/[pcre]/code/branches/pcre16/pcre_compile.c
ViewVC logotype

Diff of /code/branches/pcre16/pcre_compile.c

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

revision 309 by ph10, Mon Jan 21 15:22:09 2008 UTC revision 598 by ph10, Sat May 7 15:37:31 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  static verbitem verbs[] = {    STRING_THEN;
206    { 6, OP_ACCEPT },  
207    { 6, OP_COMMIT },  static const verbitem verbs[] = {
208    { 1, OP_FAIL },    { 0, -1,        OP_MARK },
209    { 4, OP_FAIL },    { 4, -1,        OP_MARK },
210    { 5, OP_PRUNE },    { 6, OP_ACCEPT, -1 },
211    { 4, OP_SKIP  },    { 6, OP_COMMIT, -1 },
212    { 4, OP_THEN  }    { 1, OP_FAIL,   -1 },
213      { 4, OP_FAIL,   -1 },
214      { 5, OP_PRUNE,  OP_PRUNE_ARG },
215      { 4, OP_SKIP,   OP_SKIP_ARG  },
216      { 4, OP_THEN,   OP_THEN_ARG  }
217  };  };
218    
219  static int verbcount = sizeof(verbs)/sizeof(verbitem);  static const int verbcount = sizeof(verbs)/sizeof(verbitem);
220    
221    
222  /* Tables of names of POSIX character classes and their lengths. The names are  /* Tables of names of POSIX character classes and their lengths. The names are
# Line 178  length entry. The first three must be al Line 226  length entry. The first three must be al
226  for handling case independence. */  for handling case independence. */
227    
228  static const char posix_names[] =  static const char posix_names[] =
229    "alpha\0"  "lower\0"  "upper\0"  "alnum\0"  "ascii\0"  "blank\0"    STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
230    "cntrl\0"  "digit\0"  "graph\0"  "print\0"  "punct\0"  "space\0"    STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
231    "word\0"   "xdigit";    STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
232      STRING_word0  STRING_xdigit;
233    
234  static const uschar posix_name_lengths[] = {  static const uschar posix_name_lengths[] = {
235    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
# Line 212  static const int posix_class_maps[] = { Line 261  static const int posix_class_maps[] = {
261    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
262  };  };
263    
264    /* Table of substitutes for \d etc when PCRE_UCP is set. The POSIX class
265    substitutes must be in the order of the names, defined above, and there are
266    both positive and negative cases. NULL means no substitute. */
267    
268    #ifdef SUPPORT_UCP
269    static const uschar *substitutes[] = {
270      (uschar *)"\\P{Nd}",    /* \D */
271      (uschar *)"\\p{Nd}",    /* \d */
272      (uschar *)"\\P{Xsp}",   /* \S */       /* NOTE: Xsp is Perl space */
273      (uschar *)"\\p{Xsp}",   /* \s */
274      (uschar *)"\\P{Xwd}",   /* \W */
275      (uschar *)"\\p{Xwd}"    /* \w */
276    };
277    
278    static const uschar *posix_substitutes[] = {
279      (uschar *)"\\p{L}",     /* alpha */
280      (uschar *)"\\p{Ll}",    /* lower */
281      (uschar *)"\\p{Lu}",    /* upper */
282      (uschar *)"\\p{Xan}",   /* alnum */
283      NULL,                   /* ascii */
284      (uschar *)"\\h",        /* blank */
285      NULL,                   /* cntrl */
286      (uschar *)"\\p{Nd}",    /* digit */
287      NULL,                   /* graph */
288      NULL,                   /* print */
289      NULL,                   /* punct */
290      (uschar *)"\\p{Xps}",   /* space */    /* NOTE: Xps is POSIX space */
291      (uschar *)"\\p{Xwd}",   /* word */
292      NULL,                   /* xdigit */
293      /* Negated cases */
294      (uschar *)"\\P{L}",     /* ^alpha */
295      (uschar *)"\\P{Ll}",    /* ^lower */
296      (uschar *)"\\P{Lu}",    /* ^upper */
297      (uschar *)"\\P{Xan}",   /* ^alnum */
298      NULL,                   /* ^ascii */
299      (uschar *)"\\H",        /* ^blank */
300      NULL,                   /* ^cntrl */
301      (uschar *)"\\P{Nd}",    /* ^digit */
302      NULL,                   /* ^graph */
303      NULL,                   /* ^print */
304      NULL,                   /* ^punct */
305      (uschar *)"\\P{Xps}",   /* ^space */   /* NOTE: Xps is POSIX space */
306      (uschar *)"\\P{Xwd}",   /* ^word */
307      NULL                    /* ^xdigit */
308    };
309    #define POSIX_SUBSIZE (sizeof(posix_substitutes)/sizeof(uschar *))
310    #endif
311    
312  #define STRING(a)  # a  #define STRING(a)  # a
313  #define XSTRING(s) STRING(s)  #define XSTRING(s) STRING(s)
# Line 224  the number of relocations needed when a Line 320  the number of relocations needed when a
320  it is now one long string. We cannot use a table of offsets, because the  it is now one long string. We cannot use a table of offsets, because the
321  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
322  simply count through to the one we want - this isn't a performance issue  simply count through to the one we want - this isn't a performance issue
323  because these strings are used only when there is a compilation error. */  because these strings are used only when there is a compilation error.
324    
325    Each substring ends with \0 to insert a null character. This includes the final
326    substring, so that the whole string ends with \0\0, which can be detected when
327    counting through. */
328    
329  static const char error_texts[] =  static const char error_texts[] =
330    "no error\0"    "no error\0"
# Line 271  static const char error_texts[] = Line 371  static const char error_texts[] =
371    /* 35 */    /* 35 */
372    "invalid condition (?(0)\0"    "invalid condition (?(0)\0"
373    "\\C not allowed in lookbehind assertion\0"    "\\C not allowed in lookbehind assertion\0"
374    "PCRE does not support \\L, \\l, \\N, \\U, or \\u\0"    "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
375    "number after (?C is > 255\0"    "number after (?C is > 255\0"
376    "closing ) for (?C expected\0"    "closing ) for (?C expected\0"
377    /* 40 */    /* 40 */
# Line 295  static const char error_texts[] = Line 395  static const char error_texts[] =
395    /* 55 */    /* 55 */
396    "repeating a DEFINE group is not allowed\0"    "repeating a DEFINE group is not allowed\0"
397    "inconsistent NEWLINE options\0"    "inconsistent NEWLINE options\0"
398    "\\g is not followed by a braced name or an optionally braced non-zero number\0"    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
399    "(?+ or (?- or (?(+ or (?(- must be followed by a non-zero number\0"    "a numbered reference must not be zero\0"
400    "(*VERB) with an argument is not supported\0"    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
401    /* 60 */    /* 60 */
402    "(*VERB) not recognized\0"    "(*VERB) not recognized\0"
403    "number is too big\0"    "number is too big\0"
404    "subpattern name expected\0"    "subpattern name expected\0"
405    "digit expected after (?+";    "digit expected after (?+\0"
406      "] is an invalid data character in JavaScript compatibility mode\0"
407      /* 65 */
408      "different names for subpatterns of the same number are not allowed\0"
409      "(*MARK) must have an argument\0"
410      "this version of PCRE is not compiled with PCRE_UCP support\0"
411      "\\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 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 a number, either plain or braced. If positive, it      /* \g must be followed by one of a number of specific things:
651      is an absolute backreference. If negative, it is a relative backreference.  
652      This is a Perl 5.10 feature. Perl 5.10 also supports \g{name} as a      (1) A number, either plain or braced. If positive, it is an absolute
653      reference to a named group. This is part of Perl's movement towards a      backreference. If negative, it is a relative backreference. This is a Perl
654      unified syntax for back references. As this is synonymous with \k{name}, we      5.10 feature.
655      fudge it up by pretending it really was \k. */  
656        (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
658        this is synonymous with \k{name}, we fudge it up by pretending it really
659        was \k.
660    
661        (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
663        (possibly recursive) subroutine calls, _not_ backreferences. Just return
664        the -ESC_g code (cf \k). */
665    
666        case CHAR_g:
667        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
668          {
669          c = -ESC_g;
670          break;
671          }
672    
673        /* Handle the Perl-compatible cases */
674    
675      case 'g':      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
     if (ptr[1] == '{')  
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 554  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 563  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)      if (c < 0)   /* Integer overflow */
702        {        {
703        *errorcodeptr = ERR61;        *errorcodeptr = ERR61;
704        break;        break;
705        }        }
706    
707      if (c == 0 || (braced && *(++ptr) != '}'))      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
708        {        {
709        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
710        break;        break;
711        }        }
712    
713        if (c == 0)
714          {
715          *errorcodeptr = ERR58;
716          break;
717          }
718    
719      if (negated)      if (negated)
720        {        {
721        if (c > bracount)        if (c > bracount)
# Line 602  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)        if (c < 0)    /* Integer overflow */
754          {          {
755          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
756          break;          break;
# Line 628  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 641  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 652  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 662  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 690  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 740  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 780  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 791  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 859  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 900  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 910  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 940  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 1008  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      else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1324        {
1325        if (dup_parens && *count < hwm_count) *count = hwm_count;
1326        goto FAIL_EXIT;
1327        }
1328    
1329      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    ptr += 2;  FAIL_EXIT:
1337    if (*ptr == 'P') ptr++;                      /* Allow optional P */  *ptrptr = ptr;
1338    return -1;
1339    }
1340    
   /* We have to disambiguate (?<! and (?<= from (?<name> */  
1341    
   if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  
        *ptr != '\'')  
     continue;  
1342    
   count++;  
1343    
1344    if (name == NULL && count == lorn) return count;  /*************************************************
1345    term = *ptr++;  *       Find forward referenced subpattern       *
1346    if (term == '<') term = '>';  *************************************************/
1347    thisname = ptr;  
1348    while (*ptr != term) ptr++;  /* This function scans along a pattern's text looking for capturing
1349    if (name != NULL && lorn == ptr - thisname &&  subpatterns, and counting them. If it finds a named pattern that matches the
1350        strncmp((const char *)name, (const char *)thisname, lorn) == 0)  name it is given, it returns its number. Alternatively, if the name is NULL, it
1351      return count;  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    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    for (;;)
1383      {
1384      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, &count);
1385      if (rc > 0 || *ptr++ == 0) break;
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  *************************************************/  *************************************************/
# Line 1100  for (;;) Line 1440  for (;;)
1440    
1441      case OP_CALLOUT:      case OP_CALLOUT:
1442      case OP_CREF:      case OP_CREF:
1443        case OP_NCREF:
1444      case OP_RREF:      case OP_RREF:
1445        case OP_NRREF:
1446      case OP_DEF:      case OP_DEF:
1447      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1448      break;      break;
# Line 1116  for (;;) Line 1458  for (;;)
1458    
1459    
1460  /*************************************************  /*************************************************
1461  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1462  *************************************************/  *************************************************/
1463    
1464  /* 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,
1465  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.
1466  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
1467    temporarily terminated with OP_END when this function is called.
1468    
1469    This function is called when a backward assertion is encountered, so that if it
1470    fails, the error message can point to the correct place in the pattern.
1471    However, we cannot do this when the assertion contains subroutine calls,
1472    because they can be forward references. We solve this by remembering this case
1473    and doing the check at the end; a flag specifies which mode we are running in.
