const uschar *code, BOOL skipassert)
   BOOL skipassert)  
1413  {  {
1414  for (;;)  for (;;)
1415    {    {
1416    switch ((int)*code)    switch ((int)*code)
1417      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1418      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1419      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1420      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
# Line 1123  for (;;) Line 1430  for (;;)
1430    
1431      case OP_CALLOUT:      case OP_CALLOUT:
1432      case OP_CREF:      case OP_CREF:
1433        case OP_NCREF:
1434      case OP_RREF:      case OP_RREF:
1435        case OP_NRREF:
1436      case OP_DEF:      case OP_DEF:
1437      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1438      break;      break;
# Line 1139  for (;;) Line 1448  for (;;)
1448    
1449    
1450  /*************************************************  /*************************************************
1451  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1452  *************************************************/  *************************************************/
1453    
1454  /* 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,
1455  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.
1456  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
1457    temporarily terminated with OP_END when this function is called.
1458    
1459    This function is called when a backward assertion is encountered, so that if it
1460    fails, the error message can point to the correct place in the pattern.
1461    However, we cannot do this when the assertion contains subroutine calls,
1462    because they can be forward references. We solve this by remembering this case
1463    and doing the check at the end; a flag specifies which mode we are running in.
1464    
1465  Arguments:  Arguments:
1466    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1467    options  the compiling options    utf8     TRUE in UTF-8 mode
1468      atend    TRUE if called when the pattern is complete
1469      cd       the "compile data" structure
1470    
1471  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1472                 or -1 if there is no fixed length,
1473               or -2 if \C was encountered               or -2 if \C was encountered
1474                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1475  */  */
1476    
1477  static int  static int
1478  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, BOOL utf8, BOOL atend, compile_data *cd)
1479  {  {
1480  int length = -1;  int length = -1;
1481    
# Line 1168  branch, check the length against that of Line 1488  branch, check the length against that of
1488  for (;;)  for (;;)
1489    {    {
1490    int d;    int d;
1491      uschar *ce, *cs;
1492    register int op = *cc;    register int op = *cc;
1493    switch (op)    switch (op)
1494      {      {
1495        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1496        OP_BRA (normal non-capturing bracket) because the other variants of these
1497        opcodes are all concerned with unlimited repeated groups, which of course
1498        are not of fixed length. They will cause a -1 response from the default
1499        case of this switch. */
1500    
1501      case OP_CBRA:      case OP_CBRA:
1502      case OP_BRA:      case OP_BRA:
1503      case OP_ONCE:      case OP_ONCE:
1504      case OP_COND:      case OP_COND:
1505      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), utf8, atend, cd);
1506      if (d < 0) return d;      if (d < 0) return d;
1507      branchlength += d;      branchlength += d;
1508      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1184  for (;;) Line 1511  for (;;)
1511    
1512      /* Reached end of a branch; if it's a ket it is the end of a nested      /* Reached end of a branch; if it's a ket it is the end of a nested
1513      call. If it's ALT it is an alternation in a nested call. If it is      call. If it's ALT it is an alternation in a nested call. If it is
1514      END it's the end of the outer call. All can be handled by the same code. */      END it's the end of the outer call. All can be handled by the same code.
1515        Note that we must not include the OP_KETRxxx opcodes here, because they
1516        all imply an unlimited repeat. */
1517    
1518      case OP_ALT:      case OP_ALT:
1519      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1520      case OP_END:      case OP_END:
1521      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1522        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
# Line 1198  for (;;) Line 1525  for (;;)
1525      branchlength = 0;      branchlength = 0;
1526      break;      break;
1527    
1528        /* A true recursion implies not fixed length, but a subroutine call may
1529        be OK. If the subroutine is a forward reference, we can't deal with
1530        it until the end of the pattern, so return -3. */
1531    
1532        case OP_RECURSE:
1533        if (!atend) return -3;
1534        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1535        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1536        if (cc > cs && cc < ce) return -1;                /* Recursion */
1537        d = find_fixedlength(cs + 2, utf8, atend, cd);
1538        if (d < 0) return d;
1539        branchlength += d;
1540        cc += 1 + LINK_SIZE;
1541        break;
1542    
1543      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1544    
1545      case OP_ASSERT:      case OP_ASSERT:
# Line 1211  for (;;) Line 1553  for (;;)
1553    
1554      case OP_REVERSE:      case OP_REVERSE:
1555      case OP_CREF:      case OP_CREF:
1556        case OP_NCREF:
1557      case OP_RREF:      case OP_RREF:
1558        case OP_NRREF:
1559      case OP_DEF:      case OP_DEF:
     case OP_OPT:  
1560      case OP_CALLOUT:      case OP_CALLOUT:
1561      case OP_SOD:      case OP_SOD:
1562      case OP_SOM:      case OP_SOM:
1563        case OP_SET_SOM:
1564      case OP_EOD:      case OP_EOD:
1565      case OP_EODN:      case OP_EODN:
1566      case OP_CIRC:      case OP_CIRC:
1567        case OP_CIRCM:
1568      case OP_DOLL:      case OP_DOLL:
1569        case OP_DOLLM:
1570      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1571      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1572      cc += _pcre_OP_lengths[*cc];      cc += _pcre_OP_lengths[*cc];
# Line 1229  for (;;) Line 1575  for (;;)
1575      /* Handle literal characters */      /* Handle literal characters */
1576    
1577      case OP_CHAR:      case OP_CHAR:
1578      case OP_CHARNC:      case OP_CHARI:
1579      case OP_NOT:      case OP_NOT:
1580        case OP_NOTI:
1581      branchlength++;      branchlength++;
1582      cc += 2;      cc += 2;
1583  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1584      if ((options & PCRE_UTF8) != 0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       {  
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1585  #endif  #endif
1586      break;      break;
1587    
# Line 1248  for (;;) Line 1592  for (;;)
1592      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1593      cc += 4;      cc += 4;
1594  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1595      if ((options & PCRE_UTF8) != 0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       {  
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1596  #endif  #endif
1597      break;      break;
1598    
# Line 1275  for (;;) Line 1616  for (;;)
1616      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1617      case OP_WORDCHAR:      case OP_WORDCHAR:
1618      case OP_ANY:      case OP_ANY:
1619        case OP_ALLANY:
1620      branchlength++;      branchlength++;
1621      cc++;      cc++;
1622      break;      break;
# Line 1329  for (;;) Line 1671  for (;;)
1671    
1672    
1673  /*************************************************  /*************************************************
1674  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1675  *************************************************/  *************************************************/
1676    
1677  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1678  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1679    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1680    so that it can be called from pcre_study() when finding the minimum matching
1681    length.
1682    
1683  Arguments:  Arguments:
1684    code        points to start of expression    code        points to start of expression
1685    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1686    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1687    
1688  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
1689  */  */
1690    
1691  static const uschar *  const uschar *
1692  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1693  {  {
1694  for (;;)  for (;;)
1695    {    {
# Line 1357  for (;;) Line 1702  for (;;)
1702    
1703    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1704    
1705      /* Handle recursion */
1706    
1707      else if (c == OP_REVERSE)
1708        {
1709        if (number < 0) return (uschar *)code;
1710        code += _pcre_OP_lengths[c];
1711        }
1712    
1713    /* Handle capturing bracket */    /* Handle capturing bracket */
1714    
1715    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
1716               c == OP_CBRAPOS || c == OP_SCBRAPOS)
1717      {      {
1718      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
1719      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
# Line 1368  for (;;) Line 1722  for (;;)
1722    
1723    /* 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
1724    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
1725    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1726      must add in its length. */
1727    
1728    else    else
1729      {      {
# Line 1392  for (;;) Line 1747  for (;;)
1747        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
1748        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1749        break;        break;
1750    
1751          case OP_MARK:
1752          case OP_PRUNE_ARG:
1753          case OP_SKIP_ARG:
1754          code += code[1];
1755          break;
1756    
1757          case OP_THEN_ARG:
1758          code += code[1+LINK_SIZE];
1759          break;
1760        }        }
1761    
1762      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1406  for (;;) Line 1771  for (;;)
1771      if (utf8) switch(c)      if (utf8) switch(c)
1772        {        {
1773        case OP_CHAR:        case OP_CHAR:
1774        case OP_CHARNC:        case OP_CHARI:
1775        case OP_EXACT:        case OP_EXACT:
1776          case OP_EXACTI:
1777        case OP_UPTO:        case OP_UPTO:
1778          case OP_UPTOI:
1779        case OP_MINUPTO:        case OP_MINUPTO:
1780          case OP_MINUPTOI:
1781        case OP_POSUPTO:        case OP_POSUPTO:
1782          case OP_POSUPTOI:
1783        case OP_STAR:        case OP_STAR:
1784          case OP_STARI:
1785        case OP_MINSTAR:        case OP_MINSTAR:
1786          case OP_MINSTARI:
1787        case OP_POSSTAR:        case OP_POSSTAR:
1788          case OP_POSSTARI:
1789        case OP_PLUS:        case OP_PLUS:
1790          case OP_PLUSI:
1791        case OP_MINPLUS:        case OP_MINPLUS:
1792          case OP_MINPLUSI:
1793        case OP_POSPLUS:        case OP_POSPLUS:
1794          case OP_POSPLUSI:
1795        case OP_QUERY:        case OP_QUERY:
1796          case OP_QUERYI:
1797        case OP_MINQUERY:        case OP_MINQUERY:
1798          case OP_MINQUERYI:
1799        case OP_POSQUERY:        case OP_POSQUERY:
1800          case OP_POSQUERYI:
1801        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1802        break;        break;
1803        }        }
1804    #else
1805        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1806  #endif  #endif
1807      }      }
1808    }    }
# Line 1461  for (;;) Line 1841  for (;;)
1841    
1842    /* 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
1843    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
1844    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1845      must add in its length. */
1846    
1847    else    else
1848      {      {
# Line 1485  for (;;) Line 1866  for (;;)
1866        case OP_TYPEEXACT:        case OP_TYPEEXACT:
1867        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1868        break;        break;
1869    
1870          case OP_MARK:
1871          case OP_PRUNE_ARG:
1872          case OP_SKIP_ARG:
1873          code += code[1];
1874          break;
1875    
1876          case OP_THEN_ARG:
1877          code += code[1+LINK_SIZE];
1878          break;
1879        }        }
1880    
1881      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1499  for (;;) Line 1890  for (;;)
1890      if (utf8) switch(c)      if (utf8) switch(c)
1891        {        {
1892        case OP_CHAR:        case OP_CHAR:
1893        case OP_CHARNC:        case OP_CHARI:
1894        case OP_EXACT:        case OP_EXACT:
1895          case OP_EXACTI:
1896        case OP_UPTO:        case OP_UPTO:
1897          case OP_UPTOI:
1898        case OP_MINUPTO:        case OP_MINUPTO:
1899          case OP_MINUPTOI:
1900        case OP_POSUPTO:        case OP_POSUPTO:
1901          case OP_POSUPTOI:
1902        case OP_STAR:        case OP_STAR:
1903          case OP_STARI:
1904        case OP_MINSTAR:        case OP_MINSTAR:
1905          case OP_MINSTARI:
1906        case OP_POSSTAR:        case OP_POSSTAR:
1907          case OP_POSSTARI:
1908        case OP_PLUS:        case OP_PLUS:
1909          case OP_PLUSI:
1910        case OP_MINPLUS:        case OP_MINPLUS:
1911          case OP_MINPLUSI:
1912        case OP_POSPLUS:        case OP_POSPLUS:
1913          case OP_POSPLUSI:
1914        case OP_QUERY:        case OP_QUERY:
1915          case OP_QUERYI:
1916        case OP_MINQUERY:        case OP_MINQUERY:
1917          case OP_MINQUERYI:
1918        case OP_POSQUERY:        case OP_POSQUERY:
1919          case OP_POSQUERYI:
1920        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1921        break;        break;
1922        }        }
1923    #else
1924        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1925  #endif  #endif
1926      }      }
1927    }    }
# Line 1539  Arguments: Line 1945  Arguments:
1945    code        points to start of search    code        points to start of search
1946    endcode     points to where to stop    endcode     points to where to stop
1947    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
1948      cd          contains pointers to tables etc.
