/[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 240 by ph10, Tue Sep 11 15:47:20 2007 UTC revision 438 by ph10, Sun Sep 6 20:00:47 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 140  static const short int escapes[] = { Line 175  static const short int escapes[] = {
175  #endif  #endif
176    
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
212  now all in a single string, to reduce the number of relocations when a shared  now all in a single string, to reduce the number of relocations when a shared
213  library is dynamically loaded. The list of lengths is terminated by a zero  library is dynamically loaded. The list of lengths is terminated by a zero
214  length entry. The first three must be alpha, lower, upper, as this is assumed  length entry. The first three must be alpha, lower, upper, as this is assumed
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 219  static const int posix_class_maps[] = { Line 257  static const int posix_class_maps[] = {
257  /* The texts of compile-time error messages. These are "char *" because they  /* The texts of compile-time error messages. These are "char *" because they
258  are passed to the outside world. Do not ever re-use any error number, because  are passed to the outside world. Do not ever re-use any error number, because
259  they are documented. Always add a new error instead. Messages marked DEAD below  they are documented. Always add a new error instead. Messages marked DEAD below
260  are no longer used. This used to be a table of strings, but in order to reduce  are no longer used. This used to be a table of strings, but in order to reduce
261  the number of relocations needed when a shared library is loaded dynamically,  the number of relocations needed when a shared library is loaded dynamically,
262  it is now one long string. We cannot use a table of offsets, because the  it is now one long string. We cannot use a table of offsets, because the
263  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
264  simply count through to the one we want - this isn't a performance issue  simply count through to the one we want - this isn't a performance issue
265  because these strings are used only when there is a compilation error. */  because these strings are used only when there is a compilation error. */
266    
267  static const char error_texts[] =  static const char error_texts[] =
# Line 241  static const char error_texts[] = Line 279  static const char error_texts[] =
279    /* 10 */    /* 10 */
280    "operand of unlimited repeat could match the empty string\0"  /** DEAD **/    "operand of unlimited repeat could match the empty string\0"  /** DEAD **/
281    "internal error: unexpected repeat\0"    "internal error: unexpected repeat\0"
282    "unrecognized character after (?\0"    "unrecognized character after (? or (?-\0"
283    "POSIX named classes are supported only within a class\0"    "POSIX named classes are supported only within a class\0"
284    "missing )\0"    "missing )\0"
285    /* 15 */    /* 15 */
# 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";    "number is too big\0"
342      "subpattern name expected\0"
343      "digit expected after (?+\0"
344      "] is an invalid data character in JavaScript compatibility mode";
345    
346    
347  /* 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 319  For convenience, we use the same bit def Line 360  For convenience, we use the same bit def
360    
361  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
362    
363  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC
364    
365    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
366    UTF-8 mode. */
367    
368  static const unsigned char digitab[] =  static const unsigned char digitab[] =
369    {    {
370    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
# Line 355  static const unsigned char digitab[] = Line 400  static const unsigned char digitab[] =
400    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
401    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
402    
403  #else           /* This is the "abnormal" case, for EBCDIC systems */  #else
404    
405    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
406    
407  static const unsigned char digitab[] =  static const unsigned char digitab[] =
408    {    {
409    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
# Line 439  static BOOL Line 487  static BOOL
487  *            Find an error text                  *  *            Find an error text                  *
488  *************************************************/  *************************************************/
489    
490  /* The error texts are now all in one long string, to save on relocations. As  /* The error texts are now all in one long string, to save on relocations. As
491  some of the text is of unknown length, we can't use a table of offsets.  some of the text is of unknown length, we can't use a table of offsets.
492  Instead, just count through the strings. This is not a performance issue  Instead, just count through the strings. This is not a performance issue
493  because it happens only when there has been a compilation error.  because it happens only when there has been a compilation error.
494    
495  Argument:   the error number  Argument:   the error number
# Line 452  static const char * Line 500  static const char *
500  find_error_text(int n)  find_error_text(int n)
501  {  {
502  const char *s = error_texts;  const char *s = error_texts;
503  for (; n > 0; n--) while (*s++ != 0);  for (; n > 0; n--) while (*s++ != 0) {};
504  return s;  return s;
505  }  }
506    
# Line 496  ptr--;                            /* Set Line 544  ptr--;                            /* Set
544    
545  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
546    
547  /* Non-alphamerics are literals. For digits or letters, do an initial lookup in  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
548  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.
549  Otherwise further processing may be required. */  Otherwise further processing may be required. */
550    
551  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
552  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */
553  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0) c = i;
554    
555  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
556  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */
557  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
558  #endif  #endif
559    
# Line 521  else Line 569  else
569      /* 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
570      error. */      error. */
571    
572      case 'l':      case CHAR_l:
573      case 'L':      case CHAR_L:
574      case 'N':      case CHAR_N:
575      case 'u':      case CHAR_u:
576      case 'U':      case CHAR_U:
577      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
578      break;      break;
579    
580      /* \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:
581      is an absolute backreference. If negative, it is a relative backreference.  
582      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
583      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
584      unified syntax for back references. As this is synonymous with \k{name}, we      5.10 feature.
585      fudge it up by pretending it really was \k. */  
586        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
587        is part of Perl's movement towards a unified syntax for back references. As
588        this is synonymous with \k{name}, we fudge it up by pretending it really
589        was \k.
590    
591        (3) For Oniguruma compatibility we also support \g followed by a name or a
592        number either in angle brackets or in single quotes. However, these are
593        (possibly recursive) subroutine calls, _not_ backreferences. Just return
594        the -ESC_g code (cf \k). */
595    
596      case 'g':      case CHAR_g:
597      if (ptr[1] == '{')      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
598          {
599          c = -ESC_g;
600          break;
601          }
602    
603        /* Handle the Perl-compatible cases */
604    
605        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
606        {        {
607        const uschar *p;        const uschar *p;
608        for (p = ptr+2; *p != 0 && *p != '}'; p++)        for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
609          if (*p != '-' && (digitab[*p] & ctype_digit) == 0) break;          if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
610        if (*p != 0 && *p != '}')        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
611          {          {
612          c = -ESC_k;          c = -ESC_k;
613          break;          break;
# Line 552  else Line 617  else
617        }        }
618      else braced = FALSE;      else braced = FALSE;
619    
620      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
621        {        {
622        negated = TRUE;        negated = TRUE;
623        ptr++;        ptr++;
# Line 561  else Line 626  else
626    
627      c = 0;      c = 0;
628      while ((digitab[ptr[1]] & ctype_digit) != 0)      while ((digitab[ptr[1]] & ctype_digit) != 0)
629        c = c * 10 + *(++ptr) - '0';        c = c * 10 + *(++ptr) - CHAR_0;
630    
631      if (c < 0)      if (c < 0)   /* Integer overflow */
632        {        {
633        *errorcodeptr = ERR61;        *errorcodeptr = ERR61;
634        break;        break;
635        }        }
636    
637      if (c == 0 || (braced && *(++ptr) != '}'))      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
638        {        {
639        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
640        break;        break;
641        }        }
642    
643        if (c == 0)
644          {
645          *errorcodeptr = ERR58;
646          break;
647          }
648    
649      if (negated)      if (negated)
650        {        {
651        if (c > bracount)        if (c > bracount)
# Line 600  else Line 671  else
671      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
672      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
673    
674      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:
675      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
676    
677      if (!isclass)      if (!isclass)
678        {        {
679        oldptr = ptr;        oldptr = ptr;
680        c -= '0';        c -= CHAR_0;
681        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
682          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - CHAR_0;
683        if (c < 0)        if (c < 0)    /* Integer overflow */
684          {          {
685          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
686          break;          break;
# Line 626  else Line 697  else
697      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.
698      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
699    
700      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
701        {        {
702        ptr--;        ptr--;
703        c = 0;        c = 0;
# Line 639  else Line 710  else
710      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
711      than 3 octal digits. */      than 3 octal digits. */
712    
713      case '0':      case CHAR_0:
714      c -= '0';      c -= CHAR_0;
715      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
716          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
717      if (!utf8 && c > 255) *errorcodeptr = ERR51;      if (!utf8 && c > 255) *errorcodeptr = ERR51;
718      break;      break;
719    
# Line 650  else Line 721  else
721      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
722      treated as a data character. */      treated as a data character. */
723    
724      case 'x':      case CHAR_x:
725      if (ptr[1] == '{')      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
726        {        {
727        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
728        int count = 0;        int count = 0;
# Line 660  else Line 731  else
731        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
732          {          {
733          register int cc = *pt++;          register int cc = *pt++;
734          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
735          count++;          count++;
736    
737  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
738          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
739          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
740  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
741          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
742          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
743  #endif  #endif
744          }          }
745    
746        if (*pt == '}')        if (*pt == CHAR_RIGHT_CURLY_BRACKET)
747          {          {
748          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
749          ptr = pt;          ptr = pt;
# Line 688  else Line 759  else
759      c = 0;      c = 0;
760      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
761        {        {
762        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
763        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
764  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
765        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
766        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
767  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
768        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
769        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
770  #endif  #endif
771        }        }
772      break;      break;
# Line 704  else Line 775  else
775      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
776      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
777    
778      case 'c':      case CHAR_c:
779      c = *(++ptr);      c = *(++ptr);
780      if (c == 0)      if (c == 0)
781        {        {
# Line 712  else Line 783  else
783        break;        break;
784        }        }
785    
786  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
787      if (c >= 'a' && c <= 'z') c -= 32;      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
788      c ^= 0x40;      c ^= 0x40;
789  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
790      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
791      c ^= 0xC0;      c ^= 0xC0;
792  #endif  #endif
793      break;      break;
794    
795      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
796      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
797      for Perl compatibility, it is a literal. This code looks a bit odd, but      otherwise, for Perl compatibility, it is a literal. This code looks a bit
798      there used to be some cases other than the default, and there may be again      odd, but there used to be some cases other than the default, and there may
799      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
800    
801      default:      default:
802      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 778  if (c == 0) goto ERROR_RETURN; Line 849  if (c == 0) goto ERROR_RETURN;
849  /* \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
850  negation. */  negation. */
851    
852  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
853    {    {
854    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
855      {      {
856      *negptr = TRUE;      *negptr = TRUE;
857      ptr++;      ptr++;
# Line 789  if (c == '{') Line 860  if (c == '{')
860      {      {
861      c = *(++ptr);      c = *(++ptr);
862      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
863      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
864      name[i] = c;      name[i] = c;
865      }      }
866    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
867    name[i] = 0;    name[i] = 0;
868    }    }
869    
# Line 857  is_counted_repeat(const uschar *p) Line 928  is_counted_repeat(const uschar *p)
928  {  {
929  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
930  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
931  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
932    
933  if (*p++ != ',') return FALSE;  if (*p++ != CHAR_COMMA) return FALSE;
934  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
935    
936  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
937  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
938    
939  return (*p == '}');  return (*p == CHAR_RIGHT_CURLY_BRACKET);
940  }  }
941    
942    
# Line 898  int max = -1; Line 969  int max = -1;
969  /* 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
970  an integer overflow. */  an integer overflow. */
971    
972  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;
973  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
974    {    {
975    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 908  if (min < 0 || min > 65535) Line 979  if (min < 0 || min > 65535)
979  /* 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.
