/[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 335 by ph10, Sat Apr 12 14:36:14 2008 UTC revision 507 by ph10, Wed Mar 10 16:08:01 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 91  is 4 there is plenty of room. */ Line 92  is 4 there is plenty of room. */
92    
93  #define COMPILE_WORK_SIZE (4096)  #define COMPILE_WORK_SIZE (4096)
94    
95    /* The overrun tests check for a slightly smaller size so that they detect the
96    overrun before it actually does run off the end of the data block. */
97    
98    #define WORK_SIZE_CHECK (COMPILE_WORK_SIZE - 100)
99    
100    
101  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
102  are simple data values; negative values are for special things like \d and so  are simple data values; negative values are for special things like \d and so
103  on. Zero means further processing is needed (for things like \x), or the escape  on. Zero means further processing is needed (for things like \x), or the escape
104  is invalid. */  is invalid. */
105    
106  #ifndef EBCDIC  /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC
107    
108    /* This is the "normal" table for ASCII systems or for EBCDIC systems running
109    in UTF-8 mode. */
110    
111  static const short int escapes[] = {  static const short int escapes[] = {
112       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,                       0,
113       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,                       0,
114     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */       0,                       0,
115  -ESC_H,      0,      0, -ESC_K,      0,      0,      0,      0,   /* H - O */       0,                       0,
116  -ESC_P, -ESC_Q, -ESC_R, -ESC_S,      0,      0, -ESC_V, -ESC_W,   /* P - W */       0,                       0,
117  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */       CHAR_COLON,              CHAR_SEMICOLON,
118     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
119  -ESC_h,      0,      0, -ESC_k,      0,      0,  ESC_n,      0,   /* h - o */       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
120  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0, -ESC_v, -ESC_w,   /* p - w */       CHAR_COMMERCIAL_AT,      -ESC_A,
121       0,      0, -ESC_z                                            /* x - z */       -ESC_B,                  -ESC_C,
122         -ESC_D,                  -ESC_E,
123         0,                       -ESC_G,
124         -ESC_H,                  0,
125         0,                       -ESC_K,
126         0,                       0,
127         0,                       0,
128         -ESC_P,                  -ESC_Q,
129         -ESC_R,                  -ESC_S,
130         0,                       0,
131         -ESC_V,                  -ESC_W,
132         -ESC_X,                  0,
133         -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,
134         CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,
135         CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,
136         CHAR_GRAVE_ACCENT,       7,
137         -ESC_b,                  0,
138         -ESC_d,                  ESC_e,
139         ESC_f,                   0,
140         -ESC_h,                  0,
141         0,                       -ESC_k,
142         0,                       0,
143         ESC_n,                   0,
144         -ESC_p,                  0,
145         ESC_r,                   -ESC_s,
146         ESC_tee,                 0,
147         -ESC_v,                  -ESC_w,
148         0,                       0,
149         -ESC_z
150  };  };
151    
152  #else           /* This is the "abnormal" table for EBCDIC systems */  #else
153    
154    /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
155    
156  static const short int escapes[] = {  static const short int escapes[] = {
157  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
158  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 142  static const short int escapes[] = { Line 183  static const short int escapes[] = {
183    
184  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
185  searched linearly. Put all the names into a single string, in order to reduce  searched linearly. Put all the names into a single string, in order to reduce
186  the number of relocations when a shared library is dynamically linked. */  the number of relocations when a shared library is dynamically linked. The
187    string is built from string macros so that it works in UTF-8 mode on EBCDIC
188    platforms. */
189    
190  typedef struct verbitem {  typedef struct verbitem {
191    int   len;    int   len;
# Line 150  typedef struct verbitem { Line 193  typedef struct verbitem {
193  } verbitem;  } verbitem;
194    
195  static const char verbnames[] =  static const char verbnames[] =
196    "ACCEPT\0"    STRING_ACCEPT0
197    "COMMIT\0"    STRING_COMMIT0
198    "F\0"    STRING_F0
199    "FAIL\0"    STRING_FAIL0
200    "PRUNE\0"    STRING_PRUNE0
201    "SKIP\0"    STRING_SKIP0
202    "THEN";    STRING_THEN;
203    
204  static const verbitem verbs[] = {  static const verbitem verbs[] = {
205    { 6, OP_ACCEPT },    { 6, OP_ACCEPT },
# Line 178  length entry. The first three must be al Line 221  length entry. The first three must be al
221  for handling case independence. */  for handling case independence. */
222    
223  static const char posix_names[] =  static const char posix_names[] =
224    "alpha\0"  "lower\0"  "upper\0"  "alnum\0"  "ascii\0"  "blank\0"    STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
225    "cntrl\0"  "digit\0"  "graph\0"  "print\0"  "punct\0"  "space\0"    STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
226    "word\0"   "xdigit";    STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
227      STRING_word0  STRING_xdigit;
228    
229  static const uschar posix_name_lengths[] = {  static const uschar posix_name_lengths[] = {
230    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 268  the number of relocations needed when a
268  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
269  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
270  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
271  because these strings are used only when there is a compilation error. */  because these strings are used only when there is a compilation error.