1474    
1475  Arguments:  Arguments:
1476    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1477    options  the compiling options    options  the compiling options
1478      atend    TRUE if called when the pattern is complete
1479      cd       the "compile data" structure
1480    
1481  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1482                 or -1 if there is no fixed length,
1483               or -2 if \C was encountered               or -2 if \C was encountered
1484                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1485  */  */
1486    
1487  static int  static int
1488  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)
1489  {  {
1490  int length = -1;  int length = -1;
1491    
# Line 1145  branch, check the length against that of Line 1498  branch, check the length against that of
1498  for (;;)  for (;;)
1499    {    {
1500    int d;    int d;
1501      uschar *ce, *cs;
1502    register int op = *cc;    register int op = *cc;
1503    switch (op)    switch (op)
1504      {      {
# Line 1152  for (;;) Line 1506  for (;;)
1506      case OP_BRA:      case OP_BRA:
1507      case OP_ONCE:      case OP_ONCE:
1508      case OP_COND:      case OP_COND:
1509      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);
1510      if (d < 0) return d;      if (d < 0) return d;
1511      branchlength += d;      branchlength += d;
1512      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1175  for (;;) Line 1529  for (;;)
1529      branchlength = 0;      branchlength = 0;
1530      break;      break;
1531    
1532        /* A true recursion implies not fixed length, but a subroutine call may
1533        be OK. If the subroutine is a forward reference, we can't deal with
1534        it until the end of the pattern, so return -3. */
1535    
1536        case OP_RECURSE:
1537        if (!atend) return -3;
1538        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1539        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1540        if (cc > cs && cc < ce) return -1;                /* Recursion */
1541        d = find_fixedlength(cs + 2, options, atend, cd);
1542        if (d < 0) return d;
1543        branchlength += d;
1544        cc += 1 + LINK_SIZE;
1545        break;
1546    
1547      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1548    
1549      case OP_ASSERT:      case OP_ASSERT:
# Line 1188  for (;;) Line 1557  for (;;)
1557    
1558      case OP_REVERSE:      case OP_REVERSE:
1559      case OP_CREF:      case OP_CREF:
1560        case OP_NCREF:
1561      case OP_RREF:      case OP_RREF:
1562        case OP_NRREF:
1563      case OP_DEF:      case OP_DEF:
1564      case OP_OPT:      case OP_OPT:
1565      case OP_CALLOUT:      case OP_CALLOUT:
1566      case OP_SOD:      case OP_SOD:
1567      case OP_SOM:      case OP_SOM:
1568        case OP_SET_SOM:
1569      case OP_EOD:      case OP_EOD:
1570      case OP_EODN:      case OP_EODN:
1571      case OP_CIRC:      case OP_CIRC:
# Line 1211  for (;;) Line 1583  for (;;)
1583      branchlength++;      branchlength++;
1584      cc += 2;      cc += 2;
1585  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1586      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1587        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1588  #endif  #endif
1589      break;      break;
1590    
# Line 1225  for (;;) Line 1595  for (;;)
1595      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1596      cc += 4;      cc += 4;
1597  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1598      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1599        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1600  #endif  #endif
1601      break;      break;
1602    
# Line 1252  for (;;) Line 1620  for (;;)
1620      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1621      case OP_WORDCHAR:      case OP_WORDCHAR:
1622      case OP_ANY:      case OP_ANY:
1623        case OP_ALLANY:
1624      branchlength++;      branchlength++;
1625      cc++;      cc++;
1626      break;      break;
# Line 1306  for (;;) Line 1675  for (;;)
1675    
1676    
1677  /*************************************************  /*************************************************
1678  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1679  *************************************************/  *************************************************/
1680    
1681  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1682  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1683    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1684    so that it can be called from pcre_study() when finding the minimum matching
1685    length.
1686    
1687  Arguments:  Arguments:
1688    code        points to start of expression    code        points to start of expression
1689    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1690    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1691    
1692  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
1693  */  */
1694    
1695  static const uschar *  const uschar *
1696  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1697  {  {
1698  for (;;)  for (;;)
1699    {    {
# Line 1334  for (;;) Line 1706  for (;;)
1706    
1707    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1708    
1709      /* Handle recursion */
1710    
1711      else if (c == OP_REVERSE)
1712        {
1713        if (number < 0) return (uschar *)code;
1714        code += _pcre_OP_lengths[c];
1715        }
1716    
1717    /* Handle capturing bracket */    /* Handle capturing bracket */
1718    
1719    else if (c == OP_CBRA)    else if (c == OP_CBRA)
# Line 1345  for (;;) Line 1725  for (;;)
1725    
1726    /* 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
1727    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
1728    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1729      must add in its length. */
1730    
1731    else    else
1732      {      {
# Line 1369  for (;;) Line 1750  for (;;)
1750        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
1751        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1752        break;        break;
1753    
1754          case OP_MARK:
1755          case OP_PRUNE_ARG:
1756          case OP_SKIP_ARG:
1757          code += code[1];
1758          break;
1759    
1760          case OP_THEN_ARG:
1761          code += code[1+LINK_SIZE];
1762          break;
1763        }        }
1764    
1765      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1400  for (;;) Line 1791  for (;;)
1791        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1792        break;        break;
1793        }        }
1794    #else
1795        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1796  #endif  #endif
1797      }      }
1798    }    }
# Line 1438  for (;;) Line 1831  for (;;)
1831    
1832    /* 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
1833    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
1834    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1835      must add in its length. */
1836    
1837    else    else
1838      {      {
# Line 1462  for (;;) Line 1856  for (;;)
1856        case OP_TYPEEXACT:        case OP_TYPEEXACT:
1857        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1858        break;        break;
1859    
1860          case OP_MARK:
1861          case OP_PRUNE_ARG:
1862          case OP_SKIP_ARG:
1863          code += code[1];
1864          break;
1865    
1866          case OP_THEN_ARG:
1867          code += code[1+LINK_SIZE];
1868          break;
1869        }        }
1870    
1871      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1493  for (;;) Line 1897  for (;;)
1897        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1898        break;        break;
1899        }        }
1900    #else
1901        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1902  #endif  #endif
1903      }      }
1904    }    }
# Line 1516  Arguments: Line 1922  Arguments:
1922    code        points to start of search    code        points to start of search
1923    endcode     points to where to stop    endcode     points to where to stop
1924    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
1925      cd          contains pointers to tables etc.
1926    
1927  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
1928  */  */
1929    
1930  static BOOL  static BOOL
1931  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
1932      compile_data *cd)
1933  {  {
1934  register int c;  register int c;
1935  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);
# Line 1544  for (code = first_significant_code(code Line 1952  for (code = first_significant_code(code
1952    
1953    /* Groups with zero repeats can of course be empty; skip them. */    /* Groups with zero repeats can of course be empty; skip them. */
1954    
1955    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)
1956      {      {
1957      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1958      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
# Line 1552  for (code = first_significant_code(code Line 1960  for (code = first_significant_code(code
1960      continue;      continue;
1961      }      }
1962    
1963      /* For a recursion/subroutine call, if its end has been reached, which
1964      implies a subroutine call, we can scan it. */
1965    
1966      if (c == OP_RECURSE)
1967        {
1968        BOOL empty_branch = FALSE;
1969        const uschar *scode = cd->start_code + GET(code, 1);
1970        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
1971        do
1972          {
1973          if (could_be_empty_branch(scode, endcode, utf8, cd))
1974            {
1975            empty_branch = TRUE;
1976            break;
1977            }
1978          scode += GET(scode, 1);
1979          }
1980        while (*scode == OP_ALT);
1981        if (!empty_branch) return FALSE;  /* All branches are non-empty */
1982        continue;
1983        }
1984    
1985    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
1986    
1987    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)
# Line 1559  for (code = first_significant_code(code Line 1989  for (code = first_significant_code(code
1989      BOOL empty_branch;      BOOL empty_branch;
1990      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
1991    
1992      /* Scan a closed bracket */      /* If a conditional group has only one branch, there is a second, implied,
1993        empty branch, so just skip over the conditional, because it could be empty.
1994        Otherwise, scan the individual branches of the group. */
1995    
1996      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;  
1997        code += GET(code, 1);        code += GET(code, 1);
1998        else
1999          {
2000          empty_branch = FALSE;
2001          do
2002            {
2003            if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
2004              empty_branch = TRUE;
2005            code += GET(code, 1);
2006            }
2007          while (*code == OP_ALT);
2008          if (!empty_branch) return FALSE;   /* All branches are non-empty */
2009        }        }
2010      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
2011      c = *code;      c = *code;
2012      continue;      continue;
2013      }      }
# Line 1630  for (code = first_significant_code(code Line 2068  for (code = first_significant_code(code
2068      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2069      case OP_WORDCHAR:      case OP_WORDCHAR:
2070      case OP_ANY:      case OP_ANY:
2071        case OP_ALLANY:
2072      case OP_ANYBYTE:      case OP_ANYBYTE:
2073      case OP_CHAR:      case OP_CHAR:
2074      case OP_CHARNC:      case OP_CHARNC:
# Line 1686  for (code = first_significant_code(code Line 2125  for (code = first_significant_code(code
2125      case OP_QUERY:      case OP_QUERY:
2126      case OP_MINQUERY:      case OP_MINQUERY:
2127      case OP_POSQUERY:      case OP_POSQUERY:
2128        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2129        break;
2130    
2131      case OP_UPTO:      case OP_UPTO:
2132      case OP_MINUPTO:      case OP_MINUPTO:
2133      case OP_POSUPTO:      case OP_POSUPTO:
2134      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2135      break;      break;
2136  #endif  #endif
2137    
2138        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2139        string. */
2140    
2141        case OP_MARK:
2142        case OP_PRUNE_ARG:
2143        case OP_SKIP_ARG:
2144        code += code[1];
2145        break;
2146    
2147        case OP_THEN_ARG:
2148        code += code[1+LINK_SIZE];
2149        break;
2150    
2151        /* None of the remaining opcodes are required to match a character. */
2152    
2153        default:
2154        break;
2155      }      }
2156    }    }
2157    
# Line 1714  Arguments: Line 2174  Arguments:
2174    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2175    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2176    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2177      cd          pointers to tables etc
2178    
2179  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2180  */  */
2181    
2182  static BOOL  static BOOL
2183  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2184    BOOL utf8)    BOOL utf8, compile_data *cd)
2185  {  {
2186  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2187    {    {
2188    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2189        return FALSE;
2190    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2191    }    }
2192  return TRUE;  return TRUE;
# Line 1770  int terminator; /* Don't combin Line 2232  int terminator; /* Don't combin
2232  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2233  for (++ptr; *ptr != 0; ptr++)  for (++ptr; *ptr != 0; ptr++)
2234    {    {
2235    if (*ptr == '\\' && ptr[1] == ']') ptr++; else    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else
2236      {      {
2237      if (*ptr == ']') return FALSE;      if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2238      if (*ptr == terminator && ptr[1] == ']')      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2239        {        {
2240        *endptr = ptr;        *endptr = ptr;
2241        return TRUE;        return TRUE;
# Line 1824  return -1; Line 2286  return -1;
2286  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2287  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2288  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
2289  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
2290  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
2291  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
2292  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
2293  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2294    OP_END.
2295    
2296  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2297  recursions and subroutine calls. It must also check the list of such references  recursions and subroutine calls. It must also check the list of such references
# Line 1905  auto_callout(uschar *code, const uschar Line 2368  auto_callout(uschar *code, const uschar
2368  {  {
2369  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2370  *code++ = 255;  *code++ = 255;
2371  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2372  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2373  return code + 2*LINK_SIZE;  return code + 2*LINK_SIZE;
2374  }  }
2375    
# Line 1931  Returns: nothing Line 2394  Returns: nothing
2394  static void  static void
2395  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)
2396  {  {
2397  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2398  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2399  }  }
2400    
# Line 1963  get_othercase_range(unsigned int *cptr, Line 2426  get_othercase_range(unsigned int *cptr,
2426  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2427    
2428  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2429    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2430    
2431  if (c > d) return FALSE;  if (c > d) return FALSE;
2432    
# Line 1972  next = othercase + 1; Line 2435  next = othercase + 1;
2435    
2436  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2437    {    {
2438    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2439    next++;    next++;
2440    }    }
2441    
# Line 1981  for (++c; c <= d; c++) Line 2444  for (++c; c <= d; c++)
2444    
2445  return TRUE;  return TRUE;
2446  }  }
2447    
2448    
2449    
2450    /*************************************************
2451    *        Check a character and a property        *
2452    *************************************************/
2453    
2454    /* This function is called by check_auto_possessive() when a property item
2455    is adjacent to a fixed character.