1949    
1950  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
1951  */  */
1952    
1953  static BOOL  static BOOL
1954  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
1955      compile_data *cd)
1956  {  {
1957  register int c;  register int c;
1958  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], TRUE);
1959       code < endcode;       code < endcode;
1960       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], TRUE))
1961    {    {
1962    const uschar *ccode;    const uschar *ccode;
1963    
# Line 1565  for (code = first_significant_code(code Line 1973  for (code = first_significant_code(code
1973      continue;      continue;
1974      }      }
1975    
1976      /* For a recursion/subroutine call, if its end has been reached, which
1977      implies a subroutine call, we can scan it. */
1978    
1979      if (c == OP_RECURSE)
1980        {
1981        BOOL empty_branch = FALSE;
1982        const uschar *scode = cd->start_code + GET(code, 1);
1983        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
1984        do
1985          {
1986          if (could_be_empty_branch(scode, endcode, utf8, cd))
1987            {
1988            empty_branch = TRUE;
1989            break;
1990            }
1991          scode += GET(scode, 1);
1992          }
1993        while (*scode == OP_ALT);
1994        if (!empty_branch) return FALSE;  /* All branches are non-empty */
1995        continue;
1996        }
1997    
1998    /* Groups with zero repeats can of course be empty; skip them. */    /* Groups with zero repeats can of course be empty; skip them. */
1999    
2000    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2001          c == OP_BRAPOSZERO)
2002      {      {
2003      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
2004      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
# Line 1575  for (code = first_significant_code(code Line 2006  for (code = first_significant_code(code
2006      continue;      continue;
2007      }      }
2008    
2009      /* A nested group that is already marked as "could be empty" can just be
2010      skipped. */
2011    
2012      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2013          c == OP_SCBRA || c == OP_SCBRAPOS)
2014        {
2015        do code += GET(code, 1); while (*code == OP_ALT);
2016        c = *code;
2017        continue;
2018        }
2019    
2020    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
2021    
2022    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_BRA  || c == OP_BRAPOS ||
2023          c == OP_CBRA || c == OP_CBRAPOS ||
2024          c == OP_ONCE || c == OP_COND)
2025      {      {
2026      BOOL empty_branch;      BOOL empty_branch;
2027      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
2028    
2029      /* Scan a closed bracket */      /* If a conditional group has only one branch, there is a second, implied,
2030        empty branch, so just skip over the conditional, because it could be empty.
2031        Otherwise, scan the individual branches of the group. */
2032    
2033      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;  
2034        code += GET(code, 1);        code += GET(code, 1);
2035        else
2036          {
2037          empty_branch = FALSE;
2038          do
2039            {
2040            if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
2041              empty_branch = TRUE;
2042            code += GET(code, 1);
2043            }
2044          while (*code == OP_ALT);
2045          if (!empty_branch) return FALSE;   /* All branches are non-empty */
2046        }        }
2047      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
2048      c = *code;      c = *code;
2049      continue;      continue;
2050      }      }
# Line 1653  for (code = first_significant_code(code Line 2105  for (code = first_significant_code(code
2105      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2106      case OP_WORDCHAR:      case OP_WORDCHAR:
2107      case OP_ANY:      case OP_ANY:
2108        case OP_ALLANY:
2109      case OP_ANYBYTE:      case OP_ANYBYTE:
2110      case OP_CHAR:      case OP_CHAR:
2111      case OP_CHARNC:      case OP_CHARI:
2112      case OP_NOT:      case OP_NOT:
2113        case OP_NOTI:
2114      case OP_PLUS:      case OP_PLUS:
2115      case OP_MINPLUS:      case OP_MINPLUS:
2116      case OP_POSPLUS:      case OP_POSPLUS:
# Line 1696  for (code = first_significant_code(code Line 2150  for (code = first_significant_code(code
2150      case OP_KET:      case OP_KET:
2151      case OP_KETRMAX:      case OP_KETRMAX:
2152      case OP_KETRMIN:      case OP_KETRMIN:
2153        case OP_KETRPOS:
2154      case OP_ALT:      case OP_ALT:
2155      return TRUE;      return TRUE;
2156    
# Line 1704  for (code = first_significant_code(code Line 2159  for (code = first_significant_code(code
2159    
2160  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2161      case OP_STAR:      case OP_STAR:
2162        case OP_STARI:
2163      case OP_MINSTAR:      case OP_MINSTAR:
2164        case OP_MINSTARI:
2165      case OP_POSSTAR:      case OP_POSSTAR:
2166        case OP_POSSTARI:
2167      case OP_QUERY:      case OP_QUERY:
2168        case OP_QUERYI:
2169      case OP_MINQUERY:      case OP_MINQUERY:
2170        case OP_MINQUERYI:
2171      case OP_POSQUERY:      case OP_POSQUERY:
2172        case OP_POSQUERYI:
2173        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2174        break;
2175    
2176      case OP_UPTO:      case OP_UPTO:
2177        case OP_UPTOI:
2178      case OP_MINUPTO:      case OP_MINUPTO:
2179        case OP_MINUPTOI:
2180      case OP_POSUPTO:      case OP_POSUPTO:
2181      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      case OP_POSUPTOI:
2182        if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2183      break;      break;
2184  #endif  #endif
2185    
2186        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2187        string. */
2188    
2189        case OP_MARK:
2190        case OP_PRUNE_ARG:
2191        case OP_SKIP_ARG:
2192        code += code[1];
2193        break;
2194    
2195        case OP_THEN_ARG:
2196        code += code[1+LINK_SIZE];
2197        break;
2198    
2199        /* None of the remaining opcodes are required to match a character. */
2200    
2201        default:
2202        break;
2203      }      }
2204    }    }
2205    
# Line 1737  Arguments: Line 2222  Arguments:
2222    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2223    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2224    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2225      cd          pointers to tables etc
2226    
2227  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2228  */  */
2229    
2230  static BOOL  static BOOL
2231  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2232    BOOL utf8)    BOOL utf8, compile_data *cd)
2233  {  {
2234  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2235    {    {
2236    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2237        return FALSE;
2238    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2239    }    }
2240  return TRUE;  return TRUE;
# Line 1793  int terminator; /* Don't combin Line 2280  int terminator; /* Don't combin
2280  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2281  for (++ptr; *ptr != 0; ptr++)  for (++ptr; *ptr != 0; ptr++)
2282    {    {
2283    if (*ptr == '\\' && ptr[1] == ']') ptr++; else    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else
2284      {      {
2285      if (*ptr == ']') return FALSE;      if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2286      if (*ptr == terminator && ptr[1] == ']')      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2287        {        {
2288        *endptr = ptr;        *endptr = ptr;
2289        return TRUE;        return TRUE;
# Line 1847  return -1; Line 2334  return -1;
2334  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2335  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2336  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
2337  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
2338  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
2339  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
2340  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
2341  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2342    OP_END.
2343    
2344  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2345  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 1928  auto_callout(uschar *code, const uschar Line 2416  auto_callout(uschar *code, const uschar
2416  {  {
2417  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2418  *code++ = 255;  *code++ = 255;
2419  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2420  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2421  return code + 2*LINK_SIZE;  return code + 2*LINK_SIZE;
2422  }  }
2423    
# Line 1954  Returns: nothing Line 2442  Returns: nothing
2442  static void  static void
2443  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)
2444  {  {
2445  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2446  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2447  }  }
2448    
# Line 1986  get_othercase_range(unsigned int *cptr, Line 2474  get_othercase_range(unsigned int *cptr,
2474  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2475    
2476  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2477    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2478    
2479  if (c > d) return FALSE;  if (c > d) return FALSE;
2480    
# Line 1995  next = othercase + 1; Line 2483  next = othercase + 1;
2483    
2484  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2485    {    {
2486    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2487    next++;    next++;
2488    }    }
2489    
# Line 2004  for (++c; c <= d; c++) Line 2492  for (++c; c <= d; c++)
2492    
2493  return TRUE;  return TRUE;
2494  }  }
2495    
2496    
2497    
2498    /*************************************************
2499    *        Check a character and a property        *
2500    *************************************************/
2501    
2502    /* This function is called by check_auto_possessive() when a property item
2503    is adjacent to a fixed character.