272    
273    Each substring ends with \0 to insert a null character. This includes the final
274    substring, so that the whole string ends with \0\0, which can be detected when
275    counting through. */
276    
277  static const char error_texts[] =  static const char error_texts[] =
278    "no error\0"    "no error\0"
# Line 302  static const char error_texts[] = Line 350  static const char error_texts[] =
350    "(*VERB) not recognized\0"    "(*VERB) not recognized\0"
351    "number is too big\0"    "number is too big\0"
352    "subpattern name expected\0"    "subpattern name expected\0"
353    "digit expected after (?+";    "digit expected after (?+\0"
354      "] is an invalid data character in JavaScript compatibility mode\0"
355      /* 65 */
356      "different names for subpatterns of the same number are not allowed\0";
357    
358  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
359  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 371  For convenience, we use the same bit def
371    
372  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
373    
374  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC
375    
376    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
377    UTF-8 mode. */
378    
379  static const unsigned char digitab[] =  static const unsigned char digitab[] =
380    {    {
381    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 411  static const unsigned char digitab[] =
411    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
412    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
413    
414  #else           /* This is the "abnormal" case, for EBCDIC systems */  #else
415    
416    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
417    
418  static const unsigned char digitab[] =  static const unsigned char digitab[] =
419    {    {
420    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 511  static const char *
511  find_error_text(int n)  find_error_text(int n)
512  {  {
513  const char *s = error_texts;  const char *s = error_texts;
514  for (; n > 0; n--) while (*s++ != 0);  for (; n > 0; n--)
515      {
516      while (*s++ != 0) {};
517      if (*s == 0) return "Error text not found (please report)";
518      }
519  return s;  return s;
520  }  }
521    
# Line 502  if (c == 0) *errorcodeptr = ERR1; Line 563  if (c == 0) *errorcodeptr = ERR1;
563  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.
564  Otherwise further processing may be required. */  Otherwise further processing may be required. */
565    
566  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
567  else if (c < '0' || c > 'z') {}                           /* Not alphanumeric */  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */
568  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0) c = i;
569    
570  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
571  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 584  else
584      /* 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
585      error. */      error. */
586    
587      case 'l':      case CHAR_l:
588      case 'L':      case CHAR_L:
589      case 'N':      case CHAR_N:
590      case 'u':      case CHAR_u:
591      case 'U':      case CHAR_U:
592      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
593      break;      break;
594    
595      /* \g must be followed by one of a number of specific things:      /* \g must be followed by one of a number of specific things:
596    
597      (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
598      backreference. If negative, it is a relative backreference. This is a Perl      backreference. If negative, it is a relative backreference. This is a Perl
599      5.10 feature.      5.10 feature.
600    
601      (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
602      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
603      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
604      was \k.      was \k.
605    
606      (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
607      number either in angle brackets or in single quotes. However, these are      number either in angle brackets or in single quotes. However, these are
608      (possibly recursive) subroutine calls, _not_ backreferences. Just return      (possibly recursive) subroutine calls, _not_ backreferences. Just return
609      the -ESC_g code (cf \k). */      the -ESC_g code (cf \k). */
610    
611      case 'g':      case CHAR_g:
612      if (ptr[1] == '<' || ptr[1] == '\'')      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
613        {        {
614        c = -ESC_g;        c = -ESC_g;
615        break;        break;
616        }        }
617    
618      /* Handle the Perl-compatible cases */      /* Handle the Perl-compatible cases */
619    
620      if (ptr[1] == '{')      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
621        {        {
622        const uschar *p;        const uschar *p;
623        for (p = ptr+2; *p != 0 && *p != '}'; p++)        for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
624          if (*p != '-' && (digitab[*p] & ctype_digit) == 0) break;          if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
625        if (*p != 0 && *p != '}')        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
626          {          {
627          c = -ESC_k;          c = -ESC_k;
628          break;          break;
# Line 571  else Line 632  else
632        }        }
633      else braced = FALSE;      else braced = FALSE;
634    
635      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
636        {        {
637        negated = TRUE;        negated = TRUE;
638        ptr++;        ptr++;
# Line 580  else Line 641  else
641    
642      c = 0;      c = 0;
643      while ((digitab[ptr[1]] & ctype_digit) != 0)      while ((digitab[ptr[1]] & ctype_digit) != 0)
644        c = c * 10 + *(++ptr) - '0';        c = c * 10 + *(++ptr) - CHAR_0;
645    
646      if (c < 0)   /* Integer overflow */      if (c < 0)   /* Integer overflow */
647        {        {
648        *errorcodeptr = ERR61;        *errorcodeptr = ERR61;
649        break;        break;
650        }        }
651    
652      if (braced && *(++ptr) != '}')      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
653        {        {
654        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
655        break;        break;
656        }        }
657    
658      if (c == 0)      if (c == 0)
659        {        {
660        *errorcodeptr = ERR58;        *errorcodeptr = ERR58;
661        break;        break;
662        }        }
663    
664      if (negated)      if (negated)
665        {        {
# Line 625  else Line 686  else
686      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
687      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
688    
689      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:
690      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
691    
692      if (!isclass)      if (!isclass)
693        {        {
694        oldptr = ptr;        oldptr = ptr;
695        c -= '0';        c -= CHAR_0;
696        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
697          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - CHAR_0;
698        if (c < 0)    /* Integer overflow */        if (c < 0)    /* Integer overflow */
699          {          {
700          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
# Line 651  else Line 712  else
712      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.
