/[pcre]/code/trunk/pcre_compile.c
ViewVC logotype

Diff of /code/trunk/pcre_compile.c

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

revision 274 by ph10, Tue Nov 20 10:05:23 2007 UTC revision 461 by ph10, Mon Oct 5 10:59:35 2009 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2007 University of Cambridge             Copyright (c) 1997-2009 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 97  are simple data values; negative values Line 97  are simple data values; negative values
97  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
98  is invalid. */  is invalid. */
99    
100  #ifndef EBCDIC  /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC
101    
102    /* This is the "normal" table for ASCII systems or for EBCDIC systems running
103    in UTF-8 mode. */
104    
105  static const short int escapes[] = {  static const short int escapes[] = {
106       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,                       0,
107       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,                       0,
108     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */       0,                       0,
109  -ESC_H,      0,      0, -ESC_K,      0,      0,      0,      0,   /* H - O */       0,                       0,
110  -ESC_P, -ESC_Q, -ESC_R, -ESC_S,      0,      0, -ESC_V, -ESC_W,   /* P - W */       0,                       0,
111  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */       CHAR_COLON,              CHAR_SEMICOLON,
112     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
113  -ESC_h,      0,      0, -ESC_k,      0,      0,  ESC_n,      0,   /* h - o */       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
114  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0, -ESC_v, -ESC_w,   /* p - w */       CHAR_COMMERCIAL_AT,      -ESC_A,
115       0,      0, -ESC_z                                            /* x - z */       -ESC_B,                  -ESC_C,
116         -ESC_D,                  -ESC_E,
117         0,                       -ESC_G,
118         -ESC_H,                  0,
119         0,                       -ESC_K,
120         0,                       0,
121         0,                       0,
122         -ESC_P,                  -ESC_Q,
123         -ESC_R,                  -ESC_S,
124         0,                       0,
125         -ESC_V,                  -ESC_W,
126         -ESC_X,                  0,
127         -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,
128         CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,
129         CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,
130         CHAR_GRAVE_ACCENT,       7,
131         -ESC_b,                  0,
132         -ESC_d,                  ESC_e,
133         ESC_f,                   0,
134         -ESC_h,                  0,
135         0,                       -ESC_k,
136         0,                       0,
137         ESC_n,                   0,
138         -ESC_p,                  0,
139         ESC_r,                   -ESC_s,
140         ESC_tee,                 0,
141         -ESC_v,                  -ESC_w,
142         0,                       0,
143         -ESC_z
144  };  };
145    
146  #else           /* This is the "abnormal" table for EBCDIC systems */  #else
147    
148    /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
149    
150  static const short int escapes[] = {  static const short int escapes[] = {
151  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
152  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 142  static const short int escapes[] = { Line 177  static const short int escapes[] = {
177    
178  /* 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
179  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
180  the number of relocations when a shared library is dynamically linked. */  the number of relocations when a shared library is dynamically linked. The
181    string is built from string macros so that it works in UTF-8 mode on EBCDIC
182    platforms. */
183    
184  typedef struct verbitem {  typedef struct verbitem {
185    int   len;    int   len;
# Line 150  typedef struct verbitem { Line 187  typedef struct verbitem {
187  } verbitem;  } verbitem;
188    
189  static const char verbnames[] =  static const char verbnames[] =
190    "ACCEPT\0"    STRING_ACCEPT0
191    "COMMIT\0"    STRING_COMMIT0
192    "F\0"    STRING_F0
193    "FAIL\0"    STRING_FAIL0
194    "PRUNE\0"    STRING_PRUNE0
195    "SKIP\0"    STRING_SKIP0
196    "THEN";    STRING_THEN;
197    
198  static verbitem verbs[] = {  static const verbitem verbs[] = {
199    { 6, OP_ACCEPT },    { 6, OP_ACCEPT },
200    { 6, OP_COMMIT },    { 6, OP_COMMIT },
201    { 1, OP_FAIL },    { 1, OP_FAIL },
# Line 168  static verbitem verbs[] = { Line 205  static verbitem verbs[] = {
205    { 4, OP_THEN  }    { 4, OP_THEN  }
206  };  };
207    
208  static int verbcount = sizeof(verbs)/sizeof(verbitem);  static const int verbcount = sizeof(verbs)/sizeof(verbitem);
209    
210    
211  /* 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 215  length entry. The first three must be al
215  for handling case independence. */  for handling case independence. */
216    
217  static const char posix_names[] =  static const char posix_names[] =
218    "alpha\0"  "lower\0"  "upper\0"  "alnum\0"  "ascii\0"  "blank\0"    STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
219    "cntrl\0"  "digit\0"  "graph\0"  "print\0"  "punct\0"  "space\0"    STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
220    "word\0"   "xdigit";    STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
221      STRING_word0  STRING_xdigit;
222    
223  static const uschar posix_name_lengths[] = {  static const uschar posix_name_lengths[] = {
224    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 295  static const char error_texts[] = Line 333  static const char error_texts[] =
333    /* 55 */    /* 55 */
334    "repeating a DEFINE group is not allowed\0"    "repeating a DEFINE group is not allowed\0"
335    "inconsistent NEWLINE options\0"    "inconsistent NEWLINE options\0"
336    "\\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"
337    "(?+ or (?- or (?(+ or (?(- must be followed by a non-zero number\0"    "a numbered reference must not be zero\0"
338    "(*VERB) with an argument is not supported\0"    "(*VERB) with an argument is not supported\0"
339    /* 60 */    /* 60 */
340    "(*VERB) not recognized\0"    "(*VERB) not recognized\0"
341    "number is too big\0"    "number is too big\0"
342    "subpattern name expected\0"    "subpattern name expected\0"
343    "digit expected after (?+";    "digit expected after (?+\0"
344      "] is an invalid data character in JavaScript compatibility mode\0"
345      /* 65 */
346      "different names for subpatterns of the same number are not allowed";
347    
348    
349  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
# Line 321  For convenience, we use the same bit def Line 362  For convenience, we use the same bit def
362    
363  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
364    
365  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC
366    
367    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
368    UTF-8 mode. */
369    
370  static const unsigned char digitab[] =  static const unsigned char digitab[] =
371    {    {
372    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 402  static const unsigned char digitab[] =
402    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
403    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
404    
405  #else           /* This is the "abnormal" case, for EBCDIC systems */  #else
406    
407    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
408    
409  static const unsigned char digitab[] =  static const unsigned char digitab[] =
410    {    {
411    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 502  static const char *
502  find_error_text(int n)  find_error_text(int n)
503  {  {
504  const char *s = error_texts;  const char *s = error_texts;
505  for (; n > 0; n--) while (*s++ != 0);  for (; n > 0; n--) while (*s++ != 0) {};
506  return s;  return s;
507  }  }
508    
# Line 502  if (c == 0) *errorcodeptr = ERR1; Line 550  if (c == 0) *errorcodeptr = ERR1;
550  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.
551  Otherwise further processing may be required. */  Otherwise further processing may be required. */
552    
553  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
554  else if (c < '0' || c > 'z') {}                           /* Not alphanumeric */  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */
555  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0) c = i;
556    
557  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
558  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 571  else
571      /* 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
572      error. */      error. */
573    
574      case 'l':      case CHAR_l:
575      case 'L':      case CHAR_L:
576      case 'N':      case CHAR_N:
577      case 'u':      case CHAR_u:
578      case 'U':      case CHAR_U:
579      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
580      break;      break;
581    
582      /* \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:
583      is an absolute backreference. If negative, it is a relative backreference.  
584      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
585      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
586      unified syntax for back references. As this is synonymous with \k{name}, we      5.10 feature.
587      fudge it up by pretending it really was \k. */  
588        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
589        is part of Perl's movement towards a unified syntax for back references. As
590        this is synonymous with \k{name}, we fudge it up by pretending it really
591        was \k.
592    
593        (3) For Oniguruma compatibility we also support \g followed by a name or a
594        number either in angle brackets or in single quotes. However, these are
595        (possibly recursive) subroutine calls, _not_ backreferences. Just return
596        the -ESC_g code (cf \k). */
597    
598      case 'g':      case CHAR_g:
599      if (ptr[1] == '{')      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
600          {
601          c = -ESC_g;
602          break;
603          }
604    
605        /* Handle the Perl-compatible cases */
606    
607        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
608        {        {
609        const uschar *p;        const uschar *p;
610        for (p = ptr+2; *p != 0 && *p != '}'; p++)        for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
611          if (*p != '-' && (digitab[*p] & ctype_digit) == 0) break;          if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
612        if (*p != 0 && *p != '}')        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
613          {          {
614          c = -ESC_k;          c = -ESC_k;
615          break;          break;
# Line 554  else Line 619  else
619        }        }
620      else braced = FALSE;      else braced = FALSE;
621    
622      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
623        {        {
624        negated = TRUE;        negated = TRUE;
625        ptr++;        ptr++;
# Line 563  else Line 628  else
628    
629      c = 0;      c = 0;
630      while ((digitab[ptr[1]] & ctype_digit) != 0)      while ((digitab[ptr[1]] & ctype_digit) != 0)
631        c = c * 10 + *(++ptr) - '0';        c = c * 10 + *(++ptr) - CHAR_0;
632    
633      if (c < 0)      if (c < 0)   /* Integer overflow */
634        {        {
635        *errorcodeptr = ERR61;        *errorcodeptr = ERR61;
636        break;        break;
637        }        }
638    
639      if (c == 0 || (braced && *(++ptr) != '}'))      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
640        {        {
641        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
642        break;        break;
643        }        }
644    
645        if (c == 0)
646          {
647          *errorcodeptr = ERR58;
648          break;
649          }
650    
651      if (negated)      if (negated)
652        {        {
653        if (c > bracount)        if (c > bracount)
# Line 602  else Line 673  else
673      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
674      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
675    
676      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:
677      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
678    
679      if (!isclass)      if (!isclass)
680        {        {
681        oldptr = ptr;        oldptr = ptr;
682        c -= '0';        c -= CHAR_0;
683        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
684          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - CHAR_0;
685        if (c < 0)        if (c < 0)    /* Integer overflow */
686          {          {
687          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
688          break;          break;
# Line 628  else Line 699  else
699      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.
700      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
701    
702      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
703        {        {
704        ptr--;        ptr--;
705        c = 0;        c = 0;
# Line 641  else Line 712  else
712      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
713      than 3 octal digits. */      than 3 octal digits. */
714    
715      case '0':      case CHAR_0:
716      c -= '0';      c -= CHAR_0;
717      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
718          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
719      if (!utf8 && c > 255) *errorcodeptr = ERR51;      if (!utf8 && c > 255) *errorcodeptr = ERR51;
720      break;      break;
721    
# Line 652  else Line 723  else
723      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
724      treated as a data character. */      treated as a data character. */
725    
726      case 'x':      case CHAR_x:
727      if (ptr[1] == '{')      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
728        {        {
729        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
730        int count = 0;        int count = 0;
# Line 662  else Line 733  else
733        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
734          {          {
735          register int cc = *pt++;          register int cc = *pt++;
736          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
737          count++;          count++;
738    
739  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
740          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
741          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
742  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
743          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
744          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
745  #endif  #endif
746          }          }
747    
748        if (*pt == '}')        if (*pt == CHAR_RIGHT_CURLY_BRACKET)
749          {          {
750          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
751          ptr = pt;          ptr = pt;
# Line 690  else Line 761  else
761      c = 0;      c = 0;
762      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
763        {        {
764        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
765        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
766  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
767        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
768        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
769  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
770        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
771        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
772  #endif  #endif
773        }        }
774      break;      break;
# Line 706  else Line 777  else
777      This coding is ASCII-specific, but then the whole concept of \cx is      This coding is ASCII-specific, but then the whole concept of \cx is
778      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
779    
780      case 'c':      case CHAR_c:
781      c = *(++ptr);      c = *(++ptr);
782      if (c == 0)      if (c == 0)
783        {        {
# Line 714  else Line 785  else
785        break;        break;
786        }        }
787    
788  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
789      if (c >= 'a' && c <= 'z') c -= 32;      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
790      c ^= 0x40;      c ^= 0x40;
791  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
792      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
793      c ^= 0xC0;      c ^= 0xC0;
794  #endif  #endif
795      break;      break;
# Line 780  if (c == 0) goto ERROR_RETURN; Line 851  if (c == 0) goto ERROR_RETURN;
851  /* \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
852  negation. */  negation. */
853    
854  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
855    {    {
856    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
857      {      {
858      *negptr = TRUE;      *negptr = TRUE;
859      ptr++;      ptr++;
# Line 791  if (c == '{') Line 862  if (c == '{')
862      {      {
863      c = *(++ptr);      c = *(++ptr);
864      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
865      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
866      name[i] = c;      name[i] = c;
867      }      }
868    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
869    name[i] = 0;    name[i] = 0;
870    }    }
871    
# Line 859  is_counted_repeat(const uschar *p) Line 930  is_counted_repeat(const uschar *p)
930  {  {
931  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
932  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
933  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
934    
935  if (*p++ != ',') return FALSE;  if (*p++ != CHAR_COMMA) return FALSE;
936  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
937    
938  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
939  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
940    
941  return (*p == '}');  return (*p == CHAR_RIGHT_CURLY_BRACKET);
942  }  }
943    
944    
# Line 900  int max = -1; Line 971  int max = -1;
971  /* 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
972  an integer overflow. */  an integer overflow. */
973    
974  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;
975  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
976    {    {
977    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 910  if (min < 0 || min > 65535) Line 981  if (min < 0 || min > 65535)
981  /* 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.
982  Also, max must not be less than min. */  Also, max must not be less than min. */
983    
984  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
985    {    {
986    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
987      {      {
988      max = 0;      max = 0;
989      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;
990      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
991        {        {
992        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 940  return p; Line 1011  return p;
1011    
1012    
1013  /*************************************************  /*************************************************
1014  *       Find forward referenced subpattern       *  *  Subroutine for finding forward reference      *
1015  *************************************************/  *************************************************/
1016    
1017  /* This function scans along a pattern's text looking for capturing  /* This recursive function is called only from find_parens() below. The
1018    top-level call starts at the beginning of the pattern. All other calls must
1019    start at a parenthesis. It scans along a pattern's text looking for capturing
1020  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
1021  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
1022  returns when it reaches a given numbered subpattern. This is used for forward  returns when it reaches a given numbered subpattern. We know that if (?P< is
1023  references to subpatterns. We know that if (?P< is encountered, the name will  encountered, the name will be terminated by '>' because that is checked in the
1024  be terminated by '>' because that is checked in the first pass.  first pass. Recursion is used to keep track of subpatterns that reset the
1025    capturing group numbers - the (?| feature.