2504    
2505    Arguments:
2506      c            the character
2507      ptype        the property type
2508      pdata        the data for the type
2509      negated      TRUE if it's a negated property (\P or \p{^)
2510    
2511    Returns:       TRUE if auto-possessifying is OK
2512    */
2513    
2514    static BOOL
2515    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2516    {
2517    const ucd_record *prop = GET_UCD(c);
2518    switch(ptype)
2519      {
2520      case PT_LAMP:
2521      return (prop->chartype == ucp_Lu ||
2522              prop->chartype == ucp_Ll ||
2523              prop->chartype == ucp_Lt) == negated;
2524    
2525      case PT_GC:
2526      return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;
2527    
2528      case PT_PC:
2529      return (pdata == prop->chartype) == negated;
2530    
2531      case PT_SC:
2532      return (pdata == prop->script) == negated;
2533    
2534      /* These are specials */
2535    
2536      case PT_ALNUM:
2537      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2538              _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;
2539    
2540      case PT_SPACE:    /* Perl space */
2541      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2542              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2543              == negated;
2544    
2545      case PT_PXSPACE:  /* POSIX space */
2546      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2547              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2548              c == CHAR_FF || c == CHAR_CR)
2549              == negated;
2550    
2551      case PT_WORD:
2552      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2553              _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2554              c == CHAR_UNDERSCORE) == negated;
2555      }
2556    return FALSE;
2557    }
2558  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2559    
2560    
# Line 2017  whether the next thing could possibly ma Line 2568  whether the next thing could possibly ma
2568  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
2569    
2570  Arguments:  Arguments:
2571    op_code       the repeated op code    previous      pointer to the repeated opcode
   this          data for this item, depends on the opcode  
2572    utf8          TRUE in UTF-8 mode    utf8          TRUE in UTF-8 mode
   utf8_char     used for utf8 character bytes, NULL if not relevant  
2573    ptr           next character in pattern    ptr           next character in pattern
2574    options       options bits    options       options bits
2575    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 2029  Returns: TRUE if possessifying is Line 2578  Returns: TRUE if possessifying is
2578  */  */
2579    
2580  static BOOL  static BOOL
2581  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
2582    const uschar *ptr, int options, compile_data *cd)    int options, compile_data *cd)
2583  {  {
2584  int next;  int c, next;
2585    int op_code = *previous++;
2586    
2587  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
2588    
# Line 2041  if ((options & PCRE_EXTENDED) != 0) Line 2591  if ((options & PCRE_EXTENDED) != 0)
2591    for (;;)    for (;;)
2592      {      {
2593      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2594      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2595        {        {
2596        while (*(++ptr) != 0)        ptr++;
2597          while (*ptr != 0)
2598            {
2599          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2600            ptr++;
2601    #ifdef SUPPORT_UTF8
2602            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2603    #endif
2604            }
2605        }        }
2606      else break;      else break;
2607      }      }
# Line 2053  if ((options & PCRE_EXTENDED) != 0) Line 2610  if ((options & PCRE_EXTENDED) != 0)
2610  /* 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
2611  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2612    
2613  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2614    {    {
2615    int temperrorcode = 0;    int temperrorcode = 0;
2616    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
# Line 2078  if ((options & PCRE_EXTENDED) != 0) Line 2635  if ((options & PCRE_EXTENDED) != 0)
2635    for (;;)    for (;;)
2636      {      {
2637      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2638      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2639        {        {
2640        while (*(++ptr) != 0)        ptr++;
2641          while (*ptr != 0)
2642            {
2643          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2644            ptr++;
2645    #ifdef SUPPORT_UTF8
2646            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2647    #endif
2648            }
2649        }        }
2650      else break;      else break;
2651      }      }
# Line 2089  if ((options & PCRE_EXTENDED) != 0) Line 2653  if ((options & PCRE_EXTENDED) != 0)
2653    
2654  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
2655    
2656  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2657    return FALSE;    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2658        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. */  
   
2659    
2660  /* Handle cases when the next item is a character. */  /* Now compare the next item with the previous opcode. First, handle cases when
2661    the next item is a character. */
2662    
2663  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
2664    {    {
2665    case OP_CHAR:    case OP_CHAR:
2666  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2667    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2668    #else
2669      c = *previous;
2670  #endif  #endif
2671    return item != next;    return c != next;
2672    
2673    /* For CHARNC (caseless character) we must check the other case. If we have    /* For CHARI (caseless character) we must check the other case. If we have
2674    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
2675    high-valued characters. */    high-valued characters. */
2676    
2677    case OP_CHARNC:    case OP_CHARI:
2678  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2679    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2680    #else
2681      c = *previous;
2682  #endif  #endif
2683    if (item == next) return FALSE;    if (c == next) return FALSE;
2684  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2685    if (utf8)    if (utf8)
2686      {      {
2687      unsigned int othercase;      unsigned int othercase;
2688      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2689  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2690      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
2691  #else  #else
2692      othercase = NOTACHAR;      othercase = NOTACHAR;
2693  #endif  #endif
2694      return (unsigned int)item != othercase;      return (unsigned int)c != othercase;
2695      }      }
2696    else    else
2697  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2698    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2699    
2700    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT and OP_NOTI, the data is always a single-byte character. These
2701      opcodes are not used for multi-byte characters, because they are coded using
2702      an XCLASS instead. */
2703    
2704    case OP_NOT:    case OP_NOT:
2705    if (item == next) return TRUE;    return (c = *previous) == next;
2706    if ((options & PCRE_CASELESS) == 0) return FALSE;  
2707      case OP_NOTI:
2708      if ((c = *previous) == next) return TRUE;
2709  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2710    if (utf8)    if (utf8)
2711      {      {
2712      unsigned int othercase;      unsigned int othercase;
2713      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2714  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2715      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
2716  #else  #else
2717      othercase = NOTACHAR;      othercase = NOTACHAR;
2718  #endif  #endif
2719      return (unsigned int)item == othercase;      return (unsigned int)c == othercase;
2720      }      }
2721    else    else
2722  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2723    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2724    
2725      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2726      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
2727    
2728    case OP_DIGIT:    case OP_DIGIT:
2729    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
# Line 2195  if (next >= 0) switch(op_code) Line 2766  if (next >= 0) switch(op_code)
2766      case 0x202f:      case 0x202f:
2767      case 0x205f:      case 0x205f:
2768      case 0x3000:      case 0x3000:
2769      return op_code != OP_HSPACE;      return op_code == OP_NOT_HSPACE;
2770      default:      default:
2771      return op_code == OP_HSPACE;      return op_code != OP_NOT_HSPACE;
2772      }      }
2773    
2774      case OP_ANYNL:
2775    case OP_VSPACE:    case OP_VSPACE:
2776    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
2777    switch(next)    switch(next)
# Line 2211  if (next >= 0) switch(op_code) Line 2783  if (next >= 0) switch(op_code)
2783      case 0x85:      case 0x85:
2784      case 0x2028:      case 0x2028:
2785      case 0x2029:      case 0x2029:
2786      return op_code != OP_VSPACE;      return op_code == OP_NOT_VSPACE;
2787      default:      default:
2788      return op_code == OP_VSPACE;      return op_code != OP_NOT_VSPACE;
2789      }      }
2790    
2791    #ifdef SUPPORT_UCP
2792      case OP_PROP:
2793      return check_char_prop(next, previous[0], previous[1], FALSE);
2794    
2795      case OP_NOTPROP:
2796      return check_char_prop(next, previous[0], previous[1], TRUE);
2797    #endif
2798    
2799    default:    default:
2800    return FALSE;    return FALSE;
2801    }    }
2802    
2803    
2804  /* 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
2805    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
2806    generated only when PCRE_UCP is *not* set, that is, when only ASCII
2807    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
2808    replaced by OP_PROP codes when PCRE_UCP is set. */
2809    
2810  switch(op_code)  switch(op_code)
2811    {    {
2812    case OP_CHAR:    case OP_CHAR:
2813    case OP_CHARNC:    case OP_CHARI:
2814  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2815    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2816    #else
2817      c = *previous;
2818  #endif  #endif
2819    switch(-next)    switch(-next)
2820      {      {
2821      case ESC_d:      case ESC_d:
2822      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
2823    
2824      case ESC_D:      case ESC_D:
2825      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
2826    
2827      case ESC_s:      case ESC_s:
2828      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
2829    
2830      case ESC_S:      case ESC_S:
2831      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
2832    
2833      case ESC_w:      case ESC_w:
2834      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
2835    
2836      case ESC_W:      case ESC_W:
2837      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
2838    
2839      case ESC_h:      case ESC_h:
2840      case ESC_H:      case ESC_H:
2841      switch(item)      switch(c)
2842        {        {
2843        case 0x09:        case 0x09:
2844        case 0x20:        case 0x20:
# Line 2280  switch(op_code) Line 2866  switch(op_code)
2866    
2867      case ESC_v:      case ESC_v:
2868      case ESC_V:      case ESC_V:
2869      switch(item)      switch(c)
2870        {        {
2871        case 0x0a:        case 0x0a:
2872        case 0x0b:        case 0x0b:
# Line 2294  switch(op_code) Line 2880  switch(op_code)
2880        return -next == ESC_v;        return -next == ESC_v;
2881        }        }
2882    
2883        /* When PCRE_UCP is set, these values get generated for \d etc. Find
2884        their substitutions and process them. The result will always be either
2885        -ESC_p or -ESC_P. Then fall through to process those values. */
2886    
2887    #ifdef SUPPORT_UCP
2888        case ESC_du:
2889        case ESC_DU:
2890        case ESC_wu:
2891        case ESC_WU:
2892        case ESC_su:
2893        case ESC_SU:
2894          {
2895          int temperrorcode = 0;
2896          ptr = substitutes[-next - ESC_DU];
2897          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
2898          if (temperrorcode != 0) return FALSE;
2899          ptr++;    /* For compatibility */
2900          }
2901        /* Fall through */
2902    
2903        case ESC_p:
2904        case ESC_P:
2905          {
2906          int ptype, pdata, errorcodeptr;
2907          BOOL negated;
2908    
2909          ptr--;      /* Make ptr point at the p or P */
2910          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
2911          if (ptype < 0) return FALSE;
2912          ptr++;      /* Point past the final curly ket */
2913    
2914          /* If the property item is optional, we have to give up. (When generated
2915          from \d etc by PCRE_UCP, this test will have been applied much earlier,
2916          to the original \d etc. At this point, ptr will point to a zero byte. */
2917    
2918          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2919            strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2920              return FALSE;
2921    
2922          /* Do the property check. */
2923    
2924          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
2925          }
2926    #endif
2927    
2928      default:      default:
2929      return FALSE;      return FALSE;
2930      }      }
2931    
2932      /* In principle, support for Unicode properties should be integrated here as
2933      well. It means re-organizing the above code so as to get hold of the property
2934      values before switching on the op-code. However, I wonder how many patterns
2935      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
2936      these op-codes are never generated.) */
2937    
2938    case OP_DIGIT:    case OP_DIGIT:
2939    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2940           next == -ESC_h || next == -ESC_v;           next == -ESC_h || next == -ESC_v || next == -ESC_R;
2941    
2942    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
2943    return next == -ESC_d;    return next == -ESC_d;
2944    
2945    case OP_WHITESPACE:    case OP_WHITESPACE:
2946    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
2947    
2948    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
2949    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2950    
2951    case OP_HSPACE:    case OP_HSPACE:
2952    return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
2953             next == -ESC_w || next == -ESC_v || next == -ESC_R;
2954    
2955    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
2956    return next == -ESC_h;    return next == -ESC_h;
2957    
2958    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
2959      case OP_ANYNL:
2960    case OP_VSPACE:    case OP_VSPACE:
2961    return next == -ESC_V || next == -ESC_d || next == -ESC_w;    return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2962    
2963    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
2964    return next == -ESC_v;    return next == -ESC_v || next == -ESC_R;
2965    
2966    case OP_WORDCHAR:    case OP_WORDCHAR:
2967    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
2968             next == -ESC_v || next == -ESC_R;
2969    
2970    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
2971    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2389  BOOL inescq = FALSE; Line 3029  BOOL inescq = FALSE;
3029  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
3030  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
3031  const uschar *tempptr;  const uschar *tempptr;
3032    const uschar *nestptr = NULL;
3033  uschar *previous = NULL;  uschar *previous = NULL;
3034  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
3035  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
# Line 2405  BOOL utf8 = FALSE; Line 3046  BOOL utf8 = FALSE;
3046  uschar *utf8_char = NULL;  uschar *utf8_char = NULL;
3047  #endif  #endif
3048    
3049  #ifdef DEBUG  #ifdef PCRE_DEBUG
3050  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
3051  #endif  #endif
3052    
# Line 2459  for (;; ptr++) Line 3100  for (;; ptr++)
3100    
3101    c = *ptr;    c = *ptr;
3102    
3103      /* If we are at the end of a nested substitution, revert to the outer level
3104      string. Nesting only happens one level deep. */
3105    
3106      if (c == 0 && nestptr != NULL)
3107        {
3108        ptr = nestptr;
3109        nestptr = NULL;
3110        c = *ptr;
3111        }
3112    
3113    /* 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
3114    previous cycle of this loop. */    previous cycle of this loop. */
3115    
3116    if (lengthptr != NULL)    if (lengthptr != NULL)
3117      {      {
3118  #ifdef DEBUG  #ifdef PCRE_DEBUG
3119      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3120  #endif  #endif
3121      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
3122        {        {
3123        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3124        goto FAILED;        goto FAILED;
# Line 2489  for (;; ptr++) Line 3140  for (;; ptr++)
3140        goto FAILED;        goto FAILED;
3141        }        }
3142    
3143      *lengthptr += code - last_code;      *lengthptr += (int)(code - last_code);
3144      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));
3145    
3146      /* 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 2516  for (;; ptr++) Line 3167  for (;; ptr++)
3167    /* 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
3168    reference list. */    reference list. */
3169    
3170    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
3171      {      {
3172      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3173      goto FAILED;      goto FAILED;
# Line 2526  for (;; ptr++) Line 3177  for (;; ptr++)
3177    
3178    if (inescq && c != 0)    if (inescq && c != 0)
3179      {      {
3180      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3181        {        {
3182        inescq = FALSE;        inescq = FALSE;
3183        ptr++;        ptr++;
# Line 2552  for (;; ptr++) Line 3203  for (;; ptr++)
3203    /* 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
3204    a quantifier. */    a quantifier. */
3205    
3206    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
3207      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
3208        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
3209    
3210    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
3211         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2568  for (;; ptr++) Line 3220  for (;; ptr++)
3220    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3221      {      {
3222      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
3223      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
3224        {        {
3225        while (*(++ptr) != 0)        ptr++;
3226          while (*ptr != 0)
3227          {          {
3228          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3229            ptr++;
3230    #ifdef SUPPORT_UTF8
3231            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
3232    #endif
3233          }          }
3234        if (*ptr != 0) continue;        if (*ptr != 0) continue;
3235    
# Line 2593  for (;; ptr++) Line 3250  for (;; ptr++)
3250      {      {
3251      /* ===================================================================*/      /* ===================================================================*/
3252      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
3253      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
3254      case ')':      case CHAR_RIGHT_PARENTHESIS:
3255      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
3256      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
3257      *codeptr = code;      *codeptr = code;
# Line 2606  for (;; ptr++) Line 3263  for (;; ptr++)
3263          *errorcodeptr = ERR20;          *errorcodeptr = ERR20;
3264          goto FAILED;          goto FAILED;
3265          }          }
3266        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += (int)(code - last_code);   /* To include callout length */
3267        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
3268        }        }
3269      return TRUE;      return TRUE;
# Line 2616  for (;; ptr++) Line 3273  for (;; ptr++)
3273      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
3274      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3275    
3276      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
3277        previous = NULL;
3278      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3279        {        {
3280        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3281          *code++ = OP_CIRCM;
3282        }        }
3283      previous = NULL;      else *code++ = OP_CIRC;
     *code++ = OP_CIRC;  
3284      break;      break;
3285    
3286      case '$':      case CHAR_DOLLAR_SIGN:
3287      previous = NULL;      previous = NULL;
3288      *code++ = OP_DOLL;      *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
3289      break;      break;
3290    
3291      /* 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
3292      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
3293    
3294      case '.':      case CHAR_DOT:
3295      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3296      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
3297      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3298      previous = code;      previous = code;
3299      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
3300      break;      break;
3301    
3302    
# Line 2653  for (;; ptr++) Line 3311  for (;; ptr++)
3311      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,
3312      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
3313      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.