2456    
2457    Arguments:
2458      c            the character
2459      ptype        the property type
2460      pdata        the data for the type
2461      negated      TRUE if it's a negated property (\P or \p{^)
2462    
2463    Returns:       TRUE if auto-possessifying is OK
2464    */
2465    
2466    static BOOL
2467    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2468    {
2469    const ucd_record *prop = GET_UCD(c);
2470    switch(ptype)
2471      {
2472      case PT_LAMP:
2473      return (prop->chartype == ucp_Lu ||
2474              prop->chartype == ucp_Ll ||
2475              prop->chartype == ucp_Lt) == negated;
2476    
2477      case PT_GC:
2478      return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;
2479    
2480      case PT_PC:
2481      return (pdata == prop->chartype) == negated;
2482    
2483      case PT_SC:
2484      return (pdata == prop->script) == negated;
2485    
2486      /* These are specials */
2487    
2488      case PT_ALNUM:
2489      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2490              _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;
2491    
2492      case PT_SPACE:    /* Perl space */
2493      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2494              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2495              == negated;
2496    
2497      case PT_PXSPACE:  /* POSIX space */
2498      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2499              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2500              c == CHAR_FF || c == CHAR_CR)
2501              == negated;
2502    
2503      case PT_WORD:
2504      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2505              _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2506              c == CHAR_UNDERSCORE) == negated;
2507      }
2508    return FALSE;
2509    }
2510  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2511    
2512    
# Line 1994  whether the next thing could possibly ma Line 2520  whether the next thing could possibly ma
2520  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
2521    
2522  Arguments:  Arguments:
2523    op_code       the repeated op code    previous      pointer to the repeated opcode
   this          data for this item, depends on the opcode  
2524    utf8          TRUE in UTF-8 mode    utf8          TRUE in UTF-8 mode
   utf8_char     used for utf8 character bytes, NULL if not relevant  
2525    ptr           next character in pattern    ptr           next character in pattern
2526    options       options bits    options       options bits
2527    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 2006  Returns: TRUE if possessifying is Line 2530  Returns: TRUE if possessifying is
2530  */  */
2531    
2532  static BOOL  static BOOL
2533  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
2534    const uschar *ptr, int options, compile_data *cd)    int options, compile_data *cd)
2535  {  {
2536  int next;  int c, next;
2537    int op_code = *previous++;
2538    
2539  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
2540    
# Line 2018  if ((options & PCRE_EXTENDED) != 0) Line 2543  if ((options & PCRE_EXTENDED) != 0)
2543    for (;;)    for (;;)
2544      {      {
2545      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2546      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2547        {        {
2548        while (*(++ptr) != 0)        ptr++;
2549          while (*ptr != 0)
2550            {
2551          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2552            ptr++;
2553    #ifdef SUPPORT_UTF8
2554            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2555    #endif
2556            }
2557        }        }
2558      else break;      else break;
2559      }      }
# Line 2030  if ((options & PCRE_EXTENDED) != 0) Line 2562  if ((options & PCRE_EXTENDED) != 0)
2562  /* 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
2563  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2564    
2565  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2566    {    {
2567    int temperrorcode = 0;    int temperrorcode = 0;
2568    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
# Line 2055  if ((options & PCRE_EXTENDED) != 0) Line 2587  if ((options & PCRE_EXTENDED) != 0)
2587    for (;;)    for (;;)
2588      {      {
2589      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2590      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2591        {        {
2592        while (*(++ptr) != 0)        ptr++;
2593          while (*ptr != 0)
2594            {
2595          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2596            ptr++;
2597    #ifdef SUPPORT_UTF8
2598            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2599    #endif
2600            }
2601        }        }
2602      else break;      else break;
2603      }      }
# Line 2066  if ((options & PCRE_EXTENDED) != 0) Line 2605  if ((options & PCRE_EXTENDED) != 0)
2605    
2606  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
2607    
2608  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2609    return FALSE;    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2610        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. */  
   
2611    
2612  /* Handle cases when the next item is a character. */  /* Now compare the next item with the previous opcode. First, handle cases when
2613    the next item is a character. */
2614    
2615  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
2616    {    {
2617    case OP_CHAR:    case OP_CHAR:
2618  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2619    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2620    #else
2621      c = *previous;
2622  #endif  #endif
2623    return item != next;    return c != next;
2624    
2625    /* For CHARNC (caseless character) we must check the other case. If we have    /* For CHARNC (caseless character) we must check the other case. If we have
2626    Unicode property support, we can use it to test the other case of    Unicode property support, we can use it to test the other case of
# Line 2091  if (next >= 0) switch(op_code) Line 2628  if (next >= 0) switch(op_code)
2628    
2629    case OP_CHARNC:    case OP_CHARNC:
2630  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2631    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2632    #else
2633      c = *previous;
2634  #endif  #endif
2635    if (item == next) return FALSE;    if (c == next) return FALSE;
2636  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2637    if (utf8)    if (utf8)
2638      {      {
2639      unsigned int othercase;      unsigned int othercase;
2640      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2641  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2642      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
2643  #else  #else
2644      othercase = NOTACHAR;      othercase = NOTACHAR;
2645  #endif  #endif
2646      return (unsigned int)item != othercase;      return (unsigned int)c != othercase;
2647      }      }
2648    else    else
2649  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2650    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2651    
2652    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT, its data is always a single-byte character. */
2653    
2654    case OP_NOT:    case OP_NOT:
2655    if (next < 0) return FALSE;  /* Not a character */    if ((c = *previous) == next) return TRUE;
   if (item == next) return TRUE;  
2656    if ((options & PCRE_CASELESS) == 0) return FALSE;    if ((options & PCRE_CASELESS) == 0) return FALSE;
2657  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2658    if (utf8)    if (utf8)
# Line 2122  if (next >= 0) switch(op_code) Line 2660  if (next >= 0) switch(op_code)
2660      unsigned int othercase;      unsigned int othercase;
2661      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2662  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2663      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
2664  #else  #else
2665      othercase = NOTACHAR;      othercase = NOTACHAR;
2666  #endif  #endif
2667      return (unsigned int)item == othercase;      return (unsigned int)c == othercase;
2668      }      }
2669    else    else
2670  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2671    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2672    
2673      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2674      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
2675    
2676    case OP_DIGIT:    case OP_DIGIT:
2677    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
# Line 2173  if (next >= 0) switch(op_code) Line 2714  if (next >= 0) switch(op_code)
2714      case 0x202f:      case 0x202f:
2715      case 0x205f:      case 0x205f:
2716      case 0x3000:      case 0x3000:
2717      return op_code != OP_HSPACE;      return op_code == OP_NOT_HSPACE;
2718      default:      default:
2719      return op_code == OP_HSPACE;      return op_code != OP_NOT_HSPACE;
2720      }      }
2721    
2722      case OP_ANYNL:
2723    case OP_VSPACE:    case OP_VSPACE:
2724    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
2725    switch(next)    switch(next)
# Line 2189  if (next >= 0) switch(op_code) Line 2731  if (next >= 0) switch(op_code)
2731      case 0x85:      case 0x85:
2732      case 0x2028:      case 0x2028:
2733      case 0x2029:      case 0x2029:
2734      return op_code != OP_VSPACE;      return op_code == OP_NOT_VSPACE;
2735      default:      default:
2736      return op_code == OP_VSPACE;      return op_code != OP_NOT_VSPACE;
2737      }      }
2738    
2739    #ifdef SUPPORT_UCP
2740      case OP_PROP:
2741      return check_char_prop(next, previous[0], previous[1], FALSE);
2742    
2743      case OP_NOTPROP:
2744      return check_char_prop(next, previous[0], previous[1], TRUE);
2745    #endif
2746    
2747    default:    default:
2748    return FALSE;    return FALSE;
2749    }    }
2750    
2751    
2752  /* 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
2753    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
2754    generated only when PCRE_UCP is *not* set, that is, when only ASCII
2755    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
2756    replaced by OP_PROP codes when PCRE_UCP is set. */
2757    
2758  switch(op_code)  switch(op_code)
2759    {    {
2760    case OP_CHAR:    case OP_CHAR:
2761    case OP_CHARNC:    case OP_CHARNC:
2762  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2763    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2764    #else
2765      c = *previous;
2766  #endif  #endif
2767    switch(-next)    switch(-next)
2768      {      {
2769      case ESC_d:      case ESC_d:
2770      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
2771    
2772      case ESC_D:      case ESC_D:
2773      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
2774    
2775      case ESC_s:      case ESC_s:
2776      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
2777    
2778      case ESC_S:      case ESC_S:
2779      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
2780    
2781      case ESC_w:      case ESC_w:
2782      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
2783    
2784      case ESC_W:      case ESC_W:
2785      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
2786    
2787      case ESC_h:      case ESC_h:
2788      case ESC_H:      case ESC_H:
2789      switch(item)      switch(c)
2790        {        {
2791        case 0x09:        case 0x09:
2792        case 0x20:        case 0x20:
# Line 2258  switch(op_code) Line 2814  switch(op_code)
2814    
2815      case ESC_v:      case ESC_v:
2816      case ESC_V:      case ESC_V:
2817      switch(item)      switch(c)
2818        {        {
2819        case 0x0a:        case 0x0a:
2820        case 0x0b:        case 0x0b:
# Line 2272  switch(op_code) Line 2828  switch(op_code)
2828        return -next == ESC_v;        return -next == ESC_v;
2829        }        }
2830    
2831        /* When PCRE_UCP is set, these values get generated for \d etc. Find
2832        their substitutions and process them. The result will always be either
2833        -ESC_p or -ESC_P. Then fall through to process those values. */
2834    
2835    #ifdef SUPPORT_UCP
2836        case ESC_du:
2837        case ESC_DU:
2838        case ESC_wu:
2839        case ESC_WU:
2840        case ESC_su:
2841        case ESC_SU:
2842          {
2843          int temperrorcode = 0;
2844          ptr = substitutes[-next - ESC_DU];
2845          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
2846          if (temperrorcode != 0) return FALSE;
2847          ptr++;    /* For compatibility */
2848          }
2849        /* Fall through */
2850    
2851        case ESC_p:
2852        case ESC_P:
2853          {
2854          int ptype, pdata, errorcodeptr;
2855          BOOL negated;
2856    
2857          ptr--;      /* Make ptr point at the p or P */
2858          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
2859          if (ptype < 0) return FALSE;
2860          ptr++;      /* Point past the final curly ket */
2861    
2862          /* If the property item is optional, we have to give up. (When generated
2863          from \d etc by PCRE_UCP, this test will have been applied much earlier,
2864          to the original \d etc. At this point, ptr will point to a zero byte. */
2865    
2866          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2867            strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2868              return FALSE;
2869    
2870          /* Do the property check. */
2871    
2872          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
2873          }
2874    #endif
2875    
2876      default:      default:
2877      return FALSE;      return FALSE;
2878      }      }
2879    
2880      /* In principle, support for Unicode properties should be integrated here as
2881      well. It means re-organizing the above code so as to get hold of the property
2882      values before switching on the op-code. However, I wonder how many patterns
2883      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
2884      these op-codes are never generated.) */
2885    
2886    case OP_DIGIT:    case OP_DIGIT:
2887    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2888           next == -ESC_h || next == -ESC_v;           next == -ESC_h || next == -ESC_v || next == -ESC_R;
2889    
2890    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
2891    return next == -ESC_d;    return next == -ESC_d;
2892    
2893    case OP_WHITESPACE:    case OP_WHITESPACE:
2894    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
2895    
2896    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
2897    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2898    
2899    case OP_HSPACE:    case OP_HSPACE:
2900    return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
2901             next == -ESC_w || next == -ESC_v || next == -ESC_R;
2902    
2903    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
2904    return next == -ESC_h;    return next == -ESC_h;
2905    
2906    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
2907      case OP_ANYNL:
2908    case OP_VSPACE:    case OP_VSPACE:
2909    return next == -ESC_V || next == -ESC_d || next == -ESC_w;    return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2910    
2911    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
2912    return next == -ESC_v;    return next == -ESC_v || next == -ESC_R;
2913    
2914    case OP_WORDCHAR:    case OP_WORDCHAR:
2915    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
2916             next == -ESC_v || next == -ESC_R;
2917    
2918    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
2919    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2367  BOOL inescq = FALSE; Line 2977  BOOL inescq = FALSE;
2977  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
2978  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
2979  const uschar *tempptr;  const uschar *tempptr;
2980    const uschar *nestptr = NULL;
2981  uschar *previous = NULL;  uschar *previous = NULL;
2982  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
2983  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
# Line 2383  BOOL utf8 = FALSE; Line 2994  BOOL utf8 = FALSE;
2994  uschar *utf8_char = NULL;  uschar *utf8_char = NULL;
2995  #endif  #endif
2996    
2997  #ifdef DEBUG  #ifdef PCRE_DEBUG
2998  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2999  #endif  #endif
3000    
# Line 2437  for (;; ptr++) Line 3048  for (;; ptr++)
3048    
3049    c = *ptr;    c = *ptr;
3050    
3051      /* If we are at the end of a nested substitution, revert to the outer level
3052      string. Nesting only happens one level deep. */
3053    
3054      if (c == 0 && nestptr != NULL)
3055        {
3056        ptr = nestptr;
3057        nestptr = NULL;
3058        c = *ptr;
3059        }
3060    
3061    /* 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
3062    previous cycle of this loop. */    previous cycle of this loop. */
3063    
3064    if (lengthptr != NULL)    if (lengthptr != NULL)
3065      {      {
3066  #ifdef DEBUG  #ifdef PCRE_DEBUG
3067      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3068  #endif  #endif
3069      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
3070        {        {