1026    
1027  Arguments:  Arguments:
1028    ptr          current position in the pattern    ptrptr       address of the current character pointer (updated)
1029    count        current count of capturing parens so far encountered    cd           compile background data
1030    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1031    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1032    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1033      count        pointer to the current capturing subpattern number (updated)
1034    
1035  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1036  */  */
1037    
1038  static int  static int
1039  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,
1040    BOOL xmode)    BOOL xmode, int *count)
1041  {  {
1042  const uschar *thisname;  uschar *ptr = *ptrptr;
1043    int start_count = *count;
1044    int hwm_count = start_count;
1045    BOOL dup_parens = FALSE;
1046    
1047  for (; *ptr != 0; ptr++)  /* If the first character is a parenthesis, check on the type of group we are
1048    dealing with. The very first call may not start with a parenthesis. */
1049    
1050    if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1051    {    {
1052    int term;    if (ptr[1] == CHAR_QUESTION_MARK &&
1053          ptr[2] == CHAR_VERTICAL_LINE)
1054        {
1055        ptr += 3;
1056        dup_parens = TRUE;
1057        }
1058    
1059      /* Handle a normal, unnamed capturing parenthesis */
1060    
1061      else if (ptr[1] != CHAR_QUESTION_MARK && ptr[1] != CHAR_ASTERISK)
1062        {
1063        *count += 1;
1064        if (name == NULL && *count == lorn) return *count;
1065        ptr++;
1066        }
1067    
1068      /* Handle a condition. If it is an assertion, just carry on so that it
1069      is processed as normal. If not, skip to the closing parenthesis of the
1070      condition (there can't be any nested parens. */
1071    
1072      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1073        {
1074        ptr += 2;
1075        if (ptr[1] != CHAR_QUESTION_MARK)
1076          {
1077          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1078          if (*ptr != 0) ptr++;
1079          }
1080        }
1081    
1082      /* We have either (? or (* and not a condition */
1083    
1084      else
1085        {
1086        ptr += 2;
1087        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1088    
1089        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1090    
1091        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1092            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1093          {
1094          int term;
1095          const uschar *thisname;
1096          *count += 1;
1097          if (name == NULL && *count == lorn) return *count;
1098          term = *ptr++;
1099          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1100          thisname = ptr;
1101          while (*ptr != term) ptr++;
1102          if (name != NULL && lorn == ptr - thisname &&
1103              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1104            return *count;
1105          term++;
1106          }
1107        }
1108      }
1109    
1110    /* Past any initial parenthesis handling, scan for parentheses or vertical
1111    bars. */
1112    
1113    for (; *ptr != 0; ptr++)
1114      {
1115    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1116    
1117    if (*ptr == '\\')    if (*ptr == CHAR_BACKSLASH)
1118      {      {
1119      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1120      if (*ptr == 'Q') for (;;)      if (*ptr == CHAR_Q) for (;;)
1121        {        {
1122        while (*(++ptr) != 0 && *ptr != '\\');        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1123        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1124        if (*(++ptr) == 'E') break;        if (*(++ptr) == CHAR_E) break;
1125        }        }
1126      continue;      continue;
1127      }      }
1128    
1129    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1130      are handled for real. If the first character is '^', skip it. Also, if the
1131      first few characters (either before or after ^) are \Q\E or \E we skip them
1132      too. This makes for compatibility with Perl. Note the use of STR macros to
1133      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1134    
1135    if (*ptr == '[')    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1136      {      {
1137      while (*(++ptr) != ']')      BOOL negate_class = FALSE;
1138        for (;;)
1139          {
1140          if (ptr[1] == CHAR_BACKSLASH)
1141            {
1142            if (ptr[2] == CHAR_E)
1143              ptr+= 2;
1144            else if (strncmp((const char *)ptr+2,
1145                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1146              ptr += 4;
1147            else
1148              break;
1149            }
1150          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1151            {
1152            negate_class = TRUE;
1153            ptr++;
1154            }
1155          else break;
1156          }
1157    
1158        /* If the next character is ']', it is a data character that must be
1159        skipped, except in JavaScript compatibility mode. */
1160    
1161        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1162            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1163          ptr++;
1164    
1165        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1166        {        {
1167        if (*ptr == 0) return -1;        if (*ptr == 0) return -1;
1168        if (*ptr == '\\')        if (*ptr == CHAR_BACKSLASH)
1169          {          {
1170          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) goto FAIL_EXIT;
1171          if (*ptr == 'Q') for (;;)          if (*ptr == CHAR_Q) for (;;)
1172            {            {
1173            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1174            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1175            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1176            }            }
1177          continue;          continue;
1178          }          }
# Line 1008  for (; *ptr != 0; ptr++) Line 1182  for (; *ptr != 0; ptr++)
1182    
1183    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1184    
1185    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1186      {      {
1187      while (*(++ptr) != 0 && *ptr != '\n');      while (*(++ptr) != 0 && *ptr != CHAR_NL) {};
1188      if (*ptr == 0) return -1;      if (*ptr == 0) goto FAIL_EXIT;
1189      continue;      continue;
1190      }      }
1191    
1192    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1193    
1194    if (*ptr != '(') continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != '?' && ptr[1] != '*')  
1195      {      {
1196      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, count);
1197      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1198      continue;      if (*ptr == 0) goto FAIL_EXIT;
1199        }
1200    
1201      else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1202        {
1203        if (dup_parens && *count < hwm_count) *count = hwm_count;
1204        *ptrptr = ptr;
1205        return -1;
1206      }      }
1207    
1208    ptr += 2;    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1209    if (*ptr == 'P') ptr++;                      /* Allow optional P */      {
1210        if (*count > hwm_count) hwm_count = *count;
1211        *count = start_count;
1212        }
1213      }
1214    
1215    /* We have to disambiguate (?<! and (?<= from (?<name> */  FAIL_EXIT:
1216    *ptrptr = ptr;
1217    return -1;
1218    }
1219    
   if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  
        *ptr != '\'')  
     continue;  
1220    
   count++;  
1221    
1222    if (name == NULL && count == lorn) return count;  
1223    term = *ptr++;  /*************************************************
1224    if (term == '<') term = '>';  *       Find forward referenced subpattern       *
1225    thisname = ptr;  *************************************************/
1226    while (*ptr != term) ptr++;  
1227    if (name != NULL && lorn == ptr - thisname &&  /* This function scans along a pattern's text looking for capturing
1228        strncmp((const char *)name, (const char *)thisname, lorn) == 0)  subpatterns, and counting them. If it finds a named pattern that matches the
1229      return count;  name it is given, it returns its number. Alternatively, if the name is NULL, it
1230    returns when it reaches a given numbered subpattern. This is used for forward
1231    references to subpatterns. We used to be able to start this scan from the
1232    current compiling point, using the current count value from cd->bracount, and
1233    do it all in a single loop, but the addition of the possibility of duplicate
1234    subpattern numbers means that we have to scan from the very start, in order to
1235    take account of such duplicates, and to use a recursive function to keep track
1236    of the different types of group.
1237    
1238    Arguments:
1239      cd           compile background data
1240      name         name to seek, or NULL if seeking a numbered subpattern
1241      lorn         name length, or subpattern number if name is NULL
1242      xmode        TRUE if we are in /x mode
1243    
1244    Returns:       the number of the found subpattern, or -1 if not found
1245    */
1246    
1247    static int
1248    find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode)
1249    {
1250    uschar *ptr = (uschar *)cd->start_pattern;
1251    int count = 0;
1252    int rc;
1253    
1254    /* If the pattern does not start with an opening parenthesis, the first call
1255    to find_parens_sub() will scan right to the end (if necessary). However, if it
1256    does start with a parenthesis, find_parens_sub() will return when it hits the
1257    matching closing parens. That is why we have to have a loop. */
1258    
1259    for (;;)
1260      {
1261      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, &count);
1262      if (rc > 0 || *ptr++ == 0) break;
1263    }    }
1264    
1265  return -1;  return rc;
1266  }  }
1267    
1268    
1269    
1270    
1271  /*************************************************  /*************************************************
1272  *      Find first significant op code            *  *      Find first significant op code            *
1273  *************************************************/  *************************************************/
# Line 1100  for (;;) Line 1317  for (;;)
1317    
1318      case OP_CALLOUT:      case OP_CALLOUT:
1319      case OP_CREF:      case OP_CREF:
1320        case OP_NCREF:
1321      case OP_RREF:      case OP_RREF:
1322        case OP_NRREF:
1323      case OP_DEF:      case OP_DEF:
1324      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1325      break;      break;
# Line 1116  for (;;) Line 1335  for (;;)
1335    
1336    
1337  /*************************************************  /*************************************************
1338  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1339  *************************************************/  *************************************************/
1340    
1341  /* 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,
1342  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.
1343  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
1344    temporarily terminated with OP_END when this function is called.
1345    
1346    This function is called when a backward assertion is encountered, so that if it
1347    fails, the error message can point to the correct place in the pattern.
1348    However, we cannot do this when the assertion contains subroutine calls,
1349    because they can be forward references. We solve this by remembering this case
1350    and doing the check at the end; a flag specifies which mode we are running in.
1351    
1352  Arguments:  Arguments:
1353    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1354    options  the compiling options    options  the compiling options
1355      atend    TRUE if called when the pattern is complete
1356      cd       the "compile data" structure
1357    
1358  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1359                 or -1 if there is no fixed length,
1360               or -2 if \C was encountered               or -2 if \C was encountered
1361                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1362  */  */
1363    
1364  static int  static int
1365  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)
1366  {  {
1367  int length = -1;  int length = -1;
1368    
# Line 1145  branch, check the length against that of Line 1375  branch, check the length against that of
1375  for (;;)  for (;;)
1376    {    {
1377    int d;    int d;
1378      uschar *ce, *cs;
1379    register int op = *cc;    register int op = *cc;
1380    switch (op)    switch (op)
1381      {      {
# Line 1152  for (;;) Line 1383  for (;;)
1383      case OP_BRA:      case OP_BRA:
1384      case OP_ONCE:      case OP_ONCE:
1385      case OP_COND:      case OP_COND:
1386      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);
1387      if (d < 0) return d;      if (d < 0) return d;
1388      branchlength += d;      branchlength += d;
1389      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1175  for (;;) Line 1406  for (;;)
1406      branchlength = 0;      branchlength = 0;
1407      break;      break;
1408    
1409        /* A true recursion implies not fixed length, but a subroutine call may
1410        be OK. If the subroutine is a forward reference, we can't deal with
1411        it until the end of the pattern, so return -3. */
1412    
1413        case OP_RECURSE:
1414        if (!atend) return -3;
1415        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1416        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1417        if (cc > cs && cc < ce) return -1;                /* Recursion */
1418        d = find_fixedlength(cs + 2, options, atend, cd);
1419        if (d < 0) return d;
1420        branchlength += d;
1421        cc += 1 + LINK_SIZE;
1422        break;
1423    
1424      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1425    
1426      case OP_ASSERT:      case OP_ASSERT:
# Line 1188  for (;;) Line 1434  for (;;)
1434    
1435      case OP_REVERSE:      case OP_REVERSE:
1436      case OP_CREF:      case OP_CREF:
1437        case OP_NCREF:
1438      case OP_RREF:      case OP_RREF:
1439        case OP_NRREF:
1440      case OP_DEF:      case OP_DEF:
1441      case OP_OPT:      case OP_OPT:
1442      case OP_CALLOUT:      case OP_CALLOUT:
# Line 1211  for (;;) Line 1459  for (;;)
1459      branchlength++;      branchlength++;
1460      cc += 2;      cc += 2;
1461  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1462      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1463        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1464  #endif  #endif
1465      break;      break;
1466    
# Line 1225  for (;;) Line 1471  for (;;)
1471      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1472      cc += 4;      cc += 4;
1473  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1474      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1475        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1476  #endif  #endif
1477      break;      break;
1478    
# Line 1252  for (;;) Line 1496  for (;;)
1496      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1497      case OP_WORDCHAR:      case OP_WORDCHAR:
1498      case OP_ANY:      case OP_ANY:
1499        case OP_ALLANY:
1500      branchlength++;      branchlength++;
1501      cc++;      cc++;
1502      break;      break;
# Line 1306  for (;;) Line 1551  for (;;)
1551    
1552    
1553  /*************************************************  /*************************************************
1554  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1555  *************************************************/  *************************************************/
1556    
1557  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1558  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1559    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1560    so that it can be called from pcre_study() when finding the minimum matching
1561    length.
1562    
1563  Arguments:  Arguments:
1564    code        points to start of expression    code        points to start of expression
1565    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1566    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1567    
1568  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
1569  */  */
1570    
1571  static const uschar *  const uschar *
1572  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1573  {  {
1574  for (;;)  for (;;)
1575    {    {
# Line 1334  for (;;) Line 1582  for (;;)
1582    
1583    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1584    
1585      /* Handle recursion */
1586    
1587      else if (c == OP_REVERSE)
1588        {
1589        if (number < 0) return (uschar *)code;
1590        code += _pcre_OP_lengths[c];
1591        }
1592    
1593    /* Handle capturing bracket */    /* Handle capturing bracket */
1594    
1595    else if (c == OP_CBRA)    else if (c == OP_CBRA)
# Line 1400  for (;;) Line 1656  for (;;)
1656        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1657        break;        break;
1658        }        }
1659    #else
1660        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1661  #endif  #endif
1662      }      }
1663    }    }
# Line 1493  for (;;) Line 1751  for (;;)
1751        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1752        break;        break;
1753        }        }
1754    #else
1755        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1756  #endif  #endif
1757      }      }
1758    }    }
# Line 1508  for (;;) Line 1768  for (;;)
1768  can match the empty string or not. It is called from could_be_empty()  can match the empty string or not. It is called from could_be_empty()
1769  below and from compile_branch() when checking for an unlimited repeat of a  below and from compile_branch() when checking for an unlimited repeat of a
1770  group that can match nothing. Note that first_significant_code() skips over  group that can match nothing. Note that first_significant_code() skips over
1771  assertions. If we hit an unclosed bracket, we return "empty" - this means we've  backward and negative forward assertions when its final argument is TRUE. If we
1772  struck an inner bracket whose current branch will already have been scanned.  hit an unclosed bracket, we return "empty" - this means we've struck an inner
1773    bracket whose current branch will already have been scanned.
