/[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 334 by ph10, Fri Apr 11 15:48:14 2008 UTC revision 503 by ph10, Sun Mar 7 17:35:52 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 const verbitem verbs[] = {  static const verbitem verbs[] = {
200    { 6, OP_ACCEPT },    { 6, OP_ACCEPT },
# 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 302  static const char error_texts[] = Line 345  static const char error_texts[] =
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 one of a number of specific things:      /* \g must be followed by one of a number of specific things:
591    
592      (1) A number, either plain or braced. If positive, it is an absolute      (1) A number, either plain or braced. If positive, it is an absolute
593      backreference. If negative, it is a relative backreference. This is a Perl      backreference. If negative, it is a relative backreference. This is a Perl
594      5.10 feature.      5.10 feature.
595    
596      (2) Perl 5.10 also supports \g{name} as a reference to a named group. This      (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
597      is part of Perl's movement towards a unified syntax for back references. As      is part of Perl's movement towards a unified syntax for back references. As
598      this is synonymous with \k{name}, we fudge it up by pretending it really      this is synonymous with \k{name}, we fudge it up by pretending it really
599      was \k.      was \k.
600    
601      (3) For Oniguruma compatibility we also support \g followed by a name or a      (3) For Oniguruma compatibility we also support \g followed by a name or a
602      number either in angle brackets or in single quotes. However, these are      number either in angle brackets or in single quotes. However, these are
603      (possibly recursive) subroutine calls, _not_ backreferences. Just return      (possibly recursive) subroutine calls, _not_ backreferences. Just return
604      the -ESC_g code (cf \k). */      the -ESC_g code (cf \k). */
605    
606      case 'g':      case CHAR_g:
607      if (ptr[1] == '<' || ptr[1] == '\'')      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
608        {        {
609        c = -ESC_g;        c = -ESC_g;
610        break;        break;
611        }        }
612    
613      /* Handle the Perl-compatible cases */      /* Handle the Perl-compatible cases */
614    
615      if (ptr[1] == '{')      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
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 571  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 580  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)   /* Integer overflow */      if (c < 0)   /* Integer overflow */
642        {        {
643        *errorcodeptr = ERR61;        *errorcodeptr = ERR61;
644        break;        break;
645        }        }
646    
647      if (braced && *(++ptr) != '}')      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
648        {        {
649        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
650        break;        break;
651        }        }
652    
653      if (c == 0)      if (c == 0)
654        {        {
655        *errorcodeptr = ERR58;        *errorcodeptr = ERR58;
656        break;        break;
657        }        }
658    
659      if (negated)      if (negated)
660        {        {
# Line 625  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)    /* Integer overflow */        if (c < 0)    /* Integer overflow */
694          {          {
695          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
# Line 651  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 664  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 675  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 685  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 713  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 729  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 737  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 803  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 814  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 882  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 923  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 933  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 963  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 1031  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    
   /* We have to disambiguate (?<! and (?<= from (?<name> */  
1228    
   if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  
        *ptr != '\'')  
     continue;  
1229    
   count++;  
1230    
1231    if (name == NULL && count == lorn) return count;  /*************************************************
1232    term = *ptr++;  *       Find forward referenced subpattern       *
1233    if (term == '<') term = '>';  *************************************************/
1234    thisname = ptr;  
1235    while (*ptr != term) ptr++;  /* This function scans along a pattern's text looking for capturing
1236    if (name != NULL && lorn == ptr - thisname &&  subpatterns, and counting them. If it finds a named pattern that matches the
1237        strncmp((const char *)name, (const char *)thisname, lorn) == 0)  name it is given, it returns its number. Alternatively, if the name is NULL, it
1238      return count;  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    static int
1256    find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode)
1257    {
1258    uschar *ptr = (uschar *)cd->start_pattern;
1259    int count = 0;
1260    int rc;
1261    
1262    /* 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 1123  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 1139  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 1168  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 1175  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 1198  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 1211  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 1234  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 1248  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 1275  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 1329  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 1357  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 1423  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 1516  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 1539  Arguments: Line 1785  Arguments:
1785    code        points to start of search    code        points to start of search
1786    endcode     points to where to stop    endcode     points to where to stop
1787    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
1788      cd          contains pointers to tables etc.
1789    
1790  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
1791  */  */
1792    
1793  static BOOL  static BOOL
1794  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
1795      compile_data *cd)
1796  {  {
1797  register int c;  register int c;
1798  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);
# Line 1554  for (code = first_significant_code(code Line 1802  for (code = first_significant_code(code
1802    const uschar *ccode;    const uschar *ccode;
1803    
1804    c = *code;    c = *code;
1805    
1806    /* Skip over forward assertions; the other assertions are skipped by    /* Skip over forward assertions; the other assertions are skipped by
1807    first_significant_code() with a TRUE final argument. */    first_significant_code() with a TRUE final argument. */
1808    
# Line 1567  for (code = first_significant_code(code Line 1815  for (code = first_significant_code(code
1815    
1816    /* Groups with zero repeats can of course be empty; skip them. */    /* Groups with zero repeats can of course be empty; skip them. */
1817    
1818    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)
1819      {      {
1820      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1821      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
1822      c = *code;      c = *code;
1823      continue;      continue;
1824      }      }
1825    
1826      /* For a recursion/subroutine call, if its end has been reached, which
1827      implies a subroutine call, we can scan it. */
1828    
1829      if (c == OP_RECURSE)
1830        {
1831        const uschar *scode = cd->start_code + GET(code, 1);
1832        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
1833        do
1834          {
1835          if (!could_be_empty_branch(scode, endcode, utf8, cd)) return FALSE;
1836          scode += GET(scode, 1);
1837          }
1838        while (*scode == OP_ALT);
1839        continue;
1840        }
1841    
1842    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
1843    
# Line 1582  for (code = first_significant_code(code Line 1846  for (code = first_significant_code(code
1846      BOOL empty_branch;      BOOL empty_branch;
1847      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
1848    
1849      /* Scan a closed bracket */      /* If a conditional group has only one branch, there is a second, implied,
1850        empty branch, so just skip over the conditional, because it could be empty.
1851        Otherwise, scan the individual branches of the group. */
1852    
1853      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;  
1854        code += GET(code, 1);        code += GET(code, 1);
1855        else
1856          {
1857          empty_branch = FALSE;
1858          do
1859            {
1860            if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
1861              empty_branch = TRUE;
1862            code += GET(code, 1);
1863            }
1864          while (*code == OP_ALT);
1865          if (!empty_branch) return FALSE;   /* All branches are non-empty */
1866        }        }
1867      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
1868      c = *code;      c = *code;
1869      continue;      continue;
1870      }      }
# Line 1653  for (code = first_significant_code(code Line 1925  for (code = first_significant_code(code
1925      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1926      case OP_WORDCHAR:      case OP_WORDCHAR:
1927      case OP_ANY:      case OP_ANY:
1928        case OP_ALLANY:
1929      case OP_ANYBYTE:      case OP_ANYBYTE:
1930      case OP_CHAR:      case OP_CHAR:
1931      case OP_CHARNC:      case OP_CHARNC:
# Line 1709  for (code = first_significant_code(code Line 1982  for (code = first_significant_code(code
1982      case OP_QUERY:      case OP_QUERY:
1983      case OP_MINQUERY:      case OP_MINQUERY:
1984      case OP_POSQUERY:      case OP_POSQUERY:
1985        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
1986        break;
1987    
1988      case OP_UPTO:      case OP_UPTO:
1989      case OP_MINUPTO:      case OP_MINUPTO:
1990      case OP_POSUPTO:      case OP_POSUPTO:
1991      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
1992      break;      break;
1993  #endif  #endif
1994    
1995        /* None of the remaining opcodes are required to match a character. */
1996    
1997        default:
1998        break;
1999      }      }
2000    }    }
2001    
# Line 1737  Arguments: Line 2018  Arguments:
2018    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2019    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2020    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2021      cd          pointers to tables etc
2022    
2023  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2024  */  */
2025    
2026  static BOOL  static BOOL
2027  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2028    BOOL utf8)    BOOL utf8, compile_data *cd)
2029  {  {
2030  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2031    {    {
2032    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2033        return FALSE;
2034    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2035    }    }
2036  return TRUE;  return TRUE;
# Line 1793  int terminator;          /* Don't combin Line 2076  int terminator;          /* Don't combin
2076  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2077  for (++ptr; *ptr != 0; ptr++)  for (++ptr; *ptr != 0; ptr++)
2078    {    {
2079    if (*ptr == '\\' && ptr[1] == ']') ptr++; else    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else
2080      {      {
2081      if (*ptr == ']') return FALSE;      if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2082      if (*ptr == terminator && ptr[1] == ']')      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2083        {        {
2084        *endptr = ptr;        *endptr = ptr;
2085        return TRUE;        return TRUE;
# Line 1847  return -1; Line 2130  return -1;
2130  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2131  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2132  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
2133  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
2134  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
2135  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
2136  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
2137  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2138    OP_END.