980  Also, max must not be less than min. */  Also, max must not be less than min. */
981    
982  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
983    {    {
984    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
985      {      {
986      max = 0;      max = 0;
987      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;
988      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
989        {        {
990        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 938  return p; Line 1009  return p;
1009    
1010    
1011  /*************************************************  /*************************************************
1012  *       Find forward referenced subpattern       *  *  Subroutine for finding forward reference      *
1013  *************************************************/  *************************************************/
1014    
1015  /* This function scans along a pattern's text looking for capturing  /* This recursive function is called only from find_parens() below. The
1016    top-level call starts at the beginning of the pattern. All other calls must
1017    start at a parenthesis. It scans along a pattern's text looking for capturing
1018  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
1019  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
1020  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
1021  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
1022  be terminated by '>' because that is checked in the first pass.  first pass. Recursion is used to keep track of subpatterns that reset the
1023    capturing group numbers - the (?| feature.
1024    
1025  Arguments:  Arguments:
1026    ptr          current position in the pattern    ptrptr       address of the current character pointer (updated)
1027    count        current count of capturing parens so far encountered    cd           compile background data
1028    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1029    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1030    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1031      count        pointer to the current capturing subpattern number (updated)
1032    
1033  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1034  */  */
1035    
1036  static int  static int
1037  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,
1038    BOOL xmode)    BOOL xmode, int *count)
1039  {  {
1040  const uschar *thisname;  uschar *ptr = *ptrptr;
1041    int start_count = *count;
1042    int hwm_count = start_count;
1043    BOOL dup_parens = FALSE;
1044    
1045  for (; *ptr != 0; ptr++)  /* If the first character is a parenthesis, check on the type of group we are
1046    dealing with. The very first call may not start with a parenthesis. */
1047    
1048    if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1049    {    {
1050    int term;    if (ptr[1] == CHAR_QUESTION_MARK &&
1051          ptr[2] == CHAR_VERTICAL_LINE)
1052        {
1053        ptr += 3;
1054        dup_parens = TRUE;
1055        }
1056    
1057      /* Handle a normal, unnamed capturing parenthesis */
1058    
1059      else if (ptr[1] != CHAR_QUESTION_MARK && ptr[1] != CHAR_ASTERISK)
1060        {
1061        *count += 1;
1062        if (name == NULL && *count == lorn) return *count;
1063        ptr++;
1064        }
1065    
1066      /* Handle a condition. If it is an assertion, just carry on so that it
1067      is processed as normal. If not, skip to the closing parenthesis of the
1068      condition (there can't be any nested parens. */
1069    
1070      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1071        {
1072        ptr += 2;
1073        if (ptr[1] != CHAR_QUESTION_MARK)
1074          {
1075          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1076          if (*ptr != 0) ptr++;
1077          }
1078        }
1079    
1080      /* We have either (? or (* and not a condition */
1081    
1082      else
1083        {
1084        ptr += 2;
1085        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1086    
1087        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1088    
1089        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1090            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1091          {
1092          int term;
1093          const uschar *thisname;
1094          *count += 1;
1095          if (name == NULL && *count == lorn) return *count;
1096          term = *ptr++;
1097          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1098          thisname = ptr;
1099          while (*ptr != term) ptr++;
1100          if (name != NULL && lorn == ptr - thisname &&
1101              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1102            return *count;
1103          term++;
1104          }
1105        }
1106      }
1107    
1108    /* Past any initial parenthesis handling, scan for parentheses or vertical
1109    bars. */
1110    
1111    for (; *ptr != 0; ptr++)
1112      {
1113    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1114    
1115    if (*ptr == '\\')    if (*ptr == CHAR_BACKSLASH)
1116      {      {
1117      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1118      if (*ptr == 'Q') for (;;)      if (*ptr == CHAR_Q) for (;;)
1119        {        {
1120        while (*(++ptr) != 0 && *ptr != '\\');        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1121        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1122        if (*(++ptr) == 'E') break;        if (*(++ptr) == CHAR_E) break;
1123        }        }
1124      continue;      continue;
1125      }      }
1126    
1127    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1128      are handled for real. If the first character is '^', skip it. Also, if the
1129      first few characters (either before or after ^) are \Q\E or \E we skip them
1130      too. This makes for compatibility with Perl. Note the use of STR macros to
1131      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1132    
1133    if (*ptr == '[')    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1134      {      {
1135      while (*(++ptr) != ']')      BOOL negate_class = FALSE;
1136        for (;;)
1137          {
1138          if (ptr[1] == CHAR_BACKSLASH)
1139            {
1140            if (ptr[2] == CHAR_E)
1141              ptr+= 2;
1142            else if (strncmp((const char *)ptr+2,
1143                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1144              ptr += 4;
1145            else
1146              break;
1147            }
1148          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1149            {
1150            negate_class = TRUE;
1151            ptr++;
1152            }
1153          else break;
1154          }
1155    
1156        /* If the next character is ']', it is a data character that must be
1157        skipped, except in JavaScript compatibility mode. */
1158    
1159        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1160            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1161          ptr++;
1162    
1163        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1164        {        {
1165        if (*ptr == 0) return -1;        if (*ptr == 0) return -1;
1166        if (*ptr == '\\')        if (*ptr == CHAR_BACKSLASH)
1167          {          {
1168          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) goto FAIL_EXIT;
1169          if (*ptr == 'Q') for (;;)          if (*ptr == CHAR_Q) for (;;)
1170            {            {
1171            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1172            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1173            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1174            }            }
1175          continue;          continue;
1176          }          }
# Line 1006  for (; *ptr != 0; ptr++) Line 1180  for (; *ptr != 0; ptr++)
1180    
1181    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1182    
1183    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1184      {      {
1185      while (*(++ptr) != 0 && *ptr != '\n');      while (*(++ptr) != 0 && *ptr != CHAR_NL) {};
1186      if (*ptr == 0) return -1;      if (*ptr == 0) goto FAIL_EXIT;
1187      continue;      continue;
1188      }      }
1189    
1190    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1191    
1192    if (*ptr != '(') continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != '?' && ptr[1] != '*')  
1193      {      {
1194      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, count);
1195      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1196      continue;      if (*ptr == 0) goto FAIL_EXIT;
1197        }
1198    
1199      else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1200        {
1201        if (dup_parens && *count < hwm_count) *count = hwm_count;
1202        *ptrptr = ptr;
1203        return -1;
1204      }      }
1205    
1206    ptr += 2;    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1207    if (*ptr == 'P') ptr++;                      /* Allow optional P */      {
1208        if (*count > hwm_count) hwm_count = *count;
1209        *count = start_count;
1210        }
1211      }
1212    
1213    FAIL_EXIT:
1214    *ptrptr = ptr;
1215    return -1;
1216    }
1217    
1218    
   /* We have to disambiguate (?<! and (?<= from (?<name> */  
1219    
   if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  
        *ptr != '\'')  
     continue;  
1220    
1221    count++;  /*************************************************
1222    *       Find forward referenced subpattern       *
1223    *************************************************/
1224    
1225    /* This function scans along a pattern's text looking for capturing
1226    subpatterns, and counting them. If it finds a named pattern that matches the
1227    name it is given, it returns its number. Alternatively, if the name is NULL, it
1228    returns when it reaches a given numbered subpattern. This is used for forward
1229    references to subpatterns. We used to be able to start this scan from the
1230    current compiling point, using the current count value from cd->bracount, and
1231    do it all in a single loop, but the addition of the possibility of duplicate
1232    subpattern numbers means that we have to scan from the very start, in order to
1233    take account of such duplicates, and to use a recursive function to keep track
1234    of the different types of group.
1235    
1236    Arguments:
1237      cd           compile background data
1238      name         name to seek, or NULL if seeking a numbered subpattern
1239      lorn         name length, or subpattern number if name is NULL
1240      xmode        TRUE if we are in /x mode
1241    
1242    Returns:       the number of the found subpattern, or -1 if not found
1243    */
1244    
1245    static int
1246    find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode)
1247    {
1248    uschar *ptr = (uschar *)cd->start_pattern;
1249    int count = 0;
1250    int rc;
1251    
1252    /* If the pattern does not start with an opening parenthesis, the first call
1253    to find_parens_sub() will scan right to the end (if necessary). However, if it
1254    does start with a parenthesis, find_parens_sub() will return when it hits the
1255    matching closing parens. That is why we have to have a loop. */
1256    
1257    if (name == NULL && count == lorn) return count;  for (;;)
1258    term = *ptr++;    {
1259    if (term == '<') term = '>';    rc = find_parens_sub(&ptr, cd, name, lorn, xmode, &count);
1260    thisname = ptr;    if (rc > 0 || *ptr++ == 0) break;
   while (*ptr != term) ptr++;  
   if (name != NULL && lorn == ptr - thisname &&  
       strncmp((const char *)name, (const char *)thisname, lorn) == 0)  
     return count;  
1261    }    }
1262    
1263  return -1;  return rc;
1264  }  }
1265    
1266    
1267    
1268    
1269  /*************************************************  /*************************************************
1270  *      Find first significant op code            *  *      Find first significant op code            *
1271  *************************************************/  *************************************************/
# Line 1209  for (;;) Line 1426  for (;;)
1426      branchlength++;      branchlength++;
1427      cc += 2;      cc += 2;
1428  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1429      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1430        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1431  #endif  #endif
1432      break;      break;
1433    
# Line 1223  for (;;) Line 1438  for (;;)
1438      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1439      cc += 4;      cc += 4;
1440  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1441      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1442        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1443  #endif  #endif
1444      break;      break;
1445    
# Line 1250  for (;;) Line 1463  for (;;)
1463      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1464      case OP_WORDCHAR:      case OP_WORDCHAR:
1465      case OP_ANY:      case OP_ANY:
1466        case OP_ALLANY:
1467      branchlength++;      branchlength++;
1468      cc++;      cc++;
1469      break;      break;
# Line 1398  for (;;) Line 1612  for (;;)
1612        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1613        break;        break;
1614        }        }
1615    #else
1616        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1617  #endif  #endif
1618      }      }
1619    }    }
# Line 1491  for (;;) Line 1707  for (;;)
1707        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1708        break;        break;
1709        }        }
1710    #else
1711        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1712  #endif  #endif
1713      }      }
1714    }    }
# Line 1506  for (;;) Line 1724  for (;;)
1724  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()
1725  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
1726  group that can match nothing. Note that first_significant_code() skips over  group that can match nothing. Note that first_significant_code() skips over
1727  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
1728  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
1729    bracket whose current branch will already have been scanned.
1730    
1731  Arguments:  Arguments:
1732    code        points to start of search    code        points to start of search
# Line 1529  for (code = first_significant_code(code Line 1748  for (code = first_significant_code(code
1748    
1749    c = *code;    c = *code;
1750    
1751      /* Skip over forward assertions; the other assertions are skipped by
1752      first_significant_code() with a TRUE final argument. */
1753    
1754      if (c == OP_ASSERT)
1755        {
1756        do code += GET(code, 1); while (*code == OP_ALT);
1757        c = *code;
1758        continue;
1759        }
1760    
1761    /* Groups with zero repeats can of course be empty; skip them. */    /* Groups with zero repeats can of course be empty; skip them. */
1762    
1763    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)
1764      {      {
1765      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1766      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
# Line 1546  for (code = first_significant_code(code Line 1775  for (code = first_significant_code(code
1775      BOOL empty_branch;      BOOL empty_branch;
1776      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
1777    
1778      /* Scan a closed bracket */      /* If a conditional group has only one branch, there is a second, implied,
1779        empty branch, so just skip over the conditional, because it could be empty.