3071        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3072        goto FAILED;        goto FAILED;
# Line 2467  for (;; ptr++) Line 3088  for (;; ptr++)
3088        goto FAILED;        goto FAILED;
3089        }        }
3090    
3091      *lengthptr += code - last_code;      *lengthptr += (int)(code - last_code);
3092      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));
3093    
3094      /* 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 2494  for (;; ptr++) Line 3115  for (;; ptr++)
3115    /* 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
3116    reference list. */    reference list. */
3117    
3118    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
3119      {      {
3120      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3121      goto FAILED;      goto FAILED;
# Line 2504  for (;; ptr++) Line 3125  for (;; ptr++)
3125    
3126    if (inescq && c != 0)    if (inescq && c != 0)
3127      {      {
3128      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3129        {        {
3130        inescq = FALSE;        inescq = FALSE;
3131        ptr++;        ptr++;
# Line 2530  for (;; ptr++) Line 3151  for (;; ptr++)
3151    /* 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
3152    a quantifier. */    a quantifier. */
3153    
3154    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
3155      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
3156        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
3157    
3158    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
3159         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2546  for (;; ptr++) Line 3168  for (;; ptr++)
3168    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3169      {      {
3170      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
3171      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
3172        {        {
3173        while (*(++ptr) != 0)        ptr++;
3174          while (*ptr != 0)
3175          {          {
3176          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3177            ptr++;
3178    #ifdef SUPPORT_UTF8
3179            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
3180    #endif
3181          }          }
3182        if (*ptr != 0) continue;        if (*ptr != 0) continue;
3183    
# Line 2571  for (;; ptr++) Line 3198  for (;; ptr++)
3198      {      {
3199      /* ===================================================================*/      /* ===================================================================*/
3200      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
3201      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
3202      case ')':      case CHAR_RIGHT_PARENTHESIS:
3203      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
3204      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
3205      *codeptr = code;      *codeptr = code;
# Line 2584  for (;; ptr++) Line 3211  for (;; ptr++)
3211          *errorcodeptr = ERR20;          *errorcodeptr = ERR20;
3212          goto FAILED;          goto FAILED;
3213          }          }
3214        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += (int)(code - last_code);   /* To include callout length */
3215        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
3216        }        }
3217      return TRUE;      return TRUE;
# Line 2594  for (;; ptr++) Line 3221  for (;; ptr++)
3221      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
3222      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3223    
3224      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
3225      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3226        {        {
3227        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
# Line 2603  for (;; ptr++) Line 3230  for (;; ptr++)
3230      *code++ = OP_CIRC;      *code++ = OP_CIRC;
3231      break;      break;
3232    
3233      case '$':      case CHAR_DOLLAR_SIGN:
3234      previous = NULL;      previous = NULL;
3235      *code++ = OP_DOLL;      *code++ = OP_DOLL;
3236      break;      break;
# Line 2611  for (;; ptr++) Line 3238  for (;; ptr++)
3238      /* 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
3239      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
3240    
3241      case '.':      case CHAR_DOT:
3242      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3243      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
3244      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3245      previous = code;      previous = code;
3246      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
3247      break;      break;
3248    
3249    
# Line 2631  for (;; ptr++) Line 3258  for (;; ptr++)
3258      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,
3259      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
3260      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.
     */  
3261    
3262      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
3263        default (Perl) mode, it is treated as a data character. */
3264    
3265        case CHAR_RIGHT_SQUARE_BRACKET:
3266        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3267          {
3268          *errorcodeptr = ERR64;
3269          goto FAILED;
3270          }
3271        goto NORMAL_CHAR;
3272    
3273        case CHAR_LEFT_SQUARE_BRACKET:
3274      previous = code;      previous = code;
3275    
3276      /* 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
3277      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. */
3278    
3279      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3280             ptr[1] == CHAR_EQUALS_SIGN) &&
3281          check_posix_syntax(ptr, &tempptr))          check_posix_syntax(ptr, &tempptr))
3282        {        {
3283        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
3284        goto FAILED;        goto FAILED;
3285        }        }
3286    
# Line 2654  for (;; ptr++) Line 3292  for (;; ptr++)
3292      for (;;)      for (;;)
3293        {        {
3294        c = *(++ptr);        c = *(++ptr);
3295        if (c == '\\')        if (c == CHAR_BACKSLASH)
3296          {          {
3297          if (ptr[1] == 'E') ptr++;          if (ptr[1] == CHAR_E)
3298            else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;            ptr++;
3299              else break;          else if (strncmp((const char *)ptr+1,
3300                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
3301              ptr += 3;
3302            else
3303              break;
3304          }          }
3305        else if (!negate_class && c == '^')        else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
3306          negate_class = TRUE;          negate_class = TRUE;
3307        else break;        else break;
3308        }        }
3309    
3310        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
3311        an initial ']' is taken as a data character -- the code below handles
3312        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3313        [^] must match any character, so generate OP_ALLANY. */
3314    
3315        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3316            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3317          {
3318          *code++ = negate_class? OP_ALLANY : OP_FAIL;
3319          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3320          zerofirstbyte = firstbyte;
3321          break;
3322          }
3323    
3324      /* 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
3325      negation flag at the end, so that support for characters > 255 works      negation flag at the end, so that support for characters > 255 works
3326      correctly (they are all included in the class). */      correctly (they are all included in the class). */
# Line 2722  for (;; ptr++) Line 3378  for (;; ptr++)
3378    
3379        if (inescq)        if (inescq)
3380          {          {
3381          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3382            {            {
3383            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
3384            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2737  for (;; ptr++) Line 3393  for (;; ptr++)
3393        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3394        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3395    
3396        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3397            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3398            check_posix_syntax(ptr, &tempptr))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3399          {          {
3400          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3401          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3402          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3403          uschar pbits[32];          uschar pbits[32];
3404    
3405          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3406            {            {
3407            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3408            goto FAILED;            goto FAILED;
3409            }            }
3410    
3411          ptr += 2;          ptr += 2;
3412          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3413            {            {
3414            local_negate = TRUE;            local_negate = TRUE;
3415            should_flip_negation = TRUE;  /* Note negative special */            should_flip_negation = TRUE;  /* Note negative special */
3416            ptr++;            ptr++;
3417            }            }
3418    
3419          posix_class = check_posix_name(ptr, tempptr - ptr);          posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3420          if (posix_class < 0)          if (posix_class < 0)
3421            {            {
3422            *errorcodeptr = ERR30;            *errorcodeptr = ERR30;
# Line 2774  for (;; ptr++) Line 3430  for (;; ptr++)
3430          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3431            posix_class = 0;            posix_class = 0;
3432    
3433          /* 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
3434          because we may be adding and subtracting from it, and we don't want to          different escape sequences that use Unicode properties. */
3435          subtract bits that may be in the main map already. At the end we or the  
3436          result into the bit map that is being built. */  #ifdef SUPPORT_UCP
3437            if ((options & PCRE_UCP) != 0)
3438              {
3439              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3440              if (posix_substitutes[pc] != NULL)
3441                {
3442                nestptr = tempptr + 1;
3443                ptr = posix_substitutes[pc] - 1;
3444                continue;
3445                }
3446              }
3447    #endif
3448            /* In the non-UCP case, we build the bit map for the POSIX class in a
3449            chunk of local store because we may be adding and subtracting from it,
3450            and we don't want to subtract bits that may be in the main map already.
3451            At the end we or the result into the bit map that is being built. */
3452    
3453          posix_class *= 3;          posix_class *= 3;
3454    
# Line 2821  for (;; ptr++) Line 3492  for (;; ptr++)
3492    
3493        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3494        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
3495        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
3496        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
3497        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
3498        character in them, so set class_charcount bigger than one. */        are either treated as literal characters (by default), or are faulted if
3499          PCRE_EXTRA is set. */
3500    
3501        if (c == '\\')        if (c == CHAR_BACKSLASH)
3502          {          {
3503          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3504          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3505    
3506          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 */  
3507          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3508            {            {
3509            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3510              {              {
3511              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3512              }              }
# Line 2850  for (;; ptr++) Line 3520  for (;; ptr++)
3520            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
3521            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3522    
3523            /* Save time by not doing this in the pre-compile phase. */            switch (-c)
   
           if (lengthptr == NULL) switch (-c)  
3524              {              {
3525    #ifdef SUPPORT_UCP
3526                case ESC_du:     /* These are the values given for \d etc */
3527                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3528                case ESC_wu:     /* escape sequence with an appropriate \p */
3529                case ESC_WU:     /* or \P to test Unicode properties instead */
3530                case ESC_su:     /* of the default ASCII testing. */
3531                case ESC_SU:
3532                nestptr = ptr;
3533                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3534                class_charcount -= 2;                /* Undo! */
3535                continue;
3536    #endif
3537              case ESC_d:              case ESC_d:
3538              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3539              continue;              continue;
# Line 2872  for (;; ptr++) Line 3552  for (;; ptr++)
3552              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3553              continue;              continue;
3554    
3555                /* Perl 5.004 onwards omits VT from \s, but we must preserve it
3556                if it was previously set by something earlier in the character
3557                class. */
3558    
3559              case ESC_s:              case ESC_s:
3560              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];              classbits[0] |= cbits[cbit_space];
3561              classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= cbits[cbit_space+1] & ~0x08;
3562                for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3563              continue;              continue;
3564    
3565              case ESC_S:              case ESC_S:
# Line 2883  for (;; ptr++) Line 3568  for (;; ptr++)
3568              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3569              continue;              continue;
3570    
3571              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)  
             {  
3572              SETBIT(classbits, 0x09); /* VT */              SETBIT(classbits, 0x09); /* VT */
3573              SETBIT(classbits, 0x20); /* SPACE */              SETBIT(classbits, 0x20); /* SPACE */
3574              SETBIT(classbits, 0xa0); /* NSBP */              SETBIT(classbits, 0xa0); /* NSBP */
# Line 2920  for (;; ptr++) Line 3592  for (;; ptr++)
3592                }                }
3593  #endif  #endif
3594              continue;              continue;
             }  
3595    
3596            if (-c == ESC_H)              case ESC_H:
             {  
3597              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3598                {                {
3599                int x = 0xff;                int x = 0xff;
# Line 2965  for (;; ptr++) Line 3635  for (;; ptr++)
3635                }                }
3636  #endif  #endif
3637              continue;              continue;
             }  
3638    
3639            if (-c == ESC_v)              case ESC_v:
             {  
3640              SETBIT(classbits, 0x0a); /* LF */              SETBIT(classbits, 0x0a); /* LF */
3641              SETBIT(classbits, 0x0b); /* VT */              SETBIT(classbits, 0x0b); /* VT */
3642              SETBIT(classbits, 0x0c); /* FF */              SETBIT(classbits, 0x0c); /* FF */
# Line 2984  for (;; ptr++) Line 3652  for (;; ptr++)
3652                }                }
3653  #endif  #endif
3654              continue;              continue;
             }  
3655    
3656            if (-c == ESC_V)              case ESC_V:
             {  
3657              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3658                {                {
3659                int x = 0xff;                int x = 0xff;
# Line 3017  for (;; ptr++) Line 3683  for (;; ptr++)
3683                }                }
3684  #endif  #endif
3685              continue;              continue;
             }  
   
           /* We need to deal with \P and \p in both phases. */  
3686    
3687  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3688            if (-c == ESC_p || -c == ESC_P)              case ESC_p:
3689              {              case ESC_P:
3690              BOOL negated;                {
3691              int pdata;                BOOL negated;
3692              int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                int pdata;
3693              if (ptype < 0) goto FAILED;                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3694              class_utf8 = TRUE;                if (ptype < 0) goto FAILED;
3695              *class_utf8data++ = ((-c == ESC_p) != negated)?                class_utf8 = TRUE;
