/[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 305 by ph10, Sun Jan 20 20:07:32 2008 UTC revision 500 by ph10, Sat Mar 6 19:00:29 2010 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2008 University of Cambridge             Copyright (c) 1997-2010 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 53  supporting internal functions that are n Line 53  supporting internal functions that are n
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56  /* When DEBUG is defined, we need the pcre_printint() function, which is also  /* When PCRE_DEBUG is defined, we need the pcre_printint() function, which is
57  used by pcretest. DEBUG is not defined when building a production library. */  also used by pcretest. PCRE_DEBUG is not defined when building a production
58    library. */
59    
60  #ifdef DEBUG  #ifdef PCRE_DEBUG
61  #include "pcre_printint.src"  #include "pcre_printint.src"
62  #endif  #endif
63    
# Line 97  are simple data values; negative values Line 98  are simple data values; negative values
98  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
99  is invalid. */  is invalid. */
100    
101  #ifndef EBCDIC  /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC
102    
103    /* This is the "normal" table for ASCII systems or for EBCDIC systems running
104    in UTF-8 mode. */
105    
106  static const short int escapes[] = {  static const short int escapes[] = {
107       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,                       0,
108       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,                       0,
109     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */       0,                       0,
110  -ESC_H,      0,      0, -ESC_K,      0,      0,      0,      0,   /* H - O */       0,                       0,
111  -ESC_P, -ESC_Q, -ESC_R, -ESC_S,      0,      0, -ESC_V, -ESC_W,   /* P - W */       0,                       0,
112  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */       CHAR_COLON,              CHAR_SEMICOLON,
113     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
114  -ESC_h,      0,      0, -ESC_k,      0,      0,  ESC_n,      0,   /* h - o */       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
115  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0, -ESC_v, -ESC_w,   /* p - w */       CHAR_COMMERCIAL_AT,      -ESC_A,
116       0,      0, -ESC_z                                            /* x - z */       -ESC_B,                  -ESC_C,
117         -ESC_D,                  -ESC_E,
118         0,                       -ESC_G,
119         -ESC_H,                  0,
120         0,                       -ESC_K,
121         0,                       0,
122         0,                       0,
123         -ESC_P,                  -ESC_Q,
124         -ESC_R,                  -ESC_S,
125         0,                       0,
126         -ESC_V,                  -ESC_W,
127         -ESC_X,                  0,
128         -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,
129         CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,
130         CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,
131         CHAR_GRAVE_ACCENT,       7,
132         -ESC_b,                  0,
133         -ESC_d,                  ESC_e,
134         ESC_f,                   0,
135         -ESC_h,                  0,
136         0,                       -ESC_k,
137         0,                       0,
138         ESC_n,                   0,
139         -ESC_p,                  0,
140         ESC_r,                   -ESC_s,
141         ESC_tee,                 0,
142         -ESC_v,                  -ESC_w,
143         0,                       0,
144         -ESC_z
145  };  };
146    
147  #else           /* This is the "abnormal" table for EBCDIC systems */  #else
148    
149    /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
150    
151  static const short int escapes[] = {  static const short int escapes[] = {
152  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
153  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 142  static const short int escapes[] = { Line 178  static const short int escapes[] = {
178    
179  /* 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
180  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
181  the number of relocations when a shared library is dynamically linked. */  the number of relocations when a shared library is dynamically linked. The
182    string is built from string macros so that it works in UTF-8 mode on EBCDIC
183    platforms. */
184    
185  typedef struct verbitem {  typedef struct verbitem {
186    int   len;    int   len;
# Line 150  typedef struct verbitem { Line 188  typedef struct verbitem {
188  } verbitem;  } verbitem;
189    
190  static const char verbnames[] =  static const char verbnames[] =
191    "ACCEPT\0"    STRING_ACCEPT0
192    "COMMIT\0"    STRING_COMMIT0
193    "F\0"    STRING_F0
194    "FAIL\0"    STRING_FAIL0
195    "PRUNE\0"    STRING_PRUNE0
196    "SKIP\0"    STRING_SKIP0
197    "THEN";    STRING_THEN;
198    
199  static verbitem verbs[] = {  static const verbitem verbs[] = {
200    { 6, OP_ACCEPT },    { 6, OP_ACCEPT },
201    { 6, OP_COMMIT },    { 6, OP_COMMIT },
202    { 1, OP_FAIL },    { 1, OP_FAIL },
# Line 168  static verbitem verbs[] = { Line 206  static verbitem verbs[] = {
206    { 4, OP_THEN  }    { 4, OP_THEN  }
207  };  };
208    
209  static int verbcount = sizeof(verbs)/sizeof(verbitem);  static const int verbcount = sizeof(verbs)/sizeof(verbitem);
210    
211    
212  /* Tables of names of POSIX character classes and their lengths. The names are  /* Tables of names of POSIX character classes and their lengths. The names are
# Line 178  length entry. The first three must be al Line 216  length entry. The first three must be al
216  for handling case independence. */  for handling case independence. */
217    
218  static const char posix_names[] =  static const char posix_names[] =
219    "alpha\0"  "lower\0"  "upper\0"  "alnum\0"  "ascii\0"  "blank\0"    STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
220    "cntrl\0"  "digit\0"  "graph\0"  "print\0"  "punct\0"  "space\0"    STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
221    "word\0"   "xdigit";    STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
222      STRING_word0  STRING_xdigit;
223    
224  static const uschar posix_name_lengths[] = {  static const uschar posix_name_lengths[] = {
225    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 224  the number of relocations needed when a Line 263  the number of relocations needed when a
263  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
264  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
265  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
266  because these strings are used only when there is a compilation error. */  because these strings are used only when there is a compilation error.
267    
268    Each substring ends with \0 to insert a null character. This includes the final
269    substring, so that the whole string ends with \0\0, which can be detected when
270    counting through. */
271    
272  static const char error_texts[] =  static const char error_texts[] =
273    "no error\0"    "no error\0"
# Line 295  static const char error_texts[] = Line 338  static const char error_texts[] =
338    /* 55 */    /* 55 */
339    "repeating a DEFINE group is not allowed\0"    "repeating a DEFINE group is not allowed\0"
340    "inconsistent NEWLINE options\0"    "inconsistent NEWLINE options\0"
341    "\\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"
342    "(?+ or (?- or (?(+ or (?(- must be followed by a non-zero number\0"    "a numbered reference must not be zero\0"
343    "(*VERB) with an argument is not supported\0"    "(*VERB) with an argument is not supported\0"
344    /* 60 */    /* 60 */
345    "(*VERB) not recognized\0"    "(*VERB) not recognized\0"
346    "number is too big\0"    "number is too big\0"
347    "subpattern name expected\0"    "subpattern name expected\0"
348    "digit expected after (?+";    "digit expected after (?+\0"
349      "] is an invalid data character in JavaScript compatibility mode\0"
350      /* 65 */
351      "different names for subpatterns of the same number are not allowed\0";
352    
353  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
354  patterns. Note that the tables in chartables are dependent on the locale, and  patterns. Note that the tables in chartables are dependent on the locale, and
# Line 321  For convenience, we use the same bit def Line 366  For convenience, we use the same bit def
366    
367  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
368    
369  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC
370    
371    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
372    UTF-8 mode. */
373    
374  static const unsigned char digitab[] =  static const unsigned char digitab[] =
375    {    {
376    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
# Line 357  static const unsigned char digitab[] = Line 406  static const unsigned char digitab[] =
406    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
407    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
408    
409  #else           /* This is the "abnormal" case, for EBCDIC systems */  #else
410    
411    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
412    
413  static const unsigned char digitab[] =  static const unsigned char digitab[] =
414    {    {
415    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
# Line 454  static const char * Line 506  static const char *
506  find_error_text(int n)  find_error_text(int n)
507  {  {
508  const char *s = error_texts;  const char *s = error_texts;
509  for (; n > 0; n--) while (*s++ != 0);  for (; n > 0; n--)
510      {
511      while (*s++ != 0) {};
512      if (*s == 0) return "Error text not found (please report)";
513      }
514  return s;  return s;
515  }  }
516    
# Line 502  if (c == 0) *errorcodeptr = ERR1; Line 558  if (c == 0) *errorcodeptr = ERR1;
558  in a table. A non-zero result is something that can be returned immediately.  in a table. A non-zero result is something that can be returned immediately.
559  Otherwise further processing may be required. */  Otherwise further processing may be required. */
560    
561  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
562  else if (c < '0' || c > 'z') {}                           /* Not alphanumeric */  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */
563  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0) c = i;
564    
565  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
566  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */
# Line 523  else Line 579  else
579      /* 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
580      error. */      error. */
581    
582      case 'l':      case CHAR_l:
583      case 'L':      case CHAR_L:
584      case 'N':      case CHAR_N:
585      case 'u':      case CHAR_u:
586      case 'U':      case CHAR_U:
587      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
588      break;      break;
589    
590      /* \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:
591      is an absolute backreference. If negative, it is a relative backreference.  
592      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
593      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
594      unified syntax for back references. As this is synonymous with \k{name}, we      5.10 feature.
595      fudge it up by pretending it really was \k. */  
596        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
597        is part of Perl's movement towards a unified syntax for back references. As
598        this is synonymous with \k{name}, we fudge it up by pretending it really
599        was \k.
600    
601        (3) For Oniguruma compatibility we also support \g followed by a name or a
602        number either in angle brackets or in single quotes. However, these are
603        (possibly recursive) subroutine calls, _not_ backreferences. Just return
604        the -ESC_g code (cf \k). */
605    
606        case CHAR_g:
607        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
608          {
609          c = -ESC_g;
610          break;
611          }
612    
613        /* Handle the Perl-compatible cases */
614    
615      case 'g':      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
     if (ptr[1] == '{')  
616        {        {
617        const uschar *p;        const uschar *p;
618        for (p = ptr+2; *p != 0 && *p != '}'; p++)        for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
619          if (*p != '-' && (digitab[*p] & ctype_digit) == 0) break;          if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
620        if (*p != 0 && *p != '}')        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
621          {          {
622          c = -ESC_k;          c = -ESC_k;
623          break;          break;
# Line 554  else Line 627  else
627        }        }
628      else braced = FALSE;      else braced = FALSE;
629    
630      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
631        {        {
632        negated = TRUE;        negated = TRUE;
633        ptr++;        ptr++;
# Line 563  else Line 636  else
636    
637      c = 0;      c = 0;
638      while ((digitab[ptr[1]] & ctype_digit) != 0)      while ((digitab[ptr[1]] & ctype_digit) != 0)
639        c = c * 10 + *(++ptr) - '0';        c = c * 10 + *(++ptr) - CHAR_0;
640    
641      if (c < 0)      if (c < 0)   /* Integer overflow */
642        {        {
643        *errorcodeptr = ERR61;        *errorcodeptr = ERR61;
644        break;        break;
645        }        }
646    
647      if (c == 0 || (braced && *(++ptr) != '}'))      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
648        {        {
649        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
650        break;        break;
651        }        }
652    
653        if (c == 0)
654          {
655          *errorcodeptr = ERR58;
656          break;
657          }
658    
659      if (negated)      if (negated)
660        {        {
661        if (c > bracount)        if (c > bracount)
# Line 602  else Line 681  else
681      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
682      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
683    
684      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:
685      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
686    
687      if (!isclass)      if (!isclass)
688        {        {
689        oldptr = ptr;        oldptr = ptr;
690        c -= '0';        c -= CHAR_0;
691        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
692          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - CHAR_0;
693        if (c < 0)        if (c < 0)    /* Integer overflow */
694          {          {
695          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
696          break;          break;
# Line 628  else Line 707  else
707      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.
708      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
709    
710      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
711        {        {
712        ptr--;        ptr--;
713        c = 0;        c = 0;
# Line 641  else Line 720  else
720      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
721      than 3 octal digits. */      than 3 octal digits. */
722    
723      case '0':      case CHAR_0:
724      c -= '0';      c -= CHAR_0;
725      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
726          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
727      if (!utf8 && c > 255) *errorcodeptr = ERR51;      if (!utf8 && c > 255) *errorcodeptr = ERR51;
728      break;      break;
729    
# Line 652  else Line 731  else
731      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
732      treated as a data character. */      treated as a data character. */
733    
734      case 'x':      case CHAR_x:
735      if (ptr[1] == '{')      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
736        {        {
737        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
738        int count = 0;        int count = 0;
# Line 662  else Line 741  else
741        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
742          {          {
743          register int cc = *pt++;          register int cc = *pt++;
744          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
745          count++;          count++;
746    
747  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
748          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
749          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
750  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
751          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
752          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
753  #endif  #endif
754          }          }
755    
756        if (*pt == '}')        if (*pt == CHAR_RIGHT_CURLY_BRACKET)
757          {          {
758          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
759          ptr = pt;          ptr = pt;
# Line 690  else Line 769  else
769      c = 0;      c = 0;
770      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
771        {        {
772        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
773        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
774  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
775        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
776        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
777  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
778        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
779        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
780  #endif  #endif
781        }        }
782      break;      break;
# Line 706  else Line 785  else
785      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
786      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
787    
788      case 'c':      case CHAR_c:
789      c = *(++ptr);      c = *(++ptr);
790      if (c == 0)      if (c == 0)
791        {        {
# Line 714  else Line 793  else
793        break;        break;
794        }        }
795    
796  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
797      if (c >= 'a' && c <= 'z') c -= 32;      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
798      c ^= 0x40;      c ^= 0x40;
799  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
800      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
801      c ^= 0xC0;      c ^= 0xC0;
802  #endif  #endif
803      break;      break;
# Line 780  if (c == 0) goto ERROR_RETURN; Line 859  if (c == 0) goto ERROR_RETURN;
859  /* \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
860  negation. */  negation. */
861    
862  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
863    {    {
864    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
865      {      {
866      *negptr = TRUE;      *negptr = TRUE;
867      ptr++;      ptr++;
# Line 791  if (c == '{') Line 870  if (c == '{')
870      {      {
871      c = *(++ptr);      c = *(++ptr);
872      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
873      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
874      name[i] = c;      name[i] = c;
875      }      }
876    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
877    name[i] = 0;    name[i] = 0;
878    }    }
879    
# Line 859  is_counted_repeat(const uschar *p) Line 938  is_counted_repeat(const uschar *p)
938  {  {
939  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
940  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
941  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
942    
943  if (*p++ != ',') return FALSE;  if (*p++ != CHAR_COMMA) return FALSE;
944  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
945    
946  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
947  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
948    
949  return (*p == '}');  return (*p == CHAR_RIGHT_CURLY_BRACKET);
950  }  }
951    
952    
# Line 900  int max = -1; Line 979  int max = -1;
979  /* 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
980  an integer overflow. */  an integer overflow. */
981    
982  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;
983  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
984    {    {
985    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 910  if (min < 0 || min > 65535) Line 989  if (min < 0 || min > 65535)
989  /* 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.