1774    
1775  Arguments:  Arguments:
1776    code        points to start of search    code        points to start of search
# Line 1531  for (code = first_significant_code(code Line 1792  for (code = first_significant_code(code
1792    
1793    c = *code;    c = *code;
1794    
1795      /* Skip over forward assertions; the other assertions are skipped by
1796      first_significant_code() with a TRUE final argument. */
1797    
1798      if (c == OP_ASSERT)
1799        {
1800        do code += GET(code, 1); while (*code == OP_ALT);
1801        c = *code;
1802        continue;
1803        }
1804    
1805    /* Groups with zero repeats can of course be empty; skip them. */    /* Groups with zero repeats can of course be empty; skip them. */
1806    
1807    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)
1808      {      {
1809      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1810      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
# Line 1548  for (code = first_significant_code(code Line 1819  for (code = first_significant_code(code
1819      BOOL empty_branch;      BOOL empty_branch;
1820      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
1821    
1822      /* Scan a closed bracket */      /* If a conditional group has only one branch, there is a second, implied,
1823        empty branch, so just skip over the conditional, because it could be empty.
1824        Otherwise, scan the individual branches of the group. */
1825    
1826      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;  
1827        code += GET(code, 1);        code += GET(code, 1);
1828        else
1829          {
1830          empty_branch = FALSE;
1831          do
1832            {
1833            if (!empty_branch && could_be_empty_branch(code, endcode, utf8))
1834              empty_branch = TRUE;
1835            code += GET(code, 1);
1836            }
1837          while (*code == OP_ALT);
1838          if (!empty_branch) return FALSE;   /* All branches are non-empty */
1839        }        }
1840      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
1841      c = *code;      c = *code;
1842      continue;      continue;
1843      }      }
# Line 1619  for (code = first_significant_code(code Line 1898  for (code = first_significant_code(code
1898      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1899      case OP_WORDCHAR:      case OP_WORDCHAR:
1900      case OP_ANY:      case OP_ANY:
1901        case OP_ALLANY:
1902      case OP_ANYBYTE:      case OP_ANYBYTE:
1903      case OP_CHAR:      case OP_CHAR:
1904      case OP_CHARNC:      case OP_CHARNC:
# Line 1675  for (code = first_significant_code(code Line 1955  for (code = first_significant_code(code
1955      case OP_QUERY:      case OP_QUERY:
1956      case OP_MINQUERY:      case OP_MINQUERY:
1957      case OP_POSQUERY:      case OP_POSQUERY:
1958        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
1959        break;
1960    
1961      case OP_UPTO:      case OP_UPTO:
1962      case OP_MINUPTO:      case OP_MINUPTO:
1963      case OP_POSUPTO:      case OP_POSUPTO:
1964      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
1965      break;      break;
1966  #endif  #endif
1967      }      }
# Line 1726  return TRUE; Line 2009  return TRUE;
2009  *************************************************/  *************************************************/
2010    
2011  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2012  encountered in a character class. It checks whether this is followed by an  encountered in a character class. It checks whether this is followed by a
2013  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2014  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2015    
2016    Originally, this function only recognized a sequence of letters between the
2017    terminators, but it seems that Perl recognizes any sequence of characters,
2018    though of course unknown POSIX names are subsequently rejected. Perl gives an
2019    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2020    didn't consider this to be a POSIX class. Likewise for [:1234:].
2021    
2022    The problem in trying to be exactly like Perl is in the handling of escapes. We
2023    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2024    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2025    below handles the special case of \], but does not try to do any other escape
2026    processing. This makes it different from Perl for cases such as [:l\ower:]
2027    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2028    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2029    I think.
2030    
2031  Argument:  Arguments:
2032    ptr      pointer to the initial [    ptr      pointer to the initial [
2033    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2034    
2035  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2036  */  */
2037    
2038  static BOOL  static BOOL
2039  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
2040  {  {
2041  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2042  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2043  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2044    {    {
2045    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else
2046    return TRUE;      {
2047        if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2048        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2049          {
2050          *endptr = ptr;
2051          return TRUE;
2052          }
2053        }
2054    }    }
2055  return FALSE;  return FALSE;
2056  }  }
# Line 1794  return -1; Line 2096  return -1;
2096  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2097  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2098  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
2099  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
2100  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
2101  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
2102  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
2103  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2104    OP_END.
2105    
2106  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2107  recursions and subroutine calls. It must also check the list of such references  recursions and subroutine calls. It must also check the list of such references
# Line 1933  get_othercase_range(unsigned int *cptr, Line 2236  get_othercase_range(unsigned int *cptr,
2236  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2237    
2238  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2239    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2240    
2241  if (c > d) return FALSE;  if (c > d) return FALSE;
2242    
# Line 1942  next = othercase + 1; Line 2245  next = othercase + 1;
2245    
2246  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2247    {    {
2248    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2249    next++;    next++;
2250    }    }
2251    
# Line 1988  if ((options & PCRE_EXTENDED) != 0) Line 2291  if ((options & PCRE_EXTENDED) != 0)
2291    for (;;)    for (;;)
2292      {      {
2293      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2294      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2295        {        {
2296        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2297          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
# Line 2000  if ((options & PCRE_EXTENDED) != 0) Line 2303  if ((options & PCRE_EXTENDED) != 0)
2303  /* 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
2304  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2305    
2306  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2307    {    {
2308    int temperrorcode = 0;    int temperrorcode = 0;
2309    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
# Line 2025  if ((options & PCRE_EXTENDED) != 0) Line 2328  if ((options & PCRE_EXTENDED) != 0)
2328    for (;;)    for (;;)
2329      {      {
2330      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2331      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2332        {        {
2333        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2334          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
# Line 2036  if ((options & PCRE_EXTENDED) != 0) Line 2339  if ((options & PCRE_EXTENDED) != 0)
2339    
2340  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
2341    
2342  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2343    return FALSE;    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2344        return FALSE;
2345    
2346  /* Now compare the next item with the previous opcode. If the previous is a  /* Now compare the next item with the previous opcode. If the previous is a
2347  positive single character match, "item" either contains the character or, if  positive single character match, "item" either contains the character or, if
# Line 2052  if (next >= 0) switch(op_code) Line 2356  if (next >= 0) switch(op_code)
2356    case OP_CHAR:    case OP_CHAR:
2357  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2358    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2359    #else
2360      (void)(utf8_char);  /* Keep compiler happy by referencing function argument */
2361  #endif  #endif
2362    return item != next;    return item != next;
2363    
# Line 2070  if (next >= 0) switch(op_code) Line 2376  if (next >= 0) switch(op_code)
2376      unsigned int othercase;      unsigned int othercase;
2377      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2378  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2379      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
2380  #else  #else
2381      othercase = NOTACHAR;      othercase = NOTACHAR;
2382  #endif  #endif
# Line 2083  if (next >= 0) switch(op_code) Line 2389  if (next >= 0) switch(op_code)
2389    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT, "item" must be a single-byte character. */
2390    
2391    case OP_NOT:    case OP_NOT:
   if (next < 0) return FALSE;  /* Not a character */  
2392    if (item == next) return TRUE;    if (item == next) return TRUE;
2393    if ((options & PCRE_CASELESS) == 0) return FALSE;    if ((options & PCRE_CASELESS) == 0) return FALSE;
2394  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 2092  if (next >= 0) switch(op_code) Line 2397  if (next >= 0) switch(op_code)
2397      unsigned int othercase;      unsigned int othercase;
2398      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2399  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2400      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
2401  #else  #else
2402      othercase = NOTACHAR;      othercase = NOTACHAR;
2403  #endif  #endif
# Line 2346  uschar classbits[32]; Line 2651  uschar classbits[32];
2651  BOOL class_utf8;  BOOL class_utf8;
2652  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
2653  uschar *class_utf8data;  uschar *class_utf8data;
2654    uschar *class_utf8data_base;
2655  uschar utf8_char[6];  uschar utf8_char[6];
2656  #else  #else
2657  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
# Line 2385  req_caseopt = ((options & PCRE_CASELESS) Line 2691  req_caseopt = ((options & PCRE_CASELESS)
2691  for (;; ptr++)  for (;; ptr++)
2692    {    {
2693    BOOL negate_class;    BOOL negate_class;
2694    BOOL should_flip_negation;    BOOL should_flip_negation;
2695    BOOL possessive_quantifier;    BOOL possessive_quantifier;
2696    BOOL is_quantifier;    BOOL is_quantifier;
2697    BOOL is_recurse;    BOOL is_recurse;
# Line 2473  for (;; ptr++) Line 2779  for (;; ptr++)
2779    
2780    if (inescq && c != 0)    if (inescq && c != 0)
2781      {      {
2782      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
2783        {        {
2784        inescq = FALSE;        inescq = FALSE;
2785        ptr++;        ptr++;
# Line 2499  for (;; ptr++) Line 2805  for (;; ptr++)
2805    /* 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
2806    a quantifier. */    a quantifier. */
2807    
2808    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
2809      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
2810        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
2811    
2812    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
2813         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2515  for (;; ptr++) Line 2822  for (;; ptr++)
2822    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
2823      {      {
2824      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
2825      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
2826        {        {
2827        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2828          {          {
# Line 2540  for (;; ptr++) Line 2847  for (;; ptr++)
2847      {      {
2848      /* ===================================================================*/      /* ===================================================================*/
2849      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
2850      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
2851      case ')':      case CHAR_RIGHT_PARENTHESIS:
2852      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
2853      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
2854      *codeptr = code;      *codeptr = code;
# Line 2563  for (;; ptr++) Line 2870  for (;; ptr++)
2870      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
2871      the setting of any following char as a first character. */      the setting of any following char as a first character. */
2872    
2873      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
2874      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
2875        {        {
2876        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
# Line 2572  for (;; ptr++) Line 2879  for (;; ptr++)
2879      *code++ = OP_CIRC;      *code++ = OP_CIRC;
2880      break;      break;
2881    
2882      case '$':      case CHAR_DOLLAR_SIGN:
2883      previous = NULL;      previous = NULL;
2884      *code++ = OP_DOLL;      *code++ = OP_DOLL;
2885      break;      break;
# Line 2580  for (;; ptr++) Line 2887  for (;; ptr++)
2887      /* 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
2888      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
2889    
2890      case '.':      case CHAR_DOT:
2891      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
2892      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
2893      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
2894      previous = code;      previous = code;
2895      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
2896      break;      break;
2897    
2898    
# Line 2600  for (;; ptr++) Line 2907  for (;; ptr++)
2907      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,
2908      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
2909      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.