3696                XCL_PROP : XCL_NOTPROP;                *class_utf8data++ = ((-c == ESC_p) != negated)?
3697              *class_utf8data++ = ptype;                  XCL_PROP : XCL_NOTPROP;
3698              *class_utf8data++ = pdata;                *class_utf8data++ = ptype;
3699              class_charcount -= 2;   /* Not a < 256 character */                *class_utf8data++ = pdata;
3700              continue;                class_charcount -= 2;   /* Not a < 256 character */
3701              }                continue;
3702                  }
3703  #endif  #endif
3704            /* Unrecognized escapes are faulted if PCRE is running in its              /* Unrecognized escapes are faulted if PCRE is running in its
3705            strict mode. By default, for compatibility with Perl, they are              strict mode. By default, for compatibility with Perl, they are
3706            treated as literals. */              treated as literals. */
3707    
3708            if ((options & PCRE_EXTRA) != 0)              default:
3709              {              if ((options & PCRE_EXTRA) != 0)
3710              *errorcodeptr = ERR7;                {
3711              goto FAILED;                *errorcodeptr = ERR7;
3712                  goto FAILED;
3713                  }
3714                class_charcount -= 2;  /* Undo the default count from above */
3715                c = *ptr;              /* Get the final character and fall through */
3716                break;
3717              }              }
   
           class_charcount -= 2;  /* Undo the default count from above */  
           c = *ptr;              /* Get the final character and fall through */  
3718            }            }
3719    
3720          /* 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 3062  for (;; ptr++) Line 3728  for (;; ptr++)
3728        entirely. The code for handling \Q and \E is messy. */        entirely. The code for handling \Q and \E is messy. */
3729    
3730        CHECK_RANGE:        CHECK_RANGE:
3731        while (ptr[1] == '\\' && ptr[2] == 'E')        while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3732          {          {
3733          inescq = FALSE;          inescq = FALSE;
3734          ptr += 2;          ptr += 2;
# Line 3072  for (;; ptr++) Line 3738  for (;; ptr++)
3738    
3739        /* Remember \r or \n */        /* Remember \r or \n */
3740    
3741        if (c == '\r' || c == '\n') cd->external_flags |= PCRE_HASCRORLF;        if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3742    
3743        /* Check for range */        /* Check for range */
3744    
3745        if (!inescq && ptr[1] == '-')        if (!inescq && ptr[1] == CHAR_MINUS)
3746          {          {
3747          int d;          int d;
3748          ptr += 2;          ptr += 2;
3749          while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3750    
3751          /* 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
3752          mode. */          mode. */
3753    
3754          while (*ptr == '\\' && ptr[1] == 'Q')          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3755            {            {
3756            ptr += 2;            ptr += 2;
3757            if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }            if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3758                { ptr += 2; continue; }
3759            inescq = TRUE;            inescq = TRUE;
3760            break;            break;
3761            }            }
3762    
3763          if (*ptr == 0 || (!inescq && *ptr == ']'))          if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3764            {            {
3765            ptr = oldptr;            ptr = oldptr;
3766            goto LONE_SINGLE_CHARACTER;            goto LONE_SINGLE_CHARACTER;
# Line 3112  for (;; ptr++) Line 3779  for (;; ptr++)
3779          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
3780          in such circumstances. */          in such circumstances. */
3781    
3782          if (!inescq && d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3783            {            {
3784            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3785            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3786    
3787            /* \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 */  
3788    
3789            if (d < 0)            if (d < 0)
3790              {              {
3791              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  
3792                {                {
3793                ptr = oldptr;                ptr = oldptr;
3794                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 3145  for (;; ptr++) Line 3809  for (;; ptr++)
3809    
3810          /* Remember \r or \n */          /* Remember \r or \n */
3811    
3812          if (d == '\r' || d == '\n') cd->external_flags |= PCRE_HASCRORLF;          if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3813    
3814          /* 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
3815          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 3265  for (;; ptr++) Line 3929  for (;; ptr++)
3929          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3930            {            {
3931            unsigned int othercase;            unsigned int othercase;
3932            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)            if ((othercase = UCD_OTHERCASE(c)) != c)
3933              {              {
3934              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3935              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 3290  for (;; ptr++) Line 3954  for (;; ptr++)
3954          }          }
3955        }        }
3956    
3957      /* 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.
3958        If we are at the end of an internal nested string, revert to the outer
3959        string. */
3960    
3961        while (((c = *(++ptr)) != 0 ||
3962               (nestptr != NULL &&
3963                 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != 0)) &&
3964               (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
3965    
3966      while ((c = *(++ptr)) != 0 && (c != ']' || inescq));      /* Check for missing terminating ']' */
3967    
3968      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)
3969        {        {
3970        *errorcodeptr = ERR6;        *errorcodeptr = ERR6;
3971        goto FAILED;        goto FAILED;
3972        }        }
3973    
   
 /* 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  
   
   
3974      /* 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
3975      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
3976      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 3382  we set the flag only if there is a liter Line 4034  we set the flag only if there is a liter
4034    
4035      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
4036      extended class, with its own opcode, unless there was a negated special      extended class, with its own opcode, unless there was a negated special
4037      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
4038      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
4039      (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
4040      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
4041        actual compiled code. */
4042    
4043  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4044      if (class_utf8 && !should_flip_negation)      if (class_utf8 && (!should_flip_negation || (options & PCRE_UCP) != 0))
4045        {        {
4046        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
4047        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 3414  we set the flag only if there is a liter Line 4067  we set the flag only if there is a liter
4067        }        }
4068  #endif  #endif
4069    
4070      /* 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
4071      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
4072      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
4073      map into the code vector, negating it if necessary. */      (non-UCP) in the class. Then copy the 32-byte map into the code vector,
4074        negating it if necessary. */
4075    
4076      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
4077      if (negate_class)      if (negate_class)
# Line 3437  we set the flag only if there is a liter Line 4091  we set the flag only if there is a liter
4091      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
4092      has been tested above. */      has been tested above. */
4093    
4094      case '{':      case CHAR_LEFT_CURLY_BRACKET:
4095      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
4096      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
4097      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
4098      goto REPEAT;      goto REPEAT;
4099    
4100      case '*':      case CHAR_ASTERISK:
4101      repeat_min = 0;      repeat_min = 0;
4102      repeat_max = -1;      repeat_max = -1;
4103      goto REPEAT;      goto REPEAT;
4104    
4105      case '+':      case CHAR_PLUS:
4106      repeat_min = 1;      repeat_min = 1;
4107      repeat_max = -1;      repeat_max = -1;
4108      goto REPEAT;      goto REPEAT;
4109    
4110      case '?':      case CHAR_QUESTION_MARK:
4111      repeat_min = 0;      repeat_min = 0;
4112      repeat_max = 1;      repeat_max = 1;
4113    
# Line 3488  we set the flag only if there is a liter Line 4142  we set the flag only if there is a liter
4142      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
4143      repeat type to the non-default. */      repeat type to the non-default. */
4144    
4145      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
4146        {        {
4147        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
4148        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
4149        ptr++;        ptr++;
4150        }        }
4151      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
4152        {        {
4153        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
4154        ptr++;        ptr++;
# Line 3541  we set the flag only if there is a liter Line 4195  we set the flag only if there is a liter
4195    
4196        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4197            repeat_max < 0 &&            repeat_max < 0 &&
4198            check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
             options, cd))  
4199          {          {
4200          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4201          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3563  we set the flag only if there is a liter Line 4216  we set the flag only if there is a liter
4216        c = previous[1];        c = previous[1];
4217        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4218            repeat_max < 0 &&            repeat_max < 0 &&
4219            check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4220          {          {
4221          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4222          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3587  we set the flag only if there is a liter Line 4240  we set the flag only if there is a liter
4240    
4241        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4242            repeat_max < 0 &&            repeat_max < 0 &&
4243            check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4244          {          {
4245          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4246          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3609  we set the flag only if there is a liter Line 4262  we set the flag only if there is a liter
4262    
4263        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
4264    
4265          /*--------------------------------------------------------------------*/
4266          /* This code is obsolete from release 8.00; the restriction was finally
4267          removed: */
4268    
4269        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4270        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4271    
4272        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4273          /*--------------------------------------------------------------------*/
4274    
4275        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
4276    
# Line 3759  we set the flag only if there is a liter Line 4417  we set the flag only if there is a liter
4417          goto END_REPEAT;          goto END_REPEAT;
4418          }          }
4419    
4420          /*--------------------------------------------------------------------*/
4421          /* This code is obsolete from release 8.00; the restriction was finally
4422          removed: */
4423    
4424        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4425        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4426    
4427        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4428          /*--------------------------------------------------------------------*/
4429    
4430        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4431          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 3787  we set the flag only if there is a liter Line 4450  we set the flag only if there is a liter
4450        {        {
4451        register int i;        register int i;
4452        int ketoffset = 0;        int ketoffset = 0;
4453        int len = code - previous;        int len = (int)(code - previous);
4454        uschar *bralink = NULL;        uschar *bralink = NULL;
4455    
4456        /* Repeating a DEFINE group is pointless */        /* Repeating a DEFINE group is pointless */
# Line 3808  we set the flag only if there is a liter Line 4471  we set the flag only if there is a liter
4471          {          {
4472          register uschar *ket = previous;          register uschar *ket = previous;
4473          do ket += GET(ket, 1); while (*ket != OP_KET);          do ket += GET(ket, 1); while (*ket != OP_KET);
4474          ketoffset = code - ket;          ketoffset = (int)(code - ket);
4475          }          }
4476    
4477        /* The case of a zero minimum is special because of the need to stick        /* The case of a zero minimum is special because of the need to stick
# Line 3820  we set the flag only if there is a liter Line 4483  we set the flag only if there is a liter
4483    
4484        if (repeat_min == 0)        if (repeat_min == 0)
4485          {          {
4486          /* 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
4487          altogether. */          output altogether, like this:
4488    
4489          if (repeat_max == 0)          ** if (repeat_max == 0)
4490            {          **   {
4491            code = previous;          **   code = previous;
4492            goto END_REPEAT;          **   goto END_REPEAT;
4493            }          **   }
4494    
4495            However, that fails when a group is referenced as a subroutine from
4496            elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it
4497            so that it is skipped on execution. As we don't have a list of which
4498            groups are referenced, we cannot do this selectively.