     */  
3314    
3315      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
3316        default (Perl) mode, it is treated as a data character. */
3317    
3318        case CHAR_RIGHT_SQUARE_BRACKET:
3319        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3320          {
3321          *errorcodeptr = ERR64;
3322          goto FAILED;
3323          }
3324        goto NORMAL_CHAR;
3325    
3326        case CHAR_LEFT_SQUARE_BRACKET:
3327      previous = code;      previous = code;
3328    
3329      /* 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
3330      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. */
3331    
3332      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3333             ptr[1] == CHAR_EQUALS_SIGN) &&
3334          check_posix_syntax(ptr, &tempptr))          check_posix_syntax(ptr, &tempptr))
3335        {        {
3336        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
3337        goto FAILED;        goto FAILED;
3338        }        }
3339    
# Line 2676  for (;; ptr++) Line 3345  for (;; ptr++)
3345      for (;;)      for (;;)
3346        {        {
3347        c = *(++ptr);        c = *(++ptr);
3348        if (c == '\\')        if (c == CHAR_BACKSLASH)
3349          {          {
3350          if (ptr[1] == 'E') ptr++;          if (ptr[1] == CHAR_E)
3351            else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;            ptr++;
3352              else break;          else if (strncmp((const char *)ptr+1,
3353                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
3354              ptr += 3;
3355            else
3356              break;
3357          }          }
3358        else if (!negate_class && c == '^')        else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
3359          negate_class = TRUE;          negate_class = TRUE;
3360        else break;        else break;
3361        }        }
3362    
3363        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
3364        an initial ']' is taken as a data character -- the code below handles
3365        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3366        [^] must match any character, so generate OP_ALLANY. */
3367    
3368        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3369            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3370          {
3371          *code++ = negate_class? OP_ALLANY : OP_FAIL;
3372          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3373          zerofirstbyte = firstbyte;
3374          break;
3375          }
3376    
3377      /* 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
3378      negation flag at the end, so that support for characters > 255 works      negation flag at the end, so that support for characters > 255 works
3379      correctly (they are all included in the class). */      correctly (they are all included in the class). */
# Line 2744  for (;; ptr++) Line 3431  for (;; ptr++)
3431    
3432        if (inescq)        if (inescq)
3433          {          {
3434          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3435            {            {
3436            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
3437            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2759  for (;; ptr++) Line 3446  for (;; ptr++)
3446        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3447        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3448    
3449        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3450            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3451            check_posix_syntax(ptr, &tempptr))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3452          {          {
3453          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3454          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3455          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3456          uschar pbits[32];          uschar pbits[32];
3457    
3458          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3459            {            {
3460            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3461            goto FAILED;            goto FAILED;
3462            }            }
3463    
3464          ptr += 2;          ptr += 2;
3465          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3466            {            {
3467            local_negate = TRUE;            local_negate = TRUE;
3468            should_flip_negation = TRUE;  /* Note negative special */            should_flip_negation = TRUE;  /* Note negative special */
3469            ptr++;            ptr++;
3470            }            }
3471    
3472          posix_class = check_posix_name(ptr, tempptr - ptr);          posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3473          if (posix_class < 0)          if (posix_class < 0)
3474            {            {
3475            *errorcodeptr = ERR30;            *errorcodeptr = ERR30;
# Line 2796  for (;; ptr++) Line 3483  for (;; ptr++)
3483          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3484            posix_class = 0;            posix_class = 0;
3485    
3486          /* 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
3487          because we may be adding and subtracting from it, and we don't want to          different escape sequences that use Unicode properties. */
3488          subtract bits that may be in the main map already. At the end we or the  
3489          result into the bit map that is being built. */  #ifdef SUPPORT_UCP
3490            if ((options & PCRE_UCP) != 0)
3491              {
3492              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3493              if (posix_substitutes[pc] != NULL)
3494                {
3495                nestptr = tempptr + 1;
3496                ptr = posix_substitutes[pc] - 1;
3497                continue;
3498                }
3499              }
3500    #endif
3501            /* In the non-UCP case, we build the bit map for the POSIX class in a
3502            chunk of local store because we may be adding and subtracting from it,
3503            and we don't want to subtract bits that may be in the main map already.
3504            At the end we or the result into the bit map that is being built. */
3505    
3506          posix_class *= 3;          posix_class *= 3;
3507    
# Line 2843  for (;; ptr++) Line 3545  for (;; ptr++)
3545    
3546        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3547        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
3548        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
3549        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
3550        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
3551        character in them, so set class_charcount bigger than one. */        are either treated as literal characters (by default), or are faulted if
3552          PCRE_EXTRA is set. */
3553    
3554        if (c == '\\')        if (c == CHAR_BACKSLASH)
3555          {          {
3556          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3557          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3558    
3559          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 */  
3560          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3561            {            {
3562            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3563              {              {
3564              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3565              }              }
# Line 2872  for (;; ptr++) Line 3573  for (;; ptr++)
3573            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
3574            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3575    
3576            /* Save time by not doing this in the pre-compile phase. */            switch (-c)
   
           if (lengthptr == NULL) switch (-c)  
3577              {              {
3578    #ifdef SUPPORT_UCP
3579                case ESC_du:     /* These are the values given for \d etc */
3580                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3581                case ESC_wu:     /* escape sequence with an appropriate \p */
3582                case ESC_WU:     /* or \P to test Unicode properties instead */
3583                case ESC_su:     /* of the default ASCII testing. */
3584                case ESC_SU:
3585                nestptr = ptr;
3586                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3587                class_charcount -= 2;                /* Undo! */
3588                continue;
3589    #endif
3590              case ESC_d:              case ESC_d:
3591              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3592              continue;              continue;
# Line 2894  for (;; ptr++) Line 3605  for (;; ptr++)
3605              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3606              continue;              continue;
3607    
3608                /* Perl 5.004 onwards omits VT from \s, but we must preserve it
3609                if it was previously set by something earlier in the character
3610                class. */
3611    
3612              case ESC_s:              case ESC_s:
3613              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];              classbits[0] |= cbits[cbit_space];
3614              classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= cbits[cbit_space+1] & ~0x08;
3615                for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3616              continue;              continue;
3617    
3618              case ESC_S:              case ESC_S:
# Line 2905  for (;; ptr++) Line 3621  for (;; ptr++)
3621              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3622              continue;              continue;
3623    
3624              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)  
             {  
3625              SETBIT(classbits, 0x09); /* VT */              SETBIT(classbits, 0x09); /* VT */
3626              SETBIT(classbits, 0x20); /* SPACE */              SETBIT(classbits, 0x20); /* SPACE */
3627              SETBIT(classbits, 0xa0); /* NSBP */              SETBIT(classbits, 0xa0); /* NSBP */
# Line 2942  for (;; ptr++) Line 3645  for (;; ptr++)
3645                }                }
3646  #endif  #endif
3647              continue;              continue;
             }  
3648    
3649            if (-c == ESC_H)              case ESC_H:
             {  
3650              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3651                {                {
3652                int x = 0xff;                int x = 0xff;
# Line 2987  for (;; ptr++) Line 3688  for (;; ptr++)
3688                }                }
3689  #endif  #endif
3690              continue;              continue;
             }  
3691    
3692            if (-c == ESC_v)              case ESC_v:
             {  
3693              SETBIT(classbits, 0x0a); /* LF */              SETBIT(classbits, 0x0a); /* LF */
3694              SETBIT(classbits, 0x0b); /* VT */              SETBIT(classbits, 0x0b); /* VT */
3695              SETBIT(classbits, 0x0c); /* FF */              SETBIT(classbits, 0x0c); /* FF */
# Line 3006  for (;; ptr++) Line 3705  for (;; ptr++)
3705                }                }
3706  #endif  #endif
3707              continue;              continue;
             }  
3708    
3709            if (-c == ESC_V)              case ESC_V:
             {  
3710              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3711                {                {
3712                int x = 0xff;                int x = 0xff;
# Line 3039  for (;; ptr++) Line 3736  for (;; ptr++)
3736                }                }
3737  #endif  #endif
3738              continue;              continue;
             }  
   
           /* We need to deal with \P and \p in both phases. */  
3739    
3740  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3741            if (-c == ESC_p || -c == ESC_P)              case ESC_p:
3742              {              case ESC_P:
3743              BOOL negated;                {
3744              int pdata;                BOOL negated;
3745              int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                int pdata;
3746              if (ptype < 0) goto FAILED;                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3747              class_utf8 = TRUE;                if (ptype < 0) goto FAILED;
3748              *class_utf8data++ = ((-c == ESC_p) != negated)?                class_utf8 = TRUE;