     */  
2910    
2911      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
2912        default (Perl) mode, it is treated as a data character. */
2913    
2914        case CHAR_RIGHT_SQUARE_BRACKET:
2915        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2916          {
2917          *errorcodeptr = ERR64;
2918          goto FAILED;
2919          }
2920        goto NORMAL_CHAR;
2921    
2922        case CHAR_LEFT_SQUARE_BRACKET:
2923      previous = code;      previous = code;
2924    
2925      /* 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
2926      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. */
2927    
2928      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2929          check_posix_syntax(ptr, &tempptr, cd))           ptr[1] == CHAR_EQUALS_SIGN) &&
2930            check_posix_syntax(ptr, &tempptr))
2931        {        {
2932        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
2933        goto FAILED;        goto FAILED;
2934        }        }
2935    
# Line 2623  for (;; ptr++) Line 2941  for (;; ptr++)
2941      for (;;)      for (;;)
2942        {        {
2943        c = *(++ptr);        c = *(++ptr);
2944        if (c == '\\')        if (c == CHAR_BACKSLASH)
2945          {          {
2946          if (ptr[1] == 'E') ptr++;          if (ptr[1] == CHAR_E)
2947            else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;            ptr++;
2948              else break;          else if (strncmp((const char *)ptr+1,
2949                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
2950              ptr += 3;
2951            else
2952              break;
2953          }          }
2954        else if (!negate_class && c == '^')        else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
2955          negate_class = TRUE;          negate_class = TRUE;
2956        else break;        else break;
2957        }        }
2958    
2959      /* If a class contains a negative special such as \S, we need to flip the      /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
2960      negation flag at the end, so that support for characters > 255 works      an initial ']' is taken as a data character -- the code below handles
2961        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
2962        [^] must match any character, so generate OP_ALLANY. */
2963    
2964        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
2965            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2966          {
2967          *code++ = negate_class? OP_ALLANY : OP_FAIL;
2968          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
2969          zerofirstbyte = firstbyte;
2970          break;
2971          }
2972    
2973        /* If a class contains a negative special such as \S, we need to flip the
2974        negation flag at the end, so that support for characters > 255 works
2975      correctly (they are all included in the class). */      correctly (they are all included in the class). */
2976    
2977      should_flip_negation = FALSE;      should_flip_negation = FALSE;
# Line 2657  for (;; ptr++) Line 2993  for (;; ptr++)
2993  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2994      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
2995      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
2996        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
2997  #endif  #endif
2998    
2999      /* Process characters until ] is reached. By writing this as a "do" it      /* Process characters until ] is reached. By writing this as a "do" it
# Line 2672  for (;; ptr++) Line 3009  for (;; ptr++)
3009          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3010          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3011          }          }
3012    
3013          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
3014          data and reset the pointer. This is so that very large classes that
3015          contain a zillion UTF-8 characters no longer overwrite the work space
3016          (which is on the stack). */
3017    
3018          if (lengthptr != NULL)
3019            {
3020            *lengthptr += class_utf8data - class_utf8data_base;
3021            class_utf8data = class_utf8data_base;
3022            }
3023    
3024  #endif  #endif
3025    
3026        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
3027    
3028        if (inescq)        if (inescq)
3029          {          {
3030          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3031            {            {
3032            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
3033            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2693  for (;; ptr++) Line 3042  for (;; ptr++)
3042        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3043        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3044    
3045        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3046            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3047            check_posix_syntax(ptr, &tempptr, cd))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3048          {          {
3049          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3050          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3051          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3052          uschar pbits[32];          uschar pbits[32];
3053    
3054          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3055            {            {
3056            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3057            goto FAILED;            goto FAILED;
3058            }            }
3059    
3060          ptr += 2;          ptr += 2;
3061          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3062            {            {
3063            local_negate = TRUE;            local_negate = TRUE;
3064            should_flip_negation = TRUE;  /* Note negative special */            should_flip_negation = TRUE;  /* Note negative special */
3065            ptr++;            ptr++;
3066            }            }
3067    
# Line 2782  for (;; ptr++) Line 3131  for (;; ptr++)
3131        to 'or' into the one we are building. We assume they have more than one        to 'or' into the one we are building. We assume they have more than one
3132        character in them, so set class_charcount bigger than one. */        character in them, so set class_charcount bigger than one. */
3133    
3134        if (c == '\\')        if (c == CHAR_BACKSLASH)
3135          {          {
3136          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3137          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3138    
3139          if (-c == ESC_b) c = '\b';       /* \b is backslash in a class */          if (-c == ESC_b) c = CHAR_BS;       /* \b is backspace in a class */
3140          else if (-c == ESC_X) c = 'X';   /* \X is literal X in a class */          else if (-c == ESC_X) c = CHAR_X;   /* \X is literal X in a class */
3141          else if (-c == ESC_R) c = 'R';   /* \R is literal R in a class */          else if (-c == ESC_R) c = CHAR_R;   /* \R is literal R in a class */
3142          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3143            {            {
3144            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3145              {              {
3146              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3147              }              }
# Line 2815  for (;; ptr++) Line 3164  for (;; ptr++)
3164              continue;              continue;
3165    
3166              case ESC_D:              case ESC_D:
3167              should_flip_negation = TRUE;              should_flip_negation = TRUE;
3168              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
3169              continue;              continue;
3170    
# Line 2824  for (;; ptr++) Line 3173  for (;; ptr++)
3173              continue;              continue;
3174    
3175              case ESC_W:              case ESC_W:
3176              should_flip_negation = TRUE;              should_flip_negation = TRUE;
3177              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3178              continue;              continue;
3179    
# Line 2834  for (;; ptr++) Line 3183  for (;; ptr++)
3183              continue;              continue;
3184    
3185              case ESC_S:              case ESC_S:
3186              should_flip_negation = TRUE;              should_flip_negation = TRUE;
3187              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
3188              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3189              continue;              continue;
3190    
             case ESC_E: /* Perl ignores an orphan \E */  
             continue;  
   
3191              default:    /* Not recognized; fall through */              default:    /* Not recognized; fall through */
3192              break;      /* Need "default" setting to stop compiler warning. */              break;      /* Need "default" setting to stop compiler warning. */
3193              }              }
# Line 3021  for (;; ptr++) Line 3367  for (;; ptr++)
3367        entirely. The code for handling \Q and \E is messy. */        entirely. The code for handling \Q and \E is messy. */
3368    
3369        CHECK_RANGE:        CHECK_RANGE:
3370        while (ptr[1] == '\\' && ptr[2] == 'E')        while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3371          {          {
3372          inescq = FALSE;          inescq = FALSE;
3373          ptr += 2;          ptr += 2;
# Line 3031  for (;; ptr++) Line 3377  for (;; ptr++)
3377    
3378        /* Remember \r or \n */        /* Remember \r or \n */
3379    
3380        if (c == '\r' || c == '\n') cd->external_flags |= PCRE_HASCRORLF;        if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3381    
3382        /* Check for range */        /* Check for range */
3383    
3384        if (!inescq && ptr[1] == '-')        if (!inescq && ptr[1] == CHAR_MINUS)
3385          {          {
3386          int d;          int d;
3387          ptr += 2;          ptr += 2;
3388          while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3389    
3390          /* 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
3391          mode. */          mode. */
3392    
3393          while (*ptr == '\\' && ptr[1] == 'Q')          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3394            {            {
3395            ptr += 2;            ptr += 2;
3396            if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }            if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3397                { ptr += 2; continue; }
3398            inescq = TRUE;            inescq = TRUE;
3399            break;            break;
3400            }            }
3401    
3402          if (*ptr == 0 || (!inescq && *ptr == ']'))          if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3403            {            {
3404            ptr = oldptr;            ptr = oldptr;
3405            goto LONE_SINGLE_CHARACTER;            goto LONE_SINGLE_CHARACTER;
# Line 3071  for (;; ptr++) Line 3418  for (;; ptr++)
3418          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
3419          in such circumstances. */          in such circumstances. */
3420    
3421          if (!inescq && d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3422            {            {
3423            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3424            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3425    
3426            /* \b is backslash; \X is literal X; \R is literal R; any other            /* \b is backspace; \X is literal X; \R is literal R; any other
3427            special means the '-' was literal */            special means the '-' was literal */
3428    
3429            if (d < 0)            if (d < 0)
3430              {              {
3431              if (d == -ESC_b) d = '\b';              if (d == -ESC_b) d = CHAR_BS;
3432              else if (d == -ESC_X) d = 'X';              else if (d == -ESC_X) d = CHAR_X;
3433              else if (d == -ESC_R) d = 'R'; else              else if (d == -ESC_R) d = CHAR_R; else
3434                {                {
3435                ptr = oldptr;                ptr = oldptr;
3436                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 3104  for (;; ptr++) Line 3451  for (;; ptr++)
3451    
3452          /* Remember \r or \n */          /* Remember \r or \n */
3453    
3454          if (d == '\r' || d == '\n') cd->external_flags |= PCRE_HASCRORLF;          if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3455    
3456          /* 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
3457          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 3224  for (;; ptr++) Line 3571  for (;; ptr++)
3571          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3572            {            {
3573            unsigned int othercase;            unsigned int othercase;
3574            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)            if ((othercase = UCD_OTHERCASE(c)) != c)
3575              {              {
3576              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3577              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 3251  for (;; ptr++) Line 3598  for (;; ptr++)
3598    
3599      /* Loop until ']' reached. This "while" is the end of the "do" above. */      /* Loop until ']' reached. This "while" is the end of the "do" above. */
3600    
3601      while ((c = *(++ptr)) != 0 && (c != ']' || inescq));      while ((c = *(++ptr)) != 0 && (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
3602    
3603      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)                          /* Missing terminating ']' */
3604        {        {
# Line 3340  we set the flag only if there is a liter Line 3687  we set the flag only if there is a liter
3687      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3688    
3689      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
3690      extended class, with its own opcode, unless there was a negated special      extended class, with its own opcode, unless there was a negated special
3691      such as \S in the class, because in that case all characters > 255 are in      such as \S in the class, because in that case all characters > 255 are in
3692      the class, so any that were explicitly given as well can be ignored. If      the class, so any that were explicitly given as well can be ignored. If
3693      (when there are explicit characters > 255 that must be listed) there are no      (when there are explicit characters > 255 that must be listed) there are no
3694      characters < 256, we can omit the bitmap in the actual compiled code. */      characters < 256, we can omit the bitmap in the actual compiled code. */
3695    
# Line 3373  we set the flag only if there is a liter Line 3720  we set the flag only if there is a liter
3720        }        }
3721  #endif  #endif
3722    
3723      /* If there are no characters > 255, set the opcode to OP_CLASS or      /* If there are no characters > 255, set the opcode to OP_CLASS or
3724      OP_NCLASS, depending on whether the whole class was negated and whether      OP_NCLASS, depending on whether the whole class was negated and whether
3725      there were negative specials such as \S in the class. Then copy the 32-byte      there were negative specials such as \S in the class. Then copy the 32-byte
3726      map into the code vector, negating it if necessary. */      map into the code vector, negating it if necessary. */
3727    
3728      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;      *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
3729      if (negate_class)      if (negate_class)
3730        {        {
# Line 3396  we set the flag only if there is a liter Line 3743  we set the flag only if there is a liter
3743      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
3744      has been tested above. */      has been tested above. */
3745    
3746      case '{':      case CHAR_LEFT_CURLY_BRACKET:
3747      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
3748      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
3749      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
3750      goto REPEAT;      goto REPEAT;
3751    
3752      case '*':      case CHAR_ASTERISK:
3753      repeat_min = 0;      repeat_min = 0;
3754      repeat_max = -1;      repeat_max = -1;
3755      goto REPEAT;      goto REPEAT;
3756    
3757      case '+':      case CHAR_PLUS:
3758      repeat_min = 1;      repeat_min = 1;
3759      repeat_max = -1;      repeat_max = -1;
3760      goto REPEAT;      goto REPEAT;
3761    
3762      case '?':      case CHAR_QUESTION_MARK:
3763      repeat_min = 0;      repeat_min = 0;
3764      repeat_max = 1;      repeat_max = 1;
3765    
# Line 3447  we set the flag only if there is a liter Line 3794  we set the flag only if there is a liter
3794      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
3795      repeat type to the non-default. */      repeat type to the non-default. */
3796    
3797      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
3798        {        {
3799        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
3800        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
3801        ptr++;        ptr++;
3802        }        }
3803      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
3804        {        {
3805        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
3806        ptr++;        ptr++;
# Line 3568  we set the flag only if there is a liter Line 3915  we set the flag only if there is a liter
3915    
3916        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
3917    
3918          /*--------------------------------------------------------------------*/
3919          /* This code is obsolete from release 8.00; the restriction was finally
3920          removed: */
3921    
3922        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
3923        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
3924    
3925        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
3926          /*--------------------------------------------------------------------*/
3927    
3928        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
3929    
# Line 3718  we set the flag only if there is a liter Line 4070  we set the flag only if there is a liter
4070          goto END_REPEAT;          goto END_REPEAT;
4071          }          }
4072    
4073          /*--------------------------------------------------------------------*/
4074          /* This code is obsolete from release 8.00; the restriction was finally
4075          removed: */
4076    
4077        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4078        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4079    
4080        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4081          /*--------------------------------------------------------------------*/
4082    
4083        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4084          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 3779  we set the flag only if there is a liter Line 4136  we set the flag only if there is a liter
4136    
4137        if (repeat_min == 0)        if (repeat_min == 0)
4138          {          {
4139          /* 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
4140          altogether. */          output altogether, like this:
4141    
4142          if (repeat_max == 0)          ** if (repeat_max == 0)
4143            {          **   {
4144            code = previous;          **   code = previous;
4145            goto END_REPEAT;          **   goto END_REPEAT;
4146            }          **   }
4147    
4148          /* If the maximum is 1 or unlimited, we just have to stick in the          However, that fails when a group is referenced as a subroutine from
4149          BRAZERO and do no more at this point. However, we do need to adjust          elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it
4150          any OP_RECURSE calls inside the group that refer to the group itself or          so that it is skipped on execution. As we don't have a list of which
4151          any internal or forward referenced group, because the offset is from          groups are referenced, we cannot do this selectively.