4499    
4500            If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4501            and do no more at this point. However, we do need to adjust any
4502            OP_RECURSE calls inside the group that refer to the group itself or any
4503            internal or forward referenced group, because the offset is from the
4504            start of the whole regex. Temporarily terminate the pattern while doing
4505            this. */
4506    
4507          /* If the maximum is 1 or unlimited, we just have to stick in the          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
         BRAZERO and do no more at this point. However, we do need to adjust  
         any OP_RECURSE calls inside the group that refer to the group itself or  
         any internal or forward referenced group, because the offset is from  
         the start of the whole regex. Temporarily terminate the pattern while  
         doing this. */  
   
         if (repeat_max <= 1)  
4508            {            {
4509            *code = OP_END;            *code = OP_END;
4510            adjust_recurse(previous, 1, utf8, cd, save_hwm);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
4511            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
4512            code++;            code++;
4513              if (repeat_max == 0)
4514                {
4515                *previous++ = OP_SKIPZERO;
4516                goto END_REPEAT;
4517                }
4518            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4519            }            }
4520    
# Line 3866  we set the flag only if there is a liter Line 4539  we set the flag only if there is a liter
4539            /* We chain together the bracket offset fields that have to be            /* We chain together the bracket offset fields that have to be
4540            filled in later when the ends of the brackets are reached. */            filled in later when the ends of the brackets are reached. */
4541    
4542            offset = (bralink == NULL)? 0 : previous - bralink;            offset = (bralink == NULL)? 0 : (int)(previous - bralink);
4543            bralink = previous;            bralink = previous;
4544            PUTINC(previous, 0, offset);            PUTINC(previous, 0, offset);
4545            }            }
# Line 3887  we set the flag only if there is a liter Line 4560  we set the flag only if there is a liter
4560            {            {
4561            /* 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
4562            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
4563            potential integer overflow. */            potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
4564              integer type when available, otherwise double. */
4565    
4566            if (lengthptr != NULL)            if (lengthptr != NULL)
4567              {              {
4568              int delta = (repeat_min - 1)*length_prevgroup;              int delta = (repeat_min - 1)*length_prevgroup;
4569              if ((double)(repeat_min - 1)*(double)length_prevgroup >              if ((INT64_OR_DOUBLE)(repeat_min - 1)*
4570                                                              (double)INT_MAX ||                    (INT64_OR_DOUBLE)length_prevgroup >
4571                        (INT64_OR_DOUBLE)INT_MAX ||
4572                  OFLOW_MAX - *lengthptr < delta)                  OFLOW_MAX - *lengthptr < delta)
4573                {                {
4574                *errorcodeptr = ERR20;                *errorcodeptr = ERR20;
# Line 3939  we set the flag only if there is a liter Line 4614  we set the flag only if there is a liter
4614          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
4615          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
4616          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
4617          paranoid checks to avoid integer overflow. */          paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
4618            a 64-bit integer type when available, otherwise double. */
4619    
4620          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4621            {            {
4622            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4623                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4624            if ((double)repeat_max *            if ((INT64_OR_DOUBLE)repeat_max *
4625                  (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)                  (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4626                    > (double)INT_MAX ||                    > (INT64_OR_DOUBLE)INT_MAX ||
4627                OFLOW_MAX - *lengthptr < delta)                OFLOW_MAX - *lengthptr < delta)
4628              {              {
4629              *errorcodeptr = ERR20;              *errorcodeptr = ERR20;
# Line 3972  we set the flag only if there is a liter Line 4648  we set the flag only if there is a liter
4648              {              {
4649              int offset;              int offset;
4650              *code++ = OP_BRA;              *code++ = OP_BRA;
4651              offset = (bralink == NULL)? 0 : code - bralink;              offset = (bralink == NULL)? 0 : (int)(code - bralink);
4652              bralink = code;              bralink = code;
4653              PUTINC(code, 0, offset);              PUTINC(code, 0, offset);
4654              }              }
# Line 3993  we set the flag only if there is a liter Line 4669  we set the flag only if there is a liter
4669          while (bralink != NULL)          while (bralink != NULL)
4670            {            {
4671            int oldlinkoffset;            int oldlinkoffset;
4672            int offset = code - bralink + 1;            int offset = (int)(code - bralink + 1);
4673            uschar *bra = code - offset;            uschar *bra = code - offset;
4674            oldlinkoffset = GET(bra, 1);            oldlinkoffset = GET(bra, 1);
4675            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;
# Line 4024  we set the flag only if there is a liter Line 4700  we set the flag only if there is a liter
4700            uschar *scode = bracode;            uschar *scode = bracode;
4701            do            do
4702              {              {
4703              if (could_be_empty_branch(scode, ketcode, utf8))              if (could_be_empty_branch(scode, ketcode, utf8, cd))
4704                {                {
4705                *bracode += OP_SBRA - OP_BRA;                *bracode += OP_SBRA - OP_BRA;
4706                break;                break;
# Line 4036  we set the flag only if there is a liter Line 4712  we set the flag only if there is a liter
4712          }          }
4713        }        }
4714    
4715        /* If previous is OP_FAIL, it was generated by an empty class [] in
4716        JavaScript mode. The other ways in which OP_FAIL can be generated, that is
4717        by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
4718        error above. We can just ignore the repeat in JS case. */
4719    
4720        else if (*previous == OP_FAIL) goto END_REPEAT;
4721    
4722      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
4723    
4724      else      else
# Line 4060  we set the flag only if there is a liter Line 4743  we set the flag only if there is a liter
4743      if (possessive_quantifier)      if (possessive_quantifier)
4744        {        {
4745        int len;        int len;
4746        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||  
4747            *tempcode == OP_NOTEXACT)        if (*tempcode == OP_TYPEEXACT)
4748          tempcode += _pcre_OP_lengths[*tempcode] +          tempcode += _pcre_OP_lengths[*tempcode] +
4749            ((*tempcode == OP_TYPEEXACT &&            ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
4750               (tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP))? 2:0);  
4751        len = code - tempcode;        else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
4752            {
4753            tempcode += _pcre_OP_lengths[*tempcode];
4754    #ifdef SUPPORT_UTF8
4755            if (utf8 && tempcode[-1] >= 0xc0)
4756              tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];
4757    #endif
4758            }
4759    
4760          len = (int)(code - tempcode);
4761        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4762          {          {
4763          case OP_STAR:  *tempcode = OP_POSSTAR; break;          case OP_STAR:  *tempcode = OP_POSSTAR; break;
# Line 4083  we set the flag only if there is a liter Line 4775  we set the flag only if there is a liter
4775          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4776          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4777    
4778            /* Because we are moving code along, we must ensure that any
4779            pending recursive references are updated. */
4780    
4781          default:          default:
4782            *code = OP_END;
4783            adjust_recurse(tempcode, 1 + LINK_SIZE, utf8, cd, save_hwm);
4784          memmove(tempcode + 1+LINK_SIZE, tempcode, len);          memmove(tempcode + 1+LINK_SIZE, tempcode, len);
4785          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
4786          len += 1 + LINK_SIZE;          len += 1 + LINK_SIZE;
# Line 4110  we set the flag only if there is a liter Line 4807  we set the flag only if there is a liter
4807      lookbehind or option setting or condition or all the other extended      lookbehind or option setting or condition or all the other extended
4808      parenthesis forms.  */      parenthesis forms.  */
4809    
4810      case '(':      case CHAR_LEFT_PARENTHESIS:
4811      newoptions = options;      newoptions = options;
4812      skipbytes = 0;      skipbytes = 0;
4813      bravalue = OP_CBRA;      bravalue = OP_CBRA;
# Line 4119  we set the flag only if there is a liter Line 4816  we set the flag only if there is a liter
4816    
4817      /* First deal with various "verbs" that can be introduced by '*'. */      /* First deal with various "verbs" that can be introduced by '*'. */
4818    
4819      if (*(++ptr) == '*' && (cd->ctypes[ptr[1]] & ctype_letter) != 0)      if (*(++ptr) == CHAR_ASTERISK &&
4820             ((cd->ctypes[ptr[1]] & ctype_letter) != 0 || ptr[1] == ':'))
4821        {        {
4822        int i, namelen;        int i, namelen;
4823          int arglen = 0;
4824        const char *vn = verbnames;        const char *vn = verbnames;
4825        const uschar *name = ++ptr;        const uschar *name = ptr + 1;
4826          const uschar *arg = NULL;
4827        previous = NULL;        previous = NULL;
4828        while ((cd->ctypes[*++ptr] & ctype_letter) != 0);        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
4829        if (*ptr == ':')        namelen = (int)(ptr - name);
4830    
4831          if (*ptr == CHAR_COLON)
4832          {          {
4833          *errorcodeptr = ERR59;   /* Not supported */          arg = ++ptr;
4834          goto FAILED;          while ((cd->ctypes[*ptr] & (ctype_letter|ctype_digit)) != 0
4835              || *ptr == '_') ptr++;
4836            arglen = (int)(ptr - arg);
4837          }          }
4838        if (*ptr != ')')  
4839          if (*ptr != CHAR_RIGHT_PARENTHESIS)
4840          {          {
4841          *errorcodeptr = ERR60;          *errorcodeptr = ERR60;
4842          goto FAILED;          goto FAILED;
4843          }          }
4844        namelen = ptr - name;  
4845          /* Scan the table of verb names */
4846    
4847        for (i = 0; i < verbcount; i++)        for (i = 0; i < verbcount; i++)
4848          {          {
4849          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
4850              strncmp((char *)name, vn, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
4851            {            {
4852            *code = verbs[i].op;            /* Check for open captures before ACCEPT */
4853            if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;  
4854            break;            if (verbs[i].op == OP_ACCEPT)
4855                {
4856                open_capitem *oc;
4857                cd->had_accept = TRUE;
4858                for (oc = cd->open_caps; oc != NULL; oc = oc->next)
4859                  {
4860                  *code++ = OP_CLOSE;
4861                  PUT2INC(code, 0, oc->number);
4862                  }
4863                }
4864    
4865              /* Handle the cases with/without an argument */
4866    
4867              if (arglen == 0)
4868                {
4869                if (verbs[i].op < 0)   /* Argument is mandatory */
4870                  {
4871                  *errorcodeptr = ERR66;
4872                  goto FAILED;
4873                  }
4874                *code = verbs[i].op;
4875                if (*code++ == OP_THEN)
4876                  {
4877                  PUT(code, 0, code - bcptr->current_branch - 1);
4878                  code += LINK_SIZE;
4879                  }
4880                }
4881    
4882              else
4883                {
4884                if (verbs[i].op_arg < 0)   /* Argument is forbidden */
4885                  {
4886                  *errorcodeptr = ERR59;
4887                  goto FAILED;
4888                  }
4889                *code = verbs[i].op_arg;
4890                if (*code++ == OP_THEN_ARG)
4891                  {
4892                  PUT(code, 0, code - bcptr->current_branch - 1);
4893                  code += LINK_SIZE;
4894                  }
4895                *code++ = arglen;
4896                memcpy(code, arg, arglen);
4897                code += arglen;
4898                *code++ = 0;
4899                }
4900    
4901              break;  /* Found verb, exit loop */
4902            }            }
4903    
4904          vn += verbs[i].len + 1;          vn += verbs[i].len + 1;
4905          }          }
4906        if (i < verbcount) continue;  
4907        *errorcodeptr = ERR60;        if (i < verbcount) continue;    /* Successfully handled a verb */
4908          *errorcodeptr = ERR60;          /* Verb not recognized */
4909        goto FAILED;        goto FAILED;
4910        }        }
4911    
4912      /* Deal with the extended parentheses; all are introduced by '?', and the      /* Deal with the extended parentheses; all are introduced by '?', and the
4913      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. */
4914    
4915      else if (*ptr == '?')      else if (*ptr == CHAR_QUESTION_MARK)
4916        {        {
4917        int i, set, unset, namelen;        int i, set, unset, namelen;
4918        int *optset;        int *optset;
# Line 4165  we set the flag only if there is a liter Line 4921  we set the flag only if there is a liter
4921    
4922        switch (*(++ptr))        switch (*(++ptr))
4923          {          {
4924          case '#':                 /* Comment; skip to ket */          case CHAR_NUMBER_SIGN:                 /* Comment; skip to ket */
4925          ptr++;          ptr++;
4926          while (*ptr != 0 && *ptr != ')') ptr++;          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
4927          if (*ptr == 0)          if (*ptr == 0)
4928            {            {
4929            *errorcodeptr = ERR18;            *errorcodeptr = ERR18;
# Line 4177  we set the flag only if there is a liter Line 4933  we set the flag only if there is a liter
4933    
4934    
4935          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4936          case '|':                 /* Reset capture count for each branch */          case CHAR_VERTICAL_LINE:  /* Reset capture count for each branch */
4937          reset_bracount = TRUE;          reset_bracount = TRUE;
4938          /* Fall through */          /* Fall through */
4939    
4940          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4941          case ':':                 /* Non-capturing bracket */          case CHAR_COLON:          /* Non-capturing bracket */
4942          bravalue = OP_BRA;          bravalue = OP_BRA;
4943          ptr++;          ptr++;
4944          break;          break;
4945    
4946    
4947          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4948          case '(':          case CHAR_LEFT_PARENTHESIS:
4949          bravalue = OP_COND;       /* Conditional group */          bravalue = OP_COND;       /* Conditional group */
4950    
4951          /* 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 4209  we set the flag only if there is a liter Line 4965  we set the flag only if there is a liter
4965          the switch. This will take control down to where bracketed groups,          the switch. This will take control down to where bracketed groups,
4966          including assertions, are processed. */          including assertions, are processed. */
4967    
4968          if (ptr[1] == '?' && (ptr[2] == '=' || ptr[2] == '!' || ptr[2] == '<'))          if (ptr[1] == CHAR_QUESTION_MARK && (ptr[2] == CHAR_EQUALS_SIGN ||
4969                ptr[2] == CHAR_EXCLAMATION_MARK || ptr[2] == CHAR_LESS_THAN_SIGN))
4970            break;            break;
4971    
4972          /* 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 4221  we set the flag only if there is a liter Line 4978  we set the flag only if there is a liter
4978    
4979          /* Check for a test for recursion in a named group. */          /* Check for a test for recursion in a named group. */
4980    
4981          if (ptr[1] == 'R' && ptr[2] == '&')          if (ptr[1] == CHAR_R && ptr[2] == CHAR_AMPERSAND)
4982            {            {
4983            terminator = -1;            terminator = -1;
4984            ptr += 2;            ptr += 2;
# Line 4231  we set the flag only if there is a liter Line 4988  we set the flag only if there is a liter
4988          /* 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
4989          syntax (?(<name>) or (?('name') */          syntax (?(<name>) or (?('name') */
4990    
4991          else if (ptr[1] == '<')          else if (ptr[1] == CHAR_LESS_THAN_SIGN)
4992            {            {
4993            terminator = '>';            terminator = CHAR_GREATER_THAN_SIGN;
4994            ptr++;            ptr++;
4995            }            }
4996          else if (ptr[1] == '\'')          else if (ptr[1] == CHAR_APOSTROPHE)
4997            {            {
4998            terminator = '\'';            terminator = CHAR_APOSTROPHE;
4999            ptr++;            ptr++;
5000            }            }
5001          else          else
5002            {            {
5003            terminator = 0;            terminator = 0;
5004            if (ptr[1] == '-' || ptr[1] == '+') refsign = *(++ptr);            if (ptr[1] == CHAR_MINUS || ptr[1] == CHAR_PLUS) refsign = *(++ptr);
5005            }            }
5006    
5007          /* 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 4264  we set the flag only if there is a liter Line 5021  we set the flag only if there is a liter
5021            {            {
5022            if (recno >= 0)            if (recno >= 0)
5023              recno = ((digitab[*ptr] & ctype_digit) != 0)?              recno = ((digitab[*ptr] & ctype_digit) != 0)?