2139    
2140  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2141  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 1882  while ((ptr = (uschar *)find_recurse(ptr Line 2166  while ((ptr = (uschar *)find_recurse(ptr
2166    
2167    /* See if this recursion is on the forward reference list. If so, adjust the    /* See if this recursion is on the forward reference list. If so, adjust the
2168    reference. */    reference. */
2169    
2170    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
2171      {      {
2172      offset = GET(hc, 0);      offset = GET(hc, 0);
# Line 1986  get_othercase_range(unsigned int *cptr, Line 2270  get_othercase_range(unsigned int *cptr,
2270  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2271    
2272  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2273    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2274    
2275  if (c > d) return FALSE;  if (c > d) return FALSE;
2276    
# Line 1995  next = othercase + 1; Line 2279  next = othercase + 1;
2279    
2280  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2281    {    {
2282    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2283    next++;    next++;
2284    }    }
2285    
# Line 2041  if ((options & PCRE_EXTENDED) != 0) Line 2325  if ((options & PCRE_EXTENDED) != 0)
2325    for (;;)    for (;;)
2326      {      {
2327      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2328      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2329        {        {
2330        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2331          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
# Line 2053  if ((options & PCRE_EXTENDED) != 0) Line 2337  if ((options & PCRE_EXTENDED) != 0)
2337  /* 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
2338  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2339    
2340  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2341    {    {
2342    int temperrorcode = 0;    int temperrorcode = 0;
2343    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
# Line 2078  if ((options & PCRE_EXTENDED) != 0) Line 2362  if ((options & PCRE_EXTENDED) != 0)
2362    for (;;)    for (;;)
2363      {      {
2364      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2365      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2366        {        {
2367        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2368          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
# Line 2089  if ((options & PCRE_EXTENDED) != 0) Line 2373  if ((options & PCRE_EXTENDED) != 0)
2373    
2374  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
2375    
2376  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2377    return FALSE;    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2378        return FALSE;
2379    
2380  /* 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
2381  positive single character match, "item" either contains the character or, if  positive single character match, "item" either contains the character or, if
# Line 2105  if (next >= 0) switch(op_code) Line 2390  if (next >= 0) switch(op_code)
2390    case OP_CHAR:    case OP_CHAR:
2391  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2392    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2393    #else
2394      (void)(utf8_char);  /* Keep compiler happy by referencing function argument */
2395  #endif  #endif
2396    return item != next;    return item != next;
2397    
# Line 2123  if (next >= 0) switch(op_code) Line 2410  if (next >= 0) switch(op_code)
2410      unsigned int othercase;      unsigned int othercase;
2411      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2412  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2413      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
2414  #else  #else
2415      othercase = NOTACHAR;      othercase = NOTACHAR;
2416  #endif  #endif
# Line 2144  if (next >= 0) switch(op_code) Line 2431  if (next >= 0) switch(op_code)
2431      unsigned int othercase;      unsigned int othercase;
2432      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2433  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2434      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
2435  #else  #else
2436      othercase = NOTACHAR;      othercase = NOTACHAR;
2437  #endif  #endif
# Line 2405  BOOL utf8 = FALSE; Line 2692  BOOL utf8 = FALSE;
2692  uschar *utf8_char = NULL;  uschar *utf8_char = NULL;
2693  #endif  #endif
2694    
2695  #ifdef DEBUG  #ifdef PCRE_DEBUG
2696  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2697  #endif  #endif
2698    
# Line 2458  for (;; ptr++) Line 2745  for (;; ptr++)
2745    /* Get next byte in the pattern */    /* Get next byte in the pattern */
2746    
2747    c = *ptr;    c = *ptr;
2748    
2749    /* If we are in the pre-compile phase, accumulate the length used for the    /* If we are in the pre-compile phase, accumulate the length used for the
2750    previous cycle of this loop. */    previous cycle of this loop. */
2751    
2752    if (lengthptr != NULL)    if (lengthptr != NULL)
2753      {      {
2754  #ifdef DEBUG  #ifdef PCRE_DEBUG
2755      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
2756  #endif  #endif
2757      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */
# Line 2526  for (;; ptr++) Line 2813  for (;; ptr++)
2813    
2814    if (inescq && c != 0)    if (inescq && c != 0)
2815      {      {
2816      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
2817        {        {
2818        inescq = FALSE;        inescq = FALSE;
2819        ptr++;        ptr++;
# Line 2552  for (;; ptr++) Line 2839  for (;; ptr++)
2839    /* 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
2840    a quantifier. */    a quantifier. */
2841    
2842    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
2843      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
2844        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
2845    
2846    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
2847         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2568  for (;; ptr++) Line 2856  for (;; ptr++)
2856    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
2857      {      {
2858      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
2859      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
2860        {        {
2861        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2862          {          {
# Line 2593  for (;; ptr++) Line 2881  for (;; ptr++)
2881      {      {
2882      /* ===================================================================*/      /* ===================================================================*/
2883      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
2884      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
2885      case ')':      case CHAR_RIGHT_PARENTHESIS:
2886      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
2887      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
2888      *codeptr = code;      *codeptr = code;
# Line 2616  for (;; ptr++) Line 2904  for (;; ptr++)
2904      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
2905      the setting of any following char as a first character. */      the setting of any following char as a first character. */
2906    
2907      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
2908      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
2909        {        {
2910        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
# Line 2625  for (;; ptr++) Line 2913  for (;; ptr++)
2913      *code++ = OP_CIRC;      *code++ = OP_CIRC;
2914      break;      break;
2915    
2916      case '$':      case CHAR_DOLLAR_SIGN:
2917      previous = NULL;      previous = NULL;
2918      *code++ = OP_DOLL;      *code++ = OP_DOLL;
2919      break;      break;
# Line 2633  for (;; ptr++) Line 2921  for (;; ptr++)
2921      /* 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
2922      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
2923    
2924      case '.':      case CHAR_DOT:
2925      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
2926      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
2927      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
2928      previous = code;      previous = code;
2929      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
2930      break;      break;
2931    
2932    
# Line 2653  for (;; ptr++) Line 2941  for (;; ptr++)
2941      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,
2942      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
2943      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.
     */  
2944    
2945      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
2946        default (Perl) mode, it is treated as a data character. */
2947    
2948        case CHAR_RIGHT_SQUARE_BRACKET:
2949        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2950          {
2951          *errorcodeptr = ERR64;
2952          goto FAILED;
2953          }
2954        goto NORMAL_CHAR;
2955    
2956        case CHAR_LEFT_SQUARE_BRACKET:
2957      previous = code;      previous = code;
2958    
2959      /* 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
2960      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. */
2961    
2962      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2963             ptr[1] == CHAR_EQUALS_SIGN) &&
2964          check_posix_syntax(ptr, &tempptr))          check_posix_syntax(ptr, &tempptr))
2965        {        {
2966        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
2967        goto FAILED;        goto FAILED;
2968        }        }
2969    
# Line 2676  for (;; ptr++) Line 2975  for (;; ptr++)
2975      for (;;)      for (;;)
2976        {        {
2977        c = *(++ptr);        c = *(++ptr);
2978        if (c == '\\')        if (c == CHAR_BACKSLASH)
2979          {          {
2980          if (ptr[1] == 'E') ptr++;          if (ptr[1] == CHAR_E)
2981            else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;            ptr++;
2982              else break;          else if (strncmp((const char *)ptr+1,
2983                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
2984              ptr += 3;
2985            else
2986              break;
2987          }          }
2988        else if (!negate_class && c == '^')        else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
2989          negate_class = TRUE;          negate_class = TRUE;
2990        else break;        else break;
2991        }        }
2992    
2993        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
2994        an initial ']' is taken as a data character -- the code below handles
2995        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
2996        [^] must match any character, so generate OP_ALLANY. */
2997    
2998        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
2999            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3000          {
3001          *code++ = negate_class? OP_ALLANY : OP_FAIL;
3002          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3003          zerofirstbyte = firstbyte;
3004          break;
3005          }
3006    
3007      /* 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
3008      negation flag at the end, so that support for characters > 255 works      negation flag at the end, so that support for characters > 255 works
3009      correctly (they are all included in the class). */      correctly (they are all included in the class). */
# Line 2744  for (;; ptr++) Line 3061  for (;; ptr++)
3061    
3062        if (inescq)        if (inescq)
3063          {          {
3064          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3065            {            {
3066            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
3067            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2759  for (;; ptr++) Line 3076  for (;; ptr++)
3076        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3077        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3078    
3079        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3080            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3081            check_posix_syntax(ptr, &tempptr))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3082          {          {
3083          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3084          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3085          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3086          uschar pbits[32];          uschar pbits[32];
3087    
3088          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3089            {            {
3090            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3091            goto FAILED;            goto FAILED;
3092            }            }
3093    
3094          ptr += 2;          ptr += 2;
3095          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3096            {            {
3097            local_negate = TRUE;            local_negate = TRUE;
3098            should_flip_negation = TRUE;  /* Note negative special */            should_flip_negation = TRUE;  /* Note negative special */
# Line 2848  for (;; ptr++) Line 3165  for (;; ptr++)
3165        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
3166        character in them, so set class_charcount bigger than one. */        character in them, so set class_charcount bigger than one. */
3167    
3168        if (c == '\\')        if (c == CHAR_BACKSLASH)
3169          {          {
3170          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3171          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3172    
3173          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 */
3174          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 */
3175          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 */
3176          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3177            {            {
3178            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3179              {              {
3180              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3181              }              }
# Line 3084  for (;; ptr++) Line 3401  for (;; ptr++)
3401        entirely. The code for handling \Q and \E is messy. */        entirely. The code for handling \Q and \E is messy. */
3402    
3403        CHECK_RANGE:        CHECK_RANGE:
3404        while (ptr[1] == '\\' && ptr[2] == 'E')        while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3405          {          {
3406          inescq = FALSE;          inescq = FALSE;
3407          ptr += 2;          ptr += 2;
# Line 3094  for (;; ptr++) Line 3411  for (;; ptr++)
3411    
3412        /* Remember \r or \n */        /* Remember \r or \n */
3413    
3414        if (c == '\r' || c == '\n') cd->external_flags |= PCRE_HASCRORLF;        if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3415    
3416        /* Check for range */        /* Check for range */
3417    
3418        if (!inescq && ptr[1] == '-')        if (!inescq && ptr[1] == CHAR_MINUS)
3419          {          {
3420          int d;          int d;
3421          ptr += 2;          ptr += 2;
3422          while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3423    
3424          /* 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
3425          mode. */          mode. */
3426    
3427          while (*ptr == '\\' && ptr[1] == 'Q')          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3428            {            {
3429            ptr += 2;            ptr += 2;
3430            if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }            if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3431                { ptr += 2; continue; }
3432            inescq = TRUE;            inescq = TRUE;
3433            break;            break;
3434            }            }
3435    
3436          if (*ptr == 0 || (!inescq && *ptr == ']'))          if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3437            {            {
3438            ptr = oldptr;            ptr = oldptr;
3439            goto LONE_SINGLE_CHARACTER;            goto LONE_SINGLE_CHARACTER;
# Line 3134  for (;; ptr++) Line 3452  for (;; ptr++)
3452          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
3453          in such circumstances. */          in such circumstances. */
3454    
3455          if (!inescq && d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3456            {            {
3457            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3458            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
# Line 3144  for (;; ptr++) Line 3462  for (;; ptr++)
3462    
3463            if (d < 0)            if (d < 0)
3464              {              {
3465              if (d == -ESC_b) d = '\b';              if (d == -ESC_b) d = CHAR_BS;
3466              else if (d == -ESC_X) d = 'X';              else if (d == -ESC_X) d = CHAR_X;
3467              else if (d == -ESC_R) d = 'R'; else              else if (d == -ESC_R) d = CHAR_R; else
3468                {                {
3469                ptr = oldptr;                ptr = oldptr;
3470                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 3167  for (;; ptr++) Line 3485  for (;; ptr++)
3485    
3486          /* Remember \r or \n */          /* Remember \r or \n */
3487    
3488          if (d == '\r' || d == '\n') cd->external_flags |= PCRE_HASCRORLF;          if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3489    
3490          /* 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
3491          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
# Line 3287  for (;; ptr++) Line 3605  for (;; ptr++)
3605          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3606            {            {
3607            unsigned int othercase;            unsigned int othercase;
3608            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)            if ((othercase = UCD_OTHERCASE(c)) != c)
3609              {              {
3610              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3611              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 3314  for (;; ptr++) Line 3632  for (;; ptr++)
3632    
3633      /* Loop until ']' reached. This "while" is the end of the "do" above. */      /* Loop until ']' reached. This "while" is the end of the "do" above. */
3634    
3635      while ((c = *(++ptr)) != 0 && (c != ']' || inescq));      while ((c = *(++ptr)) != 0 && (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
3636    
3637      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)                          /* Missing terminating ']' */
3638        {        {
# Line 3459  we set the flag only if there is a liter Line 3777  we set the flag only if there is a liter
3777      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
3778      has been tested above. */      has been tested above. */
3779    
3780      case '{':      case CHAR_LEFT_CURLY_BRACKET:
3781      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
3782      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
3783      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
3784      goto REPEAT;      goto REPEAT;
3785    
3786      case '*':      case CHAR_ASTERISK:
3787      repeat_min = 0;      repeat_min = 0;
3788      repeat_max = -1;      repeat_max = -1;
3789      goto REPEAT;      goto REPEAT;
3790    
3791      case '+':      case CHAR_PLUS:
3792      repeat_min = 1;      repeat_min = 1;
3793      repeat_max = -1;      repeat_max = -1;
3794      goto REPEAT;      goto REPEAT;
3795    
3796      case '?':      case CHAR_QUESTION_MARK:
3797      repeat_min = 0;      repeat_min = 0;
3798      repeat_max = 1;      repeat_max = 1;
3799    
# Line 3510  we set the flag only if there is a liter Line 3828  we set the flag only if there is a liter
3828      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
3829      repeat type to the non-default. */      repeat type to the non-default. */
3830    
3831      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
3832        {        {
3833        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
3834        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
3835        ptr++;        ptr++;
3836        }        }
3837      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
3838        {        {
3839        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
3840        ptr++;        ptr++;
# Line 3631  we set the flag only if there is a liter Line 3949  we set the flag only if there is a liter
3949    
3950        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
3951    
3952          /*--------------------------------------------------------------------*/
3953          /* This code is obsolete from release 8.00; the restriction was finally
3954          removed: */
3955    
3956        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
3957        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
3958    
3959        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
3960          /*--------------------------------------------------------------------*/
3961    
3962        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
3963    
# Line 3781  we set the flag only if there is a liter Line 4104  we set the flag only if there is a liter
4104          goto END_REPEAT;          goto END_REPEAT;
4105          }          }
4106    
4107          /*--------------------------------------------------------------------*/
4108          /* This code is obsolete from release 8.00; the restriction was finally
4109          removed: */
4110    
4111        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4112        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4113    
4114        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4115          /*--------------------------------------------------------------------*/
4116    
4117        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4118          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 3842  we set the flag only if there is a liter Line 4170  we set the flag only if there is a liter
4170    
4171        if (repeat_min == 0)        if (repeat_min == 0)
4172          {          {
4173          /* 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
4174          altogether. */          output altogether, like this:
4175    
4176          if (repeat_max == 0)          ** if (repeat_max == 0)
4177            {          **   {
4178            code = previous;          **   code = previous;
4179            goto END_REPEAT;          **   goto END_REPEAT;
4180            }          **   }
4181    
4182            However, that fails when a group is referenced as a subroutine from
4183            elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it
4184            so that it is skipped on execution. As we don't have a list of which
4185            groups are referenced, we cannot do this selectively.