4152          the start of the whole regex. Temporarily terminate the pattern while  
4153          doing this. */          If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4154            and do no more at this point. However, we do need to adjust any
4155            OP_RECURSE calls inside the group that refer to the group itself or any
4156            internal or forward referenced group, because the offset is from the
4157            start of the whole regex. Temporarily terminate the pattern while doing
4158            this. */
4159    
4160          if (repeat_max <= 1)          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
4161            {            {
4162            *code = OP_END;            *code = OP_END;
4163            adjust_recurse(previous, 1, utf8, cd, save_hwm);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
4164            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
4165            code++;            code++;
4166              if (repeat_max == 0)
4167                {
4168                *previous++ = OP_SKIPZERO;
4169                goto END_REPEAT;
4170                }
4171            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4172            }            }
4173    
# Line 3995  we set the flag only if there is a liter Line 4362  we set the flag only if there is a liter
4362          }          }
4363        }        }
4364    
4365        /* If previous is OP_FAIL, it was generated by an empty class [] in
4366        JavaScript mode. The other ways in which OP_FAIL can be generated, that is
4367        by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
4368        error above. We can just ignore the repeat in JS case. */
4369    
4370        else if (*previous == OP_FAIL) goto END_REPEAT;
4371    
4372      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
4373    
4374      else      else
# Line 4019  we set the flag only if there is a liter Line 4393  we set the flag only if there is a liter
4393      if (possessive_quantifier)      if (possessive_quantifier)
4394        {        {
4395        int len;        int len;
4396        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||  
4397            *tempcode == OP_NOTEXACT)        if (*tempcode == OP_TYPEEXACT)
4398            tempcode += _pcre_OP_lengths[*tempcode] +
4399              ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
4400    
4401          else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
4402            {
4403          tempcode += _pcre_OP_lengths[*tempcode];          tempcode += _pcre_OP_lengths[*tempcode];
4404    #ifdef SUPPORT_UTF8
4405            if (utf8 && tempcode[-1] >= 0xc0)
4406              tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];
4407    #endif
4408            }
4409    
4410        len = code - tempcode;        len = code - tempcode;
4411        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4412          {          {
# Line 4067  we set the flag only if there is a liter Line 4452  we set the flag only if there is a liter
4452      lookbehind or option setting or condition or all the other extended      lookbehind or option setting or condition or all the other extended
4453      parenthesis forms.  */      parenthesis forms.  */
4454    
4455      case '(':      case CHAR_LEFT_PARENTHESIS:
4456      newoptions = options;      newoptions = options;
4457      skipbytes = 0;      skipbytes = 0;
4458      bravalue = OP_CBRA;      bravalue = OP_CBRA;
# Line 4076  we set the flag only if there is a liter Line 4461  we set the flag only if there is a liter
4461    
4462      /* First deal with various "verbs" that can be introduced by '*'. */      /* First deal with various "verbs" that can be introduced by '*'. */
4463    
4464      if (*(++ptr) == '*' && (cd->ctypes[ptr[1]] & ctype_letter) != 0)      if (*(++ptr) == CHAR_ASTERISK && (cd->ctypes[ptr[1]] & ctype_letter) != 0)
4465        {        {
4466        int i, namelen;        int i, namelen;
4467        const char *vn = verbnames;        const char *vn = verbnames;
4468        const uschar *name = ++ptr;        const uschar *name = ++ptr;
4469        previous = NULL;        previous = NULL;
4470        while ((cd->ctypes[*++ptr] & ctype_letter) != 0);        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
4471        if (*ptr == ':')        if (*ptr == CHAR_COLON)
4472          {          {
4473          *errorcodeptr = ERR59;   /* Not supported */          *errorcodeptr = ERR59;   /* Not supported */
4474          goto FAILED;          goto FAILED;
4475          }          }
4476        if (*ptr != ')')        if (*ptr != CHAR_RIGHT_PARENTHESIS)
4477          {          {
4478          *errorcodeptr = ERR60;          *errorcodeptr = ERR60;
4479          goto FAILED;          goto FAILED;
# Line 4099  we set the flag only if there is a liter Line 4484  we set the flag only if there is a liter
4484          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
4485              strncmp((char *)name, vn, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
4486            {            {
4487            *code = verbs[i].op;            /* Check for open captures before ACCEPT */
4488            if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;  
4489              if (verbs[i].op == OP_ACCEPT)
4490                {
4491                open_capitem *oc;
4492                cd->had_accept = TRUE;
4493                for (oc = cd->open_caps; oc != NULL; oc = oc->next)
4494                  {
4495                  *code++ = OP_CLOSE;
4496                  PUT2INC(code, 0, oc->number);
4497                  }
4498                }
4499              *code++ = verbs[i].op;
4500            break;            break;
4501            }            }
4502          vn += verbs[i].len + 1;          vn += verbs[i].len + 1;
# Line 4113  we set the flag only if there is a liter Line 4509  we set the flag only if there is a liter
4509      /* Deal with the extended parentheses; all are introduced by '?', and the      /* Deal with the extended parentheses; all are introduced by '?', and the
4510      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. */
4511    
4512      else if (*ptr == '?')      else if (*ptr == CHAR_QUESTION_MARK)
4513        {        {
4514        int i, set, unset, namelen;        int i, set, unset, namelen;
4515        int *optset;        int *optset;
# Line 4122  we set the flag only if there is a liter Line 4518  we set the flag only if there is a liter
4518    
4519        switch (*(++ptr))        switch (*(++ptr))
4520          {          {
4521          case '#':                 /* Comment; skip to ket */          case CHAR_NUMBER_SIGN:                 /* Comment; skip to ket */
4522          ptr++;          ptr++;
4523          while (*ptr != 0 && *ptr != ')') ptr++;          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
4524          if (*ptr == 0)          if (*ptr == 0)
4525            {            {
4526            *errorcodeptr = ERR18;            *errorcodeptr = ERR18;
# Line 4134  we set the flag only if there is a liter Line 4530  we set the flag only if there is a liter
4530    
4531    
4532          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4533          case '|':                 /* Reset capture count for each branch */          case CHAR_VERTICAL_LINE:  /* Reset capture count for each branch */
4534          reset_bracount = TRUE;          reset_bracount = TRUE;
4535          /* Fall through */          /* Fall through */
4536    
4537          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4538          case ':':                 /* Non-capturing bracket */          case CHAR_COLON:          /* Non-capturing bracket */
4539          bravalue = OP_BRA;          bravalue = OP_BRA;
4540          ptr++;          ptr++;
4541          break;          break;
4542    
4543    
4544          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4545          case '(':          case CHAR_LEFT_PARENTHESIS:
4546          bravalue = OP_COND;       /* Conditional group */          bravalue = OP_COND;       /* Conditional group */
4547    
4548          /* 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 4166  we set the flag only if there is a liter Line 4562  we set the flag only if there is a liter
4562          the switch. This will take control down to where bracketed groups,          the switch. This will take control down to where bracketed groups,
4563          including assertions, are processed. */          including assertions, are processed. */
4564    
4565          if (ptr[1] == '?' && (ptr[2] == '=' || ptr[2] == '!' || ptr[2] == '<'))          if (ptr[1] == CHAR_QUESTION_MARK && (ptr[2] == CHAR_EQUALS_SIGN ||
4566                ptr[2] == CHAR_EXCLAMATION_MARK || ptr[2] == CHAR_LESS_THAN_SIGN))
4567            break;            break;
4568    
4569          /* 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 4178  we set the flag only if there is a liter Line 4575  we set the flag only if there is a liter
4575    
4576          /* Check for a test for recursion in a named group. */          /* Check for a test for recursion in a named group. */
4577    
4578          if (ptr[1] == 'R' && ptr[2] == '&')          if (ptr[1] == CHAR_R && ptr[2] == CHAR_AMPERSAND)
4579            {            {
4580            terminator = -1;            terminator = -1;
4581            ptr += 2;            ptr += 2;
# Line 4188  we set the flag only if there is a liter Line 4585  we set the flag only if there is a liter
4585          /* 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
4586          syntax (?(<name>) or (?('name') */          syntax (?(<name>) or (?('name') */
4587    
4588          else if (ptr[1] == '<')          else if (ptr[1] == CHAR_LESS_THAN_SIGN)
4589            {            {
4590            terminator = '>';            terminator = CHAR_GREATER_THAN_SIGN;
4591            ptr++;            ptr++;
4592            }            }
4593          else if (ptr[1] == '\'')          else if (ptr[1] == CHAR_APOSTROPHE)
4594            {            {
4595            terminator = '\'';            terminator = CHAR_APOSTROPHE;
4596            ptr++;            ptr++;
4597            }            }
4598          else          else
4599            {            {
4600            terminator = 0;            terminator = 0;
4601            if (ptr[1] == '-' || ptr[1] == '+') refsign = *(++ptr);            if (ptr[1] == CHAR_MINUS || ptr[1] == CHAR_PLUS) refsign = *(++ptr);
4602            }            }
4603    
4604          /* 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 4221  we set the flag only if there is a liter Line 4618  we set the flag only if there is a liter
4618            {            {
4619            if (recno >= 0)            if (recno >= 0)
4620              recno = ((digitab[*ptr] & ctype_digit) != 0)?              recno = ((digitab[*ptr] & ctype_digit) != 0)?
4621                recno * 10 + *ptr - '0' : -1;                recno * 10 + *ptr - CHAR_0 : -1;
4622            ptr++;            ptr++;
4623            }            }
4624          namelen = ptr - name;          namelen = ptr - name;
4625    
4626          if ((terminator > 0 && *ptr++ != terminator) || *ptr++ != ')')          if ((terminator > 0 && *ptr++ != terminator) ||
4627                *ptr++ != CHAR_RIGHT_PARENTHESIS)
4628            {            {
4629            ptr--;      /* Error offset */            ptr--;      /* Error offset */
4630            *errorcodeptr = ERR26;            *errorcodeptr = ERR26;
# Line 4248  we set the flag only if there is a liter Line 4646  we set the flag only if there is a liter
4646              *errorcodeptr = ERR58;              *errorcodeptr = ERR58;
4647              goto FAILED;              goto FAILED;
4648              }              }
4649            recno = (refsign == '-')?            recno = (refsign == CHAR_MINUS)?
4650              cd->bracount - recno + 1 : recno +cd->bracount;              cd->bracount - recno + 1 : recno +cd->bracount;
4651            if (recno <= 0 || recno > cd->final_bracount)            if (recno <= 0 || recno > cd->final_bracount)
4652              {              {
# Line 4260  we set the flag only if there is a liter Line 4658  we set the flag only if there is a liter
4658            }            }
4659    
4660          /* Otherwise (did not start with "+" or "-"), start by looking for the          /* Otherwise (did not start with "+" or "-"), start by looking for the
4661          name. */          name. If we find a name, add one to the opcode to change OP_CREF or
4662            OP_RREF into OP_NCREF or OP_NRREF. These behave exactly the same,
4663            except they record that the reference was originally to a name. The
4664            information is used to check duplicate names. */
4665    
4666          slot = cd->name_table;          slot = cd->name_table;
4667          for (i = 0; i < cd->names_found; i++)          for (i = 0; i < cd->names_found; i++)
# Line 4275  we set the flag only if there is a liter Line 4676  we set the flag only if there is a liter
4676            {            {
4677            recno = GET2(slot, 0);            recno = GET2(slot, 0);
4678            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4679              code[1+LINK_SIZE]++;
4680            }            }
4681    
4682          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
4683    
4684          else if ((i = find_parens(ptr, cd->bracount, name, namelen,          else if ((i = find_parens(cd, name, namelen,
4685                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0)) > 0)
4686            {            {
4687            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
4688              code[1+LINK_SIZE]++;
4689            }            }
4690    
4691          /* If terminator == 0 it means that the name followed directly after          /* If terminator == 0 it means that the name followed directly after
# Line 4300  we set the flag only if there is a liter Line 4703  we set the flag only if there is a liter
4703          /* Check for (?(R) for recursion. Allow digits after R to specify a          /* Check for (?(R) for recursion. Allow digits after R to specify a
4704          specific group number. */          specific group number. */
4705    
4706          else if (*name == 'R')          else if (*name == CHAR_R)
4707            {            {
4708            recno = 0;            recno = 0;
4709            for (i = 1; i < namelen; i++)            for (i = 1; i < namelen; i++)
# Line 4310  we set the flag only if there is a liter Line 4713  we set the flag only if there is a liter
4713                *errorcodeptr = ERR15;                *errorcodeptr = ERR15;
4714                goto FAILED;                goto FAILED;
4715                }                }
4716              recno = recno * 10 + name[i] - '0';              recno = recno * 10 + name[i] - CHAR_0;
4717              }              }
4718            if (recno == 0) recno = RREF_ANY;            if (recno == 0) recno = RREF_ANY;
4719            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */
# Line 4320  we set the flag only if there is a liter Line 4723  we set the flag only if there is a liter
4723          /* Similarly, check for the (?(DEFINE) "condition", which is always          /* Similarly, check for the (?(DEFINE) "condition", which is always
4724          false. */          false. */
4725    
4726          else if (namelen == 6 && strncmp((char *)name, "DEFINE", 6) == 0)          else if (namelen == 6 && strncmp((char *)name, STRING_DEFINE, 6) == 0)
4727            {            {
4728            code[1+LINK_SIZE] = OP_DEF;            code[1+LINK_SIZE] = OP_DEF;
4729            skipbytes = 1;            skipbytes = 1;
4730            }            }
4731    
4732          /* Check for the "name" actually being a subpattern number. We are          /* Check for the "name" actually being a subpattern number. We are
4733          in the second pass here, so final_bracount is set. */          in the second pass here, so final_bracount is set. */
4734    
4735          else if (recno > 0 && recno <= cd->final_bracount)          else if (recno > 0 && recno <= cd->final_bracount)
4736            {            {
# Line 4345  we set the flag only if there is a liter Line 4748  we set the flag only if there is a liter
4748    
4749    
4750          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4751          case '=':                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
4752          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
4753          ptr++;          ptr++;
4754          break;          break;
4755    
4756    
4757          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4758          case '!':                 /* Negative lookahead */          case CHAR_EXCLAMATION_MARK:            /* Negative lookahead */
4759          ptr++;          ptr++;
4760          if (*ptr == ')')          /* Optimize (?!) */          if (*ptr == CHAR_RIGHT_PARENTHESIS)    /* Optimize (?!) */
4761            {            {
4762            *code++ = OP_FAIL;            *code++ = OP_FAIL;
4763            previous = NULL;            previous = NULL;
# Line 4365  we set the flag only if there is a liter Line 4768  we set the flag only if there is a liter
4768    
4769    
4770          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4771          case '<':                 /* Lookbehind or named define */          case CHAR_LESS_THAN_SIGN:              /* Lookbehind or named define */
4772          switch (ptr[1])          switch (ptr[1])
4773            {            {
4774            case '=':               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
4775            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
4776            ptr += 2;            ptr += 2;
4777            break;            break;
4778    
4779            case '!':               /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
4780            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
4781            ptr += 2;            ptr += 2;
4782            break;            break;
# Line 4388  we set the flag only if there is a liter Line 4791  we set the flag only if there is a liter
4791    
4792    
4793          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4794          case '>':                 /* One-time brackets */          case CHAR_GREATER_THAN_SIGN:           /* One-time brackets */
4795          bravalue = OP_ONCE;          bravalue = OP_ONCE;
4796          ptr++;          ptr++;
4797          break;          break;
4798    
4799    
4800          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4801          case 'C':                 /* Callout - may be followed by digits; */          case CHAR_C:                 /* Callout - may be followed by digits; */
4802          previous_callout = code;  /* Save for later completion */          previous_callout = code;  /* Save for later completion */
4803          after_manual_callout = 1; /* Skip one item before completing */          after_manual_callout = 1; /* Skip one item before completing */
4804          *code++ = OP_CALLOUT;          *code++ = OP_CALLOUT;
4805            {            {
4806            int n = 0;            int n = 0;
4807            while ((digitab[*(++ptr)] & ctype_digit) != 0)            while ((digitab[*(++ptr)] & ctype_digit) != 0)
4808              n = n * 10 + *ptr - '0';              n = n * 10 + *ptr - CHAR_0;
4809            if (*ptr != ')')            if (*ptr != CHAR_RIGHT_PARENTHESIS)
4810              {              {
4811              *errorcodeptr = ERR39;              *errorcodeptr = ERR39;
4812              goto FAILED;              goto FAILED;
# Line 4423  we set the flag only if there is a liter Line 4826  we set the flag only if there is a liter
4826    
4827    
4828          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4829          case 'P':                 /* Python-style named subpattern handling */          case CHAR_P:              /* Python-style named subpattern handling */
4830          if (*(++ptr) == '=' || *ptr == '>')  /* Reference or recursion */          if (*(++ptr) == CHAR_EQUALS_SIGN ||
4831                *ptr == CHAR_GREATER_THAN_SIGN)  /* Reference or recursion */
4832            {            {
4833            is_recurse = *ptr == '>';            is_recurse = *ptr == CHAR_GREATER_THAN_SIGN;
4834            terminator = ')';            terminator = CHAR_RIGHT_PARENTHESIS;
4835            goto NAMED_REF_OR_RECURSE;            goto NAMED_REF_OR_RECURSE;
4836            }            }
4837          else if (*ptr != '<')    /* Test for Python-style definition */          else if (*ptr != CHAR_LESS_THAN_SIGN)  /* Test for Python-style defn */
4838            {            {
4839            *errorcodeptr = ERR41;            *errorcodeptr = ERR41;
4840            goto FAILED;            goto FAILED;
# Line 4440  we set the flag only if there is a liter Line 4844  we set the flag only if there is a liter
4844    
4845          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4846          DEFINE_NAME:    /* Come here from (?< handling */          DEFINE_NAME:    /* Come here from (?< handling */
4847          case '\'':          case CHAR_APOSTROPHE:
4848            {            {
4849            terminator = (*ptr == '<')? '>' : '\'';            terminator = (*ptr == CHAR_LESS_THAN_SIGN)?