5024                recno * 10 + *ptr - '0' : -1;                recno * 10 + *ptr - CHAR_0 : -1;
5025            ptr++;            ptr++;
5026            }            }
5027          namelen = ptr - name;          namelen = (int)(ptr - name);
5028    
5029          if ((terminator > 0 && *ptr++ != terminator) || *ptr++ != ')')          if ((terminator > 0 && *ptr++ != terminator) ||
5030                *ptr++ != CHAR_RIGHT_PARENTHESIS)
5031            {            {
5032            ptr--;      /* Error offset */            ptr--;      /* Error offset */
5033            *errorcodeptr = ERR26;            *errorcodeptr = ERR26;
# Line 4291  we set the flag only if there is a liter Line 5049  we set the flag only if there is a liter
5049              *errorcodeptr = ERR58;              *errorcodeptr = ERR58;
5050              goto FAILED;              goto FAILED;
5051              }              }
5052            recno = (refsign == '-')?            recno = (refsign == CHAR_MINUS)?
5053              cd->bracount - recno + 1 : recno +cd->bracount;              cd->bracount - recno + 1 : recno +cd->bracount;
5054            if (recno <= 0 || recno > cd->final_bracount)            if (recno <= 0 || recno > cd->final_bracount)
5055              {              {
# Line 4303  we set the flag only if there is a liter Line 5061  we set the flag only if there is a liter
5061            }            }
5062    
5063          /* Otherwise (did not start with "+" or "-"), start by looking for the          /* Otherwise (did not start with "+" or "-"), start by looking for the
5064          name. */          name. If we find a name, add one to the opcode to change OP_CREF or
5065            OP_RREF into OP_NCREF or OP_NRREF. These behave exactly the same,
5066            except they record that the reference was originally to a name. The
5067            information is used to check duplicate names. */
5068    
5069          slot = cd->name_table;          slot = cd->name_table;
5070          for (i = 0; i < cd->names_found; i++)          for (i = 0; i < cd->names_found; i++)
# Line 4318  we set the flag only if there is a liter Line 5079  we set the flag only if there is a liter
5079            {            {
5080            recno = GET2(slot, 0);            recno = GET2(slot, 0);
5081            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
5082              code[1+LINK_SIZE]++;
5083            }            }
5084    
5085          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
5086    
5087          else if ((i = find_parens(ptr, cd->bracount, name, namelen,          else if ((i = find_parens(cd, name, namelen,
5088                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0, utf8)) > 0)
5089            {            {
5090            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
5091              code[1+LINK_SIZE]++;
5092            }            }
5093    
5094          /* If terminator == 0 it means that the name followed directly after          /* If terminator == 0 it means that the name followed directly after
# Line 4343  we set the flag only if there is a liter Line 5106  we set the flag only if there is a liter
5106          /* Check for (?(R) for recursion. Allow digits after R to specify a          /* Check for (?(R) for recursion. Allow digits after R to specify a
5107          specific group number. */          specific group number. */
5108    
5109          else if (*name == 'R')          else if (*name == CHAR_R)
5110            {            {
5111            recno = 0;            recno = 0;
5112            for (i = 1; i < namelen; i++)            for (i = 1; i < namelen; i++)
# Line 4353  we set the flag only if there is a liter Line 5116  we set the flag only if there is a liter
5116                *errorcodeptr = ERR15;                *errorcodeptr = ERR15;
5117                goto FAILED;                goto FAILED;
5118                }                }
5119              recno = recno * 10 + name[i] - '0';              recno = recno * 10 + name[i] - CHAR_0;
5120              }              }
5121            if (recno == 0) recno = RREF_ANY;            if (recno == 0) recno = RREF_ANY;
5122            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */
# Line 4363  we set the flag only if there is a liter Line 5126  we set the flag only if there is a liter
5126          /* Similarly, check for the (?(DEFINE) "condition", which is always          /* Similarly, check for the (?(DEFINE) "condition", which is always
5127          false. */          false. */
5128    
5129          else if (namelen == 6 && strncmp((char *)name, "DEFINE", 6) == 0)          else if (namelen == 6 && strncmp((char *)name, STRING_DEFINE, 6) == 0)
5130            {            {
5131            code[1+LINK_SIZE] = OP_DEF;            code[1+LINK_SIZE] = OP_DEF;
5132            skipbytes = 1;            skipbytes = 1;
# Line 4388  we set the flag only if there is a liter Line 5151  we set the flag only if there is a liter
5151    
5152    
5153          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5154          case '=':                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
5155          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
5156          ptr++;          ptr++;
5157          break;          break;
5158    
5159    
5160          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5161          case '!':                 /* Negative lookahead */          case CHAR_EXCLAMATION_MARK:            /* Negative lookahead */
5162          ptr++;          ptr++;
5163          if (*ptr == ')')          /* Optimize (?!) */          if (*ptr == CHAR_RIGHT_PARENTHESIS)    /* Optimize (?!) */
5164            {            {
5165            *code++ = OP_FAIL;            *code++ = OP_FAIL;
5166            previous = NULL;            previous = NULL;
# Line 4408  we set the flag only if there is a liter Line 5171  we set the flag only if there is a liter
5171    
5172    
5173          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5174          case '<':                 /* Lookbehind or named define */          case CHAR_LESS_THAN_SIGN:              /* Lookbehind or named define */
5175          switch (ptr[1])          switch (ptr[1])
5176            {            {
5177            case '=':               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
5178            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
5179            ptr += 2;            ptr += 2;
5180            break;            break;
5181    
5182            case '!':               /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
5183            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
5184            ptr += 2;            ptr += 2;
5185            break;            break;
# Line 4431  we set the flag only if there is a liter Line 5194  we set the flag only if there is a liter
5194    
5195    
5196          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5197          case '>':                 /* One-time brackets */          case CHAR_GREATER_THAN_SIGN:           /* One-time brackets */
5198          bravalue = OP_ONCE;          bravalue = OP_ONCE;
5199          ptr++;          ptr++;
5200          break;          break;
5201    
5202    
5203          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5204          case 'C':                 /* Callout - may be followed by digits; */          case CHAR_C:                 /* Callout - may be followed by digits; */
5205          previous_callout = code;  /* Save for later completion */          previous_callout = code;  /* Save for later completion */
5206          after_manual_callout = 1; /* Skip one item before completing */          after_manual_callout = 1; /* Skip one item before completing */
5207          *code++ = OP_CALLOUT;          *code++ = OP_CALLOUT;
5208            {            {
5209            int n = 0;            int n = 0;
5210            while ((digitab[*(++ptr)] & ctype_digit) != 0)            while ((digitab[*(++ptr)] & ctype_digit) != 0)
5211              n = n * 10 + *ptr - '0';              n = n * 10 + *ptr - CHAR_0;
5212            if (*ptr != ')')            if (*ptr != CHAR_RIGHT_PARENTHESIS)
5213              {              {
5214              *errorcodeptr = ERR39;              *errorcodeptr = ERR39;
5215              goto FAILED;              goto FAILED;
# Line 4457  we set the flag only if there is a liter Line 5220  we set the flag only if there is a liter
5220              goto FAILED;              goto FAILED;
5221              }              }
5222            *code++ = n;            *code++ = n;
5223            PUT(code, 0, ptr - cd->start_pattern + 1);  /* Pattern offset */            PUT(code, 0, (int)(ptr - cd->start_pattern + 1)); /* Pattern offset */
5224            PUT(code, LINK_SIZE, 0);                    /* Default length */            PUT(code, LINK_SIZE, 0);                          /* Default length */
5225            code += 2 * LINK_SIZE;            code += 2 * LINK_SIZE;
5226            }            }
5227          previous = NULL;          previous = NULL;
# Line 4466  we set the flag only if there is a liter Line 5229  we set the flag only if there is a liter
5229    
5230    
5231          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5232          case 'P':                 /* Python-style named subpattern handling */          case CHAR_P:              /* Python-style named subpattern handling */
5233          if (*(++ptr) == '=' || *ptr == '>')  /* Reference or recursion */          if (*(++ptr) == CHAR_EQUALS_SIGN ||
5234                *ptr == CHAR_GREATER_THAN_SIGN)  /* Reference or recursion */
5235            {            {
5236            is_recurse = *ptr == '>';            is_recurse = *ptr == CHAR_GREATER_THAN_SIGN;
5237            terminator = ')';            terminator = CHAR_RIGHT_PARENTHESIS;
5238            goto NAMED_REF_OR_RECURSE;            goto NAMED_REF_OR_RECURSE;
5239            }            }
5240          else if (*ptr != '<')    /* Test for Python-style definition */          else if (*ptr != CHAR_LESS_THAN_SIGN)  /* Test for Python-style defn */
5241            {            {
5242            *errorcodeptr = ERR41;            *errorcodeptr = ERR41;
5243            goto FAILED;            goto FAILED;
# Line 4483  we set the flag only if there is a liter Line 5247  we set the flag only if there is a liter
5247    
5248          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5249          DEFINE_NAME:    /* Come here from (?< handling */          DEFINE_NAME:    /* Come here from (?< handling */
5250          case '\'':          case CHAR_APOSTROPHE:
5251            {            {
5252            terminator = (*ptr == '<')? '>' : '\'';            terminator = (*ptr == CHAR_LESS_THAN_SIGN)?