1780        Otherwise, scan the individual branches of the group. */
1781    
1782      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;  
1783        code += GET(code, 1);        code += GET(code, 1);
1784        else
1785          {
1786          empty_branch = FALSE;
1787          do
1788            {
1789            if (!empty_branch && could_be_empty_branch(code, endcode, utf8))
1790              empty_branch = TRUE;
1791            code += GET(code, 1);
1792            }
1793          while (*code == OP_ALT);
1794          if (!empty_branch) return FALSE;   /* All branches are non-empty */
1795        }        }
1796      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
1797      c = *code;      c = *code;
1798      continue;      continue;
1799      }      }
# Line 1617  for (code = first_significant_code(code Line 1854  for (code = first_significant_code(code
1854      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1855      case OP_WORDCHAR:      case OP_WORDCHAR:
1856      case OP_ANY:      case OP_ANY:
1857        case OP_ALLANY:
1858      case OP_ANYBYTE:      case OP_ANYBYTE:
1859      case OP_CHAR:      case OP_CHAR:
1860      case OP_CHARNC:      case OP_CHARNC:
# Line 1673  for (code = first_significant_code(code Line 1911  for (code = first_significant_code(code
1911      case OP_QUERY:      case OP_QUERY:
1912      case OP_MINQUERY:      case OP_MINQUERY:
1913      case OP_POSQUERY:      case OP_POSQUERY:
1914        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
1915        break;
1916    
1917      case OP_UPTO:      case OP_UPTO:
1918      case OP_MINUPTO:      case OP_MINUPTO:
1919      case OP_POSUPTO:      case OP_POSUPTO:
1920      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
1921      break;      break;
1922  #endif  #endif
1923      }      }
# Line 1724  return TRUE; Line 1965  return TRUE;
1965  *************************************************/  *************************************************/
1966    
1967  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
1968  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
1969  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
1970  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
1971    
1972    Originally, this function only recognized a sequence of letters between the
1973    terminators, but it seems that Perl recognizes any sequence of characters,
1974    though of course unknown POSIX names are subsequently rejected. Perl gives an
1975    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
1976    didn't consider this to be a POSIX class. Likewise for [:1234:].
1977    
1978    The problem in trying to be exactly like Perl is in the handling of escapes. We
1979    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
1980    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
1981    below handles the special case of \], but does not try to do any other escape
1982    processing. This makes it different from Perl for cases such as [:l\ower:]
1983    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
1984    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
1985    I think.
1986    
1987  Argument:  Arguments:
1988    ptr      pointer to the initial [    ptr      pointer to the initial [
1989    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
1990    
1991  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
1992  */  */
1993    
1994  static BOOL  static BOOL
1995  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
1996  {  {
1997  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
1998  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
1999  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2000    {    {
2001    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else
2002    return TRUE;      {
2003        if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2004        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2005          {
2006          *endptr = ptr;
2007          return TRUE;
2008          }
2009        }
2010    }    }
2011  return FALSE;  return FALSE;
2012  }  }
# Line 1777  while (posix_name_lengths[yield] != 0) Line 2037  while (posix_name_lengths[yield] != 0)
2037    {    {
2038    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2039      strncmp((const char *)ptr, pn, len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
2040    pn += posix_name_lengths[yield] + 1;    pn += posix_name_lengths[yield] + 1;
2041    yield++;    yield++;
2042    }    }
2043  return -1;  return -1;
# Line 1792  return -1; Line 2052  return -1;
2052  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2053  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2054  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
2055  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
2056  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
2057  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
2058  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
2059  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2060    OP_END.
2061    
2062  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2063  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 1931  get_othercase_range(unsigned int *cptr, Line 2192  get_othercase_range(unsigned int *cptr,
2192  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2193    
2194  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2195    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2196    
2197  if (c > d) return FALSE;  if (c > d) return FALSE;
2198    
# Line 1940  next = othercase + 1; Line 2201  next = othercase + 1;
2201    
2202  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2203    {    {
2204    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2205    next++;    next++;
2206    }    }
2207    
# Line 1986  if ((options & PCRE_EXTENDED) != 0) Line 2247  if ((options & PCRE_EXTENDED) != 0)
2247    for (;;)    for (;;)
2248      {      {
2249      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2250      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2251        {        {
2252        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2253          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
# Line 1998  if ((options & PCRE_EXTENDED) != 0) Line 2259  if ((options & PCRE_EXTENDED) != 0)
2259  /* 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
2260  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2261    
2262  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2263    {    {
2264    int temperrorcode = 0;    int temperrorcode = 0;
2265    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
# Line 2023  if ((options & PCRE_EXTENDED) != 0) Line 2284  if ((options & PCRE_EXTENDED) != 0)
2284    for (;;)    for (;;)
2285      {      {
2286      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2287      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2288        {        {
2289        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2290          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
# Line 2034  if ((options & PCRE_EXTENDED) != 0) Line 2295  if ((options & PCRE_EXTENDED) != 0)
2295    
2296  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
2297    
2298  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2299    return FALSE;    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2300        return FALSE;
2301    
2302  /* 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
2303  positive single character match, "item" either contains the character or, if  positive single character match, "item" either contains the character or, if
# Line 2050  if (next >= 0) switch(op_code) Line 2312  if (next >= 0) switch(op_code)
2312    case OP_CHAR:    case OP_CHAR:
2313  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2314    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2315    #else
2316      (void)(utf8_char);  /* Keep compiler happy by referencing function argument */
2317  #endif  #endif
2318    return item != next;    return item != next;
2319    
# Line 2068  if (next >= 0) switch(op_code) Line 2332  if (next >= 0) switch(op_code)
2332      unsigned int othercase;      unsigned int othercase;
2333      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2334  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2335      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
2336  #else  #else
2337      othercase = NOTACHAR;      othercase = NOTACHAR;
2338  #endif  #endif
# Line 2081  if (next >= 0) switch(op_code) Line 2345  if (next >= 0) switch(op_code)
2345    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT, "item" must be a single-byte character. */
2346    
2347    case OP_NOT:    case OP_NOT:
   if (next < 0) return FALSE;  /* Not a character */  
2348    if (item == next) return TRUE;    if (item == next) return TRUE;
2349    if ((options & PCRE_CASELESS) == 0) return FALSE;    if ((options & PCRE_CASELESS) == 0) return FALSE;
2350  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 2090  if (next >= 0) switch(op_code) Line 2353  if (next >= 0) switch(op_code)
2353      unsigned int othercase;      unsigned int othercase;
2354      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2355  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2356      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
2357  #else  #else
2358      othercase = NOTACHAR;      othercase = NOTACHAR;
2359  #endif  #endif
# Line 2344  uschar classbits[32]; Line 2607  uschar classbits[32];
2607  BOOL class_utf8;  BOOL class_utf8;
2608  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
2609  uschar *class_utf8data;  uschar *class_utf8data;
2610    uschar *class_utf8data_base;
2611  uschar utf8_char[6];  uschar utf8_char[6];
2612  #else  #else
2613  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
# Line 2383  req_caseopt = ((options & PCRE_CASELESS) Line 2647  req_caseopt = ((options & PCRE_CASELESS)
2647  for (;; ptr++)  for (;; ptr++)
2648    {    {
2649    BOOL negate_class;    BOOL negate_class;
2650      BOOL should_flip_negation;
2651    BOOL possessive_quantifier;    BOOL possessive_quantifier;
2652    BOOL is_quantifier;    BOOL is_quantifier;
2653    BOOL is_recurse;    BOOL is_recurse;
# Line 2470  for (;; ptr++) Line 2735  for (;; ptr++)
2735    
2736    if (inescq && c != 0)    if (inescq && c != 0)
2737      {      {
2738      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
2739        {        {
2740        inescq = FALSE;        inescq = FALSE;
2741        ptr++;        ptr++;
# Line 2496  for (;; ptr++) Line 2761  for (;; ptr++)
2761    /* 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
2762    a quantifier. */    a quantifier. */
2763    
2764    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
2765      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
2766        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
2767    
2768    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
2769         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2512  for (;; ptr++) Line 2778  for (;; ptr++)
2778    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
2779      {      {
2780      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
2781      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
2782        {        {
2783        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2784          {          {
# Line 2537  for (;; ptr++) Line 2803  for (;; ptr++)
2803      {      {
2804      /* ===================================================================*/      /* ===================================================================*/
2805      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
2806      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
2807      case ')':      case CHAR_RIGHT_PARENTHESIS:
2808      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
2809      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
2810      *codeptr = code;      *codeptr = code;
# Line 2560  for (;; ptr++) Line 2826  for (;; ptr++)
2826      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
2827      the setting of any following char as a first character. */      the setting of any following char as a first character. */
2828    
2829      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
2830      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
2831        {        {
2832        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
# Line 2569  for (;; ptr++) Line 2835  for (;; ptr++)
2835      *code++ = OP_CIRC;      *code++ = OP_CIRC;
2836      break;      break;
2837    
2838      case '$':      case CHAR_DOLLAR_SIGN:
2839      previous = NULL;      previous = NULL;
2840      *code++ = OP_DOLL;      *code++ = OP_DOLL;
2841      break;      break;
# Line 2577  for (;; ptr++) Line 2843  for (;; ptr++)
2843      /* 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
2844      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
2845    
2846      case '.':      case CHAR_DOT:
2847      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
2848      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
2849      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
2850      previous = code;      previous = code;
2851      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
2852      break;      break;
2853    
2854    
# Line 2597  for (;; ptr++) Line 2863  for (;; ptr++)
2863      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,
2864      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
2865      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.