713      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
714    
715      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
716        {        {
717        ptr--;        ptr--;
718        c = 0;        c = 0;
# Line 664  else Line 725  else
725      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
726      than 3 octal digits. */      than 3 octal digits. */
727    
728      case '0':      case CHAR_0:
729      c -= '0';      c -= CHAR_0;
730      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
731          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
732      if (!utf8 && c > 255) *errorcodeptr = ERR51;      if (!utf8 && c > 255) *errorcodeptr = ERR51;
733      break;      break;
734    
# Line 675  else Line 736  else
736      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
737      treated as a data character. */      treated as a data character. */
738    
739      case 'x':      case CHAR_x:
740      if (ptr[1] == '{')      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
741        {        {
742        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
743        int count = 0;        int count = 0;
# Line 685  else Line 746  else
746        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
747          {          {
748          register int cc = *pt++;          register int cc = *pt++;
749          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
750          count++;          count++;
751    
752  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
753          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
754          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
755  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
756          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
757          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
758  #endif  #endif
759          }          }
760    
761        if (*pt == '}')        if (*pt == CHAR_RIGHT_CURLY_BRACKET)
762          {          {
763          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
764          ptr = pt;          ptr = pt;
# Line 713  else Line 774  else
774      c = 0;      c = 0;
775      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
776        {        {
777        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
778        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
779  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
780        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
781        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
782  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
783        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
784        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
785  #endif  #endif
786        }        }
787      break;      break;
# Line 729  else Line 790  else
790      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
791      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
792    
793      case 'c':      case CHAR_c:
794      c = *(++ptr);      c = *(++ptr);
795      if (c == 0)      if (c == 0)
796        {        {
# Line 737  else Line 798  else
798        break;        break;
799        }        }
800    
801  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
802      if (c >= 'a' && c <= 'z') c -= 32;      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
803      c ^= 0x40;      c ^= 0x40;
804  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
805      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
806      c ^= 0xC0;      c ^= 0xC0;
807  #endif  #endif
808      break;      break;
# Line 803  if (c == 0) goto ERROR_RETURN; Line 864  if (c == 0) goto ERROR_RETURN;
864  /* \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
865  negation. */  negation. */
866    
867  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
868    {    {
869    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
870      {      {
871      *negptr = TRUE;      *negptr = TRUE;
872      ptr++;      ptr++;
# Line 814  if (c == '{') Line 875  if (c == '{')
875      {      {
876      c = *(++ptr);      c = *(++ptr);
877      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
878      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
879      name[i] = c;      name[i] = c;
880      }      }
881    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
882    name[i] = 0;    name[i] = 0;
883    }    }
884    
# Line 882  is_counted_repeat(const uschar *p) Line 943  is_counted_repeat(const uschar *p)
943  {  {
944  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
945  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
946  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
947    
948  if (*p++ != ',') return FALSE;  if (*p++ != CHAR_COMMA) return FALSE;
949  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
950    
951  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
952  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
953    
954  return (*p == '}');  return (*p == CHAR_RIGHT_CURLY_BRACKET);
955  }  }
956    
957    
# Line 923  int max = -1; Line 984  int max = -1;
984  /* 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
985  an integer overflow. */  an integer overflow. */
986    
987  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;
988  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
989    {    {
990    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 933  if (min < 0 || min > 65535) Line 994  if (min < 0 || min > 65535)
994  /* 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.
995  Also, max must not be less than min. */  Also, max must not be less than min. */
996    
997  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
998    {    {
999    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1000      {      {
1001      max = 0;      max = 0;
1002      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;
1003      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1004        {        {
1005        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 963  return p; Line 1024  return p;
1024    
1025    
1026  /*************************************************  /*************************************************
1027  *       Find forward referenced subpattern       *  *  Subroutine for finding forward reference      *
1028  *************************************************/  *************************************************/
1029    
1030  /* This function scans along a pattern's text looking for capturing  /* This recursive function is called only from find_parens() below. The
1031    top-level call starts at the beginning of the pattern. All other calls must
1032    start at a parenthesis. It scans along a pattern's text looking for capturing
1033  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
1034  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
1035  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
1036  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
1037  be terminated by '>' because that is checked in the first pass.  first pass. Recursion is used to keep track of subpatterns that reset the
1038    capturing group numbers - the (?| feature.