3749                XCL_PROP : XCL_NOTPROP;                *class_utf8data++ = ((-c == ESC_p) != negated)?
3750              *class_utf8data++ = ptype;                  XCL_PROP : XCL_NOTPROP;
3751              *class_utf8data++ = pdata;                *class_utf8data++ = ptype;
3752              class_charcount -= 2;   /* Not a < 256 character */                *class_utf8data++ = pdata;
3753              continue;                class_charcount -= 2;   /* Not a < 256 character */
3754              }                continue;
3755                  }
3756  #endif  #endif
3757            /* Unrecognized escapes are faulted if PCRE is running in its              /* Unrecognized escapes are faulted if PCRE is running in its
3758            strict mode. By default, for compatibility with Perl, they are              strict mode. By default, for compatibility with Perl, they are
3759            treated as literals. */              treated as literals. */
3760    
3761            if ((options & PCRE_EXTRA) != 0)              default:
3762              {              if ((options & PCRE_EXTRA) != 0)
3763              *errorcodeptr = ERR7;                {
3764              goto FAILED;                *errorcodeptr = ERR7;
3765                  goto FAILED;
3766                  }
3767                class_charcount -= 2;  /* Undo the default count from above */
3768                c = *ptr;              /* Get the final character and fall through */
3769                break;
3770              }              }
   
           class_charcount -= 2;  /* Undo the default count from above */  
           c = *ptr;              /* Get the final character and fall through */  
3771            }            }
3772    
3773          /* Fall through if we have a single character (c >= 0). This may be          /* Fall through if we have a single character (c >= 0). This may be
# Line 3084  for (;; ptr++) Line 3781  for (;; ptr++)
3781        entirely. The code for handling \Q and \E is messy. */        entirely. The code for handling \Q and \E is messy. */
3782    
3783        CHECK_RANGE:        CHECK_RANGE:
3784        while (ptr[1] == '\\' && ptr[2] == 'E')        while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3785          {          {
3786          inescq = FALSE;          inescq = FALSE;
3787          ptr += 2;          ptr += 2;
# Line 3094  for (;; ptr++) Line 3791  for (;; ptr++)
3791    
3792        /* Remember \r or \n */        /* Remember \r or \n */
3793    
3794        if (c == '\r' || c == '\n') cd->external_flags |= PCRE_HASCRORLF;        if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3795    
3796        /* Check for range */        /* Check for range */
3797    
3798        if (!inescq && ptr[1] == '-')        if (!inescq && ptr[1] == CHAR_MINUS)
3799          {          {
3800          int d;          int d;
3801          ptr += 2;          ptr += 2;
3802          while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3803    
3804          /* 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
3805          mode. */          mode. */
3806    
3807          while (*ptr == '\\' && ptr[1] == 'Q')          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3808            {            {
3809            ptr += 2;            ptr += 2;
3810            if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }            if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3811                { ptr += 2; continue; }
3812            inescq = TRUE;            inescq = TRUE;
3813            break;            break;
3814            }            }
3815    
3816          if (*ptr == 0 || (!inescq && *ptr == ']'))          if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3817            {            {
3818            ptr = oldptr;            ptr = oldptr;
3819            goto LONE_SINGLE_CHARACTER;            goto LONE_SINGLE_CHARACTER;
# Line 3134  for (;; ptr++) Line 3832  for (;; ptr++)
3832          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
3833          in such circumstances. */          in such circumstances. */
3834    
3835          if (!inescq && d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3836            {            {
3837            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3838            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3839    
3840            /* \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 */  
3841    
3842            if (d < 0)            if (d < 0)
3843              {              {
3844              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  
3845                {                {
3846                ptr = oldptr;                ptr = oldptr;
3847                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 3167  for (;; ptr++) Line 3862  for (;; ptr++)
3862    
3863          /* Remember \r or \n */          /* Remember \r or \n */
3864    
3865          if (d == '\r' || d == '\n') cd->external_flags |= PCRE_HASCRORLF;          if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3866    
3867          /* 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
3868          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
# Line 3287  for (;; ptr++) Line 3982  for (;; ptr++)
3982          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3983            {            {
3984            unsigned int othercase;            unsigned int othercase;
3985            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)            if ((othercase = UCD_OTHERCASE(c)) != c)
3986              {              {
3987              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3988              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 3312  for (;; ptr++) Line 4007  for (;; ptr++)
4007          }          }
4008        }        }
4009    
4010      /* 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.
4011        If we are at the end of an internal nested string, revert to the outer
4012        string. */
4013    
4014        while (((c = *(++ptr)) != 0 ||
4015               (nestptr != NULL &&
4016                 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != 0)) &&
4017               (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
4018    
4019      while ((c = *(++ptr)) != 0 && (c != ']' || inescq));      /* Check for missing terminating ']' */
4020    
4021      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)
4022        {        {
4023        *errorcodeptr = ERR6;        *errorcodeptr = ERR6;
4024        goto FAILED;        goto FAILED;
4025        }        }
4026    
   
 /* 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  
   
   
4027      /* 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
4028      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
4029      use of \p or \P, in other words, no use of any XCLASS features, we can      use of \p or \P, in other words, no use of any XCLASS features, we can
# Line 3348  we set the flag only if there is a liter Line 4031  we set the flag only if there is a liter
4031    
4032      In UTF-8 mode, we can optimize the negative case only if there were no      In UTF-8 mode, we can optimize the negative case only if there were no
4033      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
4034      operate on single-bytes only. This is an historical hangover. Maybe one day      operate on single-bytes characters only. This is an historical hangover.
4035      we can tidy these opcodes to handle multi-byte characters.      Maybe one day we can tidy these opcodes to handle multi-byte characters.
4036    
4037      The optimization throws away the bit map. We turn the item into a      The optimization throws away the bit map. We turn the item into a
4038      1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note      1-character OP_CHAR[I] if it's positive, or OP_NOT[I] if it's negative.
4039      that OP_NOT does not support multibyte characters. In the positive case, it      Note that OP_NOT[I] does not support multibyte characters. In the positive
4040      can cause firstbyte to be set. Otherwise, there can be no first char if      case, it can cause firstbyte to be set. Otherwise, there can be no first
4041      this item is first, whatever repeat count may follow. In the case of      char if this item is first, whatever repeat count may follow. In the case
4042      reqbyte, save the previous value for reinstating. */      of reqbyte, save the previous value for reinstating. */
4043    
4044  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4045      if (class_charcount == 1 && !class_utf8 &&      if (class_charcount == 1 && !class_utf8 &&
# Line 3367  we set the flag only if there is a liter Line 4050  we set the flag only if there is a liter
4050        {        {
4051        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
4052    
4053        /* The OP_NOT opcode works on one-byte characters only. */        /* The OP_NOT[I] opcodes work on one-byte characters only. */
4054    
4055        if (negate_class)        if (negate_class)
4056          {          {
4057          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
4058          zerofirstbyte = firstbyte;          zerofirstbyte = firstbyte;
4059          *code++ = OP_NOT;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
4060          *code++ = class_lastchar;          *code++ = class_lastchar;
4061          break;          break;
4062          }          }
# Line 3404  we set the flag only if there is a liter Line 4087  we set the flag only if there is a liter
4087    
4088      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
4089      extended class, with its own opcode, unless there was a negated special      extended class, with its own opcode, unless there was a negated special
4090      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
4091      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
4092      (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
4093      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
4094        actual compiled code. */
4095    
4096  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4097      if (class_utf8 && !should_flip_negation)      if (class_utf8 && (!should_flip_negation || (options & PCRE_UCP) != 0))
4098        {        {
4099        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
4100        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 3436  we set the flag only if there is a liter Line 4120  we set the flag only if there is a liter
4120        }        }
4121  #endif  #endif
4122    
4123      /* 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
4124      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
4125      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
4126      map into the code vector, negating it if necessary. */      (non-UCP) in the class. Then copy the 32-byte map into the code vector,
4127        negating it if necessary. */
4128    
4129      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
4130      if (negate_class)      if (negate_class)
# Line 3459  we set the flag only if there is a liter Line 4144  we set the flag only if there is a liter
4144      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
4145      has been tested above. */      has been tested above. */
4146    
4147      case '{':      case CHAR_LEFT_CURLY_BRACKET:
4148      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
4149      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
4150      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
4151      goto REPEAT;      goto REPEAT;
4152    
4153      case '*':      case CHAR_ASTERISK:
4154      repeat_min = 0;      repeat_min = 0;
4155      repeat_max = -1;      repeat_max = -1;
4156      goto REPEAT;      goto REPEAT;
4157    
4158      case '+':      case CHAR_PLUS:
4159      repeat_min = 1;      repeat_min = 1;
4160      repeat_max = -1;      repeat_max = -1;
4161      goto REPEAT;      goto REPEAT;
4162    
4163      case '?':      case CHAR_QUESTION_MARK:
4164      repeat_min = 0;      repeat_min = 0;
4165      repeat_max = 1;      repeat_max = 1;
4166    
# Line 3510  we set the flag only if there is a liter Line 4195  we set the flag only if there is a liter
4195      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
4196      repeat type to the non-default. */      repeat type to the non-default. */
4197    
4198      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
4199        {        {
4200        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
4201        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
4202        ptr++;        ptr++;
4203        }        }
4204      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
4205        {        {
4206        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
4207        ptr++;        ptr++;
# Line 3529  we set the flag only if there is a liter Line 4214  we set the flag only if there is a liter
4214      the first thing in a branch because the x will have gone into firstbyte      the first thing in a branch because the x will have gone into firstbyte
4215      instead.  */      instead.  */
4216    
4217      if (*previous == OP_CHAR || *previous == OP_CHARNC)      if (*previous == OP_CHAR || *previous == OP_CHARI)
4218        {        {
4219          op_type = (*previous == OP_CHAR)? 0 : OP_STARI - OP_STAR;
4220    
4221        /* Deal with UTF-8 characters that take up more than one byte. It's        /* Deal with UTF-8 characters that take up more than one byte. It's
4222        easier to write this out separately than try to macrify it. Use c to        easier to write this out separately than try to macrify it. Use c to
4223        hold the length of the character in bytes, plus 0x80 to flag that it's a        hold the length of the character in bytes, plus 0x80 to flag that it's a
# Line 3563  we set the flag only if there is a liter Line 4250  we set the flag only if there is a liter
4250    
4251        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4252            repeat_max < 0 &&            repeat_max < 0 &&
4253            check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
             options, cd))  
4254          {          {
4255          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4256          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3576  we set the flag only if there is a liter Line 4262  we set the flag only if there is a liter
4262      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
4263      one of the special opcodes, replacing it. The code is shared with single-      one of the special opcodes, replacing it. The code is shared with single-
4264      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
4265      repeat_type. We can also test for auto-possessification. OP_NOT is      repeat_type. We can also test for auto-possessification. OP_NOT and OP_NOTI
4266      currently used only for single-byte chars. */      are currently used only for single-byte chars. */
4267    
4268      else if (*previous == OP_NOT)      else if (*previous == OP_NOT || *previous == OP_NOTI)
4269        {        {
4270        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = ((*previous == OP_NOT)? OP_NOTSTAR : OP_NOTSTARI) - OP_STAR;
4271        c = previous[1];        c = previous[1];
4272        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4273            repeat_max < 0 &&            repeat_max < 0 &&
4274            check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4275          {          {
4276          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4277          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3609  we set the flag only if there is a liter Line 4295  we set the flag only if there is a liter
4295    
4296        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4297            repeat_max < 0 &&            repeat_max < 0 &&
4298            check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4299          {          {
4300          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4301          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3631  we set the flag only if there is a liter Line 4317  we set the flag only if there is a liter
4317    
4318        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
4319    
4320          /*--------------------------------------------------------------------*/
4321          /* This code is obsolete from release 8.00; the restriction was finally
4322          removed: */
4323    
4324        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4325        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4326    
4327        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4328          /*--------------------------------------------------------------------*/
4329    
4330        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
4331    
# Line 3773  we set the flag only if there is a liter Line 4464  we set the flag only if there is a liter
4464  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4465               *previous == OP_XCLASS ||               *previous == OP_XCLASS ||
4466  #endif  #endif
4467               *previous == OP_REF)               *previous == OP_REF ||
4468                 *previous == OP_REFI)
4469        {        {
4470        if (repeat_max == 0)        if (repeat_max == 0)
4471          {          {
# Line 3781  we set the flag only if there is a liter Line 4473  we set the flag only if there is a liter
4473          goto END_REPEAT;          goto END_REPEAT;
4474          }          }
4475    
4476          /*--------------------------------------------------------------------*/
4477          /* This code is obsolete from release 8.00; the restriction was finally
4478          removed: */
4479    
4480        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4481        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4482    
4483        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4484          /*--------------------------------------------------------------------*/
4485    
4486        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4487          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 3802  we set the flag only if there is a liter Line 4499  we set the flag only if there is a liter
4499        }        }
4500    
4501      /* If previous was a bracket group, we may have to replicate it in certain      /* If previous was a bracket group, we may have to replicate it in certain
4502      cases. */      cases. Note that at this point we can encounter only the "basic" BRA and
4503        KET opcodes, as this is the place where they get converted into the more
4504        special varieties. */
4505    
4506      else if (*previous == OP_BRA  || *previous == OP_CBRA ||      else if (*previous == OP_BRA  || *previous == OP_CBRA ||
4507               *previous == OP_ONCE || *previous == OP_COND)               *previous == OP_ONCE || *previous == OP_COND)
4508        {        {
4509        register int i;        register int i;
4510        int ketoffset = 0;        int len = (int)(code - previous);
       int len = code - previous;  
4511        uschar *bralink = NULL;        uschar *bralink = NULL;
4512          uschar *brazeroptr = NULL;
4513    
4514        /* Repeating a DEFINE group is pointless */        /* Repeating a DEFINE group is pointless */
4515    
4516        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
# Line 3820  we set the flag only if there is a liter Line 4519  we set the flag only if there is a liter
4519          goto FAILED;          goto FAILED;
4520          }          }
4521    
       /* If the maximum repeat count is unlimited, find the end of the bracket  
       by scanning through from the start, and compute the offset back to it  
       from the current code pointer. There may be an OP_OPT setting following  
       the final KET, so we can't find the end just by going back from the code  
       pointer. */  
   
       if (repeat_max == -1)  
         {  
         register uschar *ket = previous;  
         do ket += GET(ket, 1); while (*ket != OP_KET);  
         ketoffset = code - ket;  
         }  
   
4522        /* 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
4523        OP_BRAZERO in front of it, and because the group appears once in the        OP_BRAZERO in front of it, and because the group appears once in the
4524        data, whereas in other cases it appears the minimum number of times. For        data, whereas in other cases it appears the minimum number of times. For
# Line 3842  we set the flag only if there is a liter Line 4528  we set the flag only if there is a liter
4528    
4529        if (repeat_min == 0)        if (repeat_min == 0)
4530          {          {
4531          /* 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
4532          altogether. */          output altogether, like this:
   
         if (repeat_max == 0)  
           {  
           code = previous;  
           goto END_REPEAT;  
           }  
4533    
4534          /* If the maximum is 1 or unlimited, we just have to stick in the          ** if (repeat_max == 0)
4535          BRAZERO and do no more at this point. However, we do need to adjust          **   {
4536          any OP_RECURSE calls inside the group that refer to the group itself or          **   code = previous;
4537          any internal or forward referenced group, because the offset is from          **   goto END_REPEAT;
4538          the start of the whole regex. Temporarily terminate the pattern while          **   }
4539          doing this. */  
4540            However, that fails when a group is referenced as a subroutine from
4541            elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it
4542            so that it is skipped on execution. As we don't have a list of which
4543            groups are referenced, we cannot do this selectively.