5253                CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
5254            name = ++ptr;            name = ++ptr;
5255    
5256            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
5257            namelen = ptr - name;            namelen = (int)(ptr - name);
5258    
5259            /* In the pre-compile phase, just do a syntax check. */            /* In the pre-compile phase, just do a syntax check. */
5260    
# Line 4516  we set the flag only if there is a liter Line 5281  we set the flag only if there is a liter
5281                }                }
5282              }              }
5283    
5284            /* In the real compile, create the entry in the table */            /* In the real compile, create the entry in the table, maintaining
5285              alphabetical order. Duplicate names for different numbers are
5286              permitted only if PCRE_DUPNAMES is set. Duplicate names for the same
5287              number are always OK. (An existing number can be re-used if (?|
5288              appears in the pattern.) In either event, a duplicate name results in
5289              a duplicate entry in the table, even if the number is the same. This
5290              is because the number of names, and hence the table size, is computed
5291              in the pre-compile, and it affects various numbers and pointers which
5292              would all have to be modified, and the compiled code moved down, if
5293              duplicates with the same number were omitted from the table. This
5294              doesn't seem worth the hassle. However, *different* names for the
5295              same number are not permitted. */
5296    
5297            else            else
5298              {              {
5299                BOOL dupname = FALSE;
5300              slot = cd->name_table;              slot = cd->name_table;
5301    
5302              for (i = 0; i < cd->names_found; i++)              for (i = 0; i < cd->names_found; i++)
5303                {                {
5304                int crc = memcmp(name, slot+2, namelen);                int crc = memcmp(name, slot+2, namelen);
# Line 4528  we set the flag only if there is a liter Line 5306  we set the flag only if there is a liter
5306                  {                  {
5307                  if (slot[2+namelen] == 0)                  if (slot[2+namelen] == 0)
5308                    {                    {
5309                    if ((options & PCRE_DUPNAMES) == 0)                    if (GET2(slot, 0) != cd->bracount + 1 &&
5310                          (options & PCRE_DUPNAMES) == 0)
5311                      {                      {
5312                      *errorcodeptr = ERR43;                      *errorcodeptr = ERR43;
5313                      goto FAILED;                      goto FAILED;
5314                      }                      }
5315                      else dupname = TRUE;
5316                    }                    }
5317                  else crc = -1;      /* Current name is substring */                  else crc = -1;      /* Current name is a substring */
5318                  }                  }
5319    
5320                  /* Make space in the table and break the loop for an earlier
5321                  name. For a duplicate or later name, carry on. We do this for
5322                  duplicates so that in the simple case (when ?(| is not used) they
5323                  are in order of their numbers. */
5324    
5325                if (crc < 0)                if (crc < 0)
5326                  {                  {
5327                  memmove(slot + cd->name_entry_size, slot,                  memmove(slot + cd->name_entry_size, slot,
5328                    (cd->names_found - i) * cd->name_entry_size);                    (cd->names_found - i) * cd->name_entry_size);
5329                  break;                  break;
5330                  }                  }
5331    
5332                  /* Continue the loop for a later or duplicate name */
5333    
5334                slot += cd->name_entry_size;                slot += cd->name_entry_size;
5335                }                }
5336    
5337                /* For non-duplicate names, check for a duplicate number before
5338                adding the new name. */
5339    
5340                if (!dupname)
5341                  {
5342                  uschar *cslot = cd->name_table;
5343                  for (i = 0; i < cd->names_found; i++)
5344                    {
5345                    if (cslot != slot)
5346                      {
5347                      if (GET2(cslot, 0) == cd->bracount + 1)
5348                        {
5349                        *errorcodeptr = ERR65;
5350                        goto FAILED;
5351                        }
5352                      }
5353                    else i--;
5354                    cslot += cd->name_entry_size;
5355                    }
5356                  }
5357    
5358              PUT2(slot, 0, cd->bracount + 1);              PUT2(slot, 0, cd->bracount + 1);
5359              memcpy(slot + 2, name, namelen);              memcpy(slot + 2, name, namelen);
5360              slot[2+namelen] = 0;              slot[2+namelen] = 0;
5361              }              }
5362            }            }
5363    
5364          /* In both cases, count the number of names we've encountered. */          /* In both pre-compile and compile, count the number of names we've
5365            encountered. */
5366    
         ptr++;                    /* Move past > or ' */  
5367          cd->names_found++;          cd->names_found++;
5368            ptr++;                    /* Move past > or ' */
5369          goto NUMBERED_GROUP;          goto NUMBERED_GROUP;
5370    
5371    
5372          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5373          case '&':                 /* Perl recursion/subroutine syntax */          case CHAR_AMPERSAND:            /* Perl recursion/subroutine syntax */
5374          terminator = ')';          terminator = CHAR_RIGHT_PARENTHESIS;
5375          is_recurse = TRUE;          is_recurse = TRUE;
5376          /* Fall through */          /* Fall through */
5377    
# Line 4568  we set the flag only if there is a liter Line 5379  we set the flag only if there is a liter
5379          references (?P=name) and recursion (?P>name), as well as falling          references (?P=name) and recursion (?P>name), as well as falling
5380          through from the Perl recursion syntax (?&name). We also come here from          through from the Perl recursion syntax (?&name). We also come here from
5381          the Perl \k<name> or \k'name' back reference syntax and the \k{name}          the Perl \k<name> or \k'name' back reference syntax and the \k{name}
5382          .NET syntax. */          .NET syntax, and the Oniguruma \g<...> and \g'...' subroutine syntax. */
5383    
5384          NAMED_REF_OR_RECURSE:          NAMED_REF_OR_RECURSE:
5385          name = ++ptr;          name = ++ptr;
5386          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
5387          namelen = ptr - name;          namelen = (int)(ptr - name);
5388    
5389          /* In the pre-compile phase, do a syntax check and set a dummy          /* In the pre-compile phase, do a syntax check. We used to just set
5390          reference number. */          a dummy reference number, because it was not used in the first pass.
5391            However, with the change of recursive back references to be atomic,
5392            we have to look for the number so that this state can be identified, as
5393            otherwise the incorrect length is computed. If it's not a backwards
5394            reference, the dummy number will do. */
5395    
5396          if (lengthptr != NULL)          if (lengthptr != NULL)
5397            {            {
5398              const uschar *temp;
5399    
5400            if (namelen == 0)            if (namelen == 0)
5401              {              {
5402              *errorcodeptr = ERR62;              *errorcodeptr = ERR62;
# Line 4595  we set the flag only if there is a liter Line 5412  we set the flag only if there is a liter
5412              *errorcodeptr = ERR48;              *errorcodeptr = ERR48;
5413              goto FAILED;              goto FAILED;
5414              }              }
5415            recno = 0;  
5416              /* The name table does not exist in the first pass, so we cannot
5417              do a simple search as in the code below. Instead, we have to scan the
5418              pattern to find the number. It is important that we scan it only as
5419              far as we have got because the syntax of named subpatterns has not
5420              been checked for the rest of the pattern, and find_parens() assumes
5421              correct syntax. In any case, it's a waste of resources to scan
5422              further. We stop the scan at the current point by temporarily
5423              adjusting the value of cd->endpattern. */
5424    
5425              temp = cd->end_pattern;
5426              cd->end_pattern = ptr;
5427              recno = find_parens(cd, name, namelen,
5428                (options & PCRE_EXTENDED) != 0, utf8);
5429              cd->end_pattern = temp;
5430              if (recno < 0) recno = 0;    /* Forward ref; set dummy number */
5431            }            }
5432    
5433          /* In the real compile, seek the name in the table. We check the name          /* In the real compile, seek the name in the table. We check the name
# Line 4619  we set the flag only if there is a liter Line 5451  we set the flag only if there is a liter
5451              recno = GET2(slot, 0);              recno = GET2(slot, 0);
5452              }              }
5453            else if ((recno =                /* Forward back reference */            else if ((recno =                /* Forward back reference */
5454                      find_parens(ptr, cd->bracount, name, namelen,                      find_parens(cd, name, namelen,
5455                        (options & PCRE_EXTENDED) != 0)) <= 0)                        (options & PCRE_EXTENDED) != 0, utf8)) <= 0)
5456              {              {
5457              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
5458              goto FAILED;              goto FAILED;
# Line 4635  we set the flag only if there is a liter Line 5467  we set the flag only if there is a liter
5467    
5468    
5469          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5470          case 'R':                 /* Recursion */          case CHAR_R:              /* Recursion */
5471          ptr++;                    /* Same as (?0)      */          ptr++;                    /* Same as (?0)      */
5472          /* Fall through */          /* Fall through */
5473    
5474    
5475          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5476          case '-': case '+':          case CHAR_MINUS: case CHAR_PLUS:  /* Recursion or subroutine */
5477          case '0': case '1': case '2': case '3': case '4':   /* Recursion or */          case CHAR_0: case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4:
5478          case '5': case '6': case '7': case '8': case '9':   /* subroutine */          case CHAR_5: case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
5479            {            {
5480            const uschar *called;            const uschar *called;
5481              terminator = CHAR_RIGHT_PARENTHESIS;
5482    
5483            if ((refsign = *ptr) == '+')            /* Come here from the \g<...> and \g'...' code (Oniguruma
5484              compatibility). However, the syntax has been checked to ensure that
5485              the ... are a (signed) number, so that neither ERR63 nor ERR29 will
5486              be called on this path, nor with the jump to OTHER_CHAR_AFTER_QUERY
5487              ever be taken. */
5488    
5489              HANDLE_NUMERICAL_RECURSION:
5490    
5491              if ((refsign = *ptr) == CHAR_PLUS)
5492              {              {
5493              ptr++;              ptr++;
5494              if ((digitab[*ptr] & ctype_digit) == 0)              if ((digitab[*ptr] & ctype_digit) == 0)
# Line 4656  we set the flag only if there is a liter Line 5497  we set the flag only if there is a liter
5497                goto FAILED;                goto FAILED;
5498                }                }
5499              }              }
5500            else if (refsign == '-')            else if (refsign == CHAR_MINUS)
5501              {              {
5502              if ((digitab[ptr[1]] & ctype_digit) == 0)              if ((digitab[ptr[1]] & ctype_digit) == 0)
5503                goto OTHER_CHAR_AFTER_QUERY;                goto OTHER_CHAR_AFTER_QUERY;
# Line 4665  we set the flag only if there is a liter Line 5506  we set the flag only if there is a liter
5506    
5507            recno = 0;            recno = 0;
5508            while((digitab[*ptr] & ctype_digit) != 0)            while((digitab[*ptr] & ctype_digit) != 0)
5509              recno = recno * 10 + *ptr++ - '0';              recno = recno * 10 + *ptr++ - CHAR_0;
5510    
5511            if (*ptr != ')')            if (*ptr != terminator)
5512              {              {
5513              *errorcodeptr = ERR29;              *errorcodeptr = ERR29;
5514              goto FAILED;              goto FAILED;
5515              }              }
5516    
5517            if (refsign == '-')            if (refsign == CHAR_MINUS)
5518              {              {
5519              if (recno == 0)              if (recno == 0)
5520                {                {
# Line 4687  we set the flag only if there is a liter Line 5528  we set the flag only if there is a liter
5528                goto FAILED;                goto FAILED;
5529                }                }
5530              }              }
5531            else if (refsign == '+')            else if (refsign == CHAR_PLUS)
5532              {              {
5533              if (recno == 0)              if (recno == 0)
5534                {                {
# Line 4714  we set the flag only if there is a liter Line 5555  we set the flag only if there is a liter
5555            if (lengthptr == NULL)            if (lengthptr == NULL)
5556              {              {
5557              *code = OP_END;              *code = OP_END;
5558              if (recno != 0) called = find_bracket(cd->start_code, utf8, recno);              if (recno != 0)
5559                  called = _pcre_find_bracket(cd->start_code, utf8, recno);
5560    
5561              /* Forward reference */              /* Forward reference */
5562    
5563              if (called == NULL)              if (called == NULL)
5564                {                {
5565                if (find_parens(ptr, cd->bracount, NULL, recno,                if (find_parens(cd, NULL, recno,
5566                     (options & PCRE_EXTENDED) != 0) < 0)                      (options & PCRE_EXTENDED) != 0, utf8) < 0)
5567                  {                  {
5568                  *errorcodeptr = ERR15;                  *errorcodeptr = ERR15;
5569                  goto FAILED;                  goto FAILED;
5570                  }                  }
5571    
5572                  /* Fudge the value of "called" so that when it is inserted as an
5573                  offset below, what it actually inserted is the reference number
5574                  of the group. */
5575    
5576                called = cd->start_code + recno;                called = cd->start_code + recno;
5577                PUTINC(cd->hwm, 0, code + 2 + LINK_SIZE - cd->start_code);                PUTINC(cd->hwm, 0, (int)(code + 2 + LINK_SIZE - cd->start_code));
5578                }                }
5579    
5580              /* If not a forward reference, and the subpattern is still open,              /* If not a forward reference, and the subpattern is still open,
# Line 4735  we set the flag only if there is a liter Line 5582  we set the flag only if there is a liter
5582              recursion that could loop for ever, and diagnose that case. */              recursion that could loop for ever, and diagnose that case. */
5583    
5584              else if (GET(called, 1) == 0 &&              else if (GET(called, 1) == 0 &&
5585                       could_be_empty(called, code, bcptr, utf8))                       could_be_empty(called, code, bcptr, utf8, cd))
5586                {                {
5587                *errorcodeptr = ERR40;                *errorcodeptr = ERR40;
5588                goto FAILED;                goto FAILED;
# Line 4751  we set the flag only if there is a liter Line 5598  we set the flag only if there is a liter
5598            code += 1 + LINK_SIZE;            code += 1 + LINK_SIZE;
5599