4850                CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
4851            name = ++ptr;            name = ++ptr;
4852    
4853            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
# Line 4473  we set the flag only if there is a liter Line 4878  we set the flag only if there is a liter
4878                }                }
4879              }              }
4880    
4881            /* In the real compile, create the entry in the table */            /* In the real compile, create the entry in the table, maintaining
4882              alphabetical order. Duplicate names for different numbers are
4883              permitted only if PCRE_DUPNAMES is set. Duplicate names for the same
4884              number are always OK. (An existing number can be re-used if (?|
4885              appears in the pattern.) In either event, a duplicate name results in
4886              a duplicate entry in the table, even if the number is the same. This
4887              is because the number of names, and hence the table size, is computed
4888              in the pre-compile, and it affects various numbers and pointers which
4889              would all have to be modified, and the compiled code moved down, if
4890              duplicates with the same number were omitted from the table. This
4891              doesn't seem worth the hassle. However, *different* names for the
4892              same number are not permitted. */
4893    
4894            else            else
4895              {              {
4896                BOOL dupname = FALSE;
4897              slot = cd->name_table;              slot = cd->name_table;
4898    
4899              for (i = 0; i < cd->names_found; i++)              for (i = 0; i < cd->names_found; i++)
4900                {                {
4901                int crc = memcmp(name, slot+2, namelen);                int crc = memcmp(name, slot+2, namelen);
# Line 4485  we set the flag only if there is a liter Line 4903  we set the flag only if there is a liter
4903                  {                  {
4904                  if (slot[2+namelen] == 0)                  if (slot[2+namelen] == 0)
4905                    {                    {
4906                    if ((options & PCRE_DUPNAMES) == 0)                    if (GET2(slot, 0) != cd->bracount + 1 &&
4907                          (options & PCRE_DUPNAMES) == 0)
4908                      {                      {
4909                      *errorcodeptr = ERR43;                      *errorcodeptr = ERR43;
4910                      goto FAILED;                      goto FAILED;
4911                      }                      }
4912                      else dupname = TRUE;
4913                    }                    }
4914                  else crc = -1;      /* Current name is substring */                  else crc = -1;      /* Current name is a substring */
4915                  }                  }
4916    
4917                  /* Make space in the table and break the loop for an earlier
4918                  name. For a duplicate or later name, carry on. We do this for
4919                  duplicates so that in the simple case (when ?(| is not used) they
4920                  are in order of their numbers. */
4921    
4922                if (crc < 0)                if (crc < 0)
4923                  {                  {
4924                  memmove(slot + cd->name_entry_size, slot,                  memmove(slot + cd->name_entry_size, slot,
4925                    (cd->names_found - i) * cd->name_entry_size);                    (cd->names_found - i) * cd->name_entry_size);
4926                  break;                  break;
4927                  }                  }
4928    
4929                  /* Continue the loop for a later or duplicate name */
4930    
4931                slot += cd->name_entry_size;                slot += cd->name_entry_size;
4932                }                }
4933    
4934                /* For non-duplicate names, check for a duplicate number before
4935                adding the new name. */
4936    
4937                if (!dupname)
4938                  {
4939                  uschar *cslot = cd->name_table;
4940                  for (i = 0; i < cd->names_found; i++)
4941                    {
4942                    if (cslot != slot)
4943                      {
4944                      if (GET2(cslot, 0) == cd->bracount + 1)
4945                        {
4946                        *errorcodeptr = ERR65;
4947                        goto FAILED;
4948                        }
4949                      }
4950                    else i--;
4951                    cslot += cd->name_entry_size;
4952                    }
4953                  }
4954    
4955              PUT2(slot, 0, cd->bracount + 1);              PUT2(slot, 0, cd->bracount + 1);
4956              memcpy(slot + 2, name, namelen);              memcpy(slot + 2, name, namelen);
4957              slot[2+namelen] = 0;              slot[2+namelen] = 0;
4958              }              }
4959            }            }
4960    
4961          /* In both cases, count the number of names we've encountered. */          /* In both pre-compile and compile, count the number of names we've
4962            encountered. */
4963    
         ptr++;                    /* Move past > or ' */  
4964          cd->names_found++;          cd->names_found++;
4965            ptr++;                    /* Move past > or ' */
4966          goto NUMBERED_GROUP;          goto NUMBERED_GROUP;
4967    
4968    
4969          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4970          case '&':                 /* Perl recursion/subroutine syntax */          case CHAR_AMPERSAND:            /* Perl recursion/subroutine syntax */
4971          terminator = ')';          terminator = CHAR_RIGHT_PARENTHESIS;
4972          is_recurse = TRUE;          is_recurse = TRUE;
4973          /* Fall through */          /* Fall through */
4974    
# Line 4525  we set the flag only if there is a liter Line 4976  we set the flag only if there is a liter
4976          references (?P=name) and recursion (?P>name), as well as falling          references (?P=name) and recursion (?P>name), as well as falling
4977          through from the Perl recursion syntax (?&name). We also come here from          through from the Perl recursion syntax (?&name). We also come here from
4978          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}
4979          .NET syntax. */          .NET syntax, and the Oniguruma \g<...> and \g'...' subroutine syntax. */
4980    
4981          NAMED_REF_OR_RECURSE:          NAMED_REF_OR_RECURSE:
4982          name = ++ptr;          name = ++ptr;
# Line 4541  we set the flag only if there is a liter Line 4992  we set the flag only if there is a liter
4992              {              {
4993              *errorcodeptr = ERR62;              *errorcodeptr = ERR62;
4994              goto FAILED;              goto FAILED;
4995              }              }
4996            if (*ptr != terminator)            if (*ptr != terminator)
4997              {              {
4998              *errorcodeptr = ERR42;              *errorcodeptr = ERR42;
# Line 4555  we set the flag only if there is a liter Line 5006  we set the flag only if there is a liter
5006            recno = 0;            recno = 0;
5007            }            }
5008    
5009          /* 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
5010          first, and then check that we have reached the end of the name in the          first, and then check that we have reached the end of the name in the
5011          table. That way, if the name that is longer than any in the table,          table. That way, if the name that is longer than any in the table,
5012          the comparison will fail without reading beyond the table entry. */          the comparison will fail without reading beyond the table entry. */
5013    
# Line 4566  we set the flag only if there is a liter Line 5017  we set the flag only if there is a liter
5017            for (i = 0; i < cd->names_found; i++)            for (i = 0; i < cd->names_found; i++)
5018              {              {
5019              if (strncmp((char *)name, (char *)slot+2, namelen) == 0 &&              if (strncmp((char *)name, (char *)slot+2, namelen) == 0 &&
5020                  slot[2+namelen] == 0)                  slot[2+namelen] == 0)
5021                break;                break;
5022              slot += cd->name_entry_size;              slot += cd->name_entry_size;
5023              }              }
# Line 4576  we set the flag only if there is a liter Line 5027  we set the flag only if there is a liter
5027              recno = GET2(slot, 0);              recno = GET2(slot, 0);
5028              }              }
5029            else if ((recno =                /* Forward back reference */            else if ((recno =                /* Forward back reference */
5030                      find_parens(ptr, cd->bracount, name, namelen,                      find_parens(cd, name, namelen,
5031                        (options & PCRE_EXTENDED) != 0)) <= 0)                        (options & PCRE_EXTENDED) != 0)) <= 0)
5032              {              {
5033              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
# Line 4592  we set the flag only if there is a liter Line 5043  we set the flag only if there is a liter
5043    
5044    
5045          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5046          case 'R':                 /* Recursion */          case CHAR_R:              /* Recursion */
5047          ptr++;                    /* Same as (?0)      */          ptr++;                    /* Same as (?0)      */
5048          /* Fall through */          /* Fall through */
5049    
5050    
5051          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5052          case '-': case '+':          case CHAR_MINUS: case CHAR_PLUS:  /* Recursion or subroutine */
5053          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:
5054          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:
5055            {            {
5056            const uschar *called;            const uschar *called;
5057              terminator = CHAR_RIGHT_PARENTHESIS;
5058    
5059              /* Come here from the \g<...> and \g'...' code (Oniguruma
5060              compatibility). However, the syntax has been checked to ensure that
5061              the ... are a (signed) number, so that neither ERR63 nor ERR29 will
5062              be called on this path, nor with the jump to OTHER_CHAR_AFTER_QUERY
5063              ever be taken. */
5064    
5065            if ((refsign = *ptr) == '+')            HANDLE_NUMERICAL_RECURSION:
5066    
5067              if ((refsign = *ptr) == CHAR_PLUS)
5068              {              {
5069              ptr++;              ptr++;
5070              if ((digitab[*ptr] & ctype_digit) == 0)              if ((digitab[*ptr] & ctype_digit) == 0)
5071                {                {
5072                *errorcodeptr = ERR63;                *errorcodeptr = ERR63;
5073                goto FAILED;                goto FAILED;
5074                }                }
5075              }              }
5076            else if (refsign == '-')            else if (refsign == CHAR_MINUS)
5077              {              {
5078              if ((digitab[ptr[1]] & ctype_digit) == 0)              if ((digitab[ptr[1]] & ctype_digit) == 0)
5079                goto OTHER_CHAR_AFTER_QUERY;                goto OTHER_CHAR_AFTER_QUERY;
# Line 4622  we set the flag only if there is a liter Line 5082  we set the flag only if there is a liter
5082    
5083            recno = 0;            recno = 0;
5084            while((digitab[*ptr] & ctype_digit) != 0)            while((digitab[*ptr] & ctype_digit) != 0)
5085              recno = recno * 10 + *ptr++ - '0';              recno = recno * 10 + *ptr++ - CHAR_0;
5086    
5087            if (*ptr != ')')            if (*ptr != terminator)
5088              {              {
5089              *errorcodeptr = ERR29;              *errorcodeptr = ERR29;
5090              goto FAILED;              goto FAILED;
5091              }              }
5092    
5093            if (refsign == '-')            if (refsign == CHAR_MINUS)
5094              {              {
5095              if (recno == 0)              if (recno == 0)
5096                {                {
# Line 4644  we set the flag only if there is a liter Line 5104  we set the flag only if there is a liter
5104                goto FAILED;                goto FAILED;
5105                }                }
5106              }              }
5107            else if (refsign == '+')            else if (refsign == CHAR_PLUS)
5108              {              {
5109              if (recno == 0)              if (recno == 0)
5110                {                {
# Line 4671  we set the flag only if there is a liter Line 5131  we set the flag only if there is a liter
5131            if (lengthptr == NULL)            if (lengthptr == NULL)
5132              {              {
5133              *code = OP_END;              *code = OP_END;
5134              if (recno != 0) called = find_bracket(cd->start_code, utf8, recno);              if (recno != 0)
5135                  called = _pcre_find_bracket(cd->start_code, utf8, recno);
5136    
5137              /* Forward reference */              /* Forward reference */
5138    
5139              if (called == NULL)              if (called == NULL)
5140                {                {
5141                if (find_parens(ptr, cd->bracount, NULL, recno,                if (find_parens(cd, NULL, recno,
5142                     (options & PCRE_EXTENDED) != 0) < 0)                      (options & PCRE_EXTENDED) != 0) < 0)
5143                  {                  {
5144                  *errorcodeptr = ERR15;                  *errorcodeptr = ERR15;
5145                  goto FAILED;                  goto FAILED;
# Line 4730  we set the flag only if there is a liter Line 5191  we set the flag only if there is a liter
5191          set = unset = 0;          set = unset = 0;
5192          optset = &set;          optset = &set;
5193    
5194          while (*ptr != ')' && *ptr != ':')          while (*ptr != CHAR_RIGHT_PARENTHESIS && *ptr != CHAR_COLON)
5195            {            {
5196            switch (*ptr++)            switch (*ptr++)
5197              {              {
5198              case '-': optset = &unset; break;              case CHAR_MINUS: optset = &unset; break;
5199    
5200              case 'J':    /* Record that it changed in the external options */              case CHAR_J:    /* Record that it changed in the external options */
5201              *optset |= PCRE_DUPNAMES;              *optset |= PCRE_DUPNAMES;
5202              cd->external_flags |= PCRE_JCHANGED;              cd->external_flags |= PCRE_JCHANGED;
5203              break;              break;
5204    
5205              case 'i': *optset |= PCRE_CASELESS; break;              case CHAR_i: *optset |= PCRE_CASELESS; break;
5206              case 'm': *optset |= PCRE_MULTILINE; break;              case CHAR_m: *optset |= PCRE_MULTILINE; break;
5207              case 's': *optset |= PCRE_DOTALL; break;              case CHAR_s: *optset |= PCRE_DOTALL; break;
5208              case 'x': *optset |= PCRE_EXTENDED; break;              case CHAR_x: *optset |= PCRE_EXTENDED; break;
5209              case 'U': *optset |= PCRE_UNGREEDY; break;              case CHAR_U: *optset |= PCRE_UNGREEDY; break;
5210              case 'X': *optset |= PCRE_EXTRA; break;              case CHAR_X: *optset |= PCRE_EXTRA; break;
5211    
5212              default:  *errorcodeptr = ERR12;              default:  *errorcodeptr = ERR12;
5213                        ptr--;    /* Correct the offset */                        ptr--;    /* Correct the offset */
# Line 4777  we set the flag only if there is a liter Line 5238  we set the flag only if there is a liter
5238          both phases.          both phases.