4544    
4545            If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4546            and do no more at this point. However, we do need to adjust any
4547            OP_RECURSE calls inside the group that refer to the group itself or any
4548            internal or forward referenced group, because the offset is from the
4549            start of the whole regex. Temporarily terminate the pattern while doing
4550            this. */
4551    
4552          if (repeat_max <= 1)          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
4553            {            {
4554            *code = OP_END;            *code = OP_END;
4555            adjust_recurse(previous, 1, utf8, cd, save_hwm);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
4556            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
4557            code++;            code++;
4558              if (repeat_max == 0)
4559                {
4560                *previous++ = OP_SKIPZERO;
4561                goto END_REPEAT;
4562                }
4563              brazeroptr = previous;    /* Save for possessive optimizing */
4564            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4565            }            }
4566    
# Line 3888  we set the flag only if there is a liter Line 4585  we set the flag only if there is a liter
4585            /* We chain together the bracket offset fields that have to be            /* We chain together the bracket offset fields that have to be
4586            filled in later when the ends of the brackets are reached. */            filled in later when the ends of the brackets are reached. */
4587    
4588            offset = (bralink == NULL)? 0 : previous - bralink;            offset = (bralink == NULL)? 0 : (int)(previous - bralink);
4589            bralink = previous;            bralink = previous;
4590            PUTINC(previous, 0, offset);            PUTINC(previous, 0, offset);
4591            }            }
# Line 3909  we set the flag only if there is a liter Line 4606  we set the flag only if there is a liter
4606            {            {
4607            /* 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
4608            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
4609            potential integer overflow. */            potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
4610              integer type when available, otherwise double. */
4611    
4612            if (lengthptr != NULL)            if (lengthptr != NULL)
4613              {              {
4614              int delta = (repeat_min - 1)*length_prevgroup;              int delta = (repeat_min - 1)*length_prevgroup;
4615              if ((double)(repeat_min - 1)*(double)length_prevgroup >              if ((INT64_OR_DOUBLE)(repeat_min - 1)*
4616                                                              (double)INT_MAX ||                    (INT64_OR_DOUBLE)length_prevgroup >
4617                        (INT64_OR_DOUBLE)INT_MAX ||
4618                  OFLOW_MAX - *lengthptr < delta)                  OFLOW_MAX - *lengthptr < delta)
4619                {                {
4620                *errorcodeptr = ERR20;                *errorcodeptr = ERR20;
# Line 3961  we set the flag only if there is a liter Line 4660  we set the flag only if there is a liter
4660          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
4661          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
4662          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
4663          paranoid checks to avoid integer overflow. */          paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
4664            a 64-bit integer type when available, otherwise double. */
4665    
4666          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4667            {            {
4668            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4669                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4670            if ((double)repeat_max *            if ((INT64_OR_DOUBLE)repeat_max *
4671                  (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)                  (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4672                    > (double)INT_MAX ||                    > (INT64_OR_DOUBLE)INT_MAX ||
4673                OFLOW_MAX - *lengthptr < delta)                OFLOW_MAX - *lengthptr < delta)
4674              {              {
4675              *errorcodeptr = ERR20;              *errorcodeptr = ERR20;
# Line 3994  we set the flag only if there is a liter Line 4694  we set the flag only if there is a liter
4694              {              {
4695              int offset;              int offset;
4696              *code++ = OP_BRA;              *code++ = OP_BRA;
4697              offset = (bralink == NULL)? 0 : code - bralink;              offset = (bralink == NULL)? 0 : (int)(code - bralink);
4698              bralink = code;              bralink = code;
4699              PUTINC(code, 0, offset);              PUTINC(code, 0, offset);
4700              }              }
# Line 4015  we set the flag only if there is a liter Line 4715  we set the flag only if there is a liter
4715          while (bralink != NULL)          while (bralink != NULL)
4716            {            {
4717            int oldlinkoffset;            int oldlinkoffset;
4718            int offset = code - bralink + 1;            int offset = (int)(code - bralink + 1);
4719            uschar *bra = code - offset;            uschar *bra = code - offset;
4720            oldlinkoffset = GET(bra, 1);            oldlinkoffset = GET(bra, 1);
4721            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;
# Line 4025  we set the flag only if there is a liter Line 4725  we set the flag only if there is a liter
4725            }            }
4726          }          }
4727    
4728        /* If the maximum is unlimited, set a repeater in the final copy. We        /* If the maximum is unlimited, set a repeater in the final copy. For
4729        can't just offset backwards from the current code point, because we        ONCE brackets, that's all we need to do.
4730        don't know if there's been an options resetting after the ket. The  
4731        correct offset was computed above.        Otherwise, if the quantifier was possessive, we convert the BRA code to
4732          the POS form, and the KET code to KETRPOS. (It turns out to be convenient
4733          at runtime to detect this kind of subpattern at both the start and at the
4734          end.) The use of special opcodes makes it possible to reduce greatly the
4735          stack usage in pcre_exec(). If the group is preceded by OP_BRAZERO,
4736          convert this to OP_BRAPOSZERO. Then cancel the possessive flag so that
4737          the default action below, of wrapping everything inside atomic brackets,
4738          does not happen.
4739    
4740        Then, when we are doing the actual compile phase, check to see whether        Then, when we are doing the actual compile phase, check to see whether
4741        this group is a non-atomic one that could match an empty string. If so,        this group is one that could match an empty string. If so, convert the
4742        convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so        initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so that runtime
4743        that runtime checking can be done. [This check is also applied to        checking can be done. [This check is also applied to ONCE groups at
4744        atomic groups at runtime, but in a different way.] */        runtime, but in a different way.] */
4745    
4746        else        else
4747          {          {
4748          uschar *ketcode = code - ketoffset;          uschar *ketcode = code - 1 - LINK_SIZE;
4749          uschar *bracode = ketcode - GET(ketcode, 1);          uschar *bracode = ketcode - GET(ketcode, 1);
4750          *ketcode = OP_KETRMAX + repeat_type;  
4751          if (lengthptr == NULL && *bracode != OP_ONCE)          if (*bracode == OP_ONCE)
4752              *ketcode = OP_KETRMAX + repeat_type;
4753            else
4754            {            {
4755            uschar *scode = bracode;            if (possessive_quantifier)
4756            do              {
4757                *bracode += 1;                   /* Switch to xxxPOS opcodes */
4758                *ketcode = OP_KETRPOS;
4759                if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
4760                possessive_quantifier = FALSE;
4761                }
4762              else *ketcode = OP_KETRMAX + repeat_type;
4763    
4764              if (lengthptr == NULL)
4765              {              {
4766              if (could_be_empty_branch(scode, ketcode, utf8))              uschar *scode = bracode;
4767                do
4768                {                {
4769                *bracode += OP_SBRA - OP_BRA;                if (could_be_empty_branch(scode, ketcode, utf8, cd))
4770                break;                  {
4771                    *bracode += OP_SBRA - OP_BRA;
4772                    break;
4773                    }
4774                  scode += GET(scode, 1);
4775                }                }
4776              scode += GET(scode, 1);              while (*scode == OP_ALT);
4777              }              }
           while (*scode == OP_ALT);  
4778            }            }
4779          }          }
4780        }        }
4781    
4782        /* If previous is OP_FAIL, it was generated by an empty class [] in
4783        JavaScript mode. The other ways in which OP_FAIL can be generated, that is
4784        by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
4785        error above. We can just ignore the repeat in JS case. */
4786    
4787        else if (*previous == OP_FAIL) goto END_REPEAT;
4788    
4789      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
4790    
4791      else      else
# Line 4067  we set the flag only if there is a liter Line 4795  we set the flag only if there is a liter
4795        }        }
4796    
4797      /* If the character following a repeat is '+', or if certain optimization      /* If the character following a repeat is '+', or if certain optimization
4798      tests above succeeded, possessive_quantifier is TRUE. For some of the      tests above succeeded, possessive_quantifier is TRUE. For some opcodes,
4799      simpler opcodes, there is an special alternative opcode for this. For      there are special alternative opcodes for this case. For anything else, we
4800      anything else, we wrap the entire repeated item inside OP_ONCE brackets.      wrap the entire repeated item inside OP_ONCE brackets. Logically, the '+'
4801      The '+' notation is just syntactic sugar, taken from Sun's Java package,      notation is just syntactic sugar, taken from Sun's Java package, but the
4802      but the special opcodes can optimize it a bit. The repeated item starts at      special opcodes can optimize it.
4803      tempcode, not at previous, which might be the first part of a string whose  
4804      (former) last char we repeated.      Possessively repeated subpatterns have already been handled in the code
4805        just above, so possessive_quantifier is always FALSE for them at this
4806        stage.
4807    
4808        Note that the repeated item starts at tempcode, not at previous, which
4809        might be the first part of a string whose (former) last char we repeated.