4186    
4187            If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4188            and do no more at this point. However, we do need to adjust any
4189            OP_RECURSE calls inside the group that refer to the group itself or any
4190            internal or forward referenced group, because the offset is from the
4191            start of the whole regex. Temporarily terminate the pattern while doing
4192            this. */
4193    
4194          /* If the maximum is 1 or unlimited, we just have to stick in the          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
         BRAZERO and do no more at this point. However, we do need to adjust  
         any OP_RECURSE calls inside the group that refer to the group itself or  
         any internal or forward referenced group, because the offset is from  
         the start of the whole regex. Temporarily terminate the pattern while  
         doing this. */  
   
         if (repeat_max <= 1)  
4195            {            {
4196            *code = OP_END;            *code = OP_END;
4197            adjust_recurse(previous, 1, utf8, cd, save_hwm);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
4198            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
4199            code++;            code++;
4200              if (repeat_max == 0)
4201                {
4202                *previous++ = OP_SKIPZERO;
4203                goto END_REPEAT;
4204                }
4205            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4206            }            }
4207    
# Line 3909  we set the flag only if there is a liter Line 4247  we set the flag only if there is a liter
4247            {            {
4248            /* 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
4249            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
4250            potential integer overflow. */            potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
4251              integer type when available, otherwise double. */
4252    
4253            if (lengthptr != NULL)            if (lengthptr != NULL)
4254              {              {
4255              int delta = (repeat_min - 1)*length_prevgroup;              int delta = (repeat_min - 1)*length_prevgroup;
4256              if ((double)(repeat_min - 1)*(double)length_prevgroup >              if ((INT64_OR_DOUBLE)(repeat_min - 1)*
4257                                                              (double)INT_MAX ||                    (INT64_OR_DOUBLE)length_prevgroup >
4258                        (INT64_OR_DOUBLE)INT_MAX ||
4259                  OFLOW_MAX - *lengthptr < delta)                  OFLOW_MAX - *lengthptr < delta)
4260                {                {
4261                *errorcodeptr = ERR20;                *errorcodeptr = ERR20;
# Line 3961  we set the flag only if there is a liter Line 4301  we set the flag only if there is a liter
4301          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
4302          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
4303          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
4304          paranoid checks to avoid integer overflow. */          paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
4305            a 64-bit integer type when available, otherwise double. */
4306    
4307          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4308            {            {
4309            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4310                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4311            if ((double)repeat_max *            if ((INT64_OR_DOUBLE)repeat_max *
4312                  (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)                  (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4313                    > (double)INT_MAX ||                    > (INT64_OR_DOUBLE)INT_MAX ||
4314                OFLOW_MAX - *lengthptr < delta)                OFLOW_MAX - *lengthptr < delta)
4315              {              {
4316              *errorcodeptr = ERR20;              *errorcodeptr = ERR20;
# Line 4046  we set the flag only if there is a liter Line 4387  we set the flag only if there is a liter
4387            uschar *scode = bracode;            uschar *scode = bracode;
4388            do            do
4389              {              {
4390              if (could_be_empty_branch(scode, ketcode, utf8))              if (could_be_empty_branch(scode, ketcode, utf8, cd))
4391                {                {
4392                *bracode += OP_SBRA - OP_BRA;                *bracode += OP_SBRA - OP_BRA;
4393                break;                break;
# Line 4058  we set the flag only if there is a liter Line 4399  we set the flag only if there is a liter
4399          }          }
4400        }        }
4401    
4402        /* If previous is OP_FAIL, it was generated by an empty class [] in
4403        JavaScript mode. The other ways in which OP_FAIL can be generated, that is
4404        by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
4405        error above. We can just ignore the repeat in JS case. */
4406    
4407        else if (*previous == OP_FAIL) goto END_REPEAT;
4408    
4409      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
4410    
4411      else      else
# Line 4082  we set the flag only if there is a liter Line 4430  we set the flag only if there is a liter
4430      if (possessive_quantifier)      if (possessive_quantifier)
4431        {        {
4432        int len;        int len;
4433        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||  
4434            *tempcode == OP_NOTEXACT)        if (*tempcode == OP_TYPEEXACT)
4435          tempcode += _pcre_OP_lengths[*tempcode] +          tempcode += _pcre_OP_lengths[*tempcode] +
4436            ((*tempcode == OP_TYPEEXACT &&            ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
4437               (tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP))? 2:0);  
4438          else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
4439            {
4440            tempcode += _pcre_OP_lengths[*tempcode];
4441    #ifdef SUPPORT_UTF8
4442            if (utf8 && tempcode[-1] >= 0xc0)
4443              tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];
4444    #endif
4445            }
4446    
4447        len = code - tempcode;        len = code - tempcode;
4448        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4449          {          {
# Line 4105  we set the flag only if there is a liter Line 4462  we set the flag only if there is a liter
4462          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4463          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4464    
4465            /* Because we are moving code along, we must ensure that any
4466            pending recursive references are updated. */
4467    
4468          default:          default:
4469            *code = OP_END;
4470            adjust_recurse(tempcode, 1 + LINK_SIZE, utf8, cd, save_hwm);
4471          memmove(tempcode + 1+LINK_SIZE, tempcode, len);          memmove(tempcode + 1+LINK_SIZE, tempcode, len);
4472          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
4473          len += 1 + LINK_SIZE;          len += 1 + LINK_SIZE;
# Line 4132  we set the flag only if there is a liter Line 4494  we set the flag only if there is a liter
4494      lookbehind or option setting or condition or all the other extended      lookbehind or option setting or condition or all the other extended
4495      parenthesis forms.  */      parenthesis forms.  */
4496    
4497      case '(':      case CHAR_LEFT_PARENTHESIS:
4498      newoptions = options;      newoptions = options;
4499      skipbytes = 0;      skipbytes = 0;
4500      bravalue = OP_CBRA;      bravalue = OP_CBRA;
4501      save_hwm = cd->hwm;      save_hwm = cd->hwm;
4502      reset_bracount = FALSE;      reset_bracount = FALSE;
4503    
4504      /* First deal with various "verbs" that can be introduced by '*'. */      /* First deal with various "verbs" that can be introduced by '*'. */
4505    
4506      if (*(++ptr) == '*' && (cd->ctypes[ptr[1]] & ctype_letter) != 0)      if (*(++ptr) == CHAR_ASTERISK && (cd->ctypes[ptr[1]] & ctype_letter) != 0)
4507        {        {
4508        int i, namelen;        int i, namelen;
4509        const char *vn = verbnames;        const char *vn = verbnames;
4510        const uschar *name = ++ptr;        const uschar *name = ++ptr;
4511        previous = NULL;        previous = NULL;
4512        while ((cd->ctypes[*++ptr] & ctype_letter) != 0);        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
4513        if (*ptr == ':')        if (*ptr == CHAR_COLON)
4514          {          {
4515          *errorcodeptr = ERR59;   /* Not supported */          *errorcodeptr = ERR59;   /* Not supported */
4516          goto FAILED;          goto FAILED;
4517          }          }
4518        if (*ptr != ')')        if (*ptr != CHAR_RIGHT_PARENTHESIS)
4519          {          {
4520          *errorcodeptr = ERR60;          *errorcodeptr = ERR60;
4521          goto FAILED;          goto FAILED;
# Line 4164  we set the flag only if there is a liter Line 4526  we set the flag only if there is a liter
4526          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
4527              strncmp((char *)name, vn, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
4528            {            {
4529            *code = verbs[i].op;            /* Check for open captures before ACCEPT */
4530            if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;  
4531              if (verbs[i].op == OP_ACCEPT)
4532                {
4533                open_capitem *oc;
4534                cd->had_accept = TRUE;
4535                for (oc = cd->open_caps; oc != NULL; oc = oc->next)
4536                  {
4537                  *code++ = OP_CLOSE;
4538                  PUT2INC(code, 0, oc->number);
4539                  }
4540                }
4541              *code++ = verbs[i].op;
4542            break;            break;
4543            }            }
4544          vn += verbs[i].len + 1;          vn += verbs[i].len + 1;
# Line 4178  we set the flag only if there is a liter Line 4551  we set the flag only if there is a liter
4551      /* Deal with the extended parentheses; all are introduced by '?', and the      /* Deal with the extended parentheses; all are introduced by '?', and the
4552      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. */
4553    
4554      else if (*ptr == '?')      else if (*ptr == CHAR_QUESTION_MARK)
4555        {        {
4556        int i, set, unset, namelen;        int i, set, unset, namelen;
4557        int *optset;        int *optset;
# Line 4187  we set the flag only if there is a liter Line 4560  we set the flag only if there is a liter
4560    
4561        switch (*(++ptr))        switch (*(++ptr))
4562          {          {
4563          case '#':                 /* Comment; skip to ket */          case CHAR_NUMBER_SIGN:                 /* Comment; skip to ket */
4564          ptr++;          ptr++;
4565          while (*ptr != 0 && *ptr != ')') ptr++;          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
4566          if (*ptr == 0)          if (*ptr == 0)
4567            {            {
4568            *errorcodeptr = ERR18;            *errorcodeptr = ERR18;
# Line 4199  we set the flag only if there is a liter Line 4572  we set the flag only if there is a liter
4572    
4573    
4574          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4575          case '|':                 /* Reset capture count for each branch */          case CHAR_VERTICAL_LINE:  /* Reset capture count for each branch */
4576          reset_bracount = TRUE;          reset_bracount = TRUE;
4577          /* Fall through */          /* Fall through */
4578    
4579          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4580          case ':':                 /* Non-capturing bracket */          case CHAR_COLON:          /* Non-capturing bracket */
4581          bravalue = OP_BRA;          bravalue = OP_BRA;
4582          ptr++;          ptr++;
4583          break;          break;
4584    
4585    
4586          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4587          case '(':          case CHAR_LEFT_PARENTHESIS:
4588          bravalue = OP_COND;       /* Conditional group */          bravalue = OP_COND;       /* Conditional group */
4589    
4590          /* A condition can be an assertion, a number (referring to a numbered          /* A condition can be an assertion, a number (referring to a numbered
# Line 4231  we set the flag only if there is a liter Line 4604  we set the flag only if there is a liter
4604          the switch. This will take control down to where bracketed groups,          the switch. This will take control down to where bracketed groups,
4605          including assertions, are processed. */          including assertions, are processed. */
4606    
4607          if (ptr[1] == '?' && (ptr[2] == '=' || ptr[2] == '!' || ptr[2] == '<'))          if (ptr[1] == CHAR_QUESTION_MARK && (ptr[2] == CHAR_EQUALS_SIGN ||
4608                ptr[2] == CHAR_EXCLAMATION_MARK || ptr[2] == CHAR_LESS_THAN_SIGN))
4609            break;            break;
4610    
4611          /* Most other conditions use OP_CREF (a couple change to OP_RREF          /* Most other conditions use OP_CREF (a couple change to OP_RREF
# Line 4243  we set the flag only if there is a liter Line 4617  we set the flag only if there is a liter
4617    
4618          /* Check for a test for recursion in a named group. */          /* Check for a test for recursion in a named group. */
4619    
4620          if (ptr[1] == 'R' && ptr[2] == '&')          if (ptr[1] == CHAR_R && ptr[2] == CHAR_AMPERSAND)
4621            {            {
4622            terminator = -1;            terminator = -1;
4623            ptr += 2;            ptr += 2;
# Line 4253  we set the flag only if there is a liter Line 4627  we set the flag only if there is a liter
4627          /* 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
4628          syntax (?(<name>) or (?('name') */          syntax (?(<name>) or (?('name') */
4629    
4630          else if (ptr[1] == '<')          else if (ptr[1] == CHAR_LESS_THAN_SIGN)
4631            {            {
4632            terminator = '>';            terminator = CHAR_GREATER_THAN_SIGN;
4633            ptr++;            ptr++;
4634            }            }
4635          else if (ptr[1] == '\'')          else if (ptr[1] == CHAR_APOSTROPHE)
4636            {            {
4637            terminator = '\'';            terminator = CHAR_APOSTROPHE;
4638            ptr++;            ptr++;
4639            }            }
4640          else          else
4641            {            {
4642            terminator = 0;            terminator = 0;
4643            if (ptr[1] == '-' || ptr[1] == '+') refsign = *(++ptr);            if (ptr[1] == CHAR_MINUS || ptr[1] == CHAR_PLUS) refsign = *(++ptr);
4644            }            }
4645    
4646          /* We now expect to read a name; any thing else is an error */          /* We now expect to read a name; any thing else is an error */
# Line 4286  we set the flag only if there is a liter Line 4660  we set the flag only if there is a liter
4660            {            {
4661            if (recno >= 0)            if (recno >= 0)
4662              recno = ((digitab[*ptr] & ctype_digit) != 0)?              recno = ((digitab[*ptr] & ctype_digit) != 0)?