     */  
2866    
2867      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
2868        default (Perl) mode, it is treated as a data character. */
2869    
2870        case CHAR_RIGHT_SQUARE_BRACKET:
2871        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2872          {
2873          *errorcodeptr = ERR64;
2874          goto FAILED;
2875          }
2876        goto NORMAL_CHAR;
2877    
2878        case CHAR_LEFT_SQUARE_BRACKET:
2879      previous = code;      previous = code;
2880    
2881      /* 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
2882      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. */
2883    
2884      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2885          check_posix_syntax(ptr, &tempptr, cd))           ptr[1] == CHAR_EQUALS_SIGN) &&
2886            check_posix_syntax(ptr, &tempptr))
2887        {        {
2888        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
2889        goto FAILED;        goto FAILED;
2890        }        }
2891    
# Line 2620  for (;; ptr++) Line 2897  for (;; ptr++)
2897      for (;;)      for (;;)
2898        {        {
2899        c = *(++ptr);        c = *(++ptr);
2900        if (c == '\\')        if (c == CHAR_BACKSLASH)
2901          {          {
2902          if (ptr[1] == 'E') ptr++;          if (ptr[1] == CHAR_E)
2903            else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;            ptr++;
2904              else break;          else if (strncmp((const char *)ptr+1,
2905                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
2906              ptr += 3;
2907            else
2908              break;
2909          }          }
2910        else if (!negate_class && c == '^')        else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
2911          negate_class = TRUE;          negate_class = TRUE;
2912        else break;        else break;
2913        }        }
2914    
2915        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
2916        an initial ']' is taken as a data character -- the code below handles
2917        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
2918        [^] must match any character, so generate OP_ALLANY. */
2919    
2920        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
2921            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2922          {
2923          *code++ = negate_class? OP_ALLANY : OP_FAIL;
2924          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
2925          zerofirstbyte = firstbyte;
2926          break;
2927          }
2928    
2929        /* If a class contains a negative special such as \S, we need to flip the
2930        negation flag at the end, so that support for characters > 255 works
2931        correctly (they are all included in the class). */
2932    
2933        should_flip_negation = FALSE;
2934    
2935      /* Keep a count of chars with values < 256 so that we can optimize the case      /* Keep a count of chars with values < 256 so that we can optimize the case
2936      of just a single character (as long as it's < 256). However, For higher      of just a single character (as long as it's < 256). However, For higher
2937      valued UTF-8 characters, we don't yet do any optimization. */      valued UTF-8 characters, we don't yet do any optimization. */
# Line 2648  for (;; ptr++) Line 2949  for (;; ptr++)
2949  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2950      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
2951      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
2952        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
2953  #endif  #endif
2954    
2955      /* 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 2663  for (;; ptr++) Line 2965  for (;; ptr++)
2965          {                           /* Braces are required because the */          {                           /* Braces are required because the */
2966          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
2967          }          }
2968    
2969          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
2970          data and reset the pointer. This is so that very large classes that
2971          contain a zillion UTF-8 characters no longer overwrite the work space
2972          (which is on the stack). */
2973    
2974          if (lengthptr != NULL)
2975            {
2976            *lengthptr += class_utf8data - class_utf8data_base;
2977            class_utf8data = class_utf8data_base;
2978            }
2979    
2980  #endif  #endif
2981    
2982        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
2983    
2984        if (inescq)        if (inescq)
2985          {          {
2986          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
2987            {            {
2988            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
2989            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2684  for (;; ptr++) Line 2998  for (;; ptr++)
2998        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
2999        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3000    
3001        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3002            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3003            check_posix_syntax(ptr, &tempptr, cd))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3004          {          {
3005          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3006          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3007          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3008          uschar pbits[32];          uschar pbits[32];
3009    
3010          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3011            {            {
3012            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3013            goto FAILED;            goto FAILED;
3014            }            }
3015    
3016          ptr += 2;          ptr += 2;
3017          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3018            {            {
3019            local_negate = TRUE;            local_negate = TRUE;
3020              should_flip_negation = TRUE;  /* Note negative special */
3021            ptr++;            ptr++;
3022            }            }
3023    
# Line 2772  for (;; ptr++) Line 3087  for (;; ptr++)
3087        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
3088        character in them, so set class_charcount bigger than one. */        character in them, so set class_charcount bigger than one. */
3089    
3090        if (c == '\\')        if (c == CHAR_BACKSLASH)
3091          {          {
3092          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3093          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3094    
3095          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 */
3096          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 */
3097          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 */
3098          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3099            {            {
3100            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3101              {              {
3102              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3103              }              }
# Line 2805  for (;; ptr++) Line 3120  for (;; ptr++)
3120              continue;              continue;
3121    
3122              case ESC_D:              case ESC_D:
3123                should_flip_negation = TRUE;
3124              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
3125              continue;              continue;
3126    
# Line 2813  for (;; ptr++) Line 3129  for (;; ptr++)
3129              continue;              continue;
3130    
3131              case ESC_W:              case ESC_W:
3132                should_flip_negation = TRUE;
3133              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3134              continue;              continue;
3135    
# Line 2822  for (;; ptr++) Line 3139  for (;; ptr++)
3139              continue;              continue;
3140    
3141              case ESC_S:              case ESC_S:
3142                should_flip_negation = TRUE;
3143              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
3144              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3145              continue;              continue;
3146    
             case ESC_E: /* Perl ignores an orphan \E */  
             continue;  
   
3147              default:    /* Not recognized; fall through */              default:    /* Not recognized; fall through */
3148              break;      /* Need "default" setting to stop compiler warning. */              break;      /* Need "default" setting to stop compiler warning. */
3149              }              }
# Line 3008  for (;; ptr++) Line 3323  for (;; ptr++)
3323        entirely. The code for handling \Q and \E is messy. */        entirely. The code for handling \Q and \E is messy. */
3324    
3325        CHECK_RANGE:        CHECK_RANGE:
3326        while (ptr[1] == '\\' && ptr[2] == 'E')        while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3327          {          {
3328          inescq = FALSE;          inescq = FALSE;
3329          ptr += 2;          ptr += 2;
# Line 3018  for (;; ptr++) Line 3333  for (;; ptr++)
3333    
3334        /* Remember \r or \n */        /* Remember \r or \n */
3335    
3336        if (c == '\r' || c == '\n') cd->external_flags |= PCRE_HASCRORLF;        if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3337    
3338        /* Check for range */        /* Check for range */
3339    
3340        if (!inescq && ptr[1] == '-')        if (!inescq && ptr[1] == CHAR_MINUS)
3341          {          {
3342          int d;          int d;
3343          ptr += 2;          ptr += 2;
3344          while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3345    
3346          /* 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
3347          mode. */          mode. */
3348    
3349          while (*ptr == '\\' && ptr[1] == 'Q')          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3350            {            {
3351            ptr += 2;            ptr += 2;
3352            if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }            if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3353                { ptr += 2; continue; }
3354            inescq = TRUE;            inescq = TRUE;
3355            break;            break;
3356            }            }
3357    
3358          if (*ptr == 0 || (!inescq && *ptr == ']'))          if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3359            {            {
3360            ptr = oldptr;            ptr = oldptr;
3361            goto LONE_SINGLE_CHARACTER;            goto LONE_SINGLE_CHARACTER;
# Line 3058  for (;; ptr++) Line 3374  for (;; ptr++)
3374          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
3375          in such circumstances. */          in such circumstances. */
3376    
3377          if (!inescq && d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3378            {            {
3379            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3380            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3381    
3382            /* \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
3383            special means the '-' was literal */            special means the '-' was literal */
3384    
3385            if (d < 0)            if (d < 0)
3386              {              {
3387              if (d == -ESC_b) d = '\b';              if (d == -ESC_b) d = CHAR_BS;
3388              else if (d == -ESC_X) d = 'X';              else if (d == -ESC_X) d = CHAR_X;
3389              else if (d == -ESC_R) d = 'R'; else              else if (d == -ESC_R) d = CHAR_R; else
3390                {                {
3391                ptr = oldptr;                ptr = oldptr;
3392                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 3091  for (;; ptr++) Line 3407  for (;; ptr++)
3407    
3408          /* Remember \r or \n */          /* Remember \r or \n */
3409    
3410          if (d == '\r' || d == '\n') cd->external_flags |= PCRE_HASCRORLF;          if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3411    
3412          /* 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
3413          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 3211  for (;; ptr++) Line 3527  for (;; ptr++)
3527          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3528            {            {
3529            unsigned int othercase;            unsigned int othercase;
3530            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)            if ((othercase = UCD_OTHERCASE(c)) != c)
3531              {              {
3532              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3533              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 3238  for (;; ptr++) Line 3554  for (;; ptr++)
3554    
3555      /* Loop until ']' reached. This "while" is the end of the "do" above. */      /* Loop until ']' reached. This "while" is the end of the "do" above. */
3556    
3557      while ((c = *(++ptr)) != 0 && (c != ']' || inescq));      while ((c = *(++ptr)) != 0 && (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
3558    
3559      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)                          /* Missing terminating ']' */
3560        {        {
# Line 3327  we set the flag only if there is a liter Line 3643  we set the flag only if there is a liter
3643      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3644    
3645      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
3646      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode, unless there was a negated special
3647      we can omit the bitmap in the actual compiled code. */      such as \S in the class, because in that case all characters > 255 are in
3648        the class, so any that were explicitly given as well can be ignored. If
3649        (when there are explicit characters > 255 that must be listed) there are no
3650        characters < 256, we can omit the bitmap in the actual compiled code. */
3651    
3652  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3653      if (class_utf8)      if (class_utf8 && !should_flip_negation)
3654        {        {
3655        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
3656        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 3357  we set the flag only if there is a liter Line 3676  we set the flag only if there is a liter
3676        }        }
3677  #endif  #endif
3678    
3679      /* If there are no characters > 255, negate the 32-byte map if necessary,      /* If there are no characters > 255, set the opcode to OP_CLASS or
3680      and copy it into the code vector. If this is the first thing in the branch,      OP_NCLASS, depending on whether the whole class was negated and whether
3681      there can be no first char setting, whatever the repeat count. Any reqbyte      there were negative specials such as \S in the class. Then copy the 32-byte
3682      setting must remain unchanged after any kind of repeat. */      map into the code vector, negating it if necessary. */
3683    
3684        *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
3685      if (negate_class)      if (negate_class)
3686        {        {
       *code++ = OP_NCLASS;  
3687        if (lengthptr == NULL)    /* Save time in the pre-compile phase */        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
3688          for (c = 0; c < 32; c++) code[c] = ~classbits[c];          for (c = 0; c < 32; c++) code[c] = ~classbits[c];
3689        }        }
3690      else      else
3691        {        {
       *code++ = OP_CLASS;  
3692        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
3693        }        }
3694      code += 32;      code += 32;
# Line 3381  we set the flag only if there is a liter Line 3699  we set the flag only if there is a liter
3699      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
3700      has been tested above. */      has been tested above. */
3701    
3702      case '{':      case CHAR_LEFT_CURLY_BRACKET:
3703      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
3704      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
3705      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
3706      goto REPEAT;      goto REPEAT;
3707    
3708      case '*':      case CHAR_ASTERISK:
3709      repeat_min = 0;      repeat_min = 0;
3710      repeat_max = -1;      repeat_max = -1;
3711      goto REPEAT;      goto REPEAT;
3712    
3713      case '+':      case CHAR_PLUS:
3714      repeat_min = 1;      repeat_min = 1;
3715      repeat_max = -1;      repeat_max = -1;
3716      goto REPEAT;      goto REPEAT;
3717    
3718      case '?':      case CHAR_QUESTION_MARK:
3719      repeat_min = 0;      repeat_min = 0;
3720      repeat_max = 1;      repeat_max = 1;
3721    
# Line 3432  we set the flag only if there is a liter Line 3750  we set the flag only if there is a liter
3750      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
3751      repeat type to the non-default. */      repeat type to the non-default. */
3752    
3753      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
3754        {        {
3755        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
3756        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
3757        ptr++;        ptr++;
3758        }        }
3759      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
3760        {        {
3761        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
3762        ptr++;        ptr++;
# Line 3553  we set the flag only if there is a liter Line 3871  we set the flag only if there is a liter
3871    
3872        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
3873    
3874          /*--------------------------------------------------------------------*/
3875          /* This code is obsolete from release 8.00; the restriction was finally
3876          removed: */
3877    
3878        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
3879        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
3880    
3881        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
3882          /*--------------------------------------------------------------------*/
3883    
3884        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
3885    
# Line 3703  we set the flag only if there is a liter Line 4026  we set the flag only if there is a liter
4026          goto END_REPEAT;          goto END_REPEAT;
4027          }          }
4028    
4029          /*--------------------------------------------------------------------*/
4030          /* This code is obsolete from release 8.00; the restriction was finally
4031          removed: */
4032    
4033        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4034        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4035    
4036        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4037          /*--------------------------------------------------------------------*/
4038    
4039        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4040          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 3764  we set the flag only if there is a liter Line 4092  we set the flag only if there is a liter
4092    
4093        if (repeat_min == 0)        if (repeat_min == 0)
4094          {          {
4095          /* 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
4096          altogether. */          output altogether, like this:
   
         if (repeat_max == 0)  
           {  
           code = previous;  
           goto END_REPEAT;  
           }  
4097    
4098          /* If the maximum is 1 or unlimited, we just have to stick in the          ** if (repeat_max == 0)
4099          BRAZERO and do no more at this point. However, we do need to adjust          **   {
4100          any OP_RECURSE calls inside the group that refer to the group itself or          **   code = previous;
4101          any internal or forward referenced group, because the offset is from          **   goto END_REPEAT;
4102          the start of the whole regex. Temporarily terminate the pattern while          **   }
4103          doing this. */  
4104            However, that fails when a group is referenced as a subroutine from
4105            elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it
4106            so that it is skipped on execution. As we don't have a list of which
4107            groups are referenced, we cannot do this selectively.