5239    
5240          If we are not at the pattern start, compile code to change the ims          If we are not at the pattern start, compile code to change the ims
5241          options if this setting actually changes any of them. We also pass the          options if this setting actually changes any of them, and reset the
5242          new setting back so that it can be put at the start of any following          greedy defaults and the case value for firstbyte and reqbyte. */
         branches, and when this group ends (if we are in a group), a resetting  
         item can be compiled. */  
5243    
5244          if (*ptr == ')')          if (*ptr == CHAR_RIGHT_PARENTHESIS)
5245            {            {
5246            if (code == cd->start_code + 1 + LINK_SIZE &&            if (code == cd->start_code + 1 + LINK_SIZE &&
5247                 (lengthptr == NULL || *lengthptr == 2 + 2*LINK_SIZE))                 (lengthptr == NULL || *lengthptr == 2 + 2*LINK_SIZE))
5248              {              {
5249              cd->external_options = newoptions;              cd->external_options = newoptions;
             options = newoptions;  
5250              }              }
5251           else           else
5252              {              {
# Line 4797  we set the flag only if there is a liter Line 5255  we set the flag only if there is a liter
5255                *code++ = OP_OPT;                *code++ = OP_OPT;
5256                *code++ = newoptions & PCRE_IMS;                *code++ = newoptions & PCRE_IMS;
5257                }                }
   
             /* Change options at this level, and pass them back for use  
             in subsequent branches. Reset the greedy defaults and the case  
             value for firstbyte and reqbyte. */  
   
             *optionsptr = options = newoptions;  
5258              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);
5259              greedy_non_default = greedy_default ^ 1;              greedy_non_default = greedy_default ^ 1;
5260              req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
5261              }              }
5262    
5263              /* Change options at this level, and pass them back for use
5264              in subsequent branches. When not at the start of the pattern, this
5265              information is also necessary so that a resetting item can be
5266              compiled at the end of a group (if we are in a group). */
5267    
5268              *optionsptr = options = newoptions;
5269            previous = NULL;       /* This item can't be repeated */            previous = NULL;       /* This item can't be repeated */
5270            continue;              /* It is complete */            continue;              /* It is complete */
5271            }            }
# Line 4923  we set the flag only if there is a liter Line 5381  we set the flag only if there is a liter
5381    
5382      /* Error if hit end of pattern */      /* Error if hit end of pattern */
5383    
5384      if (*ptr != ')')      if (*ptr != CHAR_RIGHT_PARENTHESIS)
5385        {        {
5386        *errorcodeptr = ERR14;        *errorcodeptr = ERR14;
5387        goto FAILED;        goto FAILED;
# Line 5021  we set the flag only if there is a liter Line 5479  we set the flag only if there is a liter
5479      We can test for values between ESC_b and ESC_Z for the latter; this may      We can test for values between ESC_b and ESC_Z for the latter; this may
5480      have to change if any new ones are ever created. */      have to change if any new ones are ever created. */
5481    
5482      case '\\':      case CHAR_BACKSLASH:
5483      tempptr = ptr;      tempptr = ptr;
5484      c = check_escape(&ptr, errorcodeptr, cd->bracount, options, FALSE);      c = check_escape(&ptr, errorcodeptr, cd->bracount, options, FALSE);
5485      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
# Line 5030  we set the flag only if there is a liter Line 5488  we set the flag only if there is a liter
5488        {        {
5489        if (-c == ESC_Q)            /* Handle start of quoted string */        if (-c == ESC_Q)            /* Handle start of quoted string */
5490          {          {
5491          if (ptr[1] == '\\' && ptr[2] == 'E') ptr += 2; /* avoid empty string */          if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
5492            else inescq = TRUE;            ptr += 2;               /* avoid empty string */
5493                else inescq = TRUE;
5494          continue;          continue;
5495          }          }
5496    
# Line 5048  we set the flag only if there is a liter Line 5507  we set the flag only if there is a liter
5507        zerofirstbyte = firstbyte;        zerofirstbyte = firstbyte;
5508        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
5509    
5510          /* \g<name> or \g'name' is a subroutine call by name and \g<n> or \g'n'
5511          is a subroutine call by number (Oniguruma syntax). In fact, the value
5512          -ESC_g is returned only for these cases. So we don't need to check for <
5513          or ' if the value is -ESC_g. For the Perl syntax \g{n} the value is
5514          -ESC_REF+n, and for the Perl syntax \g{name} the result is -ESC_k (as
5515          that is a synonym for a named back reference). */
5516    
5517          if (-c == ESC_g)
5518            {
5519            const uschar *p;
5520            save_hwm = cd->hwm;   /* Normally this is set when '(' is read */
5521            terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
5522              CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
5523    
5524            /* These two statements stop the compiler for warning about possibly
5525            unset variables caused by the jump to HANDLE_NUMERICAL_RECURSION. In
5526            fact, because we actually check for a number below, the paths that
5527            would actually be in error are never taken. */
5528    
5529            skipbytes = 0;
5530            reset_bracount = FALSE;
5531    
5532            /* Test for a name */
5533    
5534            if (ptr[1] != CHAR_PLUS && ptr[1] != CHAR_MINUS)
5535              {
5536              BOOL isnumber = TRUE;
5537              for (p = ptr + 1; *p != 0 && *p != terminator; p++)
5538                {
5539                if ((cd->ctypes[*p] & ctype_digit) == 0) isnumber = FALSE;
5540                if ((cd->ctypes[*p] & ctype_word) == 0) break;
5541                }
5542              if (*p != terminator)
5543                {
5544                *errorcodeptr = ERR57;
5545                break;
5546                }
5547              if (isnumber)
5548                {
5549                ptr++;
5550                goto HANDLE_NUMERICAL_RECURSION;
5551                }
5552              is_recurse = TRUE;
5553              goto NAMED_REF_OR_RECURSE;
5554              }
5555    
5556            /* Test a signed number in angle brackets or quotes. */
5557    
5558            p = ptr + 2;
5559            while ((digitab[*p] & ctype_digit) != 0) p++;
5560            if (*p != terminator)
5561              {
5562              *errorcodeptr = ERR57;
5563              break;
5564              }
5565            ptr++;
5566            goto HANDLE_NUMERICAL_RECURSION;
5567            }
5568    
5569        /* \k<name> or \k'name' is a back reference by name (Perl syntax).        /* \k<name> or \k'name' is a back reference by name (Perl syntax).
5570        We also support \k{name} (.NET syntax) */        We also support \k{name} (.NET syntax) */
5571    
5572        if (-c == ESC_k && (ptr[1] == '<' || ptr[1] == '\'' || ptr[1] == '{'))        if (-c == ESC_k && (ptr[1] == CHAR_LESS_THAN_SIGN ||
5573              ptr[1] == CHAR_APOSTROPHE || ptr[1] == CHAR_LEFT_CURLY_BRACKET))
5574          {          {
5575          is_recurse = FALSE;          is_recurse = FALSE;
5576          terminator = (*(++ptr) == '<')? '>' : (*ptr == '\'')? '\'' : '}';          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
5577              CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?
5578              CHAR_APOSTROPHE : CHAR_RIGHT_CURLY_BRACKET;
5579          goto NAMED_REF_OR_RECURSE;          goto NAMED_REF_OR_RECURSE;
5580          }          }
5581    
# Line 5157  we set the flag only if there is a liter Line 5678  we set the flag only if there is a liter
5678    
5679      /* Remember if \r or \n were seen */      /* Remember if \r or \n were seen */
5680    
5681      if (mcbuffer[0] == '\r' || mcbuffer[0] == '\n')      if (mcbuffer[0] == CHAR_CR || mcbuffer[0] == CHAR_NL)
5682        cd->external_flags |= PCRE_HASCRORLF;        cd->external_flags |= PCRE_HASCRORLF;
5683    
5684      /* Set the first and required bytes appropriately. If no previous first      /* Set the first and required bytes appropriately. If no previous first
# Line 5255  uschar *code = *codeptr; Line 5776  uschar *code = *codeptr;
5776  uschar *last_branch = code;  uschar *last_branch = code;
5777  uschar *start_bracket = code;  uschar *start_bracket = code;
5778  uschar *reverse_count = NULL;  uschar *reverse_count = NULL;
5779    open_capitem capitem;
5780    int capnumber = 0;
5781  int firstbyte, reqbyte;  int firstbyte, reqbyte;
5782  int branchfirstbyte, branchreqbyte;  int branchfirstbyte, branchreqbyte;
5783  int length;  int length;
# Line 5281  the code that abstracts option settings Line 5804  the code that abstracts option settings
5804  them global. It tests the value of length for (2 + 2*LINK_SIZE) in the  them global. It tests the value of length for (2 + 2*LINK_SIZE) in the
5805  pre-compile phase to find out whether anything has yet been compiled or not. */  pre-compile phase to find out whether anything has yet been compiled or not. */
5806    
5807    /* If this is a capturing subpattern, add to the chain of open capturing items
5808    so that we can detect them if (*ACCEPT) is encountered. */
5809    
5810    if (*code == OP_CBRA)
5811      {
5812      capnumber = GET2(code, 1 + LINK_SIZE);
5813      capitem.number = capnumber;
5814      capitem.next = cd->open_caps;
5815      cd->open_caps = &capitem;
5816      }
5817    
5818  /* Offset is set zero to mark that this bracket is still open */  /* Offset is set zero to mark that this bracket is still open */
5819    
5820  PUT(code, 1, 0);  PUT(code, 1, 0);
# Line 5375  for (;;) Line 5909  for (;;)
5909    
5910      /* If lookbehind, check that this branch matches a fixed-length string, and      /* If lookbehind, check that this branch matches a fixed-length string, and
5911      put the length into the OP_REVERSE item. Temporarily mark the end of the      put the length into the OP_REVERSE item. Temporarily mark the end of the
5912      branch with OP_END. */      branch with OP_END. If the branch contains OP_RECURSE, the result is -3
5913        because there may be forward references that we can't check here. Set a
5914        flag to cause another lookbehind check at the end. Why not do it all at the
5915        end? Because common, erroneous checks are picked up here and the offset of
5916        the problem can be shown. */
5917    
5918      if (lookbehind)      if (lookbehind)
5919        {        {
5920        int fixed_length;        int fixed_length;
5921        *code = OP_END;        *code = OP_END;
5922        fixed_length = find_fixedlength(last_branch, options);        fixed_length = find_fixedlength(last_branch, options, FALSE, cd);
5923        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
5924        if (fixed_length < 0)        if (fixed_length == -3)
5925            {
5926            cd->check_lookbehind = TRUE;
5927            }
5928          else if (fixed_length < 0)
5929          {          {
5930          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;
5931          *ptrptr = ptr;          *ptrptr = ptr;
5932          return FALSE;          return FALSE;
5933          }          }
5934        PUT(reverse_count, 0, fixed_length);        else { PUT(reverse_count, 0, fixed_length); }
5935        }        }
5936      }      }
5937    
# Line 5402  for (;;) Line 5944  for (;;)
5944    compile a resetting op-code following, except at the very end of the pattern.    compile a resetting op-code following, except at the very end of the pattern.