990  Also, max must not be less than min. */  Also, max must not be less than min. */
991    
992  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
993    {    {
994    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
995      {      {
996      max = 0;      max = 0;
997      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;
998      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
999        {        {
1000        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 940  return p; Line 1019  return p;
1019    
1020    
1021  /*************************************************  /*************************************************
1022  *       Find forward referenced subpattern       *  *  Subroutine for finding forward reference      *
1023  *************************************************/  *************************************************/
1024    
1025  /* This function scans along a pattern's text looking for capturing  /* This recursive function is called only from find_parens() below. The
1026    top-level call starts at the beginning of the pattern. All other calls must
1027    start at a parenthesis. It scans along a pattern's text looking for capturing
1028  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
1029  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
1030  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
1031  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
1032  be terminated by '>' because that is checked in the first pass.  first pass. Recursion is used to keep track of subpatterns that reset the
1033    capturing group numbers - the (?| feature.
1034    
1035  Arguments:  Arguments:
1036    ptr          current position in the pattern    ptrptr       address of the current character pointer (updated)
1037    count        current count of capturing parens so far encountered    cd           compile background data
1038    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1039    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1040    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1041      count        pointer to the current capturing subpattern number (updated)
1042    
1043  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1044  */  */
1045    
1046  static int  static int
1047  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,
1048    BOOL xmode)    BOOL xmode, int *count)
1049  {  {
1050  const uschar *thisname;  uschar *ptr = *ptrptr;
1051    int start_count = *count;
1052    int hwm_count = start_count;
1053    BOOL dup_parens = FALSE;
1054    
1055  for (; *ptr != 0; ptr++)  /* If the first character is a parenthesis, check on the type of group we are
1056    dealing with. The very first call may not start with a parenthesis. */
1057    
1058    if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1059    {    {
1060    int term;    if (ptr[1] == CHAR_QUESTION_MARK &&
1061          ptr[2] == CHAR_VERTICAL_LINE)
1062        {
1063        ptr += 3;
1064        dup_parens = TRUE;
1065        }
1066    
1067      /* Handle a normal, unnamed capturing parenthesis */
1068    
1069      else if (ptr[1] != CHAR_QUESTION_MARK && ptr[1] != CHAR_ASTERISK)
1070        {
1071        *count += 1;
1072        if (name == NULL && *count == lorn) return *count;
1073        ptr++;
1074        }
1075    
1076      /* Handle a condition. If it is an assertion, just carry on so that it
1077      is processed as normal. If not, skip to the closing parenthesis of the
1078      condition (there can't be any nested parens. */
1079    
1080      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1081        {
1082        ptr += 2;
1083        if (ptr[1] != CHAR_QUESTION_MARK)
1084          {
1085          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1086          if (*ptr != 0) ptr++;
1087          }
1088        }
1089    
1090      /* We have either (? or (* and not a condition */
1091    
1092      else
1093        {
1094        ptr += 2;
1095        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1096    
1097        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1098    
1099        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1100            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1101          {
1102          int term;
1103          const uschar *thisname;
1104          *count += 1;
1105          if (name == NULL && *count == lorn) return *count;
1106          term = *ptr++;
1107          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1108          thisname = ptr;
1109          while (*ptr != term) ptr++;
1110          if (name != NULL && lorn == ptr - thisname &&
1111              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1112            return *count;
1113          term++;
1114          }
1115        }
1116      }
1117    
1118    /* Past any initial parenthesis handling, scan for parentheses or vertical
1119    bars. */
1120    
1121    for (; *ptr != 0; ptr++)
1122      {
1123    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1124    
1125    if (*ptr == '\\')    if (*ptr == CHAR_BACKSLASH)
1126      {      {
1127      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1128      if (*ptr == 'Q') for (;;)      if (*ptr == CHAR_Q) for (;;)
1129        {        {
1130        while (*(++ptr) != 0 && *ptr != '\\');        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1131        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1132        if (*(++ptr) == 'E') break;        if (*(++ptr) == CHAR_E) break;
1133        }        }
1134      continue;      continue;
1135      }      }
1136    
1137    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1138      are handled for real. If the first character is '^', skip it. Also, if the
1139      first few characters (either before or after ^) are \Q\E or \E we skip them
1140      too. This makes for compatibility with Perl. Note the use of STR macros to
1141      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1142    
1143    if (*ptr == '[')    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1144      {      {
1145      while (*(++ptr) != ']')      BOOL negate_class = FALSE;
1146        for (;;)
1147          {
1148          if (ptr[1] == CHAR_BACKSLASH)
1149            {
1150            if (ptr[2] == CHAR_E)
1151              ptr+= 2;
1152            else if (strncmp((const char *)ptr+2,
1153                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1154              ptr += 4;
1155            else
1156              break;
1157            }
1158          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1159            {
1160            negate_class = TRUE;
1161            ptr++;
1162            }
1163          else break;
1164          }
1165    
1166        /* If the next character is ']', it is a data character that must be
1167        skipped, except in JavaScript compatibility mode. */
1168    
1169        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1170            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1171          ptr++;
1172    
1173        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1174        {        {
1175        if (*ptr == 0) return -1;        if (*ptr == 0) return -1;
1176        if (*ptr == '\\')        if (*ptr == CHAR_BACKSLASH)
1177          {          {
1178          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) goto FAIL_EXIT;
1179          if (*ptr == 'Q') for (;;)          if (*ptr == CHAR_Q) for (;;)
1180            {            {
1181            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1182            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1183            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1184            }            }
1185          continue;          continue;
1186          }          }
# Line 1008  for (; *ptr != 0; ptr++) Line 1190  for (; *ptr != 0; ptr++)
1190    
1191    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1192    
1193    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1194      {      {
1195      while (*(++ptr) != 0 && *ptr != '\n');      while (*(++ptr) != 0 && *ptr != CHAR_NL) {};
1196      if (*ptr == 0) return -1;      if (*ptr == 0) goto FAIL_EXIT;
1197      continue;      continue;
1198      }      }
1199    
1200    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1201    
1202    if (*ptr != '(') continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != '?' && ptr[1] != '*')  
1203      {      {
1204      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, count);
1205      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1206      continue;      if (*ptr == 0) goto FAIL_EXIT;
1207      }      }
1208    
1209    ptr += 2;    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1210    if (*ptr == 'P') ptr++;                      /* Allow optional P */      {
1211        if (dup_parens && *count < hwm_count) *count = hwm_count;
1212        *ptrptr = ptr;
1213        return -1;
1214        }
1215    
1216      else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1217        {
1218        if (*count > hwm_count) hwm_count = *count;
1219        *count = start_count;
1220        }
1221      }
1222    
1223    FAIL_EXIT:
1224    *ptrptr = ptr;
1225    return -1;
1226    }
1227    
1228    
   /* We have to disambiguate (?<! and (?<= from (?<name> */  
1229    
   if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  
        *ptr != '\'')  
     continue;  
1230    
1231    count++;  /*************************************************
1232    *       Find forward referenced subpattern       *
1233    *************************************************/
1234    
1235    /* This function scans along a pattern's text looking for capturing
1236    subpatterns, and counting them. If it finds a named pattern that matches the
1237    name it is given, it returns its number. Alternatively, if the name is NULL, it
1238    returns when it reaches a given numbered subpattern. This is used for forward
1239    references to subpatterns. We used to be able to start this scan from the
1240    current compiling point, using the current count value from cd->bracount, and
1241    do it all in a single loop, but the addition of the possibility of duplicate
1242    subpattern numbers means that we have to scan from the very start, in order to
1243    take account of such duplicates, and to use a recursive function to keep track
1244    of the different types of group.
1245    
1246    Arguments:
1247      cd           compile background data
1248      name         name to seek, or NULL if seeking a numbered subpattern
1249      lorn         name length, or subpattern number if name is NULL
1250      xmode        TRUE if we are in /x mode
1251    
1252    Returns:       the number of the found subpattern, or -1 if not found
1253    */
1254    
1255    if (name == NULL && count == lorn) return count;  static int
1256    term = *ptr++;  find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode)
1257    if (term == '<') term = '>';  {
1258    thisname = ptr;  uschar *ptr = (uschar *)cd->start_pattern;
1259    while (*ptr != term) ptr++;  int count = 0;
1260    if (name != NULL && lorn == ptr - thisname &&  int rc;
1261        strncmp((const char *)name, (const char *)thisname, lorn) == 0)  
1262      return count;  /* If the pattern does not start with an opening parenthesis, the first call
1263    to find_parens_sub() will scan right to the end (if necessary). However, if it
1264    does start with a parenthesis, find_parens_sub() will return when it hits the
1265    matching closing parens. That is why we have to have a loop. */
1266    
1267    for (;;)
1268      {
1269      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, &count);
1270      if (rc > 0 || *ptr++ == 0) break;
1271    }    }
1272    
1273  return -1;  return rc;
1274  }  }
1275    
1276    
1277    
1278    
1279  /*************************************************  /*************************************************
1280  *      Find first significant op code            *  *      Find first significant op code            *
1281  *************************************************/  *************************************************/
# Line 1100  for (;;) Line 1325  for (;;)
1325    
1326      case OP_CALLOUT:      case OP_CALLOUT:
1327      case OP_CREF:      case OP_CREF:
1328        case OP_NCREF:
1329      case OP_RREF:      case OP_RREF:
1330        case OP_NRREF:
1331      case OP_DEF:      case OP_DEF:
1332      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1333      break;      break;
# Line 1116  for (;;) Line 1343  for (;;)
1343    
1344    
1345  /*************************************************  /*************************************************
1346  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1347  *************************************************/  *************************************************/
1348    
1349  /* Scan a pattern and compute the fixed length of subject that will match it,  /* Scan a branch and compute the fixed length of subject that will match it,
1350  if the length is fixed. This is needed for dealing with backward assertions.  if the length is fixed. This is needed for dealing with backward assertions.
1351  In UTF8 mode, the result is in characters rather than bytes.  In UTF8 mode, the result is in characters rather than bytes. The branch is
1352    temporarily terminated with OP_END when this function is called.
1353    
1354    This function is called when a backward assertion is encountered, so that if it
1355    fails, the error message can point to the correct place in the pattern.
1356    However, we cannot do this when the assertion contains subroutine calls,
1357    because they can be forward references. We solve this by remembering this case
1358    and doing the check at the end; a flag specifies which mode we are running in.
1359    
1360  Arguments:  Arguments:
1361    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1362    options  the compiling options    options  the compiling options
1363      atend    TRUE if called when the pattern is complete
1364      cd       the "compile data" structure
1365    
1366  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1367                 or -1 if there is no fixed length,
1368               or -2 if \C was encountered               or -2 if \C was encountered
1369                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1370  */  */
1371    
1372  static int  static int
1373  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)
1374  {  {
1375  int length = -1;  int length = -1;
1376    
# Line 1145  branch, check the length against that of Line 1383  branch, check the length against that of
1383  for (;;)  for (;;)
1384    {    {
1385    int d;    int d;
1386      uschar *ce, *cs;
1387    register int op = *cc;    register int op = *cc;
1388    switch (op)    switch (op)
1389      {      {
# Line 1152  for (;;) Line 1391  for (;;)
1391      case OP_BRA:      case OP_BRA:
1392      case OP_ONCE:      case OP_ONCE:
1393      case OP_COND:      case OP_COND:
1394      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);
1395      if (d < 0) return d;      if (d < 0) return d;
1396      branchlength += d;      branchlength += d;
1397      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1175  for (;;) Line 1414  for (;;)
1414      branchlength = 0;      branchlength = 0;
1415      break;      break;
1416    
1417        /* A true recursion implies not fixed length, but a subroutine call may
1418        be OK. If the subroutine is a forward reference, we can't deal with
1419        it until the end of the pattern, so return -3. */
1420    
1421        case OP_RECURSE:
1422        if (!atend) return -3;
1423        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1424        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1425        if (cc > cs && cc < ce) return -1;                /* Recursion */
1426        d = find_fixedlength(cs + 2, options, atend, cd);
1427        if (d < 0) return d;
1428        branchlength += d;
1429        cc += 1 + LINK_SIZE;
1430        break;
1431    
1432      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1433    
1434      case OP_ASSERT:      case OP_ASSERT:
# Line 1188  for (;;) Line 1442  for (;;)
1442    
1443      case OP_REVERSE:      case OP_REVERSE:
1444      case OP_CREF:      case OP_CREF:
1445        case OP_NCREF:
1446      case OP_RREF:      case OP_RREF:
1447        case OP_NRREF:
1448      case OP_DEF:      case OP_DEF:
1449      case OP_OPT:      case OP_OPT:
1450      case OP_CALLOUT:      case OP_CALLOUT:
1451      case OP_SOD:      case OP_SOD:
1452      case OP_SOM:      case OP_SOM:
1453        case OP_SET_SOM:
1454      case OP_EOD:      case OP_EOD:
1455      case OP_EODN:      case OP_EODN:
1456      case OP_CIRC:      case OP_CIRC:
# Line 1211  for (;;) Line 1468  for (;;)
1468      branchlength++;      branchlength++;
1469      cc += 2;      cc += 2;
1470  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1471      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1472        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1473  #endif  #endif
1474      break;      break;
1475    
# Line 1225  for (;;) Line 1480  for (;;)
1480      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1481      cc += 4;      cc += 4;
1482  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1483      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1484        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1485  #endif  #endif
1486      break;      break;
1487    
# Line 1252  for (;;) Line 1505  for (;;)
1505      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1506      case OP_WORDCHAR:      case OP_WORDCHAR:
1507      case OP_ANY:      case OP_ANY:
1508        case OP_ALLANY:
1509      branchlength++;      branchlength++;
1510      cc++;      cc++;
1511      break;      break;
# Line 1306  for (;;) Line 1560  for (;;)
1560    
1561    
1562  /*************************************************  /*************************************************
1563  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1564  *************************************************/  *************************************************/
1565    
1566  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1567  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1568    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1569    so that it can be called from pcre_study() when finding the minimum matching
1570    length.