1039    
1040  Arguments:  Arguments:
1041    ptr          current position in the pattern    ptrptr       address of the current character pointer (updated)
1042    count        current count of capturing parens so far encountered    cd           compile background data
1043    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1044    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1045    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1046      count        pointer to the current capturing subpattern number (updated)
1047    
1048  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1049  */  */
1050    
1051  static int  static int
1052  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,
1053    BOOL xmode)    BOOL xmode, int *count)
1054  {  {
1055  const uschar *thisname;  uschar *ptr = *ptrptr;
1056    int start_count = *count;
1057    int hwm_count = start_count;
1058    BOOL dup_parens = FALSE;
1059    
1060  for (; *ptr != 0; ptr++)  /* If the first character is a parenthesis, check on the type of group we are
1061    dealing with. The very first call may not start with a parenthesis. */
1062    
1063    if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1064    {    {
1065    int term;    if (ptr[1] == CHAR_QUESTION_MARK &&
1066          ptr[2] == CHAR_VERTICAL_LINE)
1067        {
1068        ptr += 3;
1069        dup_parens = TRUE;
1070        }
1071    
1072      /* Handle a normal, unnamed capturing parenthesis */
1073    
1074      else if (ptr[1] != CHAR_QUESTION_MARK && ptr[1] != CHAR_ASTERISK)
1075        {
1076        *count += 1;
1077        if (name == NULL && *count == lorn) return *count;
1078        ptr++;
1079        }
1080    
1081      /* Handle a condition. If it is an assertion, just carry on so that it
1082      is processed as normal. If not, skip to the closing parenthesis of the
1083      condition (there can't be any nested parens. */
1084    
1085      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1086        {
1087        ptr += 2;
1088        if (ptr[1] != CHAR_QUESTION_MARK)
1089          {
1090          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1091          if (*ptr != 0) ptr++;
1092          }
1093        }
1094    
1095      /* We have either (? or (* and not a condition */
1096    
1097      else
1098        {
1099        ptr += 2;
1100        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1101    
1102        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1103    
1104        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1105            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1106          {
1107          int term;
1108          const uschar *thisname;
1109          *count += 1;
1110          if (name == NULL && *count == lorn) return *count;
1111          term = *ptr++;
1112          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1113          thisname = ptr;
1114          while (*ptr != term) ptr++;
1115          if (name != NULL && lorn == ptr - thisname &&
1116              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1117            return *count;
1118          term++;
1119          }
1120        }
1121      }
1122    
1123    /* Past any initial parenthesis handling, scan for parentheses or vertical
1124    bars. */
1125    
1126    for (; *ptr != 0; ptr++)
1127      {
1128    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1129    
1130    if (*ptr == '\\')    if (*ptr == CHAR_BACKSLASH)
1131      {      {
1132      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1133      if (*ptr == 'Q') for (;;)      if (*ptr == CHAR_Q) for (;;)
1134        {        {
1135        while (*(++ptr) != 0 && *ptr != '\\');        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1136        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1137        if (*(++ptr) == 'E') break;        if (*(++ptr) == CHAR_E) break;
1138        }        }
1139      continue;      continue;
1140      }      }
1141    
1142    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1143      are handled for real. If the first character is '^', skip it. Also, if the
1144      first few characters (either before or after ^) are \Q\E or \E we skip them
1145      too. This makes for compatibility with Perl. Note the use of STR macros to
1146      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1147    
1148    if (*ptr == '[')    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1149      {      {
1150      while (*(++ptr) != ']')      BOOL negate_class = FALSE;
1151        for (;;)
1152          {
1153          if (ptr[1] == CHAR_BACKSLASH)
1154            {
1155            if (ptr[2] == CHAR_E)
1156              ptr+= 2;
1157            else if (strncmp((const char *)ptr+2,
1158                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1159              ptr += 4;
1160            else
1161              break;
1162            }
1163          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1164            {
1165            negate_class = TRUE;
1166            ptr++;
1167            }
1168          else break;
1169          }
1170    
1171        /* If the next character is ']', it is a data character that must be
1172        skipped, except in JavaScript compatibility mode. */
1173    
1174        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1175            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1176          ptr++;
1177    
1178        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1179        {        {
1180        if (*ptr == 0) return -1;        if (*ptr == 0) return -1;
1181        if (*ptr == '\\')        if (*ptr == CHAR_BACKSLASH)
1182          {          {
1183          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) goto FAIL_EXIT;
1184          if (*ptr == 'Q') for (;;)          if (*ptr == CHAR_Q) for (;;)
1185            {            {