4108    
4109            If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4110            and do no more at this point. However, we do need to adjust any
4111            OP_RECURSE calls inside the group that refer to the group itself or any
4112            internal or forward referenced group, because the offset is from the
4113            start of the whole regex. Temporarily terminate the pattern while doing
4114            this. */
4115    
4116          if (repeat_max <= 1)          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
4117            {            {
4118            *code = OP_END;            *code = OP_END;
4119            adjust_recurse(previous, 1, utf8, cd, save_hwm);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
4120            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
4121            code++;            code++;
4122              if (repeat_max == 0)
4123                {
4124                *previous++ = OP_SKIPZERO;
4125                goto END_REPEAT;
4126                }
4127            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4128            }            }
4129    
# Line 3980  we set the flag only if there is a liter Line 4318  we set the flag only if there is a liter
4318          }          }
4319        }        }
4320    
4321        /* If previous is OP_FAIL, it was generated by an empty class [] in
4322        JavaScript mode. The other ways in which OP_FAIL can be generated, that is
4323        by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
4324        error above. We can just ignore the repeat in JS case. */
4325    
4326        else if (*previous == OP_FAIL) goto END_REPEAT;
4327    
4328      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
4329    
4330      else      else
# Line 4004  we set the flag only if there is a liter Line 4349  we set the flag only if there is a liter
4349      if (possessive_quantifier)      if (possessive_quantifier)
4350        {        {
4351        int len;        int len;
4352        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||  
4353            *tempcode == OP_NOTEXACT)        if (*tempcode == OP_TYPEEXACT)
4354            tempcode += _pcre_OP_lengths[*tempcode] +
4355              ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
4356    
4357          else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
4358            {
4359          tempcode += _pcre_OP_lengths[*tempcode];          tempcode += _pcre_OP_lengths[*tempcode];
4360    #ifdef SUPPORT_UTF8
4361            if (utf8 && tempcode[-1] >= 0xc0)
4362              tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];
4363    #endif
4364            }
4365    
4366        len = code - tempcode;        len = code - tempcode;
4367        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4368          {          {
# Line 4052  we set the flag only if there is a liter Line 4408  we set the flag only if there is a liter
4408      lookbehind or option setting or condition or all the other extended      lookbehind or option setting or condition or all the other extended
4409      parenthesis forms.  */      parenthesis forms.  */
4410    
4411      case '(':      case CHAR_LEFT_PARENTHESIS:
4412      newoptions = options;      newoptions = options;
4413      skipbytes = 0;      skipbytes = 0;
4414      bravalue = OP_CBRA;      bravalue = OP_CBRA;
# Line 4061  we set the flag only if there is a liter Line 4417  we set the flag only if there is a liter
4417    
4418      /* First deal with various "verbs" that can be introduced by '*'. */      /* First deal with various "verbs" that can be introduced by '*'. */
4419    
4420      if (*(++ptr) == '*' && (cd->ctypes[ptr[1]] & ctype_letter) != 0)      if (*(++ptr) == CHAR_ASTERISK && (cd->ctypes[ptr[1]] & ctype_letter) != 0)
4421        {        {
4422        int i, namelen;        int i, namelen;
4423        const char *vn = verbnames;        const char *vn = verbnames;
4424        const uschar *name = ++ptr;        const uschar *name = ++ptr;
4425        previous = NULL;        previous = NULL;
4426        while ((cd->ctypes[*++ptr] & ctype_letter) != 0);        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
4427        if (*ptr == ':')        if (*ptr == CHAR_COLON)
4428          {          {
4429          *errorcodeptr = ERR59;   /* Not supported */          *errorcodeptr = ERR59;   /* Not supported */
4430          goto FAILED;          goto FAILED;
4431          }          }
4432        if (*ptr != ')')        if (*ptr != CHAR_RIGHT_PARENTHESIS)
4433          {          {
4434          *errorcodeptr = ERR60;          *errorcodeptr = ERR60;
4435          goto FAILED;          goto FAILED;
# Line 4088  we set the flag only if there is a liter Line 4444  we set the flag only if there is a liter
4444            if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;            if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;
4445            break;            break;
4446            }            }
4447          vn += verbs[i].len + 1;          vn += verbs[i].len + 1;
4448          }          }
4449        if (i < verbcount) continue;        if (i < verbcount) continue;
4450        *errorcodeptr = ERR60;        *errorcodeptr = ERR60;
# Line 4098  we set the flag only if there is a liter Line 4454  we set the flag only if there is a liter
4454      /* Deal with the extended parentheses; all are introduced by '?', and the      /* Deal with the extended parentheses; all are introduced by '?', and the
4455      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. */
4456    
4457      else if (*ptr == '?')      else if (*ptr == CHAR_QUESTION_MARK)
4458        {        {
4459        int i, set, unset, namelen;        int i, set, unset, namelen;
4460        int *optset;        int *optset;
# Line 4107  we set the flag only if there is a liter Line 4463  we set the flag only if there is a liter
4463    
4464        switch (*(++ptr))        switch (*(++ptr))
4465          {          {
4466          case '#':                 /* Comment; skip to ket */          case CHAR_NUMBER_SIGN:                 /* Comment; skip to ket */
4467          ptr++;          ptr++;
4468          while (*ptr != 0 && *ptr != ')') ptr++;          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
4469          if (*ptr == 0)          if (*ptr == 0)
4470            {            {
4471            *errorcodeptr = ERR18;            *errorcodeptr = ERR18;
# Line 4119  we set the flag only if there is a liter Line 4475  we set the flag only if there is a liter
4475    
4476    
4477          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4478          case '|':                 /* Reset capture count for each branch */          case CHAR_VERTICAL_LINE:  /* Reset capture count for each branch */
4479          reset_bracount = TRUE;          reset_bracount = TRUE;
4480          /* Fall through */          /* Fall through */
4481    
4482          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4483          case ':':                 /* Non-capturing bracket */          case CHAR_COLON:          /* Non-capturing bracket */
4484          bravalue = OP_BRA;          bravalue = OP_BRA;
4485          ptr++;          ptr++;
4486          break;          break;
4487    
4488    
4489          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4490          case '(':          case CHAR_LEFT_PARENTHESIS:
4491          bravalue = OP_COND;       /* Conditional group */          bravalue = OP_COND;       /* Conditional group */
4492    
4493          /* 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 4151  we set the flag only if there is a liter Line 4507  we set the flag only if there is a liter
4507          the switch. This will take control down to where bracketed groups,          the switch. This will take control down to where bracketed groups,
4508          including assertions, are processed. */          including assertions, are processed. */
4509    
4510          if (ptr[1] == '?' && (ptr[2] == '=' || ptr[2] == '!' || ptr[2] == '<'))          if (ptr[1] == CHAR_QUESTION_MARK && (ptr[2] == CHAR_EQUALS_SIGN ||
4511                ptr[2] == CHAR_EXCLAMATION_MARK || ptr[2] == CHAR_LESS_THAN_SIGN))
4512            break;            break;
4513    
4514          /* 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 4163  we set the flag only if there is a liter Line 4520  we set the flag only if there is a liter
4520    
4521          /* Check for a test for recursion in a named group. */          /* Check for a test for recursion in a named group. */
4522    
4523          if (ptr[1] == 'R' && ptr[2] == '&')          if (ptr[1] == CHAR_R && ptr[2] == CHAR_AMPERSAND)
4524            {            {
4525            terminator = -1;            terminator = -1;
4526            ptr += 2;            ptr += 2;
# Line 4173  we set the flag only if there is a liter Line 4530  we set the flag only if there is a liter
4530          /* 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
4531          syntax (?(<name>) or (?('name') */          syntax (?(<name>) or (?('name') */
4532    
4533          else if (ptr[1] == '<')          else if (ptr[1] == CHAR_LESS_THAN_SIGN)
4534            {            {
4535            terminator = '>';            terminator = CHAR_GREATER_THAN_SIGN;
4536            ptr++;            ptr++;
4537            }            }
4538          else if (ptr[1] == '\'')          else if (ptr[1] == CHAR_APOSTROPHE)
4539            {            {
4540            terminator = '\'';            terminator = CHAR_APOSTROPHE;
4541            ptr++;            ptr++;
4542            }            }
4543          else          else
4544            {            {
4545            terminator = 0;            terminator = 0;
4546            if (ptr[1] == '-' || ptr[1] == '+') refsign = *(++ptr);            if (ptr[1] == CHAR_MINUS || ptr[1] == CHAR_PLUS) refsign = *(++ptr);
4547            }            }
4548    
4549          /* 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 4206  we set the flag only if there is a liter Line 4563  we set the flag only if there is a liter
4563            {            {
4564            if (recno >= 0)            if (recno >= 0)
4565              recno = ((digitab[*ptr] & ctype_digit) != 0)?              recno = ((digitab[*ptr] & ctype_digit) != 0)?
4566                recno * 10 + *ptr - '0' : -1;                recno * 10 + *ptr - CHAR_0 : -1;
4567            ptr++;            ptr++;
4568            }            }
4569          namelen = ptr - name;          namelen = ptr - name;
4570    
4571          if ((terminator > 0 && *ptr++ != terminator) || *ptr++ != ')')          if ((terminator > 0 && *ptr++ != terminator) ||
4572                *ptr++ != CHAR_RIGHT_PARENTHESIS)
4573            {            {
4574            ptr--;      /* Error offset */            ptr--;      /* Error offset */
4575            *errorcodeptr = ERR26;            *errorcodeptr = ERR26;
# Line 4233  we set the flag only if there is a liter Line 4591  we set the flag only if there is a liter
4591              *errorcodeptr = ERR58;              *errorcodeptr = ERR58;
4592              goto FAILED;              goto FAILED;
4593              }              }
4594            if (refsign == '-')            recno = (refsign == CHAR_MINUS)?