1571    
1572  Arguments:  Arguments:
1573    code        points to start of expression    code        points to start of expression
1574    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1575    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1576    
1577  Returns:      pointer to the opcode for the bracket, or NULL if not found  Returns:      pointer to the opcode for the bracket, or NULL if not found
1578  */  */
1579    
1580  static const uschar *  const uschar *
1581  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1582  {  {
1583  for (;;)  for (;;)
1584    {    {
# Line 1334  for (;;) Line 1591  for (;;)
1591    
1592    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1593    
1594      /* Handle recursion */
1595    
1596      else if (c == OP_REVERSE)
1597        {
1598        if (number < 0) return (uschar *)code;
1599        code += _pcre_OP_lengths[c];
1600        }
1601    
1602    /* Handle capturing bracket */    /* Handle capturing bracket */
1603    
1604    else if (c == OP_CBRA)    else if (c == OP_CBRA)
# Line 1400  for (;;) Line 1665  for (;;)
1665        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1666        break;        break;
1667        }        }
1668    #else
1669        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1670  #endif  #endif
1671      }      }
1672    }    }
# Line 1493  for (;;) Line 1760  for (;;)
1760        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1761        break;        break;
1762        }        }
1763    #else
1764        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1765  #endif  #endif
1766      }      }
1767    }    }
# Line 1544  for (code = first_significant_code(code Line 1813  for (code = first_significant_code(code
1813    
1814    /* Groups with zero repeats can of course be empty; skip them. */    /* Groups with zero repeats can of course be empty; skip them. */
1815    
1816    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)
1817      {      {
1818      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1819      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
# Line 1559  for (code = first_significant_code(code Line 1828  for (code = first_significant_code(code
1828      BOOL empty_branch;      BOOL empty_branch;
1829      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
1830    
1831      /* Scan a closed bracket */      /* If a conditional group has only one branch, there is a second, implied,
1832        empty branch, so just skip over the conditional, because it could be empty.
1833        Otherwise, scan the individual branches of the group. */
1834    
1835      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;  
1836        code += GET(code, 1);        code += GET(code, 1);
1837        else
1838          {
1839          empty_branch = FALSE;
1840          do
1841            {
1842            if (!empty_branch && could_be_empty_branch(code, endcode, utf8))
1843              empty_branch = TRUE;
1844            code += GET(code, 1);
1845            }
1846          while (*code == OP_ALT);
1847          if (!empty_branch) return FALSE;   /* All branches are non-empty */
1848        }        }
1849      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
1850      c = *code;      c = *code;
1851      continue;      continue;
1852      }      }
# Line 1630  for (code = first_significant_code(code Line 1907  for (code = first_significant_code(code
1907      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1908      case OP_WORDCHAR:      case OP_WORDCHAR:
1909      case OP_ANY:      case OP_ANY:
1910        case OP_ALLANY:
1911      case OP_ANYBYTE:      case OP_ANYBYTE:
1912      case OP_CHAR:      case OP_CHAR:
1913      case OP_CHARNC:      case OP_CHARNC:
# Line 1686  for (code = first_significant_code(code Line 1964  for (code = first_significant_code(code
1964      case OP_QUERY:      case OP_QUERY:
1965      case OP_MINQUERY:      case OP_MINQUERY:
1966      case OP_POSQUERY:      case OP_POSQUERY:
1967        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
1968        break;
1969    
1970      case OP_UPTO:      case OP_UPTO:
1971      case OP_MINUPTO:      case OP_MINUPTO:
1972      case OP_POSUPTO:      case OP_POSUPTO:
1973      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
1974      break;      break;
1975  #endif  #endif
1976      }      }
# Line 1722  static BOOL Line 2003  static BOOL
2003  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2004    BOOL utf8)    BOOL utf8)
2005  {  {
2006  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2007    {    {
2008    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8))
2009        return FALSE;
2010    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2011    }    }
2012  return TRUE;  return TRUE;
# Line 1770  int terminator;          /* Don't combin Line 2052  int terminator;          /* Don't combin
2052  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2053  for (++ptr; *ptr != 0; ptr++)  for (++ptr; *ptr != 0; ptr++)
2054    {    {
2055    if (*ptr == '\\' && ptr[1] == ']') ptr++; else    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else
2056      {      {
2057      if (*ptr == ']') return FALSE;      if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2058      if (*ptr == terminator && ptr[1] == ']')      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2059        {        {
2060        *endptr = ptr;        *endptr = ptr;
2061        return TRUE;        return TRUE;
# Line 1824  return -1; Line 2106  return -1;
2106  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2107  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2108  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
2109  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
2110  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
2111  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
2112  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
2113  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2114    OP_END.
2115    
2116  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2117  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 1963  get_othercase_range(unsigned int *cptr, Line 2246  get_othercase_range(unsigned int *cptr,
2246  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2247    
2248  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2249    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2250    
2251  if (c > d) return FALSE;  if (c > d) return FALSE;
2252    
# Line 1972  next = othercase + 1; Line 2255  next = othercase + 1;
2255    
2256  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2257    {    {
2258    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2259    next++;    next++;
2260    }    }
2261    
# Line 2018  if ((options & PCRE_EXTENDED) != 0) Line 2301  if ((options & PCRE_EXTENDED) != 0)
2301    for (;;)    for (;;)
2302      {      {
2303      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2304      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2305        {        {
2306        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2307          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
# Line 2030  if ((options & PCRE_EXTENDED) != 0) Line 2313  if ((options & PCRE_EXTENDED) != 0)
2313  /* 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
2314  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2315    
2316  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2317    {    {
2318    int temperrorcode = 0;    int temperrorcode = 0;
2319    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
# Line 2055  if ((options & PCRE_EXTENDED) != 0) Line 2338  if ((options & PCRE_EXTENDED) != 0)
2338    for (;;)    for (;;)
2339      {      {
2340      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2341      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2342        {        {
2343        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2344          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
# Line 2066  if ((options & PCRE_EXTENDED) != 0) Line 2349  if ((options & PCRE_EXTENDED) != 0)
2349    
2350  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
2351    
2352  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2353    return FALSE;    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2354        return FALSE;
2355    
2356  /* 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
2357  positive single character match, "item" either contains the character or, if  positive single character match, "item" either contains the character or, if
# Line 2082  if (next >= 0) switch(op_code) Line 2366  if (next >= 0) switch(op_code)
2366    case OP_CHAR:    case OP_CHAR:
2367  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2368    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2369    #else
2370      (void)(utf8_char);  /* Keep compiler happy by referencing function argument */
2371  #endif  #endif
2372    return item != next;    return item != next;
2373    
# Line 2100  if (next >= 0) switch(op_code) Line 2386  if (next >= 0) switch(op_code)
2386      unsigned int othercase;      unsigned int othercase;
2387      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2388  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2389      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
2390  #else  #else
2391      othercase = NOTACHAR;      othercase = NOTACHAR;
2392  #endif  #endif
# Line 2113  if (next >= 0) switch(op_code) Line 2399  if (next >= 0) switch(op_code)
2399    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT, "item" must be a single-byte character. */
2400    
2401    case OP_NOT:    case OP_NOT:
   if (next < 0) return FALSE;  /* Not a character */  
2402    if (item == next) return TRUE;    if (item == next) return TRUE;
2403    if ((options & PCRE_CASELESS) == 0) return FALSE;    if ((options & PCRE_CASELESS) == 0) return FALSE;
2404  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 2122  if (next >= 0) switch(op_code) Line 2407  if (next >= 0) switch(op_code)
2407      unsigned int othercase;      unsigned int othercase;
2408      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2409  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2410      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
2411  #else  #else
2412      othercase = NOTACHAR;      othercase = NOTACHAR;
2413  #endif  #endif
# Line 2383  BOOL utf8 = FALSE; Line 2668  BOOL utf8 = FALSE;
2668  uschar *utf8_char = NULL;  uschar *utf8_char = NULL;
2669  #endif  #endif
2670    
2671  #ifdef DEBUG  #ifdef PCRE_DEBUG
2672  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2673  #endif  #endif
2674    
# Line 2442  for (;; ptr++) Line 2727  for (;; ptr++)
2727    
2728    if (lengthptr != NULL)    if (lengthptr != NULL)
2729      {      {
2730  #ifdef DEBUG  #ifdef PCRE_DEBUG
2731      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
2732  #endif  #endif
2733      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */
# Line 2504  for (;; ptr++) Line 2789  for (;; ptr++)
2789    
2790    if (inescq && c != 0)    if (inescq && c != 0)
2791      {      {
2792      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
2793        {        {
2794        inescq = FALSE;        inescq = FALSE;
2795        ptr++;        ptr++;
# Line 2530  for (;; ptr++) Line 2815  for (;; ptr++)
2815    /* 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
2816    a quantifier. */    a quantifier. */
2817    
2818    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
2819      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
2820        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
2821    
2822    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
2823         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2546  for (;; ptr++) Line 2832  for (;; ptr++)
2832    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
2833      {      {
2834      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
2835      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
2836        {        {
2837        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2838          {          {
# Line 2571  for (;; ptr++) Line 2857  for (;; ptr++)
2857      {      {
2858      /* ===================================================================*/      /* ===================================================================*/
2859      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
2860      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
2861      case ')':      case CHAR_RIGHT_PARENTHESIS:
2862      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
2863      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
2864      *codeptr = code;      *codeptr = code;
# Line 2594  for (;; ptr++) Line 2880  for (;; ptr++)
2880      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
2881      the setting of any following char as a first character. */      the setting of any following char as a first character. */
2882    
2883      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
2884      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
2885        {        {
2886        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
# Line 2603  for (;; ptr++) Line 2889  for (;; ptr++)
2889      *code++ = OP_CIRC;      *code++ = OP_CIRC;
2890      break;      break;
2891    
2892      case '$':      case CHAR_DOLLAR_SIGN:
2893      previous = NULL;      previous = NULL;
2894      *code++ = OP_DOLL;      *code++ = OP_DOLL;
2895      break;      break;
# Line 2611  for (;; ptr++) Line 2897  for (;; ptr++)
2897      /* 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
2898      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
2899    
2900      case '.':      case CHAR_DOT:
2901      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
2902      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
2903      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
2904      previous = code;      previous = code;
2905      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
2906      break;      break;
2907    
2908    
# Line 2631  for (;; ptr++) Line 2917  for (;; ptr++)
2917      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,
2918      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
2919      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.