4810    
4811      Possessifying an 'exact' quantifier has no effect, so we can ignore it. But      Possessifying an 'exact' quantifier has no effect, so we can ignore it. But
4812      an 'upto' may follow. We skip over an 'exact' item, and then test the      an 'upto' may follow. We skip over an 'exact' item, and then test the
# Line 4082  we set the flag only if there is a liter Line 4815  we set the flag only if there is a liter
4815      if (possessive_quantifier)      if (possessive_quantifier)
4816        {        {
4817        int len;        int len;
4818        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||  
4819            *tempcode == OP_NOTEXACT)        if (*tempcode == OP_TYPEEXACT)
4820          tempcode += _pcre_OP_lengths[*tempcode] +          tempcode += _pcre_OP_lengths[*tempcode] +
4821            ((*tempcode == OP_TYPEEXACT &&            ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
4822               (tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP))? 2:0);  
4823        len = code - tempcode;        else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
4824            {
4825            tempcode += _pcre_OP_lengths[*tempcode];
4826    #ifdef SUPPORT_UTF8
4827            if (utf8 && tempcode[-1] >= 0xc0)
4828              tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];
4829    #endif
4830            }
4831    
4832          len = (int)(code - tempcode);
4833        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4834          {          {
4835          case OP_STAR:  *tempcode = OP_POSSTAR; break;          case OP_STAR:  *tempcode = OP_POSSTAR; break;
# Line 4095  we set the flag only if there is a liter Line 4837  we set the flag only if there is a liter
4837          case OP_QUERY: *tempcode = OP_POSQUERY; break;          case OP_QUERY: *tempcode = OP_POSQUERY; break;
4838          case OP_UPTO:  *tempcode = OP_POSUPTO; break;          case OP_UPTO:  *tempcode = OP_POSUPTO; break;
4839    
4840          case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;          case OP_STARI:  *tempcode = OP_POSSTARI; break;
4841          case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;          case OP_PLUSI:  *tempcode = OP_POSPLUSI; break;
4842          case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;          case OP_QUERYI: *tempcode = OP_POSQUERYI; break;
4843          case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;          case OP_UPTOI:  *tempcode = OP_POSUPTOI; break;
4844    
4845          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;
4846          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;
4847          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4848          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4849    
4850            case OP_NOTSTARI:  *tempcode = OP_NOTPOSSTARI; break;
4851            case OP_NOTPLUSI:  *tempcode = OP_NOTPOSPLUSI; break;
4852            case OP_NOTQUERYI: *tempcode = OP_NOTPOSQUERYI; break;
4853            case OP_NOTUPTOI:  *tempcode = OP_NOTPOSUPTOI; break;
4854    
4855            case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;
4856            case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;
4857            case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
4858            case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;
4859    
4860            /* Because we are moving code along, we must ensure that any
4861            pending recursive references are updated. */
4862    
4863          default:          default:
4864            *code = OP_END;
4865            adjust_recurse(tempcode, 1 + LINK_SIZE, utf8, cd, save_hwm);
4866          memmove(tempcode + 1+LINK_SIZE, tempcode, len);          memmove(tempcode + 1+LINK_SIZE, tempcode, len);
4867          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
4868          len += 1 + LINK_SIZE;          len += 1 + LINK_SIZE;
# Line 4132  we set the flag only if there is a liter Line 4889  we set the flag only if there is a liter
4889      lookbehind or option setting or condition or all the other extended      lookbehind or option setting or condition or all the other extended
4890      parenthesis forms.  */      parenthesis forms.  */
4891    
4892      case '(':      case CHAR_LEFT_PARENTHESIS:
4893      newoptions = options;      newoptions = options;
4894      skipbytes = 0;      skipbytes = 0;
4895      bravalue = OP_CBRA;      bravalue = OP_CBRA;
# Line 4141  we set the flag only if there is a liter Line 4898  we set the flag only if there is a liter
4898    
4899      /* First deal with various "verbs" that can be introduced by '*'. */      /* First deal with various "verbs" that can be introduced by '*'. */
4900    
4901      if (*(++ptr) == '*' && (cd->ctypes[ptr[1]] & ctype_letter) != 0)      if (*(++ptr) == CHAR_ASTERISK &&
4902             ((cd->ctypes[ptr[1]] & ctype_letter) != 0 || ptr[1] == ':'))
4903        {        {
4904        int i, namelen;        int i, namelen;
4905          int arglen = 0;
4906        const char *vn = verbnames;        const char *vn = verbnames;
4907        const uschar *name = ++ptr;        const uschar *name = ptr + 1;
4908          const uschar *arg = NULL;
4909        previous = NULL;        previous = NULL;
4910        while ((cd->ctypes[*++ptr] & ctype_letter) != 0);        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
4911        if (*ptr == ':')        namelen = (int)(ptr - name);
4912    
4913          if (*ptr == CHAR_COLON)
4914          {          {
4915          *errorcodeptr = ERR59;   /* Not supported */          arg = ++ptr;
4916          goto FAILED;          while ((cd->ctypes[*ptr] & (ctype_letter|ctype_digit)) != 0
4917              || *ptr == '_') ptr++;
4918            arglen = (int)(ptr - arg);
4919          }          }
4920        if (*ptr != ')')  
4921          if (*ptr != CHAR_RIGHT_PARENTHESIS)
4922          {          {
4923          *errorcodeptr = ERR60;          *errorcodeptr = ERR60;
4924          goto FAILED;          goto FAILED;
4925          }          }
4926        namelen = ptr - name;  
4927          /* Scan the table of verb names */
4928    
4929        for (i = 0; i < verbcount; i++)        for (i = 0; i < verbcount; i++)
4930          {          {
4931          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
4932              strncmp((char *)name, vn, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
4933            {            {
4934            *code = verbs[i].op;            /* Check for open captures before ACCEPT */
4935            if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;  
4936            break;            if (verbs[i].op == OP_ACCEPT)
4937                {
4938                open_capitem *oc;
4939                cd->had_accept = TRUE;
4940                for (oc = cd->open_caps; oc != NULL; oc = oc->next)
4941                  {
4942                  *code++ = OP_CLOSE;
4943                  PUT2INC(code, 0, oc->number);
4944                  }
4945                }
4946    
4947              /* Handle the cases with/without an argument */
4948    
4949              if (arglen == 0)
4950                {
4951                if (verbs[i].op < 0)   /* Argument is mandatory */
4952                  {
4953                  *errorcodeptr = ERR66;
4954                  goto FAILED;
4955                  }
4956                *code = verbs[i].op;
4957                if (*code++ == OP_THEN)
4958                  {
4959                  PUT(code, 0, code - bcptr->current_branch - 1);
4960                  code += LINK_SIZE;
4961                  }
4962                }
4963    
4964              else
4965                {
4966                if (verbs[i].op_arg < 0)   /* Argument is forbidden */
4967                  {
4968                  *errorcodeptr = ERR59;
4969                  goto FAILED;
4970                  }
4971                *code = verbs[i].op_arg;
4972                if (*code++ == OP_THEN_ARG)
4973                  {
4974                  PUT(code, 0, code - bcptr->current_branch - 1);
4975                  code += LINK_SIZE;
4976                  }
4977                *code++ = arglen;
4978                memcpy(code, arg, arglen);
4979                code += arglen;
4980                *code++ = 0;
4981                }
4982    
4983              break;  /* Found verb, exit loop */
4984            }            }
4985    
4986          vn += verbs[i].len + 1;          vn += verbs[i].len + 1;
4987          }          }
4988        if (i < verbcount) continue;  
4989        *errorcodeptr = ERR60;        if (i < verbcount) continue;    /* Successfully handled a verb */
4990          *errorcodeptr = ERR60;          /* Verb not recognized */
4991        goto FAILED;        goto FAILED;
4992        }        }
4993    
4994      /* Deal with the extended parentheses; all are introduced by '?', and the      /* Deal with the extended parentheses; all are introduced by '?', and the
4995      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. */
4996    
4997      else if (*ptr == '?')      else if (*ptr == CHAR_QUESTION_MARK)
4998        {        {
4999        int i, set, unset, namelen;        int i, set, unset, namelen;
5000        int *optset;        int *optset;
# Line 4187  we set the flag only if there is a liter Line 5003  we set the flag only if there is a liter
5003    
5004        switch (*(++ptr))        switch (*(++ptr))
5005          {          {
5006          case '#':                 /* Comment; skip to ket */          case CHAR_NUMBER_SIGN:                 /* Comment; skip to ket */
5007          ptr++;          ptr++;
5008          while (*ptr != 0 && *ptr != ')') ptr++;          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
5009          if (*ptr == 0)          if (*ptr == 0)
5010            {            {
5011            *errorcodeptr = ERR18;            *errorcodeptr = ERR18;
# Line 4199  we set the flag only if there is a liter Line 5015  we set the flag only if there is a liter
5015    
5016    
5017          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5018          case '|':                 /* Reset capture count for each branch */          case CHAR_VERTICAL_LINE:  /* Reset capture count for each branch */
5019          reset_bracount = TRUE;          reset_bracount = TRUE;
5020          /* Fall through */          /* Fall through */
5021    
5022          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5023          case ':':                 /* Non-capturing bracket */          case CHAR_COLON:          /* Non-capturing bracket */
5024          bravalue = OP_BRA;          bravalue = OP_BRA;
5025          ptr++;          ptr++;
5026          break;          break;
5027    
5028    
5029          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5030          case '(':          case CHAR_LEFT_PARENTHESIS:
5031          bravalue = OP_COND;       /* Conditional group */          bravalue = OP_COND;       /* Conditional group */
5032    
5033          /* 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 4231  we set the flag only if there is a liter Line 5047  we set the flag only if there is a liter
5047          the switch. This will take control down to where bracketed groups,          the switch. This will take control down to where bracketed groups,
5048          including assertions, are processed. */          including assertions, are processed. */
5049    
5050          if (ptr[1] == '?' && (ptr[2] == '=' || ptr[2] == '!' || ptr[2] == '<'))          if (ptr[1] == CHAR_QUESTION_MARK && (ptr[2] == CHAR_EQUALS_SIGN ||
5051                ptr[2] == CHAR_EXCLAMATION_MARK || ptr[2] == CHAR_LESS_THAN_SIGN))
5052            break;            break;
5053    
5054          /* 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 4243  we set the flag only if there is a liter Line 5060  we set the flag only if there is a liter
5060    
5061          /* Check for a test for recursion in a named group. */          /* Check for a test for recursion in a named group. */
5062    
5063          if (ptr[1] == 'R' && ptr[2] == '&')          if (ptr[1] == CHAR_R && ptr[2] == CHAR_AMPERSAND)
5064            {            {
5065            terminator = -1;            terminator = -1;
5066            ptr += 2;            ptr += 2;
# Line 4253  we set the flag only if there is a liter Line 5070  we set the flag only if there is a liter
5070          /* 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
5071          syntax (?(<name>) or (?('name') */          syntax (?(<name>) or (?('name') */
5072    
5073          else if (ptr[1] == '<')          else if (ptr[1] == CHAR_LESS_THAN_SIGN)
5074            {            {
5075            terminator = '>';            terminator = CHAR_GREATER_THAN_SIGN;
5076            ptr++;            ptr++;
5077            }            }
5078          else if (ptr[1] == '\'')          else if (ptr[1] == CHAR_APOSTROPHE)
5079            {            {
5080            terminator = '\'';            terminator = CHAR_APOSTROPHE;
5081            ptr++;            ptr++;
5082            }            }
5083          else          else
5084            {            {
5085            terminator = 0;            terminator = 0;
5086            if (ptr[1] == '-' || ptr[1] == '+') refsign = *(++ptr);            if (ptr[1] == CHAR_MINUS || ptr[1] == CHAR_PLUS) refsign = *(++ptr);
5087            }            }
5088    
5089          /* 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 4286  we set the flag only if there is a liter Line 5103  we set the flag only if there is a liter
5103            {            {
5104            if (recno >= 0)            if (recno >= 0)
5105              recno = ((digitab[*ptr] & ctype_digit) != 0)?              recno = ((digitab[*ptr] & ctype_digit) != 0)?
5106                recno * 10 + *ptr - '0' : -1;                recno * 10 + *ptr - CHAR_0 : -1;
5107            ptr++;            ptr++;
5108            }            }
5109          namelen = ptr - name;          namelen = (int)(ptr - name);
5110    
5111          if ((terminator > 0 && *ptr++ != terminator) || *ptr++ != ')')          if ((terminator > 0 && *ptr++ != terminator) ||
5112                *ptr++ != CHAR_RIGHT_PARENTHESIS)
5113            {            {
5114            ptr--;      /* Error offset */            ptr--;      /* Error offset */
5115            *errorcodeptr = ERR26;            *errorcodeptr = ERR26;
# Line 4313  we set the flag only if there is a liter Line 5131  we set the flag only if there is a liter
5131              *errorcodeptr = ERR58;              *errorcodeptr = ERR58;
5132              goto FAILED;              goto FAILED;
5133              }              }
5134            recno = (refsign == '-')?            recno = (refsign == CHAR_MINUS)?