4595                cd->bracount - recno + 1 : recno +cd->bracount;
4596              if (recno <= 0 || recno > cd->final_bracount)
4597              {              {
4598              recno = cd->bracount - recno + 1;              *errorcodeptr = ERR15;
4599              if (recno <= 0)              goto FAILED;
               {  
               *errorcodeptr = ERR15;  
               goto FAILED;  
               }  
4600              }              }
           else recno += cd->bracount;  
4601            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4602            break;            break;
4603            }            }
# Line 4267  we set the flag only if there is a liter Line 4622  we set the flag only if there is a liter
4622    
4623          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
4624    
4625          else if ((i = find_parens(ptr, cd->bracount, name, namelen,          else if ((i = find_parens(cd, name, namelen,
4626                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0)) > 0)
4627            {            {
4628            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
# Line 4288  we set the flag only if there is a liter Line 4643  we set the flag only if there is a liter
4643          /* Check for (?(R) for recursion. Allow digits after R to specify a          /* Check for (?(R) for recursion. Allow digits after R to specify a
4644          specific group number. */          specific group number. */
4645    
4646          else if (*name == 'R')          else if (*name == CHAR_R)
4647            {            {
4648            recno = 0;            recno = 0;
4649            for (i = 1; i < namelen; i++)            for (i = 1; i < namelen; i++)
# Line 4298  we set the flag only if there is a liter Line 4653  we set the flag only if there is a liter
4653                *errorcodeptr = ERR15;                *errorcodeptr = ERR15;
4654                goto FAILED;                goto FAILED;
4655                }                }
4656              recno = recno * 10 + name[i] - '0';              recno = recno * 10 + name[i] - CHAR_0;
4657              }              }
4658            if (recno == 0) recno = RREF_ANY;            if (recno == 0) recno = RREF_ANY;
4659            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */
# Line 4308  we set the flag only if there is a liter Line 4663  we set the flag only if there is a liter
4663          /* Similarly, check for the (?(DEFINE) "condition", which is always          /* Similarly, check for the (?(DEFINE) "condition", which is always
4664          false. */          false. */
4665    
4666          else if (namelen == 6 && strncmp((char *)name, "DEFINE", 6) == 0)          else if (namelen == 6 && strncmp((char *)name, STRING_DEFINE, 6) == 0)
4667            {            {
4668            code[1+LINK_SIZE] = OP_DEF;            code[1+LINK_SIZE] = OP_DEF;
4669            skipbytes = 1;            skipbytes = 1;
4670            }            }
4671    
4672          /* Check for the "name" actually being a subpattern number. */          /* Check for the "name" actually being a subpattern number. We are
4673            in the second pass here, so final_bracount is set. */
4674    
4675          else if (recno > 0)          else if (recno > 0 && recno <= cd->final_bracount)
4676            {            {
4677            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4678            }            }
# Line 4332  we set the flag only if there is a liter Line 4688  we set the flag only if there is a liter
4688    
4689    
4690          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4691          case '=':                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
4692          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
4693          ptr++;          ptr++;
4694          break;          break;
4695    
4696    
4697          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4698          case '!':                 /* Negative lookahead */          case CHAR_EXCLAMATION_MARK:            /* Negative lookahead */
4699          ptr++;          ptr++;
4700          if (*ptr == ')')          /* Optimize (?!) */          if (*ptr == CHAR_RIGHT_PARENTHESIS)    /* Optimize (?!) */
4701            {            {
4702            *code++ = OP_FAIL;            *code++ = OP_FAIL;
4703            previous = NULL;            previous = NULL;
# Line 4352  we set the flag only if there is a liter Line 4708  we set the flag only if there is a liter
4708    
4709    
4710          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4711          case '<':                 /* Lookbehind or named define */          case CHAR_LESS_THAN_SIGN:              /* Lookbehind or named define */
4712          switch (ptr[1])          switch (ptr[1])
4713            {            {
4714            case '=':               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
4715            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
4716            ptr += 2;            ptr += 2;
4717            break;            break;
4718    
4719            case '!':               /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
4720            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
4721            ptr += 2;            ptr += 2;
4722            break;            break;
# Line 4375  we set the flag only if there is a liter Line 4731  we set the flag only if there is a liter
4731    
4732    
4733          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4734          case '>':                 /* One-time brackets */          case CHAR_GREATER_THAN_SIGN:           /* One-time brackets */
4735          bravalue = OP_ONCE;          bravalue = OP_ONCE;
4736          ptr++;          ptr++;
4737          break;          break;
4738    
4739    
4740          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4741          case 'C':                 /* Callout - may be followed by digits; */          case CHAR_C:                 /* Callout - may be followed by digits; */
4742          previous_callout = code;  /* Save for later completion */          previous_callout = code;  /* Save for later completion */
4743          after_manual_callout = 1; /* Skip one item before completing */          after_manual_callout = 1; /* Skip one item before completing */
4744          *code++ = OP_CALLOUT;          *code++ = OP_CALLOUT;
4745            {            {
4746            int n = 0;            int n = 0;
4747            while ((digitab[*(++ptr)] & ctype_digit) != 0)            while ((digitab[*(++ptr)] & ctype_digit) != 0)
4748              n = n * 10 + *ptr - '0';              n = n * 10 + *ptr - CHAR_0;
4749            if (*ptr != ')')            if (*ptr != CHAR_RIGHT_PARENTHESIS)
4750              {              {
4751              *errorcodeptr = ERR39;              *errorcodeptr = ERR39;
4752              goto FAILED;              goto FAILED;
# Line 4410  we set the flag only if there is a liter Line 4766  we set the flag only if there is a liter
4766    
4767    
4768          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4769          case 'P':                 /* Python-style named subpattern handling */          case CHAR_P:              /* Python-style named subpattern handling */
4770          if (*(++ptr) == '=' || *ptr == '>')  /* Reference or recursion */          if (*(++ptr) == CHAR_EQUALS_SIGN ||
4771                *ptr == CHAR_GREATER_THAN_SIGN)  /* Reference or recursion */
4772            {            {
4773            is_recurse = *ptr == '>';            is_recurse = *ptr == CHAR_GREATER_THAN_SIGN;
4774            terminator = ')';            terminator = CHAR_RIGHT_PARENTHESIS;
4775            goto NAMED_REF_OR_RECURSE;            goto NAMED_REF_OR_RECURSE;
4776            }            }
4777          else if (*ptr != '<')    /* Test for Python-style definition */          else if (*ptr != CHAR_LESS_THAN_SIGN)  /* Test for Python-style defn */
4778            {            {
4779            *errorcodeptr = ERR41;            *errorcodeptr = ERR41;
4780            goto FAILED;            goto FAILED;
# Line 4427  we set the flag only if there is a liter Line 4784  we set the flag only if there is a liter
4784    
4785          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4786          DEFINE_NAME:    /* Come here from (?< handling */          DEFINE_NAME:    /* Come here from (?< handling */
4787          case '\'':          case CHAR_APOSTROPHE:
4788            {            {
4789            terminator = (*ptr == '<')? '>' : '\'';            terminator = (*ptr == CHAR_LESS_THAN_SIGN)?
4790                CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
4791            name = ++ptr;            name = ++ptr;
4792    
4793            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
# Line 4503  we set the flag only if there is a liter Line 4861  we set the flag only if there is a liter
4861    
4862    
4863          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4864          case '&':                 /* Perl recursion/subroutine syntax */          case CHAR_AMPERSAND:            /* Perl recursion/subroutine syntax */
4865          terminator = ')';          terminator = CHAR_RIGHT_PARENTHESIS;
4866          is_recurse = TRUE;          is_recurse = TRUE;
4867          /* Fall through */          /* Fall through */
4868    
4869          /* We come here from the Python syntax above that handles both          /* We come here from the Python syntax above that handles both
4870          references (?P=name) and recursion (?P>name), as well as falling          references (?P=name) and recursion (?P>name), as well as falling
4871          through from the Perl recursion syntax (?&name). */          through from the Perl recursion syntax (?&name). We also come here from
4872            the Perl \k<name> or \k'name' back reference syntax and the \k{name}
4873            .NET syntax, and the Oniguruma \g<...> and \g'...' subroutine syntax. */
4874    
4875          NAMED_REF_OR_RECURSE:          NAMED_REF_OR_RECURSE:
4876          name = ++ptr;          name = ++ptr;
# Line 4522  we set the flag only if there is a liter Line 4882  we set the flag only if there is a liter
4882    
4883          if (lengthptr != NULL)          if (lengthptr != NULL)
4884            {            {
4885              if (namelen == 0)
4886                {
4887                *errorcodeptr = ERR62;
4888                goto FAILED;
4889                }
4890            if (*ptr != terminator)            if (*ptr != terminator)
4891              {              {
4892              *errorcodeptr = ERR42;              *errorcodeptr = ERR42;
# Line 4535  we set the flag only if there is a liter Line 4900  we set the flag only if there is a liter
4900            recno = 0;            recno = 0;
4901            }            }
4902    
4903          /* In the real compile, seek the name in the table */          /* In the real compile, seek the name in the table. We check the name
4904            first, and then check that we have reached the end of the name in the
4905            table. That way, if the name that is longer than any in the table,
4906            the comparison will fail without reading beyond the table entry. */
4907    
4908          else          else
4909            {            {
4910            slot = cd->name_table;            slot = cd->name_table;
4911            for (i = 0; i < cd->names_found; i++)            for (i = 0; i < cd->names_found; i++)
4912              {              {
4913              if (strncmp((char *)name, (char *)slot+2, namelen) == 0) break;              if (strncmp((char *)name, (char *)slot+2, namelen) == 0 &&
4914                    slot[2+namelen] == 0)
4915                  break;
4916              slot += cd->name_entry_size;              slot += cd->name_entry_size;
4917              }              }
4918    
# Line 4551  we set the flag only if there is a liter Line 4921  we set the flag only if there is a liter
4921              recno = GET2(slot, 0);              recno = GET2(slot, 0);
4922              }              }
4923            else if ((recno =                /* Forward back reference */            else if ((recno =                /* Forward back reference */
4924                      find_parens(ptr, cd->bracount, name, namelen,                      find_parens(cd, name, namelen,
4925                        (options & PCRE_EXTENDED) != 0)) <= 0)                        (options & PCRE_EXTENDED) != 0)) <= 0)
4926              {              {
4927              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
# Line 4567  we set the flag only if there is a liter Line 4937  we set the flag only if there is a liter
4937    
4938    
4939          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4940          case 'R':                 /* Recursion */          case CHAR_R:              /* Recursion */
4941          ptr++;                    /* Same as (?0)      */          ptr++;                    /* Same as (?0)      */
4942          /* Fall through */          /* Fall through */
4943    
4944    
4945          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4946          case '-': case '+':          case CHAR_MINUS: case CHAR_PLUS:  /* Recursion or subroutine */
4947          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:
4948          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:
4949            {            {
4950            const uschar *called;            const uschar *called;
4951              terminator = CHAR_RIGHT_PARENTHESIS;
4952    
4953              /* Come here from the \g<...> and \g'...' code (Oniguruma
4954              compatibility). However, the syntax has been checked to ensure that
4955              the ... are a (signed) number, so that neither ERR63 nor ERR29 will
4956              be called on this path, nor with the jump to OTHER_CHAR_AFTER_QUERY
4957              ever be taken. */
4958    
4959            if ((refsign = *ptr) == '+') ptr++;            HANDLE_NUMERICAL_RECURSION:
4960            else if (refsign == '-')  
4961              if ((refsign = *ptr) == CHAR_PLUS)
4962                {
4963                ptr++;
4964                if ((digitab[*ptr] & ctype_digit) == 0)
4965                  {
4966                  *errorcodeptr = ERR63;
4967                  goto FAILED;
4968                  }
4969                }
4970              else if (refsign == CHAR_MINUS)
4971              {              {
4972              if ((digitab[ptr[1]] & ctype_digit) == 0)              if ((digitab[ptr[1]] & ctype_digit) == 0)
4973                goto OTHER_CHAR_AFTER_QUERY;                goto OTHER_CHAR_AFTER_QUERY;
# Line 4589  we set the flag only if there is a liter Line 4976  we set the flag only if there is a liter
4976    
4977            recno = 0;            recno = 0;
4978            while((digitab[*ptr] & ctype_digit) != 0)            while((digitab[*ptr] & ctype_digit) != 0)
4979              recno = recno * 10 + *ptr++ - '0';              recno = recno * 10 + *ptr++ - CHAR_0;
4980    
4981            if (*ptr != ')')            if (*ptr != terminator)
4982              {              {
4983              *errorcodeptr = ERR29;              *errorcodeptr = ERR29;
4984              goto FAILED;              goto FAILED;
4985              }              }
4986    
4987            if (refsign == '-')            if (refsign == CHAR_MINUS)
4988              {              {
4989              if (recno == 0)              if (recno == 0)
4990                {                {
# Line 4611  we set the flag only if there is a liter Line 4998  we set the flag only if there is a liter
4998                goto FAILED;                goto FAILED;
4999                }                }
5000              }              }
5001            else if (refsign == '+')            else if (refsign == CHAR_PLUS)
5002              {              {
5003              if (recno == 0)              if (recno == 0)
5004                {                {
# Line 4644  we set the flag only if there is a liter Line 5031  we set the flag only if there is a liter
5031    
5032              if (called == NULL)              if (called == NULL)
5033                {                {
5034                if (find_parens(ptr, cd->bracount, NULL, recno,                if (find_parens(cd, NULL, recno,
5035                     (options & PCRE_EXTENDED) != 0) < 0)                      (options & PCRE_EXTENDED) != 0) < 0)
5036                  {                  {
5037                  *errorcodeptr = ERR15;                  *errorcodeptr = ERR15;
5038                  goto FAILED;                  goto FAILED;
# Line 4697  we set the flag only if there is a liter Line 5084  we set the flag only if there is a liter
5084          set = unset = 0;          set = unset = 0;
5085          optset = &set;          optset = &set;
5086    
5087          while (*ptr != ')' && *ptr != ':')          while (*ptr != CHAR_RIGHT_PARENTHESIS && *ptr != CHAR_COLON)
5088            {            {
5089            switch (*ptr++)            switch (*ptr++)
5090              {              {
5091              case '-': optset = &unset; break;              case CHAR_MINUS: optset = &unset; break;
5092    
5093              case 'J':    /* Record that it changed in the external options */              case CHAR_J:    /* Record that it changed in the external options */
5094              *optset |= PCRE_DUPNAMES;              *optset |= PCRE_DUPNAMES;
5095              cd->external_flags |= PCRE_JCHANGED;              cd->external_flags |= PCRE_JCHANGED;
5096              break;              break;
5097    
5098              case 'i': *optset |= PCRE_CASELESS; break;              case CHAR_i: *optset |= PCRE_CASELESS; break;
5099              case 'm': *optset |= PCRE_MULTILINE; break;              case CHAR_m: *optset |= PCRE_MULTILINE; break;
5100              case 's': *optset |= PCRE_DOTALL; break;              case CHAR_s: *optset |= PCRE_DOTALL; break;
5101              case 'x': *optset |= PCRE_EXTENDED; break;              case CHAR_x: *optset |= PCRE_EXTENDED; break;
5102              case 'U': *optset |= PCRE_UNGREEDY; break;              case CHAR_U: *optset |= PCRE_UNGREEDY; break;
5103              case 'X': *optset |= PCRE_EXTRA; break;              case CHAR_X: *optset |= PCRE_EXTRA; break;
5104    
5105              default:  *errorcodeptr = ERR12;              default:  *errorcodeptr = ERR12;
5106                        ptr--;    /* Correct the offset */                        ptr--;    /* Correct the offset */
# Line 4744  we set the flag only if there is a liter Line 5131  we set the flag only if there is a liter
5131          both phases.          both phases.