     */  
2920    
2921      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
2922        default (Perl) mode, it is treated as a data character. */
2923    
2924        case CHAR_RIGHT_SQUARE_BRACKET:
2925        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2926          {
2927          *errorcodeptr = ERR64;
2928          goto FAILED;
2929          }
2930        goto NORMAL_CHAR;
2931    
2932        case CHAR_LEFT_SQUARE_BRACKET:
2933      previous = code;      previous = code;
2934    
2935      /* 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
2936      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. */
2937    
2938      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2939             ptr[1] == CHAR_EQUALS_SIGN) &&
2940          check_posix_syntax(ptr, &tempptr))          check_posix_syntax(ptr, &tempptr))
2941        {        {
2942        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
2943        goto FAILED;        goto FAILED;
2944        }        }
2945    
# Line 2654  for (;; ptr++) Line 2951  for (;; ptr++)
2951      for (;;)      for (;;)
2952        {        {
2953        c = *(++ptr);        c = *(++ptr);
2954        if (c == '\\')        if (c == CHAR_BACKSLASH)
2955          {          {
2956          if (ptr[1] == 'E') ptr++;          if (ptr[1] == CHAR_E)
2957            else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;            ptr++;
2958              else break;          else if (strncmp((const char *)ptr+1,
2959                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
2960              ptr += 3;
2961            else
2962              break;
2963          }          }
2964        else if (!negate_class && c == '^')        else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
2965          negate_class = TRUE;          negate_class = TRUE;
2966        else break;        else break;
2967        }        }
2968    
2969        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
2970        an initial ']' is taken as a data character -- the code below handles
2971        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
2972        [^] must match any character, so generate OP_ALLANY. */
2973    
2974        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
2975            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2976          {
2977          *code++ = negate_class? OP_ALLANY : OP_FAIL;
2978          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
2979          zerofirstbyte = firstbyte;
2980          break;
2981          }
2982    
2983      /* If a class contains a negative special such as \S, we need to flip the      /* If a class contains a negative special such as \S, we need to flip the
2984      negation flag at the end, so that support for characters > 255 works      negation flag at the end, so that support for characters > 255 works
2985      correctly (they are all included in the class). */      correctly (they are all included in the class). */
# Line 2688  for (;; ptr++) Line 3003  for (;; ptr++)
3003  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3004      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
3005      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
3006      class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */      class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
3007  #endif  #endif
3008    
3009      /* 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 2704  for (;; ptr++) Line 3019  for (;; ptr++)
3019          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3020          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3021          }          }
3022    
3023        /* In the pre-compile phase, accumulate the length of any UTF-8 extra        /* In the pre-compile phase, accumulate the length of any UTF-8 extra
3024        data and reset the pointer. This is so that very large classes that        data and reset the pointer. This is so that very large classes that
3025        contain a zillion UTF-8 characters no longer overwrite the work space        contain a zillion UTF-8 characters no longer overwrite the work space
3026        (which is on the stack). */        (which is on the stack). */
3027    
3028        if (lengthptr != NULL)        if (lengthptr != NULL)
3029          {          {
3030          *lengthptr += class_utf8data - class_utf8data_base;          *lengthptr += class_utf8data - class_utf8data_base;
3031          class_utf8data = class_utf8data_base;          class_utf8data = class_utf8data_base;
3032          }          }
3033    
3034  #endif  #endif
3035    
3036        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
3037    
3038        if (inescq)        if (inescq)
3039          {          {
3040          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3041            {            {
3042            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
3043            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2737  for (;; ptr++) Line 3052  for (;; ptr++)
3052        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3053        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3054    
3055        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3056            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3057            check_posix_syntax(ptr, &tempptr))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3058          {          {
3059          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3060          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3061          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3062          uschar pbits[32];          uschar pbits[32];
3063    
3064          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3065            {            {
3066            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3067            goto FAILED;            goto FAILED;
3068            }            }
3069    
3070          ptr += 2;          ptr += 2;
3071          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3072            {            {
3073            local_negate = TRUE;            local_negate = TRUE;
3074            should_flip_negation = TRUE;  /* Note negative special */            should_flip_negation = TRUE;  /* Note negative special */
# Line 2826  for (;; ptr++) Line 3141  for (;; ptr++)
3141        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
3142        character in them, so set class_charcount bigger than one. */        character in them, so set class_charcount bigger than one. */
3143    
3144        if (c == '\\')        if (c == CHAR_BACKSLASH)
3145          {          {
3146          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3147          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3148    
3149          if (-c == ESC_b) c = '\b';       /* \b is backspace in a class */          if (-c == ESC_b) c = CHAR_BS;       /* \b is backspace in a class */
3150          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 */
3151          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 */
3152          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3153            {            {
3154            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3155              {              {
3156              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3157              }              }
# Line 3062  for (;; ptr++) Line 3377  for (;; ptr++)
3377        entirely. The code for handling \Q and \E is messy. */        entirely. The code for handling \Q and \E is messy. */
3378    
3379        CHECK_RANGE:        CHECK_RANGE:
3380        while (ptr[1] == '\\' && ptr[2] == 'E')        while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3381          {          {
3382          inescq = FALSE;          inescq = FALSE;
3383          ptr += 2;          ptr += 2;
# Line 3072  for (;; ptr++) Line 3387  for (;; ptr++)
3387    
3388        /* Remember \r or \n */        /* Remember \r or \n */
3389    
3390        if (c == '\r' || c == '\n') cd->external_flags |= PCRE_HASCRORLF;        if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3391    
3392        /* Check for range */        /* Check for range */
3393    
3394        if (!inescq && ptr[1] == '-')        if (!inescq && ptr[1] == CHAR_MINUS)
3395          {          {
3396          int d;          int d;
3397          ptr += 2;          ptr += 2;
3398          while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3399    
3400          /* 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
3401          mode. */          mode. */
3402    
3403          while (*ptr == '\\' && ptr[1] == 'Q')          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3404            {            {
3405            ptr += 2;            ptr += 2;
3406            if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }            if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3407                { ptr += 2; continue; }
3408            inescq = TRUE;            inescq = TRUE;
3409            break;            break;
3410            }            }
3411    
3412          if (*ptr == 0 || (!inescq && *ptr == ']'))          if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3413            {            {
3414            ptr = oldptr;            ptr = oldptr;
3415            goto LONE_SINGLE_CHARACTER;            goto LONE_SINGLE_CHARACTER;
# Line 3112  for (;; ptr++) Line 3428  for (;; ptr++)
3428          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
3429          in such circumstances. */          in such circumstances. */
3430    
3431          if (!inescq && d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3432            {            {
3433            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3434            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
# Line 3122  for (;; ptr++) Line 3438  for (;; ptr++)
3438    
3439            if (d < 0)            if (d < 0)
3440              {              {
3441              if (d == -ESC_b) d = '\b';              if (d == -ESC_b) d = CHAR_BS;
3442              else if (d == -ESC_X) d = 'X';              else if (d == -ESC_X) d = CHAR_X;
3443              else if (d == -ESC_R) d = 'R'; else              else if (d == -ESC_R) d = CHAR_R; else
3444                {                {
3445                ptr = oldptr;                ptr = oldptr;
3446                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 3145  for (;; ptr++) Line 3461  for (;; ptr++)
3461    
3462          /* Remember \r or \n */          /* Remember \r or \n */
3463    
3464          if (d == '\r' || d == '\n') cd->external_flags |= PCRE_HASCRORLF;          if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3465    
3466          /* 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
3467          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
# Line 3265  for (;; ptr++) Line 3581  for (;; ptr++)
3581          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3582            {            {
3583            unsigned int othercase;            unsigned int othercase;
3584            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)            if ((othercase = UCD_OTHERCASE(c)) != c)
3585              {              {
3586              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3587              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 3292  for (;; ptr++) Line 3608  for (;; ptr++)
3608    
3609      /* Loop until ']' reached. This "while" is the end of the "do" above. */      /* Loop until ']' reached. This "while" is the end of the "do" above. */
3610    
3611      while ((c = *(++ptr)) != 0 && (c != ']' || inescq));      while ((c = *(++ptr)) != 0 && (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
3612    
3613      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)                          /* Missing terminating ']' */
3614        {        {
# Line 3437  we set the flag only if there is a liter Line 3753  we set the flag only if there is a liter
3753      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
3754      has been tested above. */      has been tested above. */
3755    
3756      case '{':      case CHAR_LEFT_CURLY_BRACKET:
3757      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
3758      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
3759      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
3760      goto REPEAT;      goto REPEAT;
3761    
3762      case '*':      case CHAR_ASTERISK:
3763      repeat_min = 0;      repeat_min = 0;
3764      repeat_max = -1;      repeat_max = -1;
3765      goto REPEAT;      goto REPEAT;
3766    
3767      case '+':      case CHAR_PLUS:
3768      repeat_min = 1;      repeat_min = 1;
3769      repeat_max = -1;      repeat_max = -1;
3770      goto REPEAT;      goto REPEAT;
3771    
3772      case '?':      case CHAR_QUESTION_MARK:
3773      repeat_min = 0;      repeat_min = 0;
3774      repeat_max = 1;      repeat_max = 1;
3775    
# Line 3488  we set the flag only if there is a liter Line 3804  we set the flag only if there is a liter
3804      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
3805      repeat type to the non-default. */      repeat type to the non-default. */
3806    
3807      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
3808        {        {
3809        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
3810        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
3811        ptr++;        ptr++;
3812        }        }
3813      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
3814        {        {
3815        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
3816        ptr++;        ptr++;
# Line 3609  we set the flag only if there is a liter Line 3925  we set the flag only if there is a liter
3925    
3926        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
3927    
3928          /*--------------------------------------------------------------------*/
3929          /* This code is obsolete from release 8.00; the restriction was finally
3930          removed: */
3931    
3932        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
3933        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
3934    
3935        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
3936          /*--------------------------------------------------------------------*/
3937    
3938        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
3939    
# Line 3759  we set the flag only if there is a liter Line 4080  we set the flag only if there is a liter
4080          goto END_REPEAT;          goto END_REPEAT;
4081          }          }
4082    
4083          /*--------------------------------------------------------------------*/
4084          /* This code is obsolete from release 8.00; the restriction was finally
4085          removed: */
4086    
4087        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4088        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4089    
4090        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4091          /*--------------------------------------------------------------------*/
4092    
4093        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4094          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 3820  we set the flag only if there is a liter Line 4146  we set the flag only if there is a liter
4146    
4147        if (repeat_min == 0)        if (repeat_min == 0)
4148          {          {
4149          /* 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
4150          altogether. */          output altogether, like this:
4151    
4152          if (repeat_max == 0)          ** if (repeat_max == 0)
4153            {          **   {
4154            code = previous;          **   code = previous;
4155            goto END_REPEAT;          **   goto END_REPEAT;
4156            }          **   }
4157    
4158          /* If the maximum is 1 or unlimited, we just have to stick in the          However, that fails when a group is referenced as a subroutine from
4159          BRAZERO and do no more at this point. However, we do need to adjust          elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it
4160          any OP_RECURSE calls inside the group that refer to the group itself or          so that it is skipped on execution. As we don't have a list of which
4161          any internal or forward referenced group, because the offset is from          groups are referenced, we cannot do this selectively.
4162          the start of the whole regex. Temporarily terminate the pattern while  
4163          doing this. */          If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4164            and do no more at this point. However, we do need to adjust any
4165            OP_RECURSE calls inside the group that refer to the group itself or any
4166            internal or forward referenced group, because the offset is from the
4167            start of the whole regex. Temporarily terminate the pattern while doing
4168            this. */
4169    
4170          if (repeat_max <= 1)          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
4171            {            {
4172            *code = OP_END;            *code = OP_END;
4173            adjust_recurse(previous, 1, utf8, cd, save_hwm);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
4174            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
4175            code++;            code++;
4176              if (repeat_max == 0)
4177                {
4178                *previous++ = OP_SKIPZERO;
4179                goto END_REPEAT;
4180                }
4181            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4182            }            }
4183    
# Line 3887  we set the flag only if there is a liter Line 4223  we set the flag only if there is a liter
4223            {            {
4224            /* In the pre-compile phase, we don't actually do the replication. We            /* In the pre-compile phase, we don't actually do the replication. We
4225            just adjust the length as if we had. Do some paranoid checks for            just adjust the length as if we had. Do some paranoid checks for
4226            potential integer overflow. */            potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
4227              integer type when available, otherwise double. */
4228    
4229            if (lengthptr != NULL)            if (lengthptr != NULL)
4230              {              {
4231              int delta = (repeat_min - 1)*length_prevgroup;              int delta = (repeat_min - 1)*length_prevgroup;
4232              if ((double)(repeat_min - 1)*(double)length_prevgroup >              if ((INT64_OR_DOUBLE)(repeat_min - 1)*
4233                                                              (double)INT_MAX ||                    (INT64_OR_DOUBLE)length_prevgroup >
4234                        (INT64_OR_DOUBLE)INT_MAX ||
4235                  OFLOW_MAX - *lengthptr < delta)                  OFLOW_MAX - *lengthptr < delta)
4236                {                {
4237                *errorcodeptr = ERR20;                *errorcodeptr = ERR20;
# Line 3939  we set the flag only if there is a liter Line 4277  we set the flag only if there is a liter
4277          just adjust the length as if we had. For each repetition we must add 1          just adjust the length as if we had. For each repetition we must add 1
4278          to the length for BRAZERO and for all but the last repetition we must          to the length for BRAZERO and for all but the last repetition we must
4279          add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some          add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
4280          paranoid checks to avoid integer overflow. */          paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
4281            a 64-bit integer type when available, otherwise double. */
4282    
4283          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4284            {            {
4285            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4286                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4287            if ((double)repeat_max *            if ((INT64_OR_DOUBLE)repeat_max *
4288                  (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)                  (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4289                    > (double)INT_MAX ||                    > (INT64_OR_DOUBLE)INT_MAX ||
4290                OFLOW_MAX - *lengthptr < delta)                OFLOW_MAX - *lengthptr < delta)
4291              {              {
4292              *errorcodeptr = ERR20;              *errorcodeptr = ERR20;