4663                recno * 10 + *ptr - '0' : -1;                recno * 10 + *ptr - CHAR_0 : -1;
4664            ptr++;            ptr++;
4665            }            }
4666          namelen = ptr - name;          namelen = ptr - name;
4667    
4668          if ((terminator > 0 && *ptr++ != terminator) || *ptr++ != ')')          if ((terminator > 0 && *ptr++ != terminator) ||
4669                *ptr++ != CHAR_RIGHT_PARENTHESIS)
4670            {            {
4671            ptr--;      /* Error offset */            ptr--;      /* Error offset */
4672            *errorcodeptr = ERR26;            *errorcodeptr = ERR26;
# Line 4313  we set the flag only if there is a liter Line 4688  we set the flag only if there is a liter
4688              *errorcodeptr = ERR58;              *errorcodeptr = ERR58;
4689              goto FAILED;              goto FAILED;
4690              }              }
4691            recno = (refsign == '-')?            recno = (refsign == CHAR_MINUS)?
4692              cd->bracount - recno + 1 : recno +cd->bracount;              cd->bracount - recno + 1 : recno +cd->bracount;
4693            if (recno <= 0 || recno > cd->final_bracount)            if (recno <= 0 || recno > cd->final_bracount)
4694              {              {
# Line 4325  we set the flag only if there is a liter Line 4700  we set the flag only if there is a liter
4700            }            }
4701    
4702          /* Otherwise (did not start with "+" or "-"), start by looking for the          /* Otherwise (did not start with "+" or "-"), start by looking for the
4703          name. */          name. If we find a name, add one to the opcode to change OP_CREF or
4704            OP_RREF into OP_NCREF or OP_NRREF. These behave exactly the same,
4705            except they record that the reference was originally to a name. The
4706            information is used to check duplicate names. */
4707    
4708          slot = cd->name_table;          slot = cd->name_table;
4709          for (i = 0; i < cd->names_found; i++)          for (i = 0; i < cd->names_found; i++)
# Line 4340  we set the flag only if there is a liter Line 4718  we set the flag only if there is a liter
4718            {            {
4719            recno = GET2(slot, 0);            recno = GET2(slot, 0);
4720            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4721              code[1+LINK_SIZE]++;
4722            }            }
4723    
4724          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
4725    
4726          else if ((i = find_parens(ptr, cd->bracount, name, namelen,          else if ((i = find_parens(cd, name, namelen,
4727                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0)) > 0)
4728            {            {
4729            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
4730              code[1+LINK_SIZE]++;
4731            }            }
4732    
4733          /* If terminator == 0 it means that the name followed directly after          /* If terminator == 0 it means that the name followed directly after
# Line 4365  we set the flag only if there is a liter Line 4745  we set the flag only if there is a liter
4745          /* Check for (?(R) for recursion. Allow digits after R to specify a          /* Check for (?(R) for recursion. Allow digits after R to specify a
4746          specific group number. */          specific group number. */
4747    
4748          else if (*name == 'R')          else if (*name == CHAR_R)
4749            {            {
4750            recno = 0;            recno = 0;
4751            for (i = 1; i < namelen; i++)            for (i = 1; i < namelen; i++)
# Line 4375  we set the flag only if there is a liter Line 4755  we set the flag only if there is a liter
4755                *errorcodeptr = ERR15;                *errorcodeptr = ERR15;
4756                goto FAILED;                goto FAILED;
4757                }                }
4758              recno = recno * 10 + name[i] - '0';              recno = recno * 10 + name[i] - CHAR_0;
4759              }              }
4760            if (recno == 0) recno = RREF_ANY;            if (recno == 0) recno = RREF_ANY;
4761            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */
# Line 4385  we set the flag only if there is a liter Line 4765  we set the flag only if there is a liter
4765          /* Similarly, check for the (?(DEFINE) "condition", which is always          /* Similarly, check for the (?(DEFINE) "condition", which is always
4766          false. */          false. */
4767    
4768          else if (namelen == 6 && strncmp((char *)name, "DEFINE", 6) == 0)          else if (namelen == 6 && strncmp((char *)name, STRING_DEFINE, 6) == 0)
4769            {            {
4770            code[1+LINK_SIZE] = OP_DEF;            code[1+LINK_SIZE] = OP_DEF;
4771            skipbytes = 1;            skipbytes = 1;
# Line 4410  we set the flag only if there is a liter Line 4790  we set the flag only if there is a liter
4790    
4791    
4792          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4793          case '=':                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
4794          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
4795          ptr++;          ptr++;
4796          break;          break;
4797    
4798    
4799          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4800          case '!':                 /* Negative lookahead */          case CHAR_EXCLAMATION_MARK:            /* Negative lookahead */
4801          ptr++;          ptr++;
4802          if (*ptr == ')')          /* Optimize (?!) */          if (*ptr == CHAR_RIGHT_PARENTHESIS)    /* Optimize (?!) */
4803            {            {
4804            *code++ = OP_FAIL;            *code++ = OP_FAIL;
4805            previous = NULL;            previous = NULL;
# Line 4430  we set the flag only if there is a liter Line 4810  we set the flag only if there is a liter
4810    
4811    
4812          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4813          case '<':                 /* Lookbehind or named define */          case CHAR_LESS_THAN_SIGN:              /* Lookbehind or named define */
4814          switch (ptr[1])          switch (ptr[1])
4815            {            {
4816            case '=':               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
4817            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
4818            ptr += 2;            ptr += 2;
4819            break;            break;
4820    
4821            case '!':               /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
4822            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
4823            ptr += 2;            ptr += 2;
4824            break;            break;
# Line 4453  we set the flag only if there is a liter Line 4833  we set the flag only if there is a liter
4833    
4834    
4835          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4836          case '>':                 /* One-time brackets */          case CHAR_GREATER_THAN_SIGN:           /* One-time brackets */
4837          bravalue = OP_ONCE;          bravalue = OP_ONCE;
4838          ptr++;          ptr++;
4839          break;          break;
4840    
4841    
4842          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4843          case 'C':                 /* Callout - may be followed by digits; */          case CHAR_C:                 /* Callout - may be followed by digits; */
4844          previous_callout = code;  /* Save for later completion */          previous_callout = code;  /* Save for later completion */
4845          after_manual_callout = 1; /* Skip one item before completing */          after_manual_callout = 1; /* Skip one item before completing */
4846          *code++ = OP_CALLOUT;          *code++ = OP_CALLOUT;
4847            {            {
4848            int n = 0;            int n = 0;
4849            while ((digitab[*(++ptr)] & ctype_digit) != 0)            while ((digitab[*(++ptr)] & ctype_digit) != 0)
4850              n = n * 10 + *ptr - '0';              n = n * 10 + *ptr - CHAR_0;
4851            if (*ptr != ')')            if (*ptr != CHAR_RIGHT_PARENTHESIS)
4852              {              {
4853              *errorcodeptr = ERR39;              *errorcodeptr = ERR39;
4854              goto FAILED;              goto FAILED;
# Line 4488  we set the flag only if there is a liter Line 4868  we set the flag only if there is a liter
4868    
4869    
4870          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4871          case 'P':                 /* Python-style named subpattern handling */          case CHAR_P:              /* Python-style named subpattern handling */
4872          if (*(++ptr) == '=' || *ptr == '>')  /* Reference or recursion */          if (*(++ptr) == CHAR_EQUALS_SIGN ||
4873                *ptr == CHAR_GREATER_THAN_SIGN)  /* Reference or recursion */
4874            {            {
4875            is_recurse = *ptr == '>';            is_recurse = *ptr == CHAR_GREATER_THAN_SIGN;
4876            terminator = ')';            terminator = CHAR_RIGHT_PARENTHESIS;
4877            goto NAMED_REF_OR_RECURSE;            goto NAMED_REF_OR_RECURSE;
4878            }            }
4879          else if (*ptr != '<')    /* Test for Python-style definition */          else if (*ptr != CHAR_LESS_THAN_SIGN)  /* Test for Python-style defn */
4880            {            {
4881            *errorcodeptr = ERR41;            *errorcodeptr = ERR41;
4882            goto FAILED;            goto FAILED;
# Line 4505  we set the flag only if there is a liter Line 4886  we set the flag only if there is a liter
4886    
4887          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4888          DEFINE_NAME:    /* Come here from (?< handling */          DEFINE_NAME:    /* Come here from (?< handling */
4889          case '\'':          case CHAR_APOSTROPHE:
4890            {            {
4891            terminator = (*ptr == '<')? '>' : '\'';            terminator = (*ptr == CHAR_LESS_THAN_SIGN)?