5132    
5133          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
5134          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
5135          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. */  
5136    
5137          if (*ptr == ')')          if (*ptr == CHAR_RIGHT_PARENTHESIS)
5138            {            {
5139            if (code == cd->start_code + 1 + LINK_SIZE &&            if (code == cd->start_code + 1 + LINK_SIZE &&
5140                 (lengthptr == NULL || *lengthptr == 2 + 2*LINK_SIZE))                 (lengthptr == NULL || *lengthptr == 2 + 2*LINK_SIZE))
5141              {              {
5142              cd->external_options = newoptions;              cd->external_options = newoptions;
             options = newoptions;  
5143              }              }
5144           else           else
5145              {              {
# Line 4764  we set the flag only if there is a liter Line 5148  we set the flag only if there is a liter
5148                *code++ = OP_OPT;                *code++ = OP_OPT;
5149                *code++ = newoptions & PCRE_IMS;                *code++ = newoptions & PCRE_IMS;
5150                }                }
   
             /* 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;  
5151              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);
5152              greedy_non_default = greedy_default ^ 1;              greedy_non_default = greedy_default ^ 1;
5153              req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
5154              }              }
5155    
5156              /* Change options at this level, and pass them back for use
5157              in subsequent branches. When not at the start of the pattern, this
5158              information is also necessary so that a resetting item can be
5159              compiled at the end of a group (if we are in a group). */
5160    
5161              *optionsptr = options = newoptions;
5162            previous = NULL;       /* This item can't be repeated */            previous = NULL;       /* This item can't be repeated */
5163            continue;              /* It is complete */            continue;              /* It is complete */
5164            }            }
# Line 4890  we set the flag only if there is a liter Line 5274  we set the flag only if there is a liter
5274    
5275      /* Error if hit end of pattern */      /* Error if hit end of pattern */
5276    
5277      if (*ptr != ')')      if (*ptr != CHAR_RIGHT_PARENTHESIS)
5278        {        {
5279        *errorcodeptr = ERR14;        *errorcodeptr = ERR14;
5280        goto FAILED;        goto FAILED;
# Line 4988  we set the flag only if there is a liter Line 5372  we set the flag only if there is a liter
5372      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
5373      have to change if any new ones are ever created. */      have to change if any new ones are ever created. */
5374    
5375      case '\\':      case CHAR_BACKSLASH:
5376      tempptr = ptr;      tempptr = ptr;
5377      c = check_escape(&ptr, errorcodeptr, cd->bracount, options, FALSE);      c = check_escape(&ptr, errorcodeptr, cd->bracount, options, FALSE);
5378      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
# Line 4997  we set the flag only if there is a liter Line 5381  we set the flag only if there is a liter
5381        {        {
5382        if (-c == ESC_Q)            /* Handle start of quoted string */        if (-c == ESC_Q)            /* Handle start of quoted string */
5383          {          {
5384          if (ptr[1] == '\\' && ptr[2] == 'E') ptr += 2; /* avoid empty string */          if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
5385            else inescq = TRUE;            ptr += 2;               /* avoid empty string */
5386                else inescq = TRUE;
5387          continue;          continue;
5388          }          }
5389    
# Line 5015  we set the flag only if there is a liter Line 5400  we set the flag only if there is a liter
5400        zerofirstbyte = firstbyte;        zerofirstbyte = firstbyte;
5401        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
5402    
5403          /* \g<name> or \g'name' is a subroutine call by name and \g<n> or \g'n'
5404          is a subroutine call by number (Oniguruma syntax). In fact, the value
5405          -ESC_g is returned only for these cases. So we don't need to check for <
5406          or ' if the value is -ESC_g. For the Perl syntax \g{n} the value is
5407          -ESC_REF+n, and for the Perl syntax \g{name} the result is -ESC_k (as
5408          that is a synonym for a named back reference). */
5409    
5410          if (-c == ESC_g)
5411            {
5412            const uschar *p;
5413            save_hwm = cd->hwm;   /* Normally this is set when '(' is read */
5414            terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
5415              CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
5416    
5417            /* These two statements stop the compiler for warning about possibly
5418            unset variables caused by the jump to HANDLE_NUMERICAL_RECURSION. In
5419            fact, because we actually check for a number below, the paths that
5420            would actually be in error are never taken. */
5421    
5422            skipbytes = 0;
5423            reset_bracount = FALSE;
5424    
5425            /* Test for a name */
5426    
5427            if (ptr[1] != CHAR_PLUS && ptr[1] != CHAR_MINUS)
5428              {
5429              BOOL isnumber = TRUE;
5430              for (p = ptr + 1; *p != 0 && *p != terminator; p++)
5431                {
5432                if ((cd->ctypes[*p] & ctype_digit) == 0) isnumber = FALSE;
5433                if ((cd->ctypes[*p] & ctype_word) == 0) break;
5434                }
5435              if (*p != terminator)
5436                {
5437                *errorcodeptr = ERR57;
5438                break;
5439                }
5440              if (isnumber)
5441                {
5442                ptr++;
5443                goto HANDLE_NUMERICAL_RECURSION;
5444                }
5445              is_recurse = TRUE;
5446              goto NAMED_REF_OR_RECURSE;
5447              }
5448    
5449            /* Test a signed number in angle brackets or quotes. */
5450    
5451            p = ptr + 2;
5452            while ((digitab[*p] & ctype_digit) != 0) p++;
5453            if (*p != terminator)
5454              {
5455              *errorcodeptr = ERR57;
5456              break;
5457              }
5458            ptr++;
5459            goto HANDLE_NUMERICAL_RECURSION;
5460            }
5461    
5462        /* \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).
5463        We also support \k{name} (.NET syntax) */        We also support \k{name} (.NET syntax) */
5464    
5465        if (-c == ESC_k && (ptr[1] == '<' || ptr[1] == '\'' || ptr[1] == '{'))        if (-c == ESC_k && (ptr[1] == CHAR_LESS_THAN_SIGN ||
5466              ptr[1] == CHAR_APOSTROPHE || ptr[1] == CHAR_LEFT_CURLY_BRACKET))
5467          {          {
5468          is_recurse = FALSE;          is_recurse = FALSE;
5469          terminator = (*(++ptr) == '<')? '>' : (*ptr == '\'')? '\'' : '}';          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
5470              CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?
5471              CHAR_APOSTROPHE : CHAR_RIGHT_CURLY_BRACKET;
5472          goto NAMED_REF_OR_RECURSE;          goto NAMED_REF_OR_RECURSE;
5473          }          }
5474    
# Line 5124  we set the flag only if there is a liter Line 5571  we set the flag only if there is a liter
5571    
5572      /* Remember if \r or \n were seen */      /* Remember if \r or \n were seen */
5573    
5574      if (mcbuffer[0] == '\r' || mcbuffer[0] == '\n')      if (mcbuffer[0] == CHAR_CR || mcbuffer[0] == CHAR_NL)
5575        cd->external_flags |= PCRE_HASCRORLF;        cd->external_flags |= PCRE_HASCRORLF;
5576    
5577      /* Set the first and required bytes appropriately. If no previous first      /* Set the first and required bytes appropriately. If no previous first
# Line 5369  for (;;) Line 5816  for (;;)
5816    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.