# Line 4036  we set the flag only if there is a liter Line 4375  we set the flag only if there is a liter
4375          }          }
4376        }        }
4377    
4378        /* If previous is OP_FAIL, it was generated by an empty class [] in
4379        JavaScript mode. The other ways in which OP_FAIL can be generated, that is
4380        by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
4381        error above. We can just ignore the repeat in JS case. */
4382    
4383        else if (*previous == OP_FAIL) goto END_REPEAT;
4384    
4385      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
4386    
4387      else      else
# Line 4060  we set the flag only if there is a liter Line 4406  we set the flag only if there is a liter
4406      if (possessive_quantifier)      if (possessive_quantifier)
4407        {        {
4408        int len;        int len;
4409        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||  
4410            *tempcode == OP_NOTEXACT)        if (*tempcode == OP_TYPEEXACT)
4411          tempcode += _pcre_OP_lengths[*tempcode] +          tempcode += _pcre_OP_lengths[*tempcode] +
4412            ((*tempcode == OP_TYPEEXACT &&            ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
4413               (tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP))? 2:0);  
4414          else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
4415            {
4416            tempcode += _pcre_OP_lengths[*tempcode];
4417    #ifdef SUPPORT_UTF8
4418            if (utf8 && tempcode[-1] >= 0xc0)
4419              tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];
4420    #endif
4421            }
4422    
4423        len = code - tempcode;        len = code - tempcode;
4424        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4425          {          {
# Line 4083  we set the flag only if there is a liter Line 4438  we set the flag only if there is a liter
4438          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4439          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4440    
4441            /* Because we are moving code along, we must ensure that any
4442            pending recursive references are updated. */
4443    
4444          default:          default:
4445            *code = OP_END;
4446            adjust_recurse(tempcode, 1 + LINK_SIZE, utf8, cd, save_hwm);
4447          memmove(tempcode + 1+LINK_SIZE, tempcode, len);          memmove(tempcode + 1+LINK_SIZE, tempcode, len);
4448          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
4449          len += 1 + LINK_SIZE;          len += 1 + LINK_SIZE;
# Line 4110  we set the flag only if there is a liter Line 4470  we set the flag only if there is a liter
4470      lookbehind or option setting or condition or all the other extended      lookbehind or option setting or condition or all the other extended
4471      parenthesis forms.  */      parenthesis forms.  */
4472    
4473      case '(':      case CHAR_LEFT_PARENTHESIS:
4474      newoptions = options;      newoptions = options;
4475      skipbytes = 0;      skipbytes = 0;
4476      bravalue = OP_CBRA;      bravalue = OP_CBRA;
# Line 4119  we set the flag only if there is a liter Line 4479  we set the flag only if there is a liter
4479    
4480      /* First deal with various "verbs" that can be introduced by '*'. */      /* First deal with various "verbs" that can be introduced by '*'. */
4481    
4482      if (*(++ptr) == '*' && (cd->ctypes[ptr[1]] & ctype_letter) != 0)      if (*(++ptr) == CHAR_ASTERISK && (cd->ctypes[ptr[1]] & ctype_letter) != 0)
4483        {        {
4484        int i, namelen;        int i, namelen;
4485        const char *vn = verbnames;        const char *vn = verbnames;
4486        const uschar *name = ++ptr;        const uschar *name = ++ptr;
4487        previous = NULL;        previous = NULL;
4488        while ((cd->ctypes[*++ptr] & ctype_letter) != 0);        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
4489        if (*ptr == ':')        if (*ptr == CHAR_COLON)
4490          {          {
4491          *errorcodeptr = ERR59;   /* Not supported */          *errorcodeptr = ERR59;   /* Not supported */
4492          goto FAILED;          goto FAILED;
4493          }          }
4494        if (*ptr != ')')        if (*ptr != CHAR_RIGHT_PARENTHESIS)
4495          {          {
4496          *errorcodeptr = ERR60;          *errorcodeptr = ERR60;
4497          goto FAILED;          goto FAILED;
# Line 4142  we set the flag only if there is a liter Line 4502  we set the flag only if there is a liter
4502          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
4503              strncmp((char *)name, vn, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
4504            {            {
4505            *code = verbs[i].op;            /* Check for open captures before ACCEPT */
4506            if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;  
4507              if (verbs[i].op == OP_ACCEPT)
4508                {
4509                open_capitem *oc;
4510                cd->had_accept = TRUE;
4511                for (oc = cd->open_caps; oc != NULL; oc = oc->next)
4512                  {
4513                  *code++ = OP_CLOSE;
4514                  PUT2INC(code, 0, oc->number);
4515                  }
4516                }
4517              *code++ = verbs[i].op;
4518            break;            break;
4519            }            }
4520          vn += verbs[i].len + 1;          vn += verbs[i].len + 1;
# Line 4156  we set the flag only if there is a liter Line 4527  we set the flag only if there is a liter
4527      /* Deal with the extended parentheses; all are introduced by '?', and the      /* Deal with the extended parentheses; all are introduced by '?', and the
4528      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. */
4529    
4530      else if (*ptr == '?')      else if (*ptr == CHAR_QUESTION_MARK)
4531        {        {
4532        int i, set, unset, namelen;        int i, set, unset, namelen;
4533        int *optset;        int *optset;
# Line 4165  we set the flag only if there is a liter Line 4536  we set the flag only if there is a liter
4536    
4537        switch (*(++ptr))        switch (*(++ptr))
4538          {          {
4539          case '#':                 /* Comment; skip to ket */          case CHAR_NUMBER_SIGN:                 /* Comment; skip to ket */
4540          ptr++;          ptr++;
4541          while (*ptr != 0 && *ptr != ')') ptr++;          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
4542          if (*ptr == 0)          if (*ptr == 0)
4543            {            {
4544            *errorcodeptr = ERR18;            *errorcodeptr = ERR18;
# Line 4177  we set the flag only if there is a liter Line 4548  we set the flag only if there is a liter
4548    
4549    
4550          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4551          case '|':                 /* Reset capture count for each branch */          case CHAR_VERTICAL_LINE:  /* Reset capture count for each branch */
4552          reset_bracount = TRUE;          reset_bracount = TRUE;
4553          /* Fall through */          /* Fall through */
4554    
4555          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4556          case ':':                 /* Non-capturing bracket */          case CHAR_COLON:          /* Non-capturing bracket */
4557          bravalue = OP_BRA;          bravalue = OP_BRA;
4558          ptr++;          ptr++;
4559          break;          break;
4560    
4561    
4562          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4563          case '(':          case CHAR_LEFT_PARENTHESIS:
4564          bravalue = OP_COND;       /* Conditional group */          bravalue = OP_COND;       /* Conditional group */
4565    
4566          /* A condition can be an assertion, a number (referring to a numbered          /* A condition can be an assertion, a number (referring to a numbered
# Line 4209  we set the flag only if there is a liter Line 4580  we set the flag only if there is a liter
4580          the switch. This will take control down to where bracketed groups,          the switch. This will take control down to where bracketed groups,
4581          including assertions, are processed. */          including assertions, are processed. */
4582    
4583          if (ptr[1] == '?' && (ptr[2] == '=' || ptr[2] == '!' || ptr[2] == '<'))          if (ptr[1] == CHAR_QUESTION_MARK && (ptr[2] == CHAR_EQUALS_SIGN ||
4584                ptr[2] == CHAR_EXCLAMATION_MARK || ptr[2] == CHAR_LESS_THAN_SIGN))
4585            break;            break;
4586    
4587          /* Most other conditions use OP_CREF (a couple change to OP_RREF          /* Most other conditions use OP_CREF (a couple change to OP_RREF
# Line 4221  we set the flag only if there is a liter Line 4593  we set the flag only if there is a liter
4593    
4594          /* Check for a test for recursion in a named group. */          /* Check for a test for recursion in a named group. */
4595    
4596          if (ptr[1] == 'R' && ptr[2] == '&')          if (ptr[1] == CHAR_R && ptr[2] == CHAR_AMPERSAND)
4597            {            {
4598            terminator = -1;            terminator = -1;
4599            ptr += 2;            ptr += 2;
# Line 4231  we set the flag only if there is a liter Line 4603  we set the flag only if there is a liter
4603          /* 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
4604          syntax (?(<name>) or (?('name') */          syntax (?(<name>) or (?('name') */
4605    
4606          else if (ptr[1] == '<')          else if (ptr[1] == CHAR_LESS_THAN_SIGN)
4607            {            {
4608            terminator = '>';            terminator = CHAR_GREATER_THAN_SIGN;
4609            ptr++;            ptr++;
4610            }            }
4611          else if (ptr[1] == '\'')          else if (ptr[1] == CHAR_APOSTROPHE)
4612            {            {
4613            terminator = '\'';            terminator = CHAR_APOSTROPHE;
4614            ptr++;            ptr++;
4615            }            }
4616          else          else
4617            {            {
4618            terminator = 0;            terminator = 0;
4619            if (ptr[1] == '-' || ptr[1] == '+') refsign = *(++ptr);            if (ptr[1] == CHAR_MINUS || ptr[1] == CHAR_PLUS) refsign = *(++ptr);
4620            }            }
4621    
4622          /* We now expect to read a name; any thing else is an error */          /* We now expect to read a name; any thing else is an error */
# Line 4264  we set the flag only if there is a liter Line 4636  we set the flag only if there is a liter
4636            {            {
4637            if (recno >= 0)            if (recno >= 0)
4638              recno = ((digitab[*ptr] & ctype_digit) != 0)?              recno = ((digitab[*ptr] & ctype_digit) != 0)?
4639                recno * 10 + *ptr - '0' : -1;                recno * 10 + *ptr - CHAR_0 : -1;
4640            ptr++;            ptr++;
4641            }            }
4642          namelen = ptr - name;          namelen = ptr - name;
4643    
4644          if ((terminator > 0 && *ptr++ != terminator) || *ptr++ != ')')          if ((terminator > 0 && *ptr++ != terminator) ||
4645                *ptr++ != CHAR_RIGHT_PARENTHESIS)
4646            {            {
4647            ptr--;      /* Error offset */            ptr--;      /* Error offset */
4648            *errorcodeptr = ERR26;            *errorcodeptr = ERR26;
# Line 4291  we set the flag only if there is a liter Line 4664  we set the flag only if there is a liter
4664              *errorcodeptr = ERR58;              *errorcodeptr = ERR58;
4665              goto FAILED;              goto FAILED;
4666              }              }
4667            recno = (refsign == '-')?            recno = (refsign == CHAR_MINUS)?
4668              cd->bracount - recno + 1 : recno +cd->bracount;              cd->bracount - recno + 1 : recno +cd->bracount;
4669            if (recno <= 0 || recno > cd->final_bracount)            if (recno <= 0 || recno > cd->final_bracount)
4670              {              {
# Line 4303  we set the flag only if there is a liter Line 4676  we set the flag only if there is a liter
4676            }            }
4677    
4678          /* Otherwise (did not start with "+" or "-"), start by looking for the          /* Otherwise (did not start with "+" or "-"), start by looking for the
4679          name. */          name. If we find a name, add one to the opcode to change OP_CREF or
4680            OP_RREF into OP_NCREF or OP_NRREF. These behave exactly the same,
4681            except they record that the reference was originally to a name. The
4682            information is used to check duplicate names. */
4683    
4684          slot = cd->name_table;          slot = cd->name_table;
4685          for (i = 0; i < cd->names_found; i++)          for (i = 0; i < cd->names_found; i++)
# Line 4318  we set the flag only if there is a liter Line 4694  we set the flag only if there is a liter
4694            {            {
4695            recno = GET2(slot, 0);            recno = GET2(slot, 0);
4696            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4697              code[1+LINK_SIZE]++;
4698            }            }
4699    
4700          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
4701    
4702          else if ((i = find_parens(ptr, cd->bracount, name, namelen,          else if ((i = find_parens(cd, name, namelen,
4703                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0)) > 0)
4704            {            {
4705            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
4706              code[1+LINK_SIZE]++;
4707            }            }
4708    
4709          /* If terminator == 0 it means that the name followed directly after          /* If terminator == 0 it means that the name followed directly after
# Line 4343  we set the flag only if there is a liter Line 4721  we set the flag only if there is a liter
4721          /* Check for (?(R) for recursion. Allow digits after R to specify a          /* Check for (?(R) for recursion. Allow digits after R to specify a
4722          specific group number. */          specific group number. */
4723    
4724          else if (*name == 'R')          else if (*name == CHAR_R)
4725            {            {
4726            recno = 0;            recno = 0;
4727            for (i = 1; i < namelen; i++)            for (i = 1; i < namelen; i++)
# Line 4353  we set the flag only if there is a liter Line 4731  we set the flag only if there is a liter
4731                *errorcodeptr = ERR15;                *errorcodeptr = ERR15;
4732                goto FAILED;                goto FAILED;
4733                }                }
4734              recno = recno * 10 + name[i] - '0';              recno = recno * 10 + name[i] - CHAR_0;
4735              }              }
4736            if (recno == 0) recno = RREF_ANY;            if (recno == 0) recno = RREF_ANY;
4737            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */
# Line 4363  we set the flag only if there is a liter Line 4741  we set the flag only if there is a liter
4741          /* Similarly, check for the (?(DEFINE) "condition", which is always          /* Similarly, check for the (?(DEFINE) "condition", which is always
4742          false. */          false. */
4743    
4744          else if (namelen == 6 && strncmp((char *)name, "DEFINE", 6) == 0)          else if (namelen == 6 && strncmp((char *)name, STRING_DEFINE, 6) == 0)
4745            {            {
4746            code[1+LINK_SIZE] = OP_DEF;            code[1+LINK_SIZE] = OP_DEF;
4747            skipbytes = 1;            skipbytes = 1;
# Line 4388  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 '=':                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
4770          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
4771          ptr++;          ptr++;
4772          break;          break;
4773    
4774    
4775          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4776          case '!':                 /* Negative lookahead */          case CHAR_EXCLAMATION_MARK:            /* Negative lookahead */
4777          ptr++;          ptr++;
4778          if (*ptr == ')')          /* Optimize (?!) */          if (*ptr == CHAR_RIGHT_PARENTHESIS)    /* Optimize (?!) */
4779            {            {
4780            *code++ = OP_FAIL;            *code++ = OP_FAIL;
4781            previous = NULL;            previous = NULL;
# Line 4408  we set the flag only if there is a liter Line 4786  we set the flag only if there is a liter
4786    
4787    
4788          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4789          case '<':                 /* Lookbehind or named define */          case CHAR_LESS_THAN_SIGN:              /* Lookbehind or named define */
4790          switch (ptr[1])          switch (ptr[1])
4791            {            {
4792            case '=':               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
4793            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
4794            ptr += 2;            ptr += 2;
4795            break;            break;
4796    
4797            case '!':               /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
4798            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
4799            ptr += 2;            ptr += 2;
4800            break;            break;
# Line 4431  we set the flag only if there is a liter Line 4809  we set the flag only if there is a liter
4809    
4810    
4811          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4812          case '>':                 /* One-time brackets */          case CHAR_GREATER_THAN_SIGN:           /* One-time brackets */
4813          bravalue = OP_ONCE;          bravalue = OP_ONCE;
4814          ptr++;          ptr++;
4815          break;          break;
4816    
4817    
4818          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4819          case 'C':                 /* Callout - may be followed by digits; */          case CHAR_C:                 /* Callout - may be followed by digits; */
4820          previous_callout = code;  /* Save for later completion */          previous_callout = code;  /* Save for later completion */
4821          after_manual_callout = 1; /* Skip one item before completing */          after_manual_callout = 1; /* Skip one item before completing */
4822          *code++ = OP_CALLOUT;          *code++ = OP_CALLOUT;
4823            {            {
4824            int n = 0;            int n = 0;
4825            while ((digitab[*(++ptr)] & ctype_digit) != 0)            while ((digitab[*(++ptr)] & ctype_digit) != 0)
4826              n = n * 10 + *ptr - '0';              n = n * 10 + *ptr - CHAR_0;
4827            if (*ptr != ')')            if (*ptr != CHAR_RIGHT_PARENTHESIS)
4828              {              {
4829              *errorcodeptr = ERR39;              *errorcodeptr = ERR39;
4830              goto FAILED;              goto FAILED;
# Line 4466  we set the flag only if there is a liter Line 4844  we set the flag only if there is a liter
4844    
4845    
4846          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4847          case 'P':                 /* Python-style named subpattern handling */          case CHAR_P:              /* Python-style named subpattern handling */
4848          if (*(++ptr) == '=' || *ptr == '>')  /* Reference or recursion */          if (*(++ptr) == CHAR_EQUALS_SIGN ||
4849                *ptr == CHAR_GREATER_THAN_SIGN)  /* Reference or recursion */
4850            {            {
4851            is_recurse = *ptr == '>';            is_recurse = *ptr == CHAR_GREATER_THAN_SIGN;
4852            terminator = ')';            terminator = CHAR_RIGHT_PARENTHESIS;
4853            goto NAMED_REF_OR_RECURSE;            goto NAMED_REF_OR_RECURSE;
4854            }            }
4855          else if (*ptr != '<')    /* Test for Python-style definition */          else if (*ptr != CHAR_LESS_THAN_SIGN)  /* Test for Python-style defn */
4856            {            {
4857            *errorcodeptr = ERR41;            *errorcodeptr = ERR41;
4858            goto FAILED;            goto FAILED;
# Line 4483  we set the flag only if there is a liter Line 4862  we set the flag only if there is a liter
4862    
4863          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4864          DEFINE_NAME:    /* Come here from (?< handling */          DEFINE_NAME:    /* Come here from (?< handling */
4865          case '\'':          case CHAR_APOSTROPHE:
4866            {            {
4867            terminator = (*ptr == '<')? '>' : '\'';            terminator = (*ptr == CHAR_LESS_THAN_SIGN)?