4892                CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
4893            name = ++ptr;            name = ++ptr;
4894    
4895            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
# Line 4538  we set the flag only if there is a liter Line 4920  we set the flag only if there is a liter
4920                }                }
4921              }              }
4922    
4923            /* In the real compile, create the entry in the table */            /* In the real compile, create the entry in the table, maintaining
4924              alphabetical order. Duplicate names for different numbers are
4925              permitted only if PCRE_DUPNAMES is set. Duplicate names for the same
4926              number are always OK. (An existing number can be re-used if (?|
4927              appears in the pattern.) In either event, a duplicate name results in
4928              a duplicate entry in the table, even if the number is the same. This
4929              is because the number of names, and hence the table size, is computed
4930              in the pre-compile, and it affects various numbers and pointers which
4931              would all have to be modified, and the compiled code moved down, if
4932              duplicates with the same number were omitted from the table. This
4933              doesn't seem worth the hassle. However, *different* names for the
4934              same number are not permitted. */
4935    
4936            else            else
4937              {              {
4938                BOOL dupname = FALSE;
4939              slot = cd->name_table;              slot = cd->name_table;
4940    
4941              for (i = 0; i < cd->names_found; i++)              for (i = 0; i < cd->names_found; i++)
4942                {                {
4943                int crc = memcmp(name, slot+2, namelen);                int crc = memcmp(name, slot+2, namelen);
# Line 4550  we set the flag only if there is a liter Line 4945  we set the flag only if there is a liter
4945                  {                  {
4946                  if (slot[2+namelen] == 0)                  if (slot[2+namelen] == 0)
4947                    {                    {
4948                    if ((options & PCRE_DUPNAMES) == 0)                    if (GET2(slot, 0) != cd->bracount + 1 &&
4949                          (options & PCRE_DUPNAMES) == 0)
4950                      {                      {
4951                      *errorcodeptr = ERR43;                      *errorcodeptr = ERR43;
4952                      goto FAILED;                      goto FAILED;
4953                      }                      }
4954                      else dupname = TRUE;
4955                    }                    }
4956                  else crc = -1;      /* Current name is substring */                  else crc = -1;      /* Current name is a substring */
4957                  }                  }
4958    
4959                  /* Make space in the table and break the loop for an earlier
4960                  name. For a duplicate or later name, carry on. We do this for
4961                  duplicates so that in the simple case (when ?(| is not used) they
4962                  are in order of their numbers. */
4963    
4964                if (crc < 0)                if (crc < 0)
4965                  {                  {
4966                  memmove(slot + cd->name_entry_size, slot,                  memmove(slot + cd->name_entry_size, slot,
4967                    (cd->names_found - i) * cd->name_entry_size);                    (cd->names_found - i) * cd->name_entry_size);
4968                  break;                  break;
4969                  }                  }
4970    
4971                  /* Continue the loop for a later or duplicate name */
4972    
4973                slot += cd->name_entry_size;                slot += cd->name_entry_size;
4974                }                }
4975    
4976                /* For non-duplicate names, check for a duplicate number before
4977                adding the new name. */
4978    
4979                if (!dupname)
4980                  {
4981                  uschar *cslot = cd->name_table;
4982                  for (i = 0; i < cd->names_found; i++)
4983                    {
4984                    if (cslot != slot)
4985                      {
4986                      if (GET2(cslot, 0) == cd->bracount + 1)
4987                        {
4988                        *errorcodeptr = ERR65;
4989                        goto FAILED;
4990                        }
4991                      }
4992                    else i--;
4993                    cslot += cd->name_entry_size;
4994                    }
4995                  }
4996    
4997              PUT2(slot, 0, cd->bracount + 1);              PUT2(slot, 0, cd->bracount + 1);
4998              memcpy(slot + 2, name, namelen);              memcpy(slot + 2, name, namelen);
4999              slot[2+namelen] = 0;              slot[2+namelen] = 0;
5000              }              }
5001            }            }
5002    
5003          /* In both cases, count the number of names we've encountered. */          /* In both pre-compile and compile, count the number of names we've
5004            encountered. */
5005    
         ptr++;                    /* Move past > or ' */  
5006          cd->names_found++;          cd->names_found++;
5007            ptr++;                    /* Move past > or ' */
5008          goto NUMBERED_GROUP;          goto NUMBERED_GROUP;
5009    
5010    
5011          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5012          case '&':                 /* Perl recursion/subroutine syntax */          case CHAR_AMPERSAND:            /* Perl recursion/subroutine syntax */
5013          terminator = ')';          terminator = CHAR_RIGHT_PARENTHESIS;
5014          is_recurse = TRUE;          is_recurse = TRUE;
5015          /* Fall through */          /* Fall through */
5016    
# Line 4641  we set the flag only if there is a liter Line 5069  we set the flag only if there is a liter
5069              recno = GET2(slot, 0);              recno = GET2(slot, 0);
5070              }              }
5071            else if ((recno =                /* Forward back reference */            else if ((recno =                /* Forward back reference */
5072                      find_parens(ptr, cd->bracount, name, namelen,                      find_parens(cd, name, namelen,
5073                        (options & PCRE_EXTENDED) != 0)) <= 0)                        (options & PCRE_EXTENDED) != 0)) <= 0)
5074              {              {
5075              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
# Line 4657  we set the flag only if there is a liter Line 5085  we set the flag only if there is a liter
5085    
5086    
5087          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5088          case 'R':                 /* Recursion */          case CHAR_R:              /* Recursion */
5089          ptr++;                    /* Same as (?0)      */          ptr++;                    /* Same as (?0)      */
5090          /* Fall through */          /* Fall through */
5091    
5092    
5093          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5094          case '-': case '+':          case CHAR_MINUS: case CHAR_PLUS:  /* Recursion or subroutine */
5095          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:
5096          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:
5097            {            {
5098            const uschar *called;            const uschar *called;
5099            terminator = ')';            terminator = CHAR_RIGHT_PARENTHESIS;
5100    
5101            /* Come here from the \g<...> and \g'...' code (Oniguruma            /* Come here from the \g<...> and \g'...' code (Oniguruma
5102            compatibility). However, the syntax has been checked to ensure that            compatibility). However, the syntax has been checked to ensure that
5103            the ... are a (signed) number, so that neither ERR63 nor ERR29 will            the ... are a (signed) number, so that neither ERR63 nor ERR29 will
5104            be called on this path, nor with the jump to OTHER_CHAR_AFTER_QUERY            be called on this path, nor with the jump to OTHER_CHAR_AFTER_QUERY
5105            ever be taken. */            ever be taken. */
   
           HANDLE_NUMERICAL_RECURSION:  
5106    
5107            if ((refsign = *ptr) == '+')            HANDLE_NUMERICAL_RECURSION:
5108    
5109              if ((refsign = *ptr) == CHAR_PLUS)
5110              {              {
5111              ptr++;              ptr++;
5112              if ((digitab[*ptr] & ctype_digit) == 0)              if ((digitab[*ptr] & ctype_digit) == 0)
# Line 4687  we set the flag only if there is a liter Line 5115  we set the flag only if there is a liter
5115                goto FAILED;                goto FAILED;
5116                }                }
5117              }              }
5118            else if (refsign == '-')            else if (refsign == CHAR_MINUS)
5119              {              {
5120              if ((digitab[ptr[1]] & ctype_digit) == 0)              if ((digitab[ptr[1]] & ctype_digit) == 0)
5121                goto OTHER_CHAR_AFTER_QUERY;                goto OTHER_CHAR_AFTER_QUERY;
# Line 4696  we set the flag only if there is a liter Line 5124  we set the flag only if there is a liter
5124    
5125            recno = 0;            recno = 0;
5126            while((digitab[*ptr] & ctype_digit) != 0)            while((digitab[*ptr] & ctype_digit) != 0)
5127              recno = recno * 10 + *ptr++ - '0';              recno = recno * 10 + *ptr++ - CHAR_0;
5128    
5129            if (*ptr != terminator)            if (*ptr != terminator)
5130              {              {
# Line 4704  we set the flag only if there is a liter Line 5132  we set the flag only if there is a liter
5132              goto FAILED;              goto FAILED;
5133              }              }
5134    
5135            if (refsign == '-')            if (refsign == CHAR_MINUS)
5136              {              {
5137              if (recno == 0)              if (recno == 0)
5138                {                {
# Line 4718  we set the flag only if there is a liter Line 5146  we set the flag only if there is a liter
5146                goto FAILED;                goto FAILED;
5147                }                }
5148              }              }
5149            else if (refsign == '+')            else if (refsign == CHAR_PLUS)
5150              {              {
5151              if (recno == 0)              if (recno == 0)
5152                {                {
# Line 4745  we set the flag only if there is a liter Line 5173  we set the flag only if there is a liter
5173            if (lengthptr == NULL)            if (lengthptr == NULL)
5174              {              {
5175              *code = OP_END;              *code = OP_END;
5176              if (recno != 0) called = find_bracket(cd->start_code, utf8, recno);              if (recno != 0)
5177                  called = _pcre_find_bracket(cd->start_code, utf8, recno);
5178    
5179              /* Forward reference */              /* Forward reference */
5180    
5181              if (called == NULL)              if (called == NULL)
5182                {                {
5183                if (find_parens(ptr, cd->bracount, NULL, recno,                if (find_parens(cd, NULL, recno,
5184                     (options & PCRE_EXTENDED) != 0) < 0)                      (options & PCRE_EXTENDED) != 0) < 0)
5185                  {                  {
5186                  *errorcodeptr = ERR15;                  *errorcodeptr = ERR15;
5187                  goto FAILED;                  goto FAILED;
5188                  }                  }
5189    
5190                  /* Fudge the value of "called" so that when it is inserted as an
5191                  offset below, what it actually inserted is the reference number
5192                  of the group. */
5193    
5194                called = cd->start_code + recno;                called = cd->start_code + recno;
5195                PUTINC(cd->hwm, 0, code + 2 + LINK_SIZE - cd->start_code);                PUTINC(cd->hwm, 0, code + 2 + LINK_SIZE - cd->start_code);
5196                }                }
# Line 4766  we set the flag only if there is a liter Line 5200  we set the flag only if there is a liter
5200              recursion that could loop for ever, and diagnose that case. */              recursion that could loop for ever, and diagnose that case. */
5201    
5202              else if (GET(called, 1) == 0 &&              else if (GET(called, 1) == 0 &&
5203                       could_be_empty(called, code, bcptr, utf8))                       could_be_empty(called, code, bcptr, utf8, cd))
5204                {                {
5205                *errorcodeptr = ERR40;                *errorcodeptr = ERR40;
5206                goto FAILED;                goto FAILED;
# Line 4804  we set the flag only if there is a liter Line 5238  we set the flag only if there is a liter
5238          set = unset = 0;          set = unset = 0;
5239          optset = &set;          optset = &set;
5240    
5241          while (*ptr != ')' && *ptr != ':')          while (*ptr != CHAR_RIGHT_PARENTHESIS && *ptr != CHAR_COLON)
5242            {            {
5243            switch (*ptr++)            switch (*ptr++)
5244              {              {
5245              case '-': optset = &unset; break;              case CHAR_MINUS: optset = &unset; break;
5246    
5247              case 'J':    /* Record that it changed in the external options */              case CHAR_J:    /* Record that it changed in the external options */
5248              *optset |= PCRE_DUPNAMES;              *optset |= PCRE_DUPNAMES;
5249              cd->external_flags |= PCRE_JCHANGED;              cd->external_flags |= PCRE_JCHANGED;
5250              break;              break;
5251    
5252              case 'i': *optset |= PCRE_CASELESS; break;              case CHAR_i: *optset |= PCRE_CASELESS; break;
5253              case 'm': *optset |= PCRE_MULTILINE; break;              case CHAR_m: *optset |= PCRE_MULTILINE; break;
5254              case 's': *optset |= PCRE_DOTALL; break;              case CHAR_s: *optset |= PCRE_DOTALL; break;
5255              case 'x': *optset |= PCRE_EXTENDED; break;              case CHAR_x: *optset |= PCRE_EXTENDED; break;
5256              case 'U': *optset |= PCRE_UNGREEDY; break;              case CHAR_U: *optset |= PCRE_UNGREEDY; break;
5257              case 'X': *optset |= PCRE_EXTRA; break;              case CHAR_X: *optset |= PCRE_EXTRA; break;
5258    
5259              default:  *errorcodeptr = ERR12;              default:  *errorcodeptr = ERR12;
5260                        ptr--;    /* Correct the offset */                        ptr--;    /* Correct the offset */
# Line 4851  we set the flag only if there is a liter Line 5285  we set the flag only if there is a liter
5285          both phases.          both phases.