5817    Return leaving the pointer at the terminating char. */    Return leaving the pointer at the terminating char. */
5818    
5819    if (*ptr != '|')    if (*ptr != CHAR_VERTICAL_LINE)
5820      {      {
5821      if (lengthptr == NULL)      if (lengthptr == NULL)
5822        {        {
# Line 5392  for (;;) Line 5839  for (;;)
5839    
5840      /* Resetting option if needed */      /* Resetting option if needed */
5841    
5842      if ((options & PCRE_IMS) != oldims && *ptr == ')')      if ((options & PCRE_IMS) != oldims && *ptr == CHAR_RIGHT_PARENTHESIS)
5843        {        {
5844        *code++ = OP_OPT;        *code++ = OP_OPT;
5845        *code++ = oldims;        *code++ = oldims;
# Line 5521  do { Line 5968  do {
5968       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;
5969       }       }
5970    
5971     /* .* 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
5972     are or may be referenced. */     it isn't in brackets that are or may be referenced. */
5973    
5974     else if ((op == OP_TYPESTAR || op == OP_TYPEMINSTAR ||     else if ((op == OP_TYPESTAR || op == OP_TYPEMINSTAR ||
5975               op == OP_TYPEPOSSTAR) &&               op == OP_TYPEPOSSTAR))
             (*options & PCRE_DOTALL) != 0)  
5976       {       {
5977       if (scode[1] != OP_ANY || (bracket_map & backref_map) != 0) return FALSE;       if (scode[1] != OP_ALLANY || (bracket_map & backref_map) != 0)
5978           return FALSE;
5979       }       }
5980    
5981     /* Check for explicit anchoring */     /* Check for explicit anchoring */
# Line 5574  do { Line 6021  do {
6021       NULL, 0, FALSE);       NULL, 0, FALSE);
6022     register int op = *scode;     register int op = *scode;
6023    
6024       /* If we are at the start of a conditional assertion group, *both* the
6025       conditional assertion *and* what follows the condition must satisfy the test
6026       for start of line. Other kinds of condition fail. Note that there may be an
6027       auto-callout at the start of a condition. */
6028    
6029       if (op == OP_COND)
6030         {
6031         scode += 1 + LINK_SIZE;
6032         if (*scode == OP_CALLOUT) scode += _pcre_OP_lengths[OP_CALLOUT];
6033         switch (*scode)
6034           {
6035           case OP_CREF:
6036           case OP_RREF:
6037           case OP_DEF:
6038           return FALSE;
6039    
6040           default:     /* Assertion */
6041           if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
6042           do scode += GET(scode, 1); while (*scode == OP_ALT);
6043           scode += 1 + LINK_SIZE;
6044           break;
6045           }
6046         scode = first_significant_code(scode, NULL, 0, FALSE);
6047         op = *scode;
6048         }
6049    
6050     /* Non-capturing brackets */     /* Non-capturing brackets */
6051    
6052     if (op == OP_BRA)     if (op == OP_BRA)
# Line 5592  do { Line 6065  do {
6065    
6066     /* Other brackets */     /* Other brackets */
6067    
6068     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_COND)     else if (op == OP_ASSERT || op == OP_ONCE)
6069       { if (!is_startline(scode, bracket_map, backref_map)) return FALSE; }       {
6070         if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
6071         }
6072    
6073     /* .* 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
6074     may be referenced. */     may be referenced. */
# Line 5710  Returns:        pointer to compiled data Line 6185  Returns:        pointer to compiled data
6185                  with errorptr and erroroffset set                  with errorptr and erroroffset set
6186  */  */
6187    
6188  PCRE_EXP_DEFN pcre *  PCRE_EXP_DEFN pcre * PCRE_CALL_CONVENTION
6189  pcre_compile(const char *pattern, int options, const char **errorptr,  pcre_compile(const char *pattern, int options, const char **errorptr,
6190    int *erroroffset, const unsigned char *tables)    int *erroroffset, const unsigned char *tables)
6191  {  {
# Line 5718  return pcre_compile2(pattern, options, N Line 6193  return pcre_compile2(pattern, options, N
6193  }  }
6194    
6195    
6196  PCRE_EXP_DEFN pcre *  PCRE_EXP_DEFN pcre * PCRE_CALL_CONVENTION
6197  pcre_compile2(const char *pattern, int options, int *errorcodeptr,  pcre_compile2(const char *pattern, int options, int *errorcodeptr,
6198    const char **errorptr, int *erroroffset, const unsigned char *tables)    const char **errorptr, int *erroroffset, const unsigned char *tables)
6199  {  {
# Line 5745  to fill in forward references to subpatt Line 6220  to fill in forward references to subpatt
6220    
6221  uschar cworkspace[COMPILE_WORK_SIZE];  uschar cworkspace[COMPILE_WORK_SIZE];
6222    
   
6223  /* Set this early so that early errors get offset 0. */  /* Set this early so that early errors get offset 0. */
6224    
6225  ptr = (const uschar *)pattern;  ptr = (const uschar *)pattern;
# Line 5773  if (erroroffset == NULL) Line 6247  if (erroroffset == NULL)
6247    
6248  *erroroffset = 0;  *erroroffset = 0;
6249    
 /* 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;  
   }  
   
6250  /* Set up pointers to the individual character tables */  /* Set up pointers to the individual character tables */
6251    
6252  if (tables == NULL) tables = _pcre_default_tables;  if (tables == NULL) tables = _pcre_default_tables;
# Line 5805  cd->fcc = tables + fcc_offset; Line 6255  cd->fcc = tables + fcc_offset;
6255  cd->cbits = tables + cbits_offset;  cd->cbits = tables + cbits_offset;
6256  cd->ctypes = tables + ctypes_offset;  cd->ctypes = tables + ctypes_offset;
6257    
6258    /* Check that all undefined public option bits are zero */
6259    
6260    if ((options & ~PUBLIC_COMPILE_OPTIONS) != 0)
6261      {
6262      errorcode = ERR17;
6263      goto PCRE_EARLY_ERROR_RETURN;
6264      }
6265    
6266  /* 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
6267  the offset for later. */  the offset for later. */
6268    
6269  while (ptr[skipatstart] == '(' && ptr[skipatstart+1] == '*')  while (ptr[skipatstart] == CHAR_LEFT_PARENTHESIS &&
6270           ptr[skipatstart+1] == CHAR_ASTERISK)
6271    {    {
6272    int newnl = 0;    int newnl = 0;
6273    int newbsr = 0;    int newbsr = 0;
6274    
6275    if (strncmp((char *)(ptr+skipatstart+2), "CR)", 3) == 0)    if (strncmp((char *)(ptr+skipatstart+2), STRING_UTF8_RIGHTPAR, 5) == 0)
6276        { skipatstart += 7; options |= PCRE_UTF8; continue; }
6277    
6278      if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)
6279      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }
6280    else if (strncmp((char *)(ptr+skipatstart+2), "LF)", 3)  == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_LF_RIGHTPAR, 3)  == 0)
6281      { skipatstart += 5; newnl = PCRE_NEWLINE_LF; }      { skipatstart += 5; newnl = PCRE_NEWLINE_LF; }
6282    else if (strncmp((char *)(ptr+skipatstart+2), "CRLF)", 5)  == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_CRLF_RIGHTPAR, 5)  == 0)
6283      { skipatstart += 7; newnl = PCRE_NEWLINE_CR + PCRE_NEWLINE_LF; }      { skipatstart += 7; newnl = PCRE_NEWLINE_CR + PCRE_NEWLINE_LF; }
6284    else if (strncmp((char *)(ptr+skipatstart+2), "ANY)", 4) == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_ANY_RIGHTPAR, 4) == 0)
6285      { skipatstart += 6; newnl = PCRE_NEWLINE_ANY; }      { skipatstart += 6; newnl = PCRE_NEWLINE_ANY; }
6286    else if (strncmp((char *)(ptr+skipatstart+2), "ANYCRLF)", 8)  == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_ANYCRLF_RIGHTPAR, 8) == 0)
6287      { skipatstart += 10; newnl = PCRE_NEWLINE_ANYCRLF; }      { skipatstart += 10; newnl = PCRE_NEWLINE_ANYCRLF; }
6288    
6289    else if (strncmp((char *)(ptr+skipatstart+2), "BSR_ANYCRLF)", 12) == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_BSR_ANYCRLF_RIGHTPAR, 12) == 0)
6290      { skipatstart += 14; newbsr = PCRE_BSR_ANYCRLF; }      { skipatstart += 14; newbsr = PCRE_BSR_ANYCRLF; }
6291    else if (strncmp((char *)(ptr+skipatstart+2), "BSR_UNICODE)", 12) == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_BSR_UNICODE_RIGHTPAR, 12) == 0)
6292      { skipatstart += 14; newbsr = PCRE_BSR_UNICODE; }      { skipatstart += 14; newbsr = PCRE_BSR_UNICODE; }
6293    
6294    if (newnl != 0)    if (newnl != 0)
# Line 5836  while (ptr[skipatstart] == '(' && ptr[sk Line 6298  while (ptr[skipatstart] == '(' && ptr[sk
6298    else break;    else break;
6299    }    }
6300    
6301    /* Can't support UTF8 unless PCRE has been compiled to include the code. */
6302    
6303    #ifdef SUPPORT_UTF8
6304    utf8 = (options & PCRE_UTF8) != 0;
6305    if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&
6306         (*erroroffset = _pcre_valid_utf8((uschar *)pattern, -1)) >= 0)
6307      {
6308      errorcode = ERR44;
6309      goto PCRE_EARLY_ERROR_RETURN2;
6310      }
6311    #else
6312    if ((options & PCRE_UTF8) != 0)
6313      {
6314      errorcode = ERR32;
6315      goto PCRE_EARLY_ERROR_RETURN;
6316      }
6317    #endif
6318    
6319  /* Check validity of \R options. */  /* Check validity of \R options. */
6320    
6321  switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))  switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))
# Line 5854  current code allows for fixed one- or tw Line 6334  current code allows for fixed one- or tw
6334  switch (options & PCRE_NEWLINE_BITS)  switch (options & PCRE_NEWLINE_BITS)
6335    {    {
6336    case 0: newline = NEWLINE; break;   /* Build-time default */    case 0: newline = NEWLINE; break;   /* Build-time default */
6337    case PCRE_NEWLINE_CR: newline = '\r'; break;    case PCRE_NEWLINE_CR: newline = CHAR_CR; break;
6338    case PCRE_NEWLINE_LF: newline = '\n'; break;    case PCRE_NEWLINE_LF: newline = CHAR_NL; break;
6339    case PCRE_NEWLINE_CR+    case PCRE_NEWLINE_CR+
6340         PCRE_NEWLINE_LF: newline = ('\r' << 8) | '\n'; break;         PCRE_NEWLINE_LF: newline = (CHAR_CR << 8) | CHAR_NL; break;
6341    case PCRE_NEWLINE_ANY: newline = -1; break;    case PCRE_NEWLINE_ANY: newline = -1; break;
6342    case PCRE_NEWLINE_ANYCRLF: newline = -2; break;    case PCRE_NEWLINE_ANYCRLF: newline = -2; break;
6343    default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;    default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;
# Line 5906  to compile parts of the pattern into; th Line 6386  to compile parts of the pattern into; th
6386  no longer needed, so hopefully this workspace will never overflow, though there  no longer needed, so hopefully this workspace will never overflow, though there
6387  is a test for its doing so. */  is a test for its doing so. */
6388    
6389  cd->bracount = 0;  cd->bracount = cd->final_bracount = 0;
6390  cd->names_found = 0;  cd->names_found = 0;
6391  cd->name_entry_size = 0;  cd->name_entry_size = 0;
6392  cd->name_table = NULL;  cd->name_table = NULL;
# Line 5983  field. Reset the bracket count and the n Line 6463  field. Reset the bracket count and the n
6463  field; this time it's used for remembering forward references to subpatterns.  field; this time it's used for remembering forward references to subpatterns.
6464  */  */
6465    
6466    cd->final_bracount = cd->bracount;  /* Save for checking forward references */
6467  cd->bracount = 0;  cd->bracount = 0;
6468  cd->names_found = 0;  cd->names_found = 0;
6469  cd->name_table = (uschar *)re + re->name_table_offset;  cd->name_table = (uschar *)re + re->name_table_offset;

Legend:
Removed from v.240  
changed lines
  Added in v.438

  ViewVC Help
Powered by ViewVC 1.1.5