5945    Return leaving the pointer at the terminating char. */    Return leaving the pointer at the terminating char. */
5946    
5947    if (*ptr != '|')    if (*ptr != CHAR_VERTICAL_LINE)
5948      {      {
5949      if (lengthptr == NULL)      if (lengthptr == NULL)
5950        {        {
# Line 5417  for (;;) Line 5959  for (;;)
5959        while (branch_length > 0);        while (branch_length > 0);
5960        }        }
5961    
5962        /* If it was a capturing subpattern, remove it from the chain. */
5963    
5964        if (capnumber > 0) cd->open_caps = cd->open_caps->next;
5965    
5966      /* Fill in the ket */      /* Fill in the ket */
5967    
5968      *code = OP_KET;      *code = OP_KET;
# Line 5425  for (;;) Line 5971  for (;;)
5971    
5972      /* Resetting option if needed */      /* Resetting option if needed */
5973    
5974      if ((options & PCRE_IMS) != oldims && *ptr == ')')      if ((options & PCRE_IMS) != oldims && *ptr == CHAR_RIGHT_PARENTHESIS)
5975        {        {
5976        *code++ = OP_OPT;        *code++ = OP_OPT;
5977        *code++ = oldims;        *code++ = oldims;
# Line 5554  do { Line 6100  do {
6100       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;
6101       }       }
6102    
6103     /* .* is not anchored unless DOTALL is set and it isn't in brackets that     /* .* is not anchored unless DOTALL is set (which generates OP_ALLANY) and
6104     are or may be referenced. */     it isn't in brackets that are or may be referenced. */
6105    
6106     else if ((op == OP_TYPESTAR || op == OP_TYPEMINSTAR ||     else if ((op == OP_TYPESTAR || op == OP_TYPEMINSTAR ||
6107               op == OP_TYPEPOSSTAR) &&               op == OP_TYPEPOSSTAR))
             (*options & PCRE_DOTALL) != 0)  
6108       {       {
6109       if (scode[1] != OP_ANY || (bracket_map & backref_map) != 0) return FALSE;       if (scode[1] != OP_ALLANY || (bracket_map & backref_map) != 0)
6110           return FALSE;
6111       }       }
6112    
6113     /* Check for explicit anchoring */     /* Check for explicit anchoring */
# Line 5607  do { Line 6153  do {
6153       NULL, 0, FALSE);       NULL, 0, FALSE);
6154     register int op = *scode;     register int op = *scode;
6155    
6156       /* If we are at the start of a conditional assertion group, *both* the
6157       conditional assertion *and* what follows the condition must satisfy the test
6158       for start of line. Other kinds of condition fail. Note that there may be an
6159       auto-callout at the start of a condition. */
6160    
6161       if (op == OP_COND)
6162         {
6163         scode += 1 + LINK_SIZE;
6164         if (*scode == OP_CALLOUT) scode += _pcre_OP_lengths[OP_CALLOUT];
6165         switch (*scode)
6166           {
6167           case OP_CREF:
6168           case OP_NCREF:
6169           case OP_RREF:
6170           case OP_NRREF:
6171           case OP_DEF:
6172           return FALSE;
6173    
6174           default:     /* Assertion */
6175           if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
6176           do scode += GET(scode, 1); while (*scode == OP_ALT);
6177           scode += 1 + LINK_SIZE;
6178           break;
6179           }
6180         scode = first_significant_code(scode, NULL, 0, FALSE);
6181         op = *scode;
6182         }
6183    
6184     /* Non-capturing brackets */     /* Non-capturing brackets */
6185    
6186     if (op == OP_BRA)     if (op == OP_BRA)
# Line 5625  do { Line 6199  do {
6199    
6200     /* Other brackets */     /* Other brackets */
6201    
6202     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_COND)     else if (op == OP_ASSERT || op == OP_ONCE)
6203       { if (!is_startline(scode, bracket_map, backref_map)) return FALSE; }       {
6204         if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
6205         }
6206    
6207     /* .* means "start at start or after \n" if it isn't in brackets that     /* .* means "start at start or after \n" if it isn't in brackets that
6208     may be referenced. */     may be referenced. */
# Line 5743  Returns: pointer to compiled data Line 6319  Returns: pointer to compiled data
6319                  with errorptr and erroroffset set                  with errorptr and erroroffset set
6320  */  */
6321    
6322  PCRE_EXP_DEFN pcre *  PCRE_EXP_DEFN pcre * PCRE_CALL_CONVENTION
6323  pcre_compile(const char *pattern, int options, const char **errorptr,  pcre_compile(const char *pattern, int options, const char **errorptr,
6324    int *erroroffset, const unsigned char *tables)    int *erroroffset, const unsigned char *tables)
6325  {  {
# Line 5751  return pcre_compile2(pattern, options, N Line 6327  return pcre_compile2(pattern, options, N
6327  }  }
6328    
6329    
6330  PCRE_EXP_DEFN pcre *  PCRE_EXP_DEFN pcre * PCRE_CALL_CONVENTION
6331  pcre_compile2(const char *pattern, int options, int *errorcodeptr,  pcre_compile2(const char *pattern, int options, int *errorcodeptr,
6332    const char **errorptr, int *erroroffset, const unsigned char *tables)    const char **errorptr, int *erroroffset, const unsigned char *tables)
6333  {  {
# Line 5760  int length = 1; /* For final END opcode Line 6336  int length = 1; /* For final END opcode
6336  int firstbyte, reqbyte, newline;  int firstbyte, reqbyte, newline;
6337  int errorcode = 0;  int errorcode = 0;
6338  int skipatstart = 0;  int skipatstart = 0;
6339  #ifdef SUPPORT_UTF8  BOOL utf8 = (options & PCRE_UTF8) != 0;
 BOOL utf8;  
 #endif  
6340  size_t size;  size_t size;
6341  uschar *code;  uschar *code;
6342  const uschar *codestart;  const uschar *codestart;
# Line 5778  to fill in forward references to subpatt Line 6352  to fill in forward references to subpatt
6352    
6353  uschar cworkspace[COMPILE_WORK_SIZE];  uschar cworkspace[COMPILE_WORK_SIZE];
6354    
   
6355  /* Set this early so that early errors get offset 0. */  /* Set this early so that early errors get offset 0. */
6356    
6357  ptr = (const uschar *)pattern;  ptr = (const uschar *)pattern;
# Line 5806  if (erroroffset == NULL) Line 6379  if (erroroffset == NULL)
6379    
6380  *erroroffset = 0;  *erroroffset = 0;
6381    
 /* Can't support UTF8 unless PCRE has been compiled to include the code. */  
   
 #ifdef SUPPORT_UTF8  
 utf8 = (options & PCRE_UTF8) != 0;  
 if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&  
      (*erroroffset = _pcre_valid_utf8((uschar *)pattern, -1)) >= 0)  
   {  
   errorcode = ERR44;  
   goto PCRE_EARLY_ERROR_RETURN2;  
   }  
 #else  
 if ((options & PCRE_UTF8) != 0)  
   {  
   errorcode = ERR32;  
   goto PCRE_EARLY_ERROR_RETURN;  
   }  
 #endif  
   
 if ((options & ~PUBLIC_OPTIONS) != 0)  
   {  
   errorcode = ERR17;  
   goto PCRE_EARLY_ERROR_RETURN;  
   }  
   
6382  /* Set up pointers to the individual character tables */  /* Set up pointers to the individual character tables */
6383    
6384  if (tables == NULL) tables = _pcre_default_tables;  if (tables == NULL) tables = _pcre_default_tables;
# Line 5838  cd->fcc = tables + fcc_offset; Line 6387  cd->fcc = tables + fcc_offset;
6387  cd->cbits = tables + cbits_offset;  cd->cbits = tables + cbits_offset;
6388  cd->ctypes = tables + ctypes_offset;  cd->ctypes = tables + ctypes_offset;
6389    
6390    /* Check that all undefined public option bits are zero */
6391    
6392    if ((options & ~PUBLIC_COMPILE_OPTIONS) != 0)
6393      {
6394      errorcode = ERR17;
6395      goto PCRE_EARLY_ERROR_RETURN;
6396      }
6397    
6398  /* Check for global one-time settings at the start of the pattern, and remember  /* Check for global one-time settings at the start of the pattern, and remember
6399  the offset for later. */  the offset for later. */
6400    
6401  while (ptr[skipatstart] == '(' && ptr[skipatstart+1] == '*')  while (ptr[skipatstart] == CHAR_LEFT_PARENTHESIS &&
6402           ptr[skipatstart+1] == CHAR_ASTERISK)
6403    {    {
6404    int newnl = 0;    int newnl = 0;
6405    int newbsr = 0;    int newbsr = 0;
6406    
6407    if (strncmp((char *)(ptr+skipatstart+2), "CR)", 3) == 0)    if (strncmp((char *)(ptr+skipatstart+2), STRING_UTF8_RIGHTPAR, 5) == 0)
6408        { skipatstart += 7; options |= PCRE_UTF8; continue; }
6409    
6410      if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)
6411      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }
6412    else if (strncmp((char *)(ptr+skipatstart+2), "LF)", 3)  == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_LF_RIGHTPAR, 3)  == 0)
6413      { skipatstart += 5; newnl = PCRE_NEWLINE_LF; }      { skipatstart += 5; newnl = PCRE_NEWLINE_LF; }
6414    else if (strncmp((char *)(ptr+skipatstart+2), "CRLF)", 5)  == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_CRLF_RIGHTPAR, 5)  == 0)
6415      { skipatstart += 7; newnl = PCRE_NEWLINE_CR + PCRE_NEWLINE_LF; }      { skipatstart += 7; newnl = PCRE_NEWLINE_CR + PCRE_NEWLINE_LF; }
6416    else if (strncmp((char *)(ptr+skipatstart+2), "ANY)", 4) == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_ANY_RIGHTPAR, 4) == 0)
6417      { skipatstart += 6; newnl = PCRE_NEWLINE_ANY; }      { skipatstart += 6; newnl = PCRE_NEWLINE_ANY; }
6418    else if (strncmp((char *)(ptr+skipatstart+2), "ANYCRLF)", 8)  == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_ANYCRLF_RIGHTPAR, 8) == 0)
6419      { skipatstart += 10; newnl = PCRE_NEWLINE_ANYCRLF; }      { skipatstart += 10; newnl = PCRE_NEWLINE_ANYCRLF; }
6420    
6421    else if (strncmp((char *)(ptr+skipatstart+2), "BSR_ANYCRLF)", 12) == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_BSR_ANYCRLF_RIGHTPAR, 12) == 0)
6422      { skipatstart += 14; newbsr = PCRE_BSR_ANYCRLF; }      { skipatstart += 14; newbsr = PCRE_BSR_ANYCRLF; }
6423    else if (strncmp((char *)(ptr+skipatstart+2), "BSR_UNICODE)", 12) == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_BSR_UNICODE_RIGHTPAR, 12) == 0)
6424      { skipatstart += 14; newbsr = PCRE_BSR_UNICODE; }      { skipatstart += 14; newbsr = PCRE_BSR_UNICODE; }
6425    
6426    if (newnl != 0)    if (newnl != 0)
# Line 5869  while (ptr[skipatstart] == '(' && ptr[sk Line 6430  while (ptr[skipatstart] == '(' && ptr[sk
6430    else break;    else break;
6431    }    }
6432    
6433    /* Can't support UTF8 unless PCRE has been compiled to include the code. */
6434    
6435    #ifdef SUPPORT_UTF8
6436    if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&
6437         (*erroroffset = _pcre_valid_utf8((uschar *)pattern, -1)) >= 0)
6438      {
6439      errorcode = ERR44;
6440      goto PCRE_EARLY_ERROR_RETURN2;
6441      }
6442    #else
6443    if (utf8)
6444      {
6445      errorcode = ERR32;
6446      goto PCRE_EARLY_ERROR_RETURN;
6447      }
6448    #endif
6449    
6450  /* Check validity of \R options. */  /* Check validity of \R options. */
6451    
6452  switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))  switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))
# Line 5887  current code allows for fixed one- or tw Line 6465  current code allows for fixed one- or tw
6465  switch (options & PCRE_NEWLINE_BITS)  switch (options & PCRE_NEWLINE_BITS)
6466    {    {
6467    case 0: newline = NEWLINE; break;   /* Build-time default */    case 0: newline = NEWLINE; break;   /* Build-time default */
6468    case PCRE_NEWLINE_CR: newline = '\r'; break;    case PCRE_NEWLINE_CR: newline = CHAR_CR; break;
6469    case PCRE_NEWLINE_LF: newline = '\n'; break;    case PCRE_NEWLINE_LF: newline = CHAR_NL; break;
6470    case PCRE_NEWLINE_CR+    case PCRE_NEWLINE_CR+
6471         PCRE_NEWLINE_LF: newline = ('\r' << 8) | '\n'; break;         PCRE_NEWLINE_LF: newline = (CHAR_CR << 8) | CHAR_NL; break;
6472    case PCRE_NEWLINE_ANY: newline = -1; break;    case PCRE_NEWLINE_ANY: newline = -1; break;
6473    case PCRE_NEWLINE_ANYCRLF: newline = -2; break;    case PCRE_NEWLINE_ANYCRLF: newline = -2; break;
6474    default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;    default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;
# Line 5951  cd->end_pattern = (const uschar *)(patte Line 6529  cd->end_pattern = (const uschar *)(patte
6529  cd->req_varyopt = 0;  cd->req_varyopt = 0;
6530  cd->external_options = options;  cd->external_options = options;
6531  cd->external_flags = 0;  cd->external_flags = 0;
6532    cd->open_caps = NULL;
6533    
6534  /* Now do the pre-compile. On error, errorcode will be set non-zero, so we  /* Now do the pre-compile. On error, errorcode will be set non-zero, so we
6535  don't need to look at the result of the function here. The initial options have  don't need to look at the result of the function here. The initial options have
# Line 6025  cd->start_code = codestart; Line 6604  cd->start_code = codestart;
6604  cd->hwm = cworkspace;  cd->hwm = cworkspace;
6605  cd->req_varyopt = 0;  cd->req_varyopt = 0;
6606  cd->had_accept = FALSE;  cd->had_accept = FALSE;
6607    cd->check_lookbehind = FALSE;
6608    cd->open_caps = NULL;
6609    
6610  /* Set up a starting, non-extracting bracket, then compile the expression. On  /* Set up a starting, non-extracting bracket, then compile the expression. On
6611  error, errorcode will be set non-zero, so we don't need to look at the result  error, errorcode will be set non-zero, so we don't need to look at the result
# Line 6063  while (errorcode == 0 && cd->hwm > cwork Line 6644  while (errorcode == 0 && cd->hwm > cwork
6644    cd->hwm -= LINK_SIZE;    cd->hwm -= LINK_SIZE;
6645    offset = GET(cd->hwm, 0);    offset = GET(cd->hwm, 0);
6646    recno = GET(codestart, offset);    recno = GET(codestart, offset);
6647    groupptr = find_bracket(codestart, (re->options & PCRE_UTF8) != 0, recno);    groupptr = _pcre_find_bracket(codestart, utf8, recno);
6648    if (groupptr == NULL) errorcode = ERR53;    if (groupptr == NULL) errorcode = ERR53;
6649      else PUT(((uschar *)codestart), offset, groupptr - codestart);      else PUT(((uschar *)codestart), offset, groupptr - codestart);
6650    }    }
# Line 6073  subpattern. */ Line 6654  subpattern. */
6654    
6655  if (errorcode == 0 && re->top_backref > re->top_bracket) errorcode = ERR15;  if (errorcode == 0 && re->top_backref > re->top_bracket) errorcode = ERR15;
6656    
6657    /* If there were any lookbehind assertions that contained OP_RECURSE
6658    (recursions or subroutine calls), a flag is set for them to be checked here,
6659    because they may contain forward references. Actual recursions can't be fixed
6660    length, but subroutine calls can. It is done like this so that those without
6661    OP_RECURSE that are not fixed length get a diagnosic with a useful offset. The
6662    exceptional ones forgo this. We scan the pattern to check that they are fixed
6663    length, and set their lengths. */
6664    
6665    if (cd->check_lookbehind)
6666      {
6667      uschar *cc = (uschar *)codestart;
6668    
6669      /* Loop, searching for OP_REVERSE items, and process those that do not have
6670      their length set. (Actually, it will also re-process any that have a length
6671      of zero, but that is a pathological case, and it does no harm.) When we find
6672      one, we temporarily terminate the branch it is in while we scan it. */
6673    
6674      for (cc = (uschar *)_pcre_find_bracket(codestart, utf8, -1);
6675           cc != NULL;
6676           cc = (uschar *)_pcre_find_bracket(cc, utf8, -1))
6677        {
6678        if (GET(cc, 1) == 0)
6679          {
6680          int fixed_length;
6681          uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);
6682          int end_op = *be;
6683          *be = OP_END;
6684          fixed_length = find_fixedlength(cc, re->options, TRUE, cd);
6685          *be = end_op;
6686          DPRINTF(("fixed length = %d\n", fixed_length));
6687          if (fixed_length < 0)
6688            {
6689            errorcode = (fixed_length == -2)? ERR36 : ERR25;
6690            break;
6691            }
6692          PUT(cc, 1, fixed_length);
6693          }
6694        cc += 1 + LINK_SIZE;
6695        }
6696      }
6697    
6698  /* Failed to compile, or error while post-processing */  /* Failed to compile, or error while post-processing */
6699    
6700  if (errorcode != 0)  if (errorcode != 0)

Legend:
Removed from v.274  
changed lines
  Added in v.461

webmaster@exim.org
ViewVC Help
Powered by ViewVC 1.1.12