5286    
5287          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
5288          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
5289          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. */  
5290    
5291          if (*ptr == ')')          if (*ptr == CHAR_RIGHT_PARENTHESIS)
5292            {            {
5293            if (code == cd->start_code + 1 + LINK_SIZE &&            if (code == cd->start_code + 1 + LINK_SIZE &&
5294                 (lengthptr == NULL || *lengthptr == 2 + 2*LINK_SIZE))                 (lengthptr == NULL || *lengthptr == 2 + 2*LINK_SIZE))
5295              {              {
5296              cd->external_options = newoptions;              cd->external_options = newoptions;
             options = newoptions;  
5297              }              }
5298           else            else
5299              {              {
5300              if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))              if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))
5301                {                {
5302                *code++ = OP_OPT;                *code++ = OP_OPT;
5303                *code++ = newoptions & PCRE_IMS;                *code++ = newoptions & PCRE_IMS;
5304                }                }
   
             /* 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;  
5305              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);
5306              greedy_non_default = greedy_default ^ 1;              greedy_non_default = greedy_default ^ 1;
5307              req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
5308              }              }
5309    
5310              /* Change options at this level, and pass them back for use
5311              in subsequent branches. When not at the start of the pattern, this
5312              information is also necessary so that a resetting item can be
5313              compiled at the end of a group (if we are in a group). */
5314    
5315              *optionsptr = options = newoptions;
5316            previous = NULL;       /* This item can't be repeated */            previous = NULL;       /* This item can't be repeated */
5317            continue;              /* It is complete */            continue;              /* It is complete */
5318            }            }
# Line 4997  we set the flag only if there is a liter Line 5428  we set the flag only if there is a liter
5428    
5429      /* Error if hit end of pattern */      /* Error if hit end of pattern */
5430    
5431      if (*ptr != ')')      if (*ptr != CHAR_RIGHT_PARENTHESIS)
5432        {        {
5433        *errorcodeptr = ERR14;        *errorcodeptr = ERR14;
5434        goto FAILED;        goto FAILED;
# Line 5094  we set the flag only if there is a liter Line 5525  we set the flag only if there is a liter
5525      back references and those types that consume a character may be repeated.      back references and those types that consume a character may be repeated.
5526      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
5527      have to change if any new ones are ever created. */      have to change if any new ones are ever created. */
5528    
5529      case '\\':      case CHAR_BACKSLASH:
5530      tempptr = ptr;      tempptr = ptr;
5531      c = check_escape(&ptr, errorcodeptr, cd->bracount, options, FALSE);      c = check_escape(&ptr, errorcodeptr, cd->bracount, options, FALSE);
5532      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
# Line 5104  we set the flag only if there is a liter Line 5535  we set the flag only if there is a liter
5535        {        {
5536        if (-c == ESC_Q)            /* Handle start of quoted string */        if (-c == ESC_Q)            /* Handle start of quoted string */
5537          {          {
5538          if (ptr[1] == '\\' && ptr[2] == 'E') ptr += 2; /* avoid empty string */          if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
5539            else inescq = TRUE;            ptr += 2;               /* avoid empty string */
5540                else inescq = TRUE;
5541          continue;          continue;
5542          }          }
5543    
# Line 5121  we set the flag only if there is a liter Line 5553  we set the flag only if there is a liter
5553    
5554        zerofirstbyte = firstbyte;        zerofirstbyte = firstbyte;
5555        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
5556    
5557        /* \g<name> or \g'name' is a subroutine call by name and \g<n> or \g'n'        /* \g<name> or \g'name' is a subroutine call by name and \g<n> or \g'n'
5558        is a subroutine call by number (Oniguruma syntax). In fact, the value        is a subroutine call by number (Oniguruma syntax). In fact, the value
5559        -ESC_g is returned only for these cases. So we don't need to check for <        -ESC_g is returned only for these cases. So we don't need to check for <
5560        or ' if the value is -ESC_g. For the Perl syntax \g{n} the value is        or ' if the value is -ESC_g. For the Perl syntax \g{n} the value is
5561        -ESC_REF+n, and for the Perl syntax \g{name} the result is -ESC_k (as        -ESC_REF+n, and for the Perl syntax \g{name} the result is -ESC_k (as
5562        that is a synonym for a named back reference). */        that is a synonym for a named back reference). */
5563    
5564        if (-c == ESC_g)        if (-c == ESC_g)
5565          {          {
5566          const uschar *p;          const uschar *p;
5567          save_hwm = cd->hwm;   /* Normally this is set when '(' is read */          save_hwm = cd->hwm;   /* Normally this is set when '(' is read */
5568          terminator = (*(++ptr) == '<')? '>' : '\'';          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
5569              CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
5570    
5571          /* These two statements stop the compiler for warning about possibly          /* These two statements stop the compiler for warning about possibly
5572          unset variables caused by the jump to HANDLE_NUMERICAL_RECURSION. In          unset variables caused by the jump to HANDLE_NUMERICAL_RECURSION. In
5573          fact, because we actually check for a number below, the paths that          fact, because we actually check for a number below, the paths that
5574          would actually be in error are never taken. */          would actually be in error are never taken. */
5575    
5576          skipbytes = 0;          skipbytes = 0;
5577          reset_bracount = FALSE;          reset_bracount = FALSE;
5578    
5579          /* Test for a name */          /* Test for a name */
5580    
5581          if (ptr[1] != '+' && ptr[1] != '-')          if (ptr[1] != CHAR_PLUS && ptr[1] != CHAR_MINUS)
5582            {            {
5583            BOOL isnumber = TRUE;            BOOL isnumber = TRUE;
5584            for (p = ptr + 1; *p != 0 && *p != terminator; p++)            for (p = ptr + 1; *p != 0 && *p != terminator; p++)
5585              {              {
5586              if ((cd->ctypes[*p] & ctype_digit) == 0) isnumber = FALSE;              if ((cd->ctypes[*p] & ctype_digit) == 0) isnumber = FALSE;
5587              if ((cd->ctypes[*p] & ctype_word) == 0) break;              if ((cd->ctypes[*p] & ctype_word) == 0) break;
5588              }              }
5589            if (*p != terminator)            if (*p != terminator)
5590              {              {
5591              *errorcodeptr = ERR57;              *errorcodeptr = ERR57;
5592              break;              break;
5593              }              }
5594            if (isnumber)            if (isnumber)
5595              {              {
5596              ptr++;              ptr++;
5597              goto HANDLE_NUMERICAL_RECURSION;              goto HANDLE_NUMERICAL_RECURSION;
5598              }              }
5599            is_recurse = TRUE;            is_recurse = TRUE;
5600            goto NAMED_REF_OR_RECURSE;            goto NAMED_REF_OR_RECURSE;
5601            }            }
5602    
5603          /* Test a signed number in angle brackets or quotes. */          /* Test a signed number in angle brackets or quotes. */
5604    
5605          p = ptr + 2;          p = ptr + 2;
5606          while ((digitab[*p] & ctype_digit) != 0) p++;          while ((digitab[*p] & ctype_digit) != 0) p++;
5607          if (*p != terminator)          if (*p != terminator)
# Line 5176  we set the flag only if there is a liter Line 5609  we set the flag only if there is a liter
5609            *errorcodeptr = ERR57;            *errorcodeptr = ERR57;
5610            break;            break;
5611            }            }
5612          ptr++;          ptr++;
5613          goto HANDLE_NUMERICAL_RECURSION;          goto HANDLE_NUMERICAL_RECURSION;
5614          }          }
5615    
5616        /* \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).
5617        We also support \k{name} (.NET syntax) */        We also support \k{name} (.NET syntax) */
5618    
5619        if (-c == ESC_k && (ptr[1] == '<' || ptr[1] == '\'' || ptr[1] == '{'))        if (-c == ESC_k && (ptr[1] == CHAR_LESS_THAN_SIGN ||
5620              ptr[1] == CHAR_APOSTROPHE || ptr[1] == CHAR_LEFT_CURLY_BRACKET))
5621          {          {
5622          is_recurse = FALSE;          is_recurse = FALSE;
5623          terminator = (*(++ptr) == '<')? '>' : (*ptr == '\'')? '\'' : '}';          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