4868                CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
4869            name = ++ptr;            name = ++ptr;
4870    
4871            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
# Line 4516  we set the flag only if there is a liter Line 4896  we set the flag only if there is a liter
4896                }                }
4897              }              }
4898    
4899            /* In the real compile, create the entry in the table */            /* In the real compile, create the entry in the table, maintaining
4900              alphabetical order. Duplicate names for different numbers are
4901              permitted only if PCRE_DUPNAMES is set. Duplicate names for the same
4902              number are always OK. (An existing number can be re-used if (?|
4903              appears in the pattern.) In either event, a duplicate name results in
4904              a duplicate entry in the table, even if the number is the same. This
4905              is because the number of names, and hence the table size, is computed
4906              in the pre-compile, and it affects various numbers and pointers which
4907              would all have to be modified, and the compiled code moved down, if
4908              duplicates with the same number were omitted from the table. This
4909              doesn't seem worth the hassle. However, *different* names for the
4910              same number are not permitted. */
4911    
4912            else            else
4913              {              {
4914                BOOL dupname = FALSE;
4915              slot = cd->name_table;              slot = cd->name_table;
4916    
4917              for (i = 0; i < cd->names_found; i++)              for (i = 0; i < cd->names_found; i++)
4918                {                {
4919                int crc = memcmp(name, slot+2, namelen);                int crc = memcmp(name, slot+2, namelen);
# Line 4528  we set the flag only if there is a liter Line 4921  we set the flag only if there is a liter
4921                  {                  {
4922                  if (slot[2+namelen] == 0)                  if (slot[2+namelen] == 0)
4923                    {                    {
4924                    if ((options & PCRE_DUPNAMES) == 0)                    if (GET2(slot, 0) != cd->bracount + 1 &&
4925                          (options & PCRE_DUPNAMES) == 0)
4926                      {                      {
4927                      *errorcodeptr = ERR43;                      *errorcodeptr = ERR43;
4928                      goto FAILED;                      goto FAILED;
4929                      }                      }
4930                      else dupname = TRUE;
4931                    }                    }
4932                  else crc = -1;      /* Current name is substring */                  else crc = -1;      /* Current name is a substring */
4933                  }                  }
4934    
4935                  /* Make space in the table and break the loop for an earlier
4936                  name. For a duplicate or later name, carry on. We do this for
4937                  duplicates so that in the simple case (when ?(| is not used) they
4938                  are in order of their numbers. */
4939    
4940                if (crc < 0)                if (crc < 0)
4941                  {                  {
4942                  memmove(slot + cd->name_entry_size, slot,                  memmove(slot + cd->name_entry_size, slot,
4943                    (cd->names_found - i) * cd->name_entry_size);                    (cd->names_found - i) * cd->name_entry_size);
4944                  break;                  break;
4945                  }                  }
4946    
4947                  /* Continue the loop for a later or duplicate name */
4948    
4949                slot += cd->name_entry_size;                slot += cd->name_entry_size;
4950                }                }
4951    
4952                /* For non-duplicate names, check for a duplicate number before
4953                adding the new name. */
4954    
4955                if (!dupname)
4956                  {
4957                  uschar *cslot = cd->name_table;
4958                  for (i = 0; i < cd->names_found; i++)
4959                    {
4960                    if (cslot != slot)
4961                      {
4962                      if (GET2(cslot, 0) == cd->bracount + 1)
4963                        {
4964                        *errorcodeptr = ERR65;
4965                        goto FAILED;
4966                        }
4967                      }
4968                    else i--;
4969                    cslot += cd->name_entry_size;
4970                    }
4971                  }
4972    
4973              PUT2(slot, 0, cd->bracount + 1);              PUT2(slot, 0, cd->bracount + 1);
4974              memcpy(slot + 2, name, namelen);              memcpy(slot + 2, name, namelen);
4975              slot[2+namelen] = 0;              slot[2+namelen] = 0;
4976              }              }
4977            }            }
4978    
4979          /* In both cases, count the number of names we've encountered. */          /* In both pre-compile and compile, count the number of names we've
4980            encountered. */
4981    
         ptr++;                    /* Move past > or ' */  
4982          cd->names_found++;          cd->names_found++;
4983            ptr++;                    /* Move past > or ' */
4984          goto NUMBERED_GROUP;          goto NUMBERED_GROUP;
4985    
4986    
4987          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4988          case '&':                 /* Perl recursion/subroutine syntax */          case CHAR_AMPERSAND:            /* Perl recursion/subroutine syntax */
4989          terminator = ')';          terminator = CHAR_RIGHT_PARENTHESIS;
4990          is_recurse = TRUE;          is_recurse = TRUE;
4991          /* Fall through */          /* Fall through */
4992    
# Line 4568  we set the flag only if there is a liter Line 4994  we set the flag only if there is a liter
4994          references (?P=name) and recursion (?P>name), as well as falling          references (?P=name) and recursion (?P>name), as well as falling
4995          through from the Perl recursion syntax (?&name). We also come here from          through from the Perl recursion syntax (?&name). We also come here from
4996          the Perl \k<name> or \k'name' back reference syntax and the \k{name}          the Perl \k<name> or \k'name' back reference syntax and the \k{name}
4997          .NET syntax. */          .NET syntax, and the Oniguruma \g<...> and \g'...' subroutine syntax. */
4998    
4999          NAMED_REF_OR_RECURSE:          NAMED_REF_OR_RECURSE:
5000          name = ++ptr;          name = ++ptr;
# Line 4619  we set the flag only if there is a liter Line 5045  we set the flag only if there is a liter
5045              recno = GET2(slot, 0);              recno = GET2(slot, 0);
5046              }              }
5047            else if ((recno =                /* Forward back reference */            else if ((recno =                /* Forward back reference */
5048                      find_parens(ptr, cd->bracount, name, namelen,                      find_parens(cd, name, namelen,
5049                        (options & PCRE_EXTENDED) != 0)) <= 0)                        (options & PCRE_EXTENDED) != 0)) <= 0)
5050              {              {
5051              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
# Line 4635  we set the flag only if there is a liter Line 5061  we set the flag only if there is a liter
5061    
5062    
5063          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5064          case 'R':                 /* Recursion */          case CHAR_R:              /* Recursion */
5065          ptr++;                    /* Same as (?0)      */          ptr++;                    /* Same as (?0)      */
5066          /* Fall through */          /* Fall through */
5067    
5068    
5069          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5070          case '-': case '+':          case CHAR_MINUS: case CHAR_PLUS:  /* Recursion or subroutine */
5071          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:
5072          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:
5073            {            {
5074            const uschar *called;            const uschar *called;
5075              terminator = CHAR_RIGHT_PARENTHESIS;
5076    
5077              /* Come here from the \g<...> and \g'...' code (Oniguruma
5078              compatibility). However, the syntax has been checked to ensure that
5079              the ... are a (signed) number, so that neither ERR63 nor ERR29 will
5080              be called on this path, nor with the jump to OTHER_CHAR_AFTER_QUERY
5081              ever be taken. */
5082    
5083              HANDLE_NUMERICAL_RECURSION:
5084    
5085            if ((refsign = *ptr) == '+')            if ((refsign = *ptr) == CHAR_PLUS)
5086              {              {
5087              ptr++;              ptr++;
5088              if ((digitab[*ptr] & ctype_digit) == 0)              if ((digitab[*ptr] & ctype_digit) == 0)
# Line 4656  we set the flag only if there is a liter Line 5091  we set the flag only if there is a liter
5091                goto FAILED;                goto FAILED;
5092                }                }
5093              }              }
5094            else if (refsign == '-')            else if (refsign == CHAR_MINUS)
5095              {              {
5096              if ((digitab[ptr[1]] & ctype_digit) == 0)              if ((digitab[ptr[1]] & ctype_digit) == 0)
5097                goto OTHER_CHAR_AFTER_QUERY;                goto OTHER_CHAR_AFTER_QUERY;
# Line 4665  we set the flag only if there is a liter Line 5100  we set the flag only if there is a liter
5100    
5101            recno = 0;            recno = 0;
5102            while((digitab[*ptr] & ctype_digit) != 0)            while((digitab[*ptr] & ctype_digit) != 0)
5103              recno = recno * 10 + *ptr++ - '0';              recno = recno * 10 + *ptr++ - CHAR_0;
5104    
5105            if (*ptr != ')')            if (*ptr != terminator)
5106              {              {
5107              *errorcodeptr = ERR29;              *errorcodeptr = ERR29;
5108              goto FAILED;              goto FAILED;
5109              }              }
5110    
5111            if (refsign == '-')            if (refsign == CHAR_MINUS)
5112              {              {
5113              if (recno == 0)              if (recno == 0)
5114                {                {
# Line 4687  we set the flag only if there is a liter Line 5122  we set the flag only if there is a liter
5122                goto FAILED;                goto FAILED;
5123                }                }
5124              }              }
5125            else if (refsign == '+')            else if (refsign == CHAR_PLUS)
5126              {              {
5127              if (recno == 0)              if (recno == 0)
5128                {                {
# Line 4714  we set the flag only if there is a liter Line 5149  we set the flag only if there is a liter
5149            if (lengthptr == NULL)            if (lengthptr == NULL)
5150              {              {
5151              *code = OP_END;              *code = OP_END;
5152              if (recno != 0) called = find_bracket(cd->start_code, utf8, recno);              if (recno != 0)
5153                  called = _pcre_find_bracket(cd->start_code, utf8, recno);
5154    
5155              /* Forward reference */              /* Forward reference */
5156    
5157              if (called == NULL)              if (called == NULL)
5158                {                {
5159                if (find_parens(ptr, cd->bracount, NULL, recno,                if (find_parens(cd, NULL, recno,
5160                     (options & PCRE_EXTENDED) != 0) < 0)                      (options & PCRE_EXTENDED) != 0) < 0)
5161                  {                  {
5162                  *errorcodeptr = ERR15;                  *errorcodeptr = ERR15;
5163                  goto FAILED;                  goto FAILED;
5164                  }                  }
5165    
5166                  /* Fudge the value of "called" so that when it is inserted as an
5167                  offset below, what it actually inserted is the reference number
5168                  of the group. */
5169    
5170                called = cd->start_code + recno;                called = cd->start_code + recno;
5171                PUTINC(cd->hwm, 0, code + 2 + LINK_SIZE - cd->start_code);                PUTINC(cd->hwm, 0, code + 2 + LINK_SIZE - cd->start_code);
5172                }                }
# Line 4773  we set the flag only if there is a liter Line 5214  we set the flag only if there is a liter
5214          set = unset = 0;          set = unset = 0;
5215          optset = &set;          optset = &set;
5216    
5217          while (*ptr != ')' && *ptr != ':')          while (*ptr != CHAR_RIGHT_PARENTHESIS && *ptr != CHAR_COLON)
5218            {            {
5219            switch (*ptr++)            switch (*ptr++)
5220              {              {
5221              case '-': optset = &unset; break;              case CHAR_MINUS: optset = &unset; break;
5222    
5223              case 'J':    /* Record that it changed in the external options */              case CHAR_J:    /* Record that it changed in the external options */
5224              *optset |= PCRE_DUPNAMES;              *optset |= PCRE_DUPNAMES;
5225              cd->external_flags |= PCRE_JCHANGED;              cd->external_flags |= PCRE_JCHANGED;
5226              break;              break;
5227    
5228              case 'i': *optset |= PCRE_CASELESS; break;              case CHAR_i: *optset |= PCRE_CASELESS; break;
5229              case 'm': *optset |= PCRE_MULTILINE; break;              case CHAR_m: *optset |= PCRE_MULTILINE; break;
5230              case 's': *optset |= PCRE_DOTALL; break;              case CHAR_s: *optset |= PCRE_DOTALL; break;
5231              case 'x': *optset |= PCRE_EXTENDED; break;              case CHAR_x: *optset |= PCRE_EXTENDED; break;
5232              case 'U': *optset |= PCRE_UNGREEDY; break;              case CHAR_U: *optset |= PCRE_UNGREEDY; break;
5233              case 'X': *optset |= PCRE_EXTRA; break;              case CHAR_X: *optset |= PCRE_EXTRA; break;
5234    
5235              default:  *errorcodeptr = ERR12;              default:  *errorcodeptr = ERR12;
5236                        ptr--;    /* Correct the offset */                        ptr--;    /* Correct the offset */
# Line 4820  we set the flag only if there is a liter Line 5261  we set the flag only if there is a liter
5261          both phases.          both phases.