1186            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1187            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1188            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1189            }            }
1190          continue;          continue;
1191          }          }
# Line 1031  for (; *ptr != 0; ptr++) Line 1195  for (; *ptr != 0; ptr++)
1195    
1196    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1197    
1198    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1199      {      {
1200      while (*(++ptr) != 0 && *ptr != '\n');      while (*(++ptr) != 0 && *ptr != CHAR_NL) {};
1201      if (*ptr == 0) return -1;      if (*ptr == 0) goto FAIL_EXIT;
1202      continue;      continue;
1203      }      }
1204    
1205    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1206    
1207    if (*ptr != '(') continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != '?' && ptr[1] != '*')  
1208      {      {
1209      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, count);
1210      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1211      continue;      if (*ptr == 0) goto FAIL_EXIT;
1212      }      }
1213    
1214    ptr += 2;    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1215    if (*ptr == 'P') ptr++;                      /* Allow optional P */      {
1216        if (dup_parens && *count < hwm_count) *count = hwm_count;
1217        *ptrptr = ptr;
1218        return -1;
1219        }
1220    
1221      else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1222        {
1223        if (*count > hwm_count) hwm_count = *count;
1224        *count = start_count;
1225        }
1226      }
1227    
1228    FAIL_EXIT:
1229    *ptrptr = ptr;
1230    return -1;
1231    }
1232    
   /* We have to disambiguate (?<! and (?<= from (?<name> */  
1233    
   if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  
        *ptr != '\'')  
     continue;  
1234    
   count++;  
1235    
1236    if (name == NULL && count == lorn) return count;  /*************************************************
1237    term = *ptr++;  *       Find forward referenced subpattern       *
1238    if (term == '<') term = '>';  *************************************************/
1239    thisname = ptr;  
1240    while (*ptr != term) ptr++;  /* This function scans along a pattern's text looking for capturing
1241    if (name != NULL && lorn == ptr - thisname &&  subpatterns, and counting them. If it finds a named pattern that matches the
1242        strncmp((const char *)name, (const char *)thisname, lorn) == 0)  name it is given, it returns its number. Alternatively, if the name is NULL, it
1243      return count;  returns when it reaches a given numbered subpattern. This is used for forward
1244    references to subpatterns. We used to be able to start this scan from the
1245    current compiling point, using the current count value from cd->bracount, and
1246    do it all in a single loop, but the addition of the possibility of duplicate
1247    subpattern numbers means that we have to scan from the very start, in order to
1248    take account of such duplicates, and to use a recursive function to keep track
1249    of the different types of group.
1250    
1251    Arguments:
1252      cd           compile background data
1253      name         name to seek, or NULL if seeking a numbered subpattern
1254      lorn         name length, or subpattern number if name is NULL
1255      xmode        TRUE if we are in /x mode
1256    
1257    Returns:       the number of the found subpattern, or -1 if not found
1258    */
1259    
1260    static int
1261    find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode)
1262    {
1263    uschar *ptr = (uschar *)cd->start_pattern;
1264    int count = 0;
1265    int rc;
1266    
1267    /* If the pattern does not start with an opening parenthesis, the first call
1268    to find_parens_sub() will scan right to the end (if necessary). However, if it
1269    does start with a parenthesis, find_parens_sub() will return when it hits the
1270    matching closing parens. That is why we have to have a loop. */
1271    
1272    for (;;)
1273      {
1274      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, &count);
1275      if (rc > 0 || *ptr++ == 0) break;
1276    }    }
1277    
1278  return -1;  return rc;
1279  }  }
1280    
1281    
1282    
1283    
1284  /*************************************************  /*************************************************
1285  *      Find first significant op code            *  *      Find first significant op code            *
1286  *************************************************/  *************************************************/
# Line 1123  for (;;) Line 1330  for (;;)
1330    
1331      case OP_CALLOUT:      case OP_CALLOUT:
1332      case OP_CREF:      case OP_CREF:
1333        case OP_NCREF:
1334      case OP_RREF:      case OP_RREF:
1335        case OP_NRREF:
1336      case OP_DEF:      case OP_DEF:
1337      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1338      break;      break;
# Line 1139  for (;;) Line 1348  for (;;)
1348    
1349    
1350  /*************************************************  /*************************************************
1351  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1352  *************************************************/  *************************************************/
1353    
1354  /* 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,
1355  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.
1356  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
1357    temporarily terminated with OP_END when this function is called.
1358    
1359    This function is called when a backward assertion is encountered, so that if it
1360    fails, the error message can point to the correct place in the pattern.
1361    However, we cannot do this when the assertion contains subroutine calls,
1362    because they can be forward references. We solve this by remembering this case
1363    and doing the check at the end; a flag specifies which mode we are running in.