5624              CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?
5625              CHAR_APOSTROPHE : CHAR_RIGHT_CURLY_BRACKET;
5626          goto NAMED_REF_OR_RECURSE;          goto NAMED_REF_OR_RECURSE;
5627          }          }
5628    
# Line 5196  we set the flag only if there is a liter Line 5632  we set the flag only if there is a liter
5632    
5633        if (-c >= ESC_REF)        if (-c >= ESC_REF)
5634          {          {
5635            open_capitem *oc;
5636          recno = -c - ESC_REF;          recno = -c - ESC_REF;
5637    
5638          HANDLE_REFERENCE:    /* Come here from named backref handling */          HANDLE_REFERENCE:    /* Come here from named backref handling */
# Line 5205  we set the flag only if there is a liter Line 5642  we set the flag only if there is a liter
5642          PUT2INC(code, 0, recno);          PUT2INC(code, 0, recno);
5643          cd->backref_map |= (recno < 32)? (1 << recno) : 1;          cd->backref_map |= (recno < 32)? (1 << recno) : 1;
5644          if (recno > cd->top_backref) cd->top_backref = recno;          if (recno > cd->top_backref) cd->top_backref = recno;
5645    
5646            /* Check to see if this back reference is recursive, that it, it
5647            is inside the group that it references. A flag is set so that the
5648            group can be made atomic. */
5649    
5650            for (oc = cd->open_caps; oc != NULL; oc = oc->next)
5651              {
5652              if (oc->number == recno)
5653                {
5654                oc->flag = TRUE;
5655                break;
5656                }
5657              }
5658          }          }
5659    
5660        /* So are Unicode property matches, if supported. */        /* So are Unicode property matches, if supported. */
# Line 5289  we set the flag only if there is a liter Line 5739  we set the flag only if there is a liter
5739    
5740      /* Remember if \r or \n were seen */      /* Remember if \r or \n were seen */
5741    
5742      if (mcbuffer[0] == '\r' || mcbuffer[0] == '\n')      if (mcbuffer[0] == CHAR_CR || mcbuffer[0] == CHAR_NL)
5743        cd->external_flags |= PCRE_HASCRORLF;        cd->external_flags |= PCRE_HASCRORLF;
5744    
5745      /* Set the first and required bytes appropriately. If no previous first      /* Set the first and required bytes appropriately. If no previous first
# Line 5387  uschar *code = *codeptr; Line 5837  uschar *code = *codeptr;
5837  uschar *last_branch = code;  uschar *last_branch = code;
5838  uschar *start_bracket = code;  uschar *start_bracket = code;
5839  uschar *reverse_count = NULL;  uschar *reverse_count = NULL;
5840    open_capitem capitem;
5841    int capnumber = 0;
5842  int firstbyte, reqbyte;  int firstbyte, reqbyte;
5843  int branchfirstbyte, branchreqbyte;  int branchfirstbyte, branchreqbyte;
5844  int length;  int length;
5845  int orig_bracount;  int orig_bracount;
5846  int max_bracount;  int max_bracount;
5847    int old_external_options = cd->external_options;
5848  branch_chain bc;  branch_chain bc;
5849    
5850  bc.outer = bcptr;  bc.outer = bcptr;
5851  bc.current = code;  bc.current_branch = code;
5852    
5853  firstbyte = reqbyte = REQ_UNSET;  firstbyte = reqbyte = REQ_UNSET;
5854    
# Line 5413  the code that abstracts option settings Line 5866  the code that abstracts option settings
5866  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
5867  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. */
5868    
5869    /* If this is a capturing subpattern, add to the chain of open capturing items
5870    so that we can detect them if (*ACCEPT) is encountered. This is also used to
5871    detect groups that contain recursive back references to themselves. */
5872    
5873    if (*code == OP_CBRA)
5874      {
5875      capnumber = GET2(code, 1 + LINK_SIZE);
5876      capitem.number = capnumber;
5877      capitem.next = cd->open_caps;
5878      capitem.flag = FALSE;
5879      cd->open_caps = &capitem;
5880      }
5881    
5882  /* Offset is set zero to mark that this bracket is still open */  /* Offset is set zero to mark that this bracket is still open */
5883    
5884  PUT(code, 1, 0);  PUT(code, 1, 0);
# Line 5457  for (;;) Line 5923  for (;;)
5923      return FALSE;      return FALSE;
5924      }      }
5925    
5926      /* If the external options have changed during this branch, it means that we
5927      are at the top level, and a leading option setting has been encountered. We
5928      need to re-set the original option values to take account of this so that,
5929      during the pre-compile phase, we know to allow for a re-set at the start of
5930      subsequent branches. */
5931    
5932      if (old_external_options != cd->external_options)
5933        oldims = cd->external_options & PCRE_IMS;
5934    
5935    /* Keep the highest bracket count in case (?| was used and some branch    /* Keep the highest bracket count in case (?| was used and some branch
5936    has fewer than the rest. */    has fewer than the rest. */
5937    
# Line 5507  for (;;) Line 5982  for (;;)
5982    
5983      /* If lookbehind, check that this branch matches a fixed-length string, and      /* If lookbehind, check that this branch matches a fixed-length string, and
5984      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
5985      branch with OP_END. */      branch with OP_END. If the branch contains OP_RECURSE, the result is -3
5986        because there may be forward references that we can't check here. Set a
5987        flag to cause another lookbehind check at the end. Why not do it all at the
5988        end? Because common, erroneous checks are picked up here and the offset of
5989        the problem can be shown. */
5990    
5991      if (lookbehind)      if (lookbehind)
5992        {        {
5993        int fixed_length;        int fixed_length;
5994        *code = OP_END;        *code = OP_END;
5995        fixed_length = find_fixedlength(last_branch, options);        fixed_length = find_fixedlength(last_branch, options, FALSE, cd);
5996        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
5997        if (fixed_length < 0)        if (fixed_length == -3)
5998            {
5999            cd->check_lookbehind = TRUE;
6000            }
6001          else if (fixed_length < 0)
6002          {          {
6003          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;
6004          *ptrptr = ptr;          *ptrptr = ptr;
6005          return FALSE;          return FALSE;
6006          }          }
6007        PUT(reverse_count, 0, fixed_length);        else { PUT(reverse_count, 0, fixed_length); }
6008        }        }
6009      }      }
6010    
# Line 5534  for (;;) Line 6017  for (;;)
6017    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.
6018    Return leaving the pointer at the terminating char. */    Return leaving the pointer at the terminating char. */
6019    
6020    if (*ptr != '|')    if (*ptr != CHAR_VERTICAL_LINE)
6021      {      {
6022      if (lengthptr == NULL)      if (lengthptr == NULL)
6023        {        {
# Line 5555  for (;;) Line 6038  for (;;)
6038      PUT(code, 1, code - start_bracket);      PUT(code, 1, code - start_bracket);
6039      code += 1 + LINK_SIZE;      code += 1 + LINK_SIZE;
6040    
6041      /* Resetting option if needed */      /* If it was a capturing subpattern, check to see if it contained any
6042        recursive back references. If so, we must wrap it in atomic brackets.
6043        In any event, remove the block from the chain. */
6044    
6045      if ((options & PCRE_IMS) != oldims && *ptr == ')')      if (capnumber > 0)
6046          {
6047          if (cd->open_caps->flag)
6048            {
6049            memmove(start_bracket + 1 + LINK_SIZE, start_bracket,
6050              code - start_bracket);
6051            *start_bracket = OP_ONCE;
6052            code += 1 + LINK_SIZE;
6053            PUT(start_bracket, 1, code - start_bracket);
6054            *code = OP_KET;
6055            PUT(code, 1, code - start_bracket);
6056            code += 1 + LINK_SIZE;
6057            length += 2 + 2*LINK_SIZE;
6058            }
6059          cd->open_caps = cd->open_caps->next;
6060          }
6061    
6062        /* Reset options if needed. */
6063    
6064        if ((options & PCRE_IMS) != oldims && *ptr == CHAR_RIGHT_PARENTHESIS)
6065        {        {
6066        *code++ = OP_OPT;        *code++ = OP_OPT;
6067        *code++ = oldims;        *code++ = oldims;
# Line 5604  for (;;) Line 6108  for (;;)
6108      {      {
6109      *code = OP_ALT;      *code = OP_ALT;
6110      PUT(code, 1, code - last_branch);      PUT(code, 1, code - last_branch);
6111      bc.current = last_branch = code;      bc.current_branch = last_branch = code;
6112      code += 1 + LINK_SIZE;      code += 1 + LINK_SIZE;
6113      }      }
6114    
# Line 5686  do { Line 6190  do {
6190       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;
6191       }       }
6192    
6193     /* .* 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
6194     are or may be referenced. */     it isn't in brackets that are or may be referenced. */
6195    
6196     else if ((op == OP_TYPESTAR || op == OP_TYPEMINSTAR ||     else if ((op == OP_TYPESTAR || op == OP_TYPEMINSTAR ||
6197               op == OP_TYPEPOSSTAR) &&               op == OP_TYPEPOSSTAR))
             (*options & PCRE_DOTALL) != 0)  
6198       {       {
6199       if (scode[1] != OP_ANY || (bracket_map & backref_map) != 0) return FALSE;       if (scode[1] != OP_ALLANY || (bracket_map & backref_map) != 0)
6200           return FALSE;
6201       }       }
6202    
6203     /* Check for explicit anchoring */     /* Check for explicit anchoring */
# Line 5739  do { Line 6243  do {
6243       NULL, 0, FALSE);       NULL, 0, FALSE);
6244     register int op = *scode;     register int op = *scode;
6245    
6246       /* If we are at the start of a conditional assertion group, *both* the
6247       conditional assertion *and* what follows the condition must satisfy the test
6248       for start of line. Other kinds of condition fail. Note that there may be an
6249       auto-callout at the start of a condition. */
6250    
6251       if (op == OP_COND)
6252         {
6253         scode += 1 + LINK_SIZE;
6254         if (*scode == OP_CALLOUT) scode += _pcre_OP_lengths[OP_CALLOUT];
6255         switch (*scode)
6256           {
6257           case OP_CREF:
6258           case OP_NCREF:
6259           case OP_RREF:
6260           case OP_NRREF:
6261           case OP_DEF:
6262           return FALSE;
6263    
6264           default:     /* Assertion */
6265           if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
6266           do scode += GET(scode, 1); while (*scode == OP_ALT);
6267           scode += 1 + LINK_SIZE;
6268           break;
6269           }
6270         scode = first_significant_code(scode, NULL, 0, FALSE);
6271         op = *scode;
6272         }
6273    
6274     /* Non-capturing brackets */     /* Non-capturing brackets */
6275    
6276     if (op == OP_BRA)     if (op == OP_BRA)
# Line 5757  do { Line 6289  do {
6289    
6290     /* Other brackets */     /* Other brackets */
6291    
6292     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_COND)     else if (op == OP_ASSERT || op == OP_ONCE)
6293       { if (!is_startline(scode, bracket_map, backref_map)) return FALSE; }       {
6294         if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
6295         }
6296    
6297     /* .* 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
6298     may be referenced. */     may be referenced. */
# Line 5875  Returns:        pointer to compiled data Line 6409  Returns:        pointer to compiled data
6409                  with errorptr and erroroffset set                  with errorptr and erroroffset set
6410  */  */