5135              cd->bracount - recno + 1 : recno +cd->bracount;              cd->bracount - recno + 1 : recno +cd->bracount;
5136            if (recno <= 0 || recno > cd->final_bracount)            if (recno <= 0 || recno > cd->final_bracount)
5137              {              {
# Line 4325  we set the flag only if there is a liter Line 5143  we set the flag only if there is a liter
5143            }            }
5144    
5145          /* Otherwise (did not start with "+" or "-"), start by looking for the          /* Otherwise (did not start with "+" or "-"), start by looking for the
5146          name. */          name. If we find a name, add one to the opcode to change OP_CREF or
5147            OP_RREF into OP_NCREF or OP_NRREF. These behave exactly the same,
5148            except they record that the reference was originally to a name. The
5149            information is used to check duplicate names. */
5150    
5151          slot = cd->name_table;          slot = cd->name_table;
5152          for (i = 0; i < cd->names_found; i++)          for (i = 0; i < cd->names_found; i++)
# Line 4340  we set the flag only if there is a liter Line 5161  we set the flag only if there is a liter
5161            {            {
5162            recno = GET2(slot, 0);            recno = GET2(slot, 0);
5163            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
5164              code[1+LINK_SIZE]++;
5165            }            }
5166    
5167          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
5168    
5169          else if ((i = find_parens(ptr, cd->bracount, name, namelen,          else if ((i = find_parens(cd, name, namelen,
5170                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0, utf8)) > 0)
5171            {            {
5172            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
5173              code[1+LINK_SIZE]++;
5174            }            }
5175    
5176          /* If terminator == 0 it means that the name followed directly after          /* If terminator == 0 it means that the name followed directly after
# Line 4365  we set the flag only if there is a liter Line 5188  we set the flag only if there is a liter
5188          /* Check for (?(R) for recursion. Allow digits after R to specify a          /* Check for (?(R) for recursion. Allow digits after R to specify a
5189          specific group number. */          specific group number. */
5190    
5191          else if (*name == 'R')          else if (*name == CHAR_R)
5192            {            {
5193            recno = 0;            recno = 0;
5194            for (i = 1; i < namelen; i++)            for (i = 1; i < namelen; i++)
# Line 4375  we set the flag only if there is a liter Line 5198  we set the flag only if there is a liter
5198                *errorcodeptr = ERR15;                *errorcodeptr = ERR15;
5199                goto FAILED;                goto FAILED;
5200                }                }
5201              recno = recno * 10 + name[i] - '0';              recno = recno * 10 + name[i] - CHAR_0;
5202              }              }
5203            if (recno == 0) recno = RREF_ANY;            if (recno == 0) recno = RREF_ANY;
5204            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */
# Line 4385  we set the flag only if there is a liter Line 5208  we set the flag only if there is a liter
5208          /* Similarly, check for the (?(DEFINE) "condition", which is always          /* Similarly, check for the (?(DEFINE) "condition", which is always
5209          false. */          false. */
5210    
5211          else if (namelen == 6 && strncmp((char *)name, "DEFINE", 6) == 0)          else if (namelen == 6 && strncmp((char *)name, STRING_DEFINE, 6) == 0)
5212            {            {
5213            code[1+LINK_SIZE] = OP_DEF;            code[1+LINK_SIZE] = OP_DEF;
5214            skipbytes = 1;            skipbytes = 1;
# Line 4410  we set the flag only if there is a liter Line 5233  we set the flag only if there is a liter
5233    
5234    
5235          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5236          case '=':                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
5237          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
5238          ptr++;          ptr++;
5239          break;          break;
5240    
5241    
5242          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5243          case '!':                 /* Negative lookahead */          case CHAR_EXCLAMATION_MARK:            /* Negative lookahead */
5244          ptr++;          ptr++;
5245          if (*ptr == ')')          /* Optimize (?!) */          if (*ptr == CHAR_RIGHT_PARENTHESIS)    /* Optimize (?!) */
5246            {            {
5247            *code++ = OP_FAIL;            *code++ = OP_FAIL;
5248            previous = NULL;            previous = NULL;
# Line 4430  we set the flag only if there is a liter Line 5253  we set the flag only if there is a liter
5253    
5254    
5255          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5256          case '<':                 /* Lookbehind or named define */          case CHAR_LESS_THAN_SIGN:              /* Lookbehind or named define */
5257          switch (ptr[1])          switch (ptr[1])
5258            {            {
5259            case '=':               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
5260            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
5261            ptr += 2;            ptr += 2;
5262            break;            break;
5263    
5264            case '!':               /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
5265            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
5266            ptr += 2;            ptr += 2;
5267            break;            break;
# Line 4453  we set the flag only if there is a liter Line 5276  we set the flag only if there is a liter
5276    
5277    
5278          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5279          case '>':                 /* One-time brackets */          case CHAR_GREATER_THAN_SIGN:           /* One-time brackets */
5280          bravalue = OP_ONCE;          bravalue = OP_ONCE;
5281          ptr++;          ptr++;
5282          break;          break;
5283    
5284    
5285          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5286          case 'C':                 /* Callout - may be followed by digits; */          case CHAR_C:                 /* Callout - may be followed by digits; */
5287          previous_callout = code;  /* Save for later completion */          previous_callout = code;  /* Save for later completion */
5288          after_manual_callout = 1; /* Skip one item before completing */          after_manual_callout = 1; /* Skip one item before completing */
5289          *code++ = OP_CALLOUT;          *code++ = OP_CALLOUT;
5290            {            {
5291            int n = 0;            int n = 0;
5292            while ((digitab[*(++ptr)] & ctype_digit) != 0)            while ((digitab[*(++ptr)] & ctype_digit) != 0)
5293              n = n * 10 + *ptr - '0';              n = n * 10 + *ptr - CHAR_0;
5294            if (*ptr != ')')            if (*ptr != CHAR_RIGHT_PARENTHESIS)
5295              {              {
5296              *errorcodeptr = ERR39;              *errorcodeptr = ERR39;
5297              goto FAILED;              goto FAILED;
# Line 4479  we set the flag only if there is a liter Line 5302  we set the flag only if there is a liter
5302              goto FAILED;              goto FAILED;
5303              }              }
5304            *code++ = n;            *code++ = n;
5305            PUT(code, 0, ptr - cd->start_pattern + 1);  /* Pattern offset */            PUT(code, 0, (int)(ptr - cd->start_pattern + 1)); /* Pattern offset */
5306            PUT(code, LINK_SIZE, 0);                    /* Default length */            PUT(code, LINK_SIZE, 0);                          /* Default length */
5307            code += 2 * LINK_SIZE;            code += 2 * LINK_SIZE;
5308            }            }
5309          previous = NULL;          previous = NULL;
# Line 4488  we set the flag only if there is a liter Line 5311  we set the flag only if there is a liter
5311    
5312    
5313          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5314          case 'P':                 /* Python-style named subpattern handling */          case CHAR_P:              /* Python-style named subpattern handling */
5315          if (*(++ptr) == '=' || *ptr == '>')  /* Reference or recursion */          if (*(++ptr) == CHAR_EQUALS_SIGN ||
5316                *ptr == CHAR_GREATER_THAN_SIGN)  /* Reference or recursion */
5317            {            {
5318            is_recurse = *ptr == '>';            is_recurse = *ptr == CHAR_GREATER_THAN_SIGN;
5319            terminator = ')';            terminator = CHAR_RIGHT_PARENTHESIS;
5320            goto NAMED_REF_OR_RECURSE;            goto NAMED_REF_OR_RECURSE;
5321            }            }
5322          else if (*ptr != '<')    /* Test for Python-style definition */          else if (*ptr != CHAR_LESS_THAN_SIGN)  /* Test for Python-style defn */
5323            {            {
5324            *errorcodeptr = ERR41;            *errorcodeptr = ERR41;
5325            goto FAILED;            goto FAILED;
# Line 4505  we set the flag only if there is a liter Line 5329  we set the flag only if there is a liter
5329    
5330          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5331          DEFINE_NAME:    /* Come here from (?< handling */          DEFINE_NAME:    /* Come here from (?< handling */
5332          case '\'':          case CHAR_APOSTROPHE:
5333            {            {
5334            terminator = (*ptr == '<')? '>' : '\'';            terminator = (*ptr == CHAR_LESS_THAN_SIGN)?
5335                CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
5336            name = ++ptr;            name = ++ptr;
5337    
5338            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
5339            namelen = ptr - name;            namelen = (int)(ptr - name);
5340    
5341            /* In the pre-compile phase, just do a syntax check. */            /* In the pre-compile phase, just do a syntax check. */
5342    
# Line 4538  we set the flag only if there is a liter Line 5363  we set the flag only if there is a liter
5363                }                }
5364              }              }
5365    
5366            /* In the real compile, create the entry in the table */            /* In the real compile, create the entry in the table, maintaining
5367              alphabetical order. Duplicate names for different numbers are
5368              permitted only if PCRE_DUPNAMES is set. Duplicate names for the same
5369              number are always OK. (An existing number can be re-used if (?|
5370              appears in the pattern.) In either event, a duplicate name results in
5371              a duplicate entry in the table, even if the number is the same. This
5372              is because the number of names, and hence the table size, is computed
5373              in the pre-compile, and it affects various numbers and pointers which
5374              would all have to be modified, and the compiled code moved down, if
5375              duplicates with the same number were omitted from the table. This
5376              doesn't seem worth the hassle. However, *different* names for the
5377              same number are not permitted. */
5378    
5379            else            else
5380              {              {
5381                BOOL dupname = FALSE;
5382              slot = cd->name_table;              slot = cd->name_table;
5383    
5384              for (i = 0; i < cd->names_found; i++)              for (i = 0; i < cd->names_found; i++)
5385                {                {
5386                int crc = memcmp(name, slot+2, namelen);                int crc = memcmp(name, slot+2, namelen);
# Line 4550  we set the flag only if there is a liter Line 5388  we set the flag only if there is a liter
5388                  {                  {
5389                  if (slot[2+namelen] == 0)                  if (slot[2+namelen] == 0)
5390                    {                    {
5391                    if ((options & PCRE_DUPNAMES) == 0)                    if (GET2(slot, 0) != cd->bracount + 1 &&
5392                          (options & PCRE_DUPNAMES) == 0)
5393                      {                      {
5394                      *errorcodeptr = ERR43;                      *errorcodeptr = ERR43;
5395                      goto FAILED;                      goto FAILED;
5396                      }                      }
5397                      else dupname = TRUE;
5398                    }                    }
5399                  else crc = -1;      /* Current name is substring */                  else crc = -1;      /* Current name is a substring */
5400                  }                  }
5401    
5402                  /* Make space in the table and break the loop for an earlier
5403                  name. For a duplicate or later name, carry on. We do this for
5404                  duplicates so that in the simple case (when ?(| is not used) they
5405                  are in order of their numbers. */
5406    
5407                if (crc < 0)                if (crc < 0)
5408                  {                  {
5409                  memmove(slot + cd->name_entry_size, slot,                  memmove(slot + cd->name_entry_size, slot,
5410                    (cd->names_found - i) * cd->name_entry_size);                    (cd->names_found - i) * cd->name_entry_size);
5411                  break;                  break;
5412                  }                  }
5413    
5414                  /* Continue the loop for a later or duplicate name */
5415    
5416                slot += cd->name_entry_size;                slot += cd->name_entry_size;
5417                }                }
5418    
5419                /* For non-duplicate names, check for a duplicate number before
5420                adding the new name. */
5421    
5422                if (!dupname)