1364    
1365  Arguments:  Arguments:
1366    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1367    options  the compiling options    options  the compiling options
1368      atend    TRUE if called when the pattern is complete
1369      cd       the "compile data" structure
1370    
1371  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1372                 or -1 if there is no fixed length,
1373               or -2 if \C was encountered               or -2 if \C was encountered
1374                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1375  */  */
1376    
1377  static int  static int
1378  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)
1379  {  {
1380  int length = -1;  int length = -1;
1381    
# Line 1168  branch, check the length against that of Line 1388  branch, check the length against that of
1388  for (;;)  for (;;)
1389    {    {
1390    int d;    int d;
1391      uschar *ce, *cs;
1392    register int op = *cc;    register int op = *cc;
1393    switch (op)    switch (op)
1394      {      {
# Line 1175  for (;;) Line 1396  for (;;)
1396      case OP_BRA:      case OP_BRA:
1397      case OP_ONCE:      case OP_ONCE:
1398      case OP_COND:      case OP_COND:
1399      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);
1400      if (d < 0) return d;      if (d < 0) return d;
1401      branchlength += d;      branchlength += d;
1402      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1198  for (;;) Line 1419  for (;;)
1419      branchlength = 0;      branchlength = 0;
1420      break;      break;
1421    
1422        /* A true recursion implies not fixed length, but a subroutine call may
1423        be OK. If the subroutine is a forward reference, we can't deal with
1424        it until the end of the pattern, so return -3. */
1425    
1426        case OP_RECURSE:
1427        if (!atend) return -3;
1428        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1429        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1430        if (cc > cs && cc < ce) return -1;                /* Recursion */
1431        d = find_fixedlength(cs + 2, options, atend, cd);
1432        if (d < 0) return d;
1433        branchlength += d;
1434        cc += 1 + LINK_SIZE;
1435        break;
1436    
1437      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1438    
1439      case OP_ASSERT:      case OP_ASSERT:
# Line 1211  for (;;) Line 1447  for (;;)
1447    
1448      case OP_REVERSE:      case OP_REVERSE:
1449      case OP_CREF:      case OP_CREF:
1450        case OP_NCREF:
1451      case OP_RREF:      case OP_RREF:
1452        case OP_NRREF:
1453      case OP_DEF:      case OP_DEF:
1454      case OP_OPT:      case OP_OPT:
1455      case OP_CALLOUT:      case OP_CALLOUT:
1456      case OP_SOD:      case OP_SOD:
1457      case OP_SOM:      case OP_SOM:
1458        case OP_SET_SOM:
1459      case OP_EOD:      case OP_EOD:
1460      case OP_EODN:      case OP_EODN:
1461      case OP_CIRC:      case OP_CIRC:
# Line 1234  for (;;) Line 1473  for (;;)
1473      branchlength++;      branchlength++;
1474      cc += 2;      cc += 2;
1475  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1476      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1477        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1478  #endif  #endif
1479      break;      break;
1480    
# Line 1248  for (;;) Line 1485  for (;;)
1485      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1486      cc += 4;      cc += 4;
1487  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1488      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1489        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1490  #endif  #endif
1491      break;      break;
1492    
# Line 1275  for (;;) Line 1510  for (;;)
1510      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1511      case OP_WORDCHAR:      case OP_WORDCHAR:
1512      case OP_ANY:      case OP_ANY:
1513        case OP_ALLANY:
1514      branchlength++;      branchlength++;
1515      cc++;      cc++;
1516      break;      break;
# Line 1329  for (;;) Line 1565  for (;;)
1565    
1566    
1567  /*************************************************  /*************************************************
1568  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1569  *************************************************/  *************************************************/
1570    
1571  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1572  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1573    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1574    so that it can be called from pcre_study() when finding the minimum matching
1575    length.
1576    
1577  Arguments:  Arguments:
1578    code        points to start of expression    code        points to start of expression
1579    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1580    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1581    
1582  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
1583  */  */
1584    
1585  static const uschar *  const uschar *
1586  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1587  {  {
1588  for (;;)  for (;;)
1589    {    {
# Line 1357  for (;;) Line 1596  for (;;)
1596    
1597    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1598    
1599      /* Handle recursion */
1600    
1601      else if (c == OP_REVERSE)
1602        {
1603        if (number < 0) return (uschar *)code;
1604        code += _pcre_OP_lengths[c];
1605        }
1606    
1607    /* Handle capturing bracket */    /* Handle capturing bracket */
1608    
1609    else if (c == OP_CBRA)    else if (c == OP_CBRA)
# Line 1423  for (;;) Line 1670  for (;;)
1670        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1671        break;        break;
1672        }        }
1673    #else
1674        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1675  #endif  #endif
1676      }      }
1677    }    }
# Line 1516  for (;;) Line 1765  for (;;)
1765        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1766        break;        break;
1767        }        }
1768    #else
1769        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1770  #endif  #endif
1771      }      }
1772    }    }
# Line 1539  Arguments: Line 1790  Arguments:
1790    code        points to start of search    code        points to start of search
1791    endcode     points to where to stop    endcode     points to where to stop
1792    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
1793      cd          contains pointers to tables etc.
1794    
1795  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
1796  */  */
1797    
1798  static BOOL  static BOOL
1799  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
1800      compile_data *cd)
1801  {  {
1802  register int c;  register int c;
1803  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 1575  for (code = first_significant_code(code Line 1828  for (code = first_significant_code(code
1828      continue;      continue;
1829      }      }
1830    
1831      /* For a recursion/subroutine call, if its end has been reached, which
1832      implies a subroutine call, we can scan it. */
1833    
1834      if (c == OP_RECURSE)
1835        {
1836        BOOL empty_branch = FALSE;
1837        const uschar *scode = cd->start_code + GET(code, 1);
1838        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
1839        do
1840          {
1841          if (could_be_empty_branch(scode, endcode, utf8, cd))
1842            {
1843            empty_branch = TRUE;
1844            break;
1845            }
1846          scode += GET(scode, 1);
1847          }
1848        while (*scode == OP_ALT);
1849        if (!empty_branch) return FALSE;  /* All branches are non-empty */
1850        continue;
1851        }
1852    
1853    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
1854    
1855    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)
# Line 1582  for (code = first_significant_code(code Line 1857  for (code = first_significant_code(code
1857      BOOL empty_branch;      BOOL empty_branch;
1858      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
1859    
1860      /* Scan a closed bracket */      /* If a conditional group has only one branch, there is a second, implied,
1861        empty branch, so just skip over the conditional, because it could be empty.