5262    
5263          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
5264          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
5265          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. */  
5266    
5267          if (*ptr == ')')          if (*ptr == CHAR_RIGHT_PARENTHESIS)
5268            {            {
5269            if (code == cd->start_code + 1 + LINK_SIZE &&            if (code == cd->start_code + 1 + LINK_SIZE &&
5270                 (lengthptr == NULL || *lengthptr == 2 + 2*LINK_SIZE))                 (lengthptr == NULL || *lengthptr == 2 + 2*LINK_SIZE))
5271              {              {
5272              cd->external_options = newoptions;              cd->external_options = newoptions;
             options = newoptions;  
5273              }              }
5274           else            else
5275              {              {
5276              if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))              if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))
5277                {                {
5278                *code++ = OP_OPT;                *code++ = OP_OPT;
5279                *code++ = newoptions & PCRE_IMS;                *code++ = newoptions & PCRE_IMS;
5280                }                }
   
             /* 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;  
5281              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);
5282              greedy_non_default = greedy_default ^ 1;              greedy_non_default = greedy_default ^ 1;
5283              req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
5284              }              }
5285    
5286              /* Change options at this level, and pass them back for use
5287              in subsequent branches. When not at the start of the pattern, this
5288              information is also necessary so that a resetting item can be
5289              compiled at the end of a group (if we are in a group). */
5290    
5291              *optionsptr = options = newoptions;
5292            previous = NULL;       /* This item can't be repeated */            previous = NULL;       /* This item can't be repeated */
5293            continue;              /* It is complete */            continue;              /* It is complete */
5294            }            }
# Line 4966  we set the flag only if there is a liter Line 5404  we set the flag only if there is a liter
5404    
5405      /* Error if hit end of pattern */      /* Error if hit end of pattern */
5406    
5407      if (*ptr != ')')      if (*ptr != CHAR_RIGHT_PARENTHESIS)
5408        {        {
5409        *errorcodeptr = ERR14;        *errorcodeptr = ERR14;
5410        goto FAILED;        goto FAILED;
# Line 5064  we set the flag only if there is a liter Line 5502  we set the flag only if there is a liter
5502      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
5503      have to change if any new ones are ever created. */      have to change if any new ones are ever created. */
5504    
5505      case '\\':      case CHAR_BACKSLASH:
5506      tempptr = ptr;      tempptr = ptr;
5507      c = check_escape(&ptr, errorcodeptr, cd->bracount, options, FALSE);      c = check_escape(&ptr, errorcodeptr, cd->bracount, options, FALSE);
5508      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
# Line 5073  we set the flag only if there is a liter Line 5511  we set the flag only if there is a liter
5511        {        {
5512        if (-c == ESC_Q)            /* Handle start of quoted string */        if (-c == ESC_Q)            /* Handle start of quoted string */
5513          {          {
5514          if (ptr[1] == '\\' && ptr[2] == 'E') ptr += 2; /* avoid empty string */          if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
5515            else inescq = TRUE;            ptr += 2;               /* avoid empty string */
5516                else inescq = TRUE;
5517          continue;          continue;
5518          }          }
5519    
# Line 5091  we set the flag only if there is a liter Line 5530  we set the flag only if there is a liter
5530        zerofirstbyte = firstbyte;        zerofirstbyte = firstbyte;
5531        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
5532    
5533          /* \g<name> or \g'name' is a subroutine call by name and \g<n> or \g'n'
5534          is a subroutine call by number (Oniguruma syntax). In fact, the value
5535          -ESC_g is returned only for these cases. So we don't need to check for <
5536          or ' if the value is -ESC_g. For the Perl syntax \g{n} the value is
5537          -ESC_REF+n, and for the Perl syntax \g{name} the result is -ESC_k (as
5538          that is a synonym for a named back reference). */
5539    
5540          if (-c == ESC_g)
5541            {
5542            const uschar *p;
5543            save_hwm = cd->hwm;   /* Normally this is set when '(' is read */
5544            terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
5545              CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
5546    
5547            /* These two statements stop the compiler for warning about possibly
5548            unset variables caused by the jump to HANDLE_NUMERICAL_RECURSION. In
5549            fact, because we actually check for a number below, the paths that
5550            would actually be in error are never taken. */
5551    
5552            skipbytes = 0;
5553            reset_bracount = FALSE;
5554    
5555            /* Test for a name */
5556    
5557            if (ptr[1] != CHAR_PLUS && ptr[1] != CHAR_MINUS)
5558              {
5559              BOOL isnumber = TRUE;
5560              for (p = ptr + 1; *p != 0 && *p != terminator; p++)
5561                {
5562                if ((cd->ctypes[*p] & ctype_digit) == 0) isnumber = FALSE;
5563                if ((cd->ctypes[*p] & ctype_word) == 0) break;
5564                }
5565              if (*p != terminator)
5566                {
5567                *errorcodeptr = ERR57;
5568                break;
5569                }
5570              if (isnumber)
5571                {
5572                ptr++;
5573                goto HANDLE_NUMERICAL_RECURSION;
5574                }
5575              is_recurse = TRUE;
5576              goto NAMED_REF_OR_RECURSE;
5577              }
5578    
5579            /* Test a signed number in angle brackets or quotes. */
5580    
5581            p = ptr + 2;
5582            while ((digitab[*p] & ctype_digit) != 0) p++;
5583            if (*p != terminator)
5584              {
5585              *errorcodeptr = ERR57;
5586              break;
5587              }
5588            ptr++;
5589            goto HANDLE_NUMERICAL_RECURSION;
5590            }
5591    
5592        /* \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).
5593        We also support \k{name} (.NET syntax) */        We also support \k{name} (.NET syntax) */
5594    
5595        if (-c == ESC_k && (ptr[1] == '<' || ptr[1] == '\'' || ptr[1] == '{'))        if (-c == ESC_k && (ptr[1] == CHAR_LESS_THAN_SIGN ||
5596              ptr[1] == CHAR_APOSTROPHE || ptr[1] == CHAR_LEFT_CURLY_BRACKET))
5597          {          {
5598          is_recurse = FALSE;          is_recurse = FALSE;
5599          terminator = (*(++ptr) == '<')? '>' : (*ptr == '\'')? '\'' : '}';          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
5600              CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?
5601              CHAR_APOSTROPHE : CHAR_RIGHT_CURLY_BRACKET;
5602          goto NAMED_REF_OR_RECURSE;          goto NAMED_REF_OR_RECURSE;
5603          }          }
5604    
# Line 5107  we set the flag only if there is a liter Line 5608  we set the flag only if there is a liter
5608    
5609        if (-c >= ESC_REF)        if (-c >= ESC_REF)
5610          {          {
5611            open_capitem *oc;
5612          recno = -c - ESC_REF;          recno = -c - ESC_REF;
5613    
5614          HANDLE_REFERENCE:    /* Come here from named backref handling */          HANDLE_REFERENCE:    /* Come here from named backref handling */
# Line 5116  we set the flag only if there is a liter Line 5618  we set the flag only if there is a liter
5618          PUT2INC(code, 0, recno);          PUT2INC(code, 0, recno);
5619          cd->backref_map |= (recno < 32)? (1 << recno) : 1;          cd->backref_map |= (recno < 32)? (1 << recno) : 1;
5620          if (recno > cd->top_backref) cd->top_backref = recno;          if (recno > cd->top_backref) cd->top_backref = recno;
5621    
5622            /* Check to see if this back reference is recursive, that it, it
5623            is inside the group that it references. A flag is set so that the
5624            group can be made atomic. */
5625    
5626            for (oc = cd->open_caps; oc != NULL; oc = oc->next)
5627              {
5628              if (oc->number == recno)
5629                {
5630                oc->flag = TRUE;
5631                break;
5632                }
5633              }
5634          }          }
5635    
5636        /* So are Unicode property matches, if supported. */        /* So are Unicode property matches, if supported. */
# Line 5200  we set the flag only if there is a liter Line 5715  we set the flag only if there is a liter
5715    
5716      /* Remember if \r or \n were seen */      /* Remember if \r or \n were seen */
5717    
5718      if (mcbuffer[0] == '\r' || mcbuffer[0] == '\n')      if (mcbuffer[0] == CHAR_CR || mcbuffer[0] == CHAR_NL)
5719        cd->external_flags |= PCRE_HASCRORLF;        cd->external_flags |= PCRE_HASCRORLF;
5720    
5721      /* Set the first and required bytes appropriately. If no previous first      /* Set the first and required bytes appropriately. If no previous first
# Line 5298  uschar *code = *codeptr; Line 5813  uschar *code = *codeptr;
5813  uschar *last_branch = code;  uschar *last_branch = code;
5814  uschar *start_bracket = code;  uschar *start_bracket = code;
5815  uschar *reverse_count = NULL;  uschar *reverse_count = NULL;
5816    open_capitem capitem;
5817    int capnumber = 0;
5818  int firstbyte, reqbyte;  int firstbyte, reqbyte;
5819  int branchfirstbyte, branchreqbyte;  int branchfirstbyte, branchreqbyte;
5820  int length;  int length;
5821  int orig_bracount;  int orig_bracount;
5822  int max_bracount;  int max_bracount;
5823    int old_external_options = cd->external_options;
5824  branch_chain bc;  branch_chain bc;
5825    
5826  bc.outer = bcptr;  bc.outer = bcptr;
5827  bc.current = code;  bc.current_branch = code;
5828    
5829  firstbyte = reqbyte = REQ_UNSET;  firstbyte = reqbyte = REQ_UNSET;
5830    
# Line 5324  the code that abstracts option settings Line 5842  the code that abstracts option settings
5842  them global. It tests the value of length for (2 + 2*LINK_SIZE) in the  them global. It tests the value of length for (2 + 2*LINK_SIZE) in the
5843  pre-compile phase to find out whether anything has yet been compiled or not. */  pre-compile phase to find out whether anything has yet been compiled or not. */
5844    
5845    /* If this is a capturing subpattern, add to the chain of open capturing items
5846    so that we can detect them if (*ACCEPT) is encountered. This is also used to
5847    detect groups that contain recursive back references to themselves. */
5848    
5849    if (*code == OP_CBRA)
5850      {
5851      capnumber = GET2(code, 1 + LINK_SIZE);
5852      capitem.number = capnumber;
5853      capitem.next = cd->open_caps;
5854      capitem.flag = FALSE;
5855      cd->open_caps = &capitem;
5856      }
5857    
5858  /* Offset is set zero to mark that this bracket is still open */  /* Offset is set zero to mark that this bracket is still open */
5859    
5860  PUT(code, 1, 0);  PUT(code, 1, 0);
# Line 5368  for (;;) Line 5899  for (;;)
5899      return FALSE;      return FALSE;
5900      }      }
5901    
5902      /* If the external options have changed during this branch, it means that we
5903      are at the top level, and a leading option setting has been encountered. We
5904      need to re-set the original option values to take account of this so that,
5905      during the pre-compile phase, we know to allow for a re-set at the start of
5906      subsequent branches. */
5907    
5908      if (old_external_options != cd->external_options)
5909        oldims = cd->external_options & PCRE_IMS;
5910    
5911    /* Keep the highest bracket count in case (?| was used and some branch    /* Keep the highest bracket count in case (?| was used and some branch
5912    has fewer than the rest. */    has fewer than the rest. */
5913    
# Line 5418  for (;;) Line 5958  for (;;)
5958    
5959      /* If lookbehind, check that this branch matches a fixed-length string, and      /* If lookbehind, check that this branch matches a fixed-length string, and
5960      put the length into the OP_REVERSE item. Temporarily mark the end of the      put the length into the OP_REVERSE item. Temporarily mark the end of the
5961      branch with OP_END. */      branch with OP_END. If the branch contains OP_RECURSE, the result is -3
5962        because there may be forward references that we can't check here. Set a
5963        flag to cause another lookbehind check at the end. Why not do it all at the
5964        end? Because common, erroneous checks are picked up here and the offset of
5965        the problem can be shown. */
5966    
5967      if (lookbehind)      if (lookbehind)
5968        {        {
5969        int fixed_length;        int fixed_length;
5970        *code = OP_END;        *code = OP_END;
5971        fixed_length = find_fixedlength(last_branch, options);        fixed_length = find_fixedlength(last_branch, options, FALSE, cd);
5972        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
5973        if (fixed_length < 0)        if (fixed_length == -3)
5974            {
5975            cd->check_lookbehind = TRUE;
5976            }
5977          else if (fixed_length < 0)
5978          {          {
5979          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;
5980          *ptrptr = ptr;          *ptrptr = ptr;
5981          return FALSE;          return FALSE;
5982          }          }
5983        PUT(reverse_count, 0, fixed_length);        else { PUT(reverse_count, 0, fixed_length); }
5984        }        }
5985      }      }
5986    
# Line 5445  for (;;) Line 5993  for (;;)
5993    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.
5994    Return leaving the pointer at the terminating char. */    Return leaving the pointer at the terminating char. */
5995    
5996    if (*ptr != '|')    if (*ptr != CHAR_VERTICAL_LINE)
5997      {      {
5998      if (lengthptr == NULL)      if (lengthptr == NULL)
5999        {        {
# Line 5466  for (;;) Line 6014  for (;;)
6014      PUT(code, 1, code - start_bracket);      PUT(code, 1, code - start_bracket);
6015      code += 1 + LINK_SIZE;      code += 1 + LINK_SIZE;
6016    
6017      /* Resetting option if needed */      /* If it was a capturing subpattern, check to see if it contained any
6018        recursive back references. If so, we must wrap it in atomic brackets.