6411    
6412  PCRE_EXP_DEFN pcre *  PCRE_EXP_DEFN pcre * PCRE_CALL_CONVENTION
6413  pcre_compile(const char *pattern, int options, const char **errorptr,  pcre_compile(const char *pattern, int options, const char **errorptr,
6414    int *erroroffset, const unsigned char *tables)    int *erroroffset, const unsigned char *tables)
6415  {  {
# Line 5883  return pcre_compile2(pattern, options, N Line 6417  return pcre_compile2(pattern, options, N
6417  }  }
6418    
6419    
6420  PCRE_EXP_DEFN pcre *  PCRE_EXP_DEFN pcre * PCRE_CALL_CONVENTION
6421  pcre_compile2(const char *pattern, int options, int *errorcodeptr,  pcre_compile2(const char *pattern, int options, int *errorcodeptr,
6422    const char **errorptr, int *erroroffset, const unsigned char *tables)    const char **errorptr, int *erroroffset, const unsigned char *tables)
6423  {  {
# Line 5892  int length = 1;  /* For final END opcode Line 6426  int length = 1;  /* For final END opcode
6426  int firstbyte, reqbyte, newline;  int firstbyte, reqbyte, newline;
6427  int errorcode = 0;  int errorcode = 0;
6428  int skipatstart = 0;  int skipatstart = 0;
6429  #ifdef SUPPORT_UTF8  BOOL utf8 = (options & PCRE_UTF8) != 0;
 BOOL utf8;  
 #endif  
6430  size_t size;  size_t size;
6431  uschar *code;  uschar *code;
6432  const uschar *codestart;  const uschar *codestart;
# Line 5937  if (erroroffset == NULL) Line 6469  if (erroroffset == NULL)
6469    
6470  *erroroffset = 0;  *erroroffset = 0;
6471    
 /* 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;  
   }  
   
6472  /* Set up pointers to the individual character tables */  /* Set up pointers to the individual character tables */
6473    
6474  if (tables == NULL) tables = _pcre_default_tables;  if (tables == NULL) tables = _pcre_default_tables;
# Line 5969  cd->fcc = tables + fcc_offset; Line 6477  cd->fcc = tables + fcc_offset;
6477  cd->cbits = tables + cbits_offset;  cd->cbits = tables + cbits_offset;
6478  cd->ctypes = tables + ctypes_offset;  cd->ctypes = tables + ctypes_offset;
6479    
6480    /* Check that all undefined public option bits are zero */
6481    
6482    if ((options & ~PUBLIC_COMPILE_OPTIONS) != 0)
6483      {
6484      errorcode = ERR17;
6485      goto PCRE_EARLY_ERROR_RETURN;
6486      }
6487    
6488  /* 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
6489  the offset for later. */  the offset for later. */
6490    
6491  while (ptr[skipatstart] == '(' && ptr[skipatstart+1] == '*')  while (ptr[skipatstart] == CHAR_LEFT_PARENTHESIS &&
6492           ptr[skipatstart+1] == CHAR_ASTERISK)
6493    {    {
6494    int newnl = 0;    int newnl = 0;
6495    int newbsr = 0;    int newbsr = 0;
6496    
6497    if (strncmp((char *)(ptr+skipatstart+2), "CR)", 3) == 0)    if (strncmp((char *)(ptr+skipatstart+2), STRING_UTF8_RIGHTPAR, 5) == 0)
6498        { skipatstart += 7; options |= PCRE_UTF8; continue; }
6499    
6500      if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)
6501      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }
6502    else if (strncmp((char *)(ptr+skipatstart+2), "LF)", 3)  == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_LF_RIGHTPAR, 3)  == 0)
6503      { skipatstart += 5; newnl = PCRE_NEWLINE_LF; }      { skipatstart += 5; newnl = PCRE_NEWLINE_LF; }
6504    else if (strncmp((char *)(ptr+skipatstart+2), "CRLF)", 5)  == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_CRLF_RIGHTPAR, 5)  == 0)
6505      { skipatstart += 7; newnl = PCRE_NEWLINE_CR + PCRE_NEWLINE_LF; }      { skipatstart += 7; newnl = PCRE_NEWLINE_CR + PCRE_NEWLINE_LF; }
6506    else if (strncmp((char *)(ptr+skipatstart+2), "ANY)", 4) == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_ANY_RIGHTPAR, 4) == 0)
6507      { skipatstart += 6; newnl = PCRE_NEWLINE_ANY; }      { skipatstart += 6; newnl = PCRE_NEWLINE_ANY; }
6508    else if (strncmp((char *)(ptr+skipatstart+2), "ANYCRLF)", 8)  == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_ANYCRLF_RIGHTPAR, 8) == 0)
6509      { skipatstart += 10; newnl = PCRE_NEWLINE_ANYCRLF; }      { skipatstart += 10; newnl = PCRE_NEWLINE_ANYCRLF; }
6510    
6511    else if (strncmp((char *)(ptr+skipatstart+2), "BSR_ANYCRLF)", 12) == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_BSR_ANYCRLF_RIGHTPAR, 12) == 0)
6512      { skipatstart += 14; newbsr = PCRE_BSR_ANYCRLF; }      { skipatstart += 14; newbsr = PCRE_BSR_ANYCRLF; }
6513    else if (strncmp((char *)(ptr+skipatstart+2), "BSR_UNICODE)", 12) == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_BSR_UNICODE_RIGHTPAR, 12) == 0)
6514      { skipatstart += 14; newbsr = PCRE_BSR_UNICODE; }      { skipatstart += 14; newbsr = PCRE_BSR_UNICODE; }
6515    
6516    if (newnl != 0)    if (newnl != 0)
# Line 6000  while (ptr[skipatstart] == '(' && ptr[sk Line 6520  while (ptr[skipatstart] == '(' && ptr[sk
6520    else break;    else break;
6521    }    }
6522    
6523    /* Can't support UTF8 unless PCRE has been compiled to include the code. */
6524    
6525    #ifdef SUPPORT_UTF8
6526    if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&
6527         (*erroroffset = _pcre_valid_utf8((USPTR)pattern, -1)) >= 0)
6528      {
6529      errorcode = ERR44;
6530      goto PCRE_EARLY_ERROR_RETURN2;
6531      }
6532    #else
6533    if (utf8)
6534      {
6535      errorcode = ERR32;
6536      goto PCRE_EARLY_ERROR_RETURN;
6537      }
6538    #endif
6539    
6540  /* Check validity of \R options. */  /* Check validity of \R options. */
6541    
6542  switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))  switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))
# Line 6018  current code allows for fixed one- or tw Line 6555  current code allows for fixed one- or tw
6555  switch (options & PCRE_NEWLINE_BITS)  switch (options & PCRE_NEWLINE_BITS)
6556    {    {
6557    case 0: newline = NEWLINE; break;   /* Build-time default */    case 0: newline = NEWLINE; break;   /* Build-time default */
6558    case PCRE_NEWLINE_CR: newline = '\r'; break;    case PCRE_NEWLINE_CR: newline = CHAR_CR; break;
6559    case PCRE_NEWLINE_LF: newline = '\n'; break;    case PCRE_NEWLINE_LF: newline = CHAR_NL; break;
6560    case PCRE_NEWLINE_CR+    case PCRE_NEWLINE_CR+
6561         PCRE_NEWLINE_LF: newline = ('\r' << 8) | '\n'; break;         PCRE_NEWLINE_LF: newline = (CHAR_CR << 8) | CHAR_NL; break;
6562    case PCRE_NEWLINE_ANY: newline = -1; break;    case PCRE_NEWLINE_ANY: newline = -1; break;
6563    case PCRE_NEWLINE_ANYCRLF: newline = -2; break;    case PCRE_NEWLINE_ANYCRLF: newline = -2; break;
6564    default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;    default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;
# Line 6082  cd->end_pattern = (const uschar *)(patte Line 6619  cd->end_pattern = (const uschar *)(patte
6619  cd->req_varyopt = 0;  cd->req_varyopt = 0;
6620  cd->external_options = options;  cd->external_options = options;
6621  cd->external_flags = 0;  cd->external_flags = 0;
6622    cd->open_caps = NULL;
6623    
6624  /* 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
6625  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 6156  cd->start_code = codestart; Line 6694  cd->start_code = codestart;
6694  cd->hwm = cworkspace;  cd->hwm = cworkspace;
6695  cd->req_varyopt = 0;  cd->req_varyopt = 0;
6696  cd->had_accept = FALSE;  cd->had_accept = FALSE;
6697    cd->check_lookbehind = FALSE;
6698    cd->open_caps = NULL;
6699    
6700  /* Set up a starting, non-extracting bracket, then compile the expression. On  /* Set up a starting, non-extracting bracket, then compile the expression. On
6701  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 6181  if debugging, leave the test till after Line 6721  if debugging, leave the test till after
6721    
6722  *code++ = OP_END;  *code++ = OP_END;
6723    
6724  #ifndef DEBUG  #ifndef PCRE_DEBUG
6725  if (code - codestart > length) errorcode = ERR23;  if (code - codestart > length) errorcode = ERR23;
6726  #endif  #endif
6727    
# Line 6194  while (errorcode == 0 && cd->hwm > cwork Line 6734  while (errorcode == 0 && cd->hwm > cwork
6734    cd->hwm -= LINK_SIZE;    cd->hwm -= LINK_SIZE;
6735    offset = GET(cd->hwm, 0);    offset = GET(cd->hwm, 0);
6736    recno = GET(codestart, offset);    recno = GET(codestart, offset);
6737    groupptr = find_bracket(codestart, (re->options & PCRE_UTF8) != 0, recno);    groupptr = _pcre_find_bracket(codestart, utf8, recno);
6738    if (groupptr == NULL) errorcode = ERR53;    if (groupptr == NULL) errorcode = ERR53;
6739      else PUT(((uschar *)codestart), offset, groupptr - codestart);      else PUT(((uschar *)codestart), offset, groupptr - codestart);
6740    }    }
# Line 6204  subpattern. */ Line 6744  subpattern. */
6744    
6745  if (errorcode == 0 && re->top_backref > re->top_bracket) errorcode = ERR15;  if (errorcode == 0 && re->top_backref > re->top_bracket) errorcode = ERR15;
6746    
6747    /* If there were any lookbehind assertions that contained OP_RECURSE
6748    (recursions or subroutine calls), a flag is set for them to be checked here,
6749    because they may contain forward references. Actual recursions can't be fixed
6750    length, but subroutine calls can. It is done like this so that those without
6751    OP_RECURSE that are not fixed length get a diagnosic with a useful offset. The
6752    exceptional ones forgo this. We scan the pattern to check that they are fixed
6753    length, and set their lengths. */
6754    
6755    if (cd->check_lookbehind)
6756      {
6757      uschar *cc = (uschar *)codestart;
6758    
6759      /* Loop, searching for OP_REVERSE items, and process those that do not have
6760      their length set. (Actually, it will also re-process any that have a length
6761      of zero, but that is a pathological case, and it does no harm.) When we find
6762      one, we temporarily terminate the branch it is in while we scan it. */
6763    
6764      for (cc = (uschar *)_pcre_find_bracket(codestart, utf8, -1);
6765           cc != NULL;
6766           cc = (uschar *)_pcre_find_bracket(cc, utf8, -1))
6767        {
6768        if (GET(cc, 1) == 0)
6769          {
6770          int fixed_length;
6771          uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);
6772          int end_op = *be;
6773          *be = OP_END;
6774          fixed_length = find_fixedlength(cc, re->options, TRUE, cd);
6775          *be = end_op;
6776          DPRINTF(("fixed length = %d\n", fixed_length));
6777          if (fixed_length < 0)
6778            {
6779            errorcode = (fixed_length == -2)? ERR36 : ERR25;
6780            break;
6781            }
6782          PUT(cc, 1, fixed_length);
6783          }
6784        cc += 1 + LINK_SIZE;
6785        }
6786      }
6787    
6788  /* Failed to compile, or error while post-processing */  /* Failed to compile, or error while post-processing */
6789    
6790  if (errorcode != 0)  if (errorcode != 0)
# Line 6264  if (reqbyte >= 0 && Line 6845  if (reqbyte >= 0 &&
6845  /* 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
6846  case when building a production library. */  case when building a production library. */
6847    
6848  #ifdef DEBUG  #ifdef PCRE_DEBUG
   
6849  printf("Length = %d top_bracket = %d top_backref = %d\n",  printf("Length = %d top_bracket = %d top_backref = %d\n",
6850    length, re->top_bracket, re->top_backref);    length, re->top_bracket, re->top_backref);
6851    
# Line 6302  if (code - codestart > length) Line 6882  if (code - codestart > length)
6882    if (errorcodeptr != NULL) *errorcodeptr = ERR23;    if (errorcodeptr != NULL) *errorcodeptr = ERR23;
6883    return NULL;    return NULL;
6884    }    }
6885  #endif   /* DEBUG */  #endif   /* PCRE_DEBUG */
6886    
6887  return (pcre *)re;  return (pcre *)re;
6888  }  }

Legend:
Removed from v.334  
changed lines
  Added in v.503

  ViewVC Help
Powered by ViewVC 1.1.5