6019        In any event, remove the block from the chain. */
6020    
6021        if (capnumber > 0)
6022          {
6023          if (cd->open_caps->flag)
6024            {
6025            memmove(start_bracket + 1 + LINK_SIZE, start_bracket,
6026              code - start_bracket);
6027            *start_bracket = OP_ONCE;
6028            code += 1 + LINK_SIZE;
6029            PUT(start_bracket, 1, code - start_bracket);
6030            *code = OP_KET;
6031            PUT(code, 1, code - start_bracket);
6032            code += 1 + LINK_SIZE;
6033            length += 2 + 2*LINK_SIZE;
6034            }
6035          cd->open_caps = cd->open_caps->next;
6036          }
6037    
6038        /* Reset options if needed. */
6039    
6040      if ((options & PCRE_IMS) != oldims && *ptr == ')')      if ((options & PCRE_IMS) != oldims && *ptr == CHAR_RIGHT_PARENTHESIS)
6041        {        {
6042        *code++ = OP_OPT;        *code++ = OP_OPT;
6043        *code++ = oldims;        *code++ = oldims;
# Line 5515  for (;;) Line 6084  for (;;)
6084      {      {
6085      *code = OP_ALT;      *code = OP_ALT;
6086      PUT(code, 1, code - last_branch);      PUT(code, 1, code - last_branch);
6087      bc.current = last_branch = code;      bc.current_branch = last_branch = code;
6088      code += 1 + LINK_SIZE;      code += 1 + LINK_SIZE;
6089      }      }
6090    
# Line 5597  do { Line 6166  do {
6166       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;
6167       }       }
6168    
6169     /* .* 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
6170     are or may be referenced. */     it isn't in brackets that are or may be referenced. */
6171    
6172     else if ((op == OP_TYPESTAR || op == OP_TYPEMINSTAR ||     else if ((op == OP_TYPESTAR || op == OP_TYPEMINSTAR ||
6173               op == OP_TYPEPOSSTAR) &&               op == OP_TYPEPOSSTAR))
             (*options & PCRE_DOTALL) != 0)  
6174       {       {
6175       if (scode[1] != OP_ANY || (bracket_map & backref_map) != 0) return FALSE;       if (scode[1] != OP_ALLANY || (bracket_map & backref_map) != 0)
6176           return FALSE;
6177       }       }
6178    
6179     /* Check for explicit anchoring */     /* Check for explicit anchoring */
# Line 5650  do { Line 6219  do {
6219       NULL, 0, FALSE);       NULL, 0, FALSE);
6220     register int op = *scode;     register int op = *scode;
6221    
6222       /* If we are at the start of a conditional assertion group, *both* the
6223       conditional assertion *and* what follows the condition must satisfy the test
6224       for start of line. Other kinds of condition fail. Note that there may be an
6225       auto-callout at the start of a condition. */
6226    
6227       if (op == OP_COND)
6228         {
6229         scode += 1 + LINK_SIZE;
6230         if (*scode == OP_CALLOUT) scode += _pcre_OP_lengths[OP_CALLOUT];
6231         switch (*scode)
6232           {
6233           case OP_CREF:
6234           case OP_NCREF:
6235           case OP_RREF:
6236           case OP_NRREF:
6237           case OP_DEF:
6238           return FALSE;
6239    
6240           default:     /* Assertion */
6241           if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
6242           do scode += GET(scode, 1); while (*scode == OP_ALT);
6243           scode += 1 + LINK_SIZE;
6244           break;
6245           }
6246         scode = first_significant_code(scode, NULL, 0, FALSE);
6247         op = *scode;
6248         }
6249    
6250     /* Non-capturing brackets */     /* Non-capturing brackets */
6251    
6252     if (op == OP_BRA)     if (op == OP_BRA)
# Line 5668  do { Line 6265  do {
6265    
6266     /* Other brackets */     /* Other brackets */
6267    
6268     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_COND)     else if (op == OP_ASSERT || op == OP_ONCE)
6269       { if (!is_startline(scode, bracket_map, backref_map)) return FALSE; }       {
6270         if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
6271         }
6272    
6273     /* .* 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
6274     may be referenced. */     may be referenced. */
# Line 5786  Returns:        pointer to compiled data Line 6385  Returns:        pointer to compiled data
6385                  with errorptr and erroroffset set                  with errorptr and erroroffset set
6386  */  */
6387    
6388  PCRE_EXP_DEFN pcre *  PCRE_EXP_DEFN pcre * PCRE_CALL_CONVENTION
6389  pcre_compile(const char *pattern, int options, const char **errorptr,  pcre_compile(const char *pattern, int options, const char **errorptr,
6390    int *erroroffset, const unsigned char *tables)    int *erroroffset, const unsigned char *tables)
6391  {  {
# Line 5794  return pcre_compile2(pattern, options, N Line 6393  return pcre_compile2(pattern, options, N
6393  }  }
6394    
6395    
6396  PCRE_EXP_DEFN pcre *  PCRE_EXP_DEFN pcre * PCRE_CALL_CONVENTION
6397  pcre_compile2(const char *pattern, int options, int *errorcodeptr,  pcre_compile2(const char *pattern, int options, int *errorcodeptr,
6398    const char **errorptr, int *erroroffset, const unsigned char *tables)    const char **errorptr, int *erroroffset, const unsigned char *tables)
6399  {  {
# Line 5803  int length = 1;  /* For final END opcode Line 6402  int length = 1;  /* For final END opcode
6402  int firstbyte, reqbyte, newline;  int firstbyte, reqbyte, newline;
6403  int errorcode = 0;  int errorcode = 0;
6404  int skipatstart = 0;  int skipatstart = 0;
6405  #ifdef SUPPORT_UTF8  BOOL utf8 = (options & PCRE_UTF8) != 0;
 BOOL utf8;  
 #endif  
6406  size_t size;  size_t size;
6407  uschar *code;  uschar *code;
6408  const uschar *codestart;  const uschar *codestart;
# Line 5848  if (erroroffset == NULL) Line 6445  if (erroroffset == NULL)
6445    
6446  *erroroffset = 0;  *erroroffset = 0;
6447    
 /* 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;  
   }  
   
6448  /* Set up pointers to the individual character tables */  /* Set up pointers to the individual character tables */
6449    
6450  if (tables == NULL) tables = _pcre_default_tables;  if (tables == NULL) tables = _pcre_default_tables;
# Line 5880  cd->fcc = tables + fcc_offset; Line 6453  cd->fcc = tables + fcc_offset;
6453  cd->cbits = tables + cbits_offset;  cd->cbits = tables + cbits_offset;
6454  cd->ctypes = tables + ctypes_offset;  cd->ctypes = tables + ctypes_offset;
6455    
6456    /* Check that all undefined public option bits are zero */
6457    
6458    if ((options & ~PUBLIC_COMPILE_OPTIONS) != 0)
6459      {
6460      errorcode = ERR17;
6461      goto PCRE_EARLY_ERROR_RETURN;
6462      }
6463    
6464  /* 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
6465  the offset for later. */  the offset for later. */
6466    
6467  while (ptr[skipatstart] == '(' && ptr[skipatstart+1] == '*')  while (ptr[skipatstart] == CHAR_LEFT_PARENTHESIS &&
6468           ptr[skipatstart+1] == CHAR_ASTERISK)
6469    {    {
6470    int newnl = 0;    int newnl = 0;
6471    int newbsr = 0;    int newbsr = 0;
6472    
6473    if (strncmp((char *)(ptr+skipatstart+2), "CR)", 3) == 0)    if (strncmp((char *)(ptr+skipatstart+2), STRING_UTF8_RIGHTPAR, 5) == 0)
6474        { skipatstart += 7; options |= PCRE_UTF8; continue; }
6475    
6476      if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)
6477      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }
6478    else if (strncmp((char *)(ptr+skipatstart+2), "LF)", 3)  == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_LF_RIGHTPAR, 3)  == 0)
6479      { skipatstart += 5; newnl = PCRE_NEWLINE_LF; }      { skipatstart += 5; newnl = PCRE_NEWLINE_LF; }
6480    else if (strncmp((char *)(ptr+skipatstart+2), "CRLF)", 5)  == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_CRLF_RIGHTPAR, 5)  == 0)
6481      { skipatstart += 7; newnl = PCRE_NEWLINE_CR + PCRE_NEWLINE_LF; }      { skipatstart += 7; newnl = PCRE_NEWLINE_CR + PCRE_NEWLINE_LF; }
6482    else if (strncmp((char *)(ptr+skipatstart+2), "ANY)", 4) == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_ANY_RIGHTPAR, 4) == 0)
6483      { skipatstart += 6; newnl = PCRE_NEWLINE_ANY; }      { skipatstart += 6; newnl = PCRE_NEWLINE_ANY; }
6484    else if (strncmp((char *)(ptr+skipatstart+2), "ANYCRLF)", 8)  == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_ANYCRLF_RIGHTPAR, 8) == 0)
6485      { skipatstart += 10; newnl = PCRE_NEWLINE_ANYCRLF; }      { skipatstart += 10; newnl = PCRE_NEWLINE_ANYCRLF; }
6486    
6487    else if (strncmp((char *)(ptr+skipatstart+2), "BSR_ANYCRLF)", 12) == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_BSR_ANYCRLF_RIGHTPAR, 12) == 0)
6488      { skipatstart += 14; newbsr = PCRE_BSR_ANYCRLF; }      { skipatstart += 14; newbsr = PCRE_BSR_ANYCRLF; }
6489    else if (strncmp((char *)(ptr+skipatstart+2), "BSR_UNICODE)", 12) == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_BSR_UNICODE_RIGHTPAR, 12) == 0)
6490      { skipatstart += 14; newbsr = PCRE_BSR_UNICODE; }      { skipatstart += 14; newbsr = PCRE_BSR_UNICODE; }
6491    
6492    if (newnl != 0)    if (newnl != 0)
# Line 5911  while (ptr[skipatstart] == '(' && ptr[sk Line 6496  while (ptr[skipatstart] == '(' && ptr[sk
6496    else break;    else break;
6497    }    }
6498    
6499    /* Can't support UTF8 unless PCRE has been compiled to include the code. */
6500    
6501    #ifdef SUPPORT_UTF8
6502    if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&
6503         (*erroroffset = _pcre_valid_utf8((USPTR)pattern, -1)) >= 0)
6504      {
6505      errorcode = ERR44;
6506      goto PCRE_EARLY_ERROR_RETURN2;
6507      }
6508    #else
6509    if (utf8)
6510      {
6511      errorcode = ERR32;
6512      goto PCRE_EARLY_ERROR_RETURN;
6513      }
6514    #endif
6515    
6516  /* Check validity of \R options. */  /* Check validity of \R options. */
6517    
6518  switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))  switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))
# Line 5929  current code allows for fixed one- or tw Line 6531  current code allows for fixed one- or tw
6531  switch (options & PCRE_NEWLINE_BITS)  switch (options & PCRE_NEWLINE_BITS)
6532    {    {
6533    case 0: newline = NEWLINE; break;   /* Build-time default */    case 0: newline = NEWLINE; break;   /* Build-time default */
6534    case PCRE_NEWLINE_CR: newline = '\r'; break;    case PCRE_NEWLINE_CR: newline = CHAR_CR; break;
6535    case PCRE_NEWLINE_LF: newline = '\n'; break;    case PCRE_NEWLINE_LF: newline = CHAR_NL; break;
6536    case PCRE_NEWLINE_CR+    case PCRE_NEWLINE_CR+
6537         PCRE_NEWLINE_LF: newline = ('\r' << 8) | '\n'; break;         PCRE_NEWLINE_LF: newline = (CHAR_CR << 8) | CHAR_NL; break;
6538    case PCRE_NEWLINE_ANY: newline = -1; break;    case PCRE_NEWLINE_ANY: newline = -1; break;
6539    case PCRE_NEWLINE_ANYCRLF: newline = -2; break;    case PCRE_NEWLINE_ANYCRLF: newline = -2; break;
6540    default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;    default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;
# Line 5993  cd->end_pattern = (const uschar *)(patte Line 6595  cd->end_pattern = (const uschar *)(patte
6595  cd->req_varyopt = 0;  cd->req_varyopt = 0;
6596  cd->external_options = options;  cd->external_options = options;
6597  cd->external_flags = 0;  cd->external_flags = 0;
6598    cd->open_caps = NULL;
6599    
6600  /* Now do the pre-compile. On error, errorcode will be set non-zero, so we  /* Now do the pre-compile. On error, errorcode will be set non-zero, so we
6601  don't need to look at the result of the function here. The initial options have  don't need to look at the result of the function here. The initial options have
# Line 6067  cd->start_code = codestart; Line 6670  cd->start_code = codestart;
6670  cd->hwm = cworkspace;  cd->hwm = cworkspace;
6671  cd->req_varyopt = 0;  cd->req_varyopt = 0;
6672  cd->had_accept = FALSE;  cd->had_accept = FALSE;
6673    cd->check_lookbehind = FALSE;
6674    cd->open_caps = NULL;
6675    
6676  /* Set up a starting, non-extracting bracket, then compile the expression. On  /* Set up a starting, non-extracting bracket, then compile the expression. On
6677  error, errorcode will be set non-zero, so we don't need to look at the result  error, errorcode will be set non-zero, so we don't need to look at the result
# Line 6092  if debugging, leave the test till after Line 6697  if debugging, leave the test till after
6697    
6698  *code++ = OP_END;  *code++ = OP_END;
6699    
6700  #ifndef DEBUG  #ifndef PCRE_DEBUG
6701  if (code - codestart > length) errorcode = ERR23;  if (code - codestart > length) errorcode = ERR23;
6702  #endif  #endif
6703    
# Line 6105  while (errorcode == 0 && cd->hwm > cwork Line 6710  while (errorcode == 0 && cd->hwm > cwork
6710    cd->hwm -= LINK_SIZE;    cd->hwm -= LINK_SIZE;
6711    offset = GET(cd->hwm, 0);    offset = GET(cd->hwm, 0);
6712    recno = GET(codestart, offset);    recno = GET(codestart, offset);
6713    groupptr = find_bracket(codestart, (re->options & PCRE_UTF8) != 0, recno);    groupptr = _pcre_find_bracket(codestart, utf8, recno);
6714    if (groupptr == NULL) errorcode = ERR53;    if (groupptr == NULL) errorcode = ERR53;
6715      else PUT(((uschar *)codestart), offset, groupptr - codestart);      else PUT(((uschar *)codestart), offset, groupptr - codestart);
6716    }    }
# Line 6115  subpattern. */ Line 6720  subpattern. */
6720    
6721  if (errorcode == 0 && re->top_backref > re->top_bracket) errorcode = ERR15;  if (errorcode == 0 && re->top_backref > re->top_bracket) errorcode = ERR15;
6722    
6723    /* If there were any lookbehind assertions that contained OP_RECURSE
6724    (recursions or subroutine calls), a flag is set for them to be checked here,
6725    because they may contain forward references. Actual recursions can't be fixed
6726    length, but subroutine calls can. It is done like this so that those without
6727    OP_RECURSE that are not fixed length get a diagnosic with a useful offset. The
6728    exceptional ones forgo this. We scan the pattern to check that they are fixed
6729    length, and set their lengths. */
6730    
6731    if (cd->check_lookbehind)
6732      {
6733      uschar *cc = (uschar *)codestart;
6734    
6735      /* Loop, searching for OP_REVERSE items, and process those that do not have
6736      their length set. (Actually, it will also re-process any that have a length
6737      of zero, but that is a pathological case, and it does no harm.) When we find
6738      one, we temporarily terminate the branch it is in while we scan it. */
6739    
6740      for (cc = (uschar *)_pcre_find_bracket(codestart, utf8, -1);
6741           cc != NULL;
6742           cc = (uschar *)_pcre_find_bracket(cc, utf8, -1))
6743        {
6744        if (GET(cc, 1) == 0)
6745          {
6746          int fixed_length;
6747          uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);
6748          int end_op = *be;
6749          *be = OP_END;
6750          fixed_length = find_fixedlength(cc, re->options, TRUE, cd);
6751          *be = end_op;
6752          DPRINTF(("fixed length = %d\n", fixed_length));
6753          if (fixed_length < 0)
6754            {
6755            errorcode = (fixed_length == -2)? ERR36 : ERR25;
6756            break;
6757            }
6758          PUT(cc, 1, fixed_length);
6759          }
6760        cc += 1 + LINK_SIZE;
6761        }
6762      }
6763    
6764  /* Failed to compile, or error while post-processing */  /* Failed to compile, or error while post-processing */
6765    
6766  if (errorcode != 0)  if (errorcode != 0)
# Line 6175  if (reqbyte >= 0 && Line 6821  if (reqbyte >= 0 &&
6821  /* Print out the compiled data if debugging is enabled. This is never the  /* Print out the compiled data if debugging is enabled. This is never the
6822  case when building a production library. */  case when building a production library. */
6823    
6824  #ifdef DEBUG  #ifdef PCRE_DEBUG
   
6825  printf("Length = %d top_bracket = %d top_backref = %d\n",  printf("Length = %d top_bracket = %d top_backref = %d\n",
6826    length, re->top_bracket, re->top_backref);    length, re->top_bracket, re->top_backref);
6827    
# Line 6213  if (code - codestart > length) Line 6858  if (code - codestart > length)
6858    if (errorcodeptr != NULL) *errorcodeptr = ERR23;    if (errorcodeptr != NULL) *errorcodeptr = ERR23;
6859    return NULL;    return NULL;
6860    }    }
6861  #endif   /* DEBUG */  #endif   /* PCRE_DEBUG */
6862    
6863  return (pcre *)re;  return (pcre *)re;
6864  }  }

Legend:
Removed from v.305  
changed lines
  Added in v.500

  ViewVC Help
Powered by ViewVC 1.1.5