1862        Otherwise, scan the individual branches of the group. */
1863    
1864      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;  
1865        code += GET(code, 1);        code += GET(code, 1);
1866        else
1867          {
1868          empty_branch = FALSE;
1869          do
1870            {
1871            if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
1872              empty_branch = TRUE;
1873            code += GET(code, 1);
1874            }
1875          while (*code == OP_ALT);
1876          if (!empty_branch) return FALSE;   /* All branches are non-empty */
1877        }        }
1878      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
1879      c = *code;      c = *code;
1880      continue;      continue;
1881      }      }
# Line 1653  for (code = first_significant_code(code Line 1936  for (code = first_significant_code(code
1936      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1937      case OP_WORDCHAR:      case OP_WORDCHAR:
1938      case OP_ANY:      case OP_ANY:
1939        case OP_ALLANY:
1940      case OP_ANYBYTE:      case OP_ANYBYTE:
1941      case OP_CHAR:      case OP_CHAR:
1942      case OP_CHARNC:      case OP_CHARNC:
# Line 1709  for (code = first_significant_code(code Line 1993  for (code = first_significant_code(code
1993      case OP_QUERY:      case OP_QUERY:
1994      case OP_MINQUERY:      case OP_MINQUERY:
1995      case OP_POSQUERY:      case OP_POSQUERY:
1996        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
1997        break;
1998    
1999      case OP_UPTO:      case OP_UPTO:
2000      case OP_MINUPTO:      case OP_MINUPTO:
2001      case OP_POSUPTO:      case OP_POSUPTO:
2002      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2003      break;      break;
2004  #endif  #endif
2005    
2006        /* None of the remaining opcodes are required to match a character. */
2007    
2008        default:
2009        break;
2010      }      }
2011    }    }
2012    
# Line 1737  Arguments: Line 2029  Arguments:
2029    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2030    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2031    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2032      cd          pointers to tables etc
2033    
2034  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2035  */  */
2036    
2037  static BOOL  static BOOL
2038  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2039    BOOL utf8)    BOOL utf8, compile_data *cd)
2040  {  {
2041  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2042    {    {
2043    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2044        return FALSE;
2045    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2046    }    }
2047  return TRUE;  return TRUE;
# Line 1793  int terminator;          /* Don't combin Line 2087  int terminator;          /* Don't combin
2087  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2088  for (++ptr; *ptr != 0; ptr++)  for (++ptr; *ptr != 0; ptr++)
2089    {    {
2090    if (*ptr == '\\' && ptr[1] == ']') ptr++; else    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else
2091      {      {
2092      if (*ptr == ']') return FALSE;      if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2093      if (*ptr == terminator && ptr[1] == ']')      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2094        {        {
2095        *endptr = ptr;        *endptr = ptr;
2096        return TRUE;        return TRUE;
# Line 1883  while ((ptr = (uschar *)find_recurse(ptr Line 2177  while ((ptr = (uschar *)find_recurse(ptr
2177    
2178    /* 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
2179    reference. */    reference. */
2180    
2181    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
2182      {      {
2183      offset = GET(hc, 0);      offset = GET(hc, 0);
# Line 1987  get_othercase_range(unsigned int *cptr, Line 2281  get_othercase_range(unsigned int *cptr,
2281  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2282    
2283  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2284    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2285    
2286  if (c > d) return FALSE;  if (c > d) return FALSE;
2287    
# Line 1996  next = othercase + 1; Line 2290  next = othercase + 1;
2290    
2291  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2292    {    {
2293    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2294    next++;    next++;
2295    }    }
2296    
# Line 2042  if ((options & PCRE_EXTENDED) != 0) Line 2336  if ((options & PCRE_EXTENDED) != 0)
2336    for (;;)    for (;;)
2337      {      {
2338      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2339      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2340        {        {
2341        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2342          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
# Line 2054  if ((options & PCRE_EXTENDED) != 0) Line 2348  if ((options & PCRE_EXTENDED) != 0)
2348  /* 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
2349  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2350    
2351  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2352    {    {
2353    int temperrorcode = 0;    int temperrorcode = 0;
2354    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
# Line 2079  if ((options & PCRE_EXTENDED) != 0) Line 2373  if ((options & PCRE_EXTENDED) != 0)
2373    for (;;)    for (;;)
2374      {      {
2375      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2376      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2377        {        {
2378        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2379          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
# Line 2090  if ((options & PCRE_EXTENDED) != 0) Line 2384  if ((options & PCRE_EXTENDED) != 0)
2384    
2385  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
2386    
2387  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2388    return FALSE;    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2389        return FALSE;
2390    
2391  /* 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
2392  positive single character match, "item" either contains the character or, if  positive single character match, "item" either contains the character or, if
# Line 2106  if (next >= 0) switch(op_code) Line 2401  if (next >= 0) switch(op_code)
2401    case OP_CHAR:    case OP_CHAR:
2402  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2403    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2404    #else
2405      (void)(utf8_char);  /* Keep compiler happy by referencing function argument */
2406  #endif  #endif
2407    return item != next;    return item != next;
2408    
# Line 2124  if (next >= 0) switch(op_code) Line 2421  if (next >= 0) switch(op_code)
2421      unsigned int othercase;      unsigned int othercase;
2422      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2423  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2424      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
2425  #else  #else
2426      othercase = NOTACHAR;      othercase = NOTACHAR;
2427  #endif  #endif
# Line 2145  if (next >= 0) switch(op_code) Line 2442  if (next >= 0) switch(op_code)
2442      unsigned int othercase;      unsigned int othercase;
2443      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2444  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2445      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
2446  #else  #else
2447      othercase = NOTACHAR;      othercase = NOTACHAR;
2448  #endif  #endif
# Line 2406  BOOL utf8 = FALSE; Line 2703  BOOL utf8 = FALSE;
2703  uschar *utf8_char = NULL;  uschar *utf8_char = NULL;
2704  #endif  #endif
2705    
2706  #ifdef DEBUG  #ifdef PCRE_DEBUG
2707  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2708  #endif  #endif
2709    
# Line 2459  for (;; ptr++) Line 2756  for (;; ptr++)
2756    /* Get next byte in the pattern */    /* Get next byte in the pattern */
2757    
2758    c = *ptr;    c = *ptr;
2759    
2760    /* 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
2761    previous cycle of this loop. */    previous cycle of this loop. */
2762    
2763    if (lengthptr != NULL)    if (lengthptr != NULL)
2764      {      {
2765  #ifdef DEBUG  #ifdef PCRE_DEBUG
2766      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
2767  #endif  #endif
2768      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
2769        {        {
2770        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
2771        goto FAILED;        goto FAILED;
# Line 2517  for (;; ptr++) Line 2814  for (;; ptr++)
2814    /* In the real compile phase, just check the workspace used by the forward    /* In the real compile phase, just check the workspace used by the forward
2815    reference list. */    reference list. */
2816    
2817    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
2818      {      {
2819      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
2820      goto FAILED;      goto FAILED;
# Line 2527  for (;; ptr++) Line 2824  for (;; ptr++)
2824    
2825    if (inescq && c != 0)    if (inescq && c != 0)
2826      {      {
2827      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
2828        {        {
2829        inescq = FALSE;        inescq = FALSE;
2830        ptr++;        ptr++;
# Line 2553  for (;; ptr++) Line 2850  for (;; ptr++)
2850    /* 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
2851    a quantifier. */    a quantifier. */
2852    
2853    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
2854      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
2855        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
2856    
2857    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
2858         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2569  for (;; ptr++) Line 2867  for (;; ptr++)
2867    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
2868      {      {
2869      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
2870      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
2871        {        {
2872        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2873          {          {
# Line 2594  for (;; ptr++) Line 2892  for (;; ptr++)
2892      {      {
2893      /* ===================================================================*/      /* ===================================================================*/
2894      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
2895      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
2896      case ')':      case CHAR_RIGHT_PARENTHESIS:
2897      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
2898      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
2899      *codeptr = code;      *codeptr = code;
# Line 2617  for (;; ptr++) Line 2915  for (;; ptr++)
2915      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
2916      the setting of any following char as a first character. */      the setting of any following char as a first character. */
2917    
2918      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
2919      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
2920        {        {
2921        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
# Line 2626  for (;; ptr++) Line 2924  for (;; ptr++)
2924      *code++ = OP_CIRC;      *code++ = OP_CIRC;
2925      break;      break;
2926    
2927      case '$':      case CHAR_DOLLAR_SIGN:
2928      previous = NULL;      previous = NULL;
2929      *code++ = OP_DOLL;      *code++ = OP_DOLL;
2930      break;      break;
# Line 2634  for (;; ptr++) Line 2932  for (;; ptr++)
2932      /* 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
2933      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
2934    
2935      case '.':      case CHAR_DOT:
2936      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
2937      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
2938      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
2939      previous = code;      previous = code;
2940      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
2941      break;      break;
2942    
2943    
# Line 2654  for (;; ptr++) Line 2952  for (;; ptr++)
2952      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,
2953      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
2954      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.
     */  
2955    
2956      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
2957        default (Perl) mode, it is treated as a data character. */
2958    
2959        case CHAR_RIGHT_SQUARE_BRACKET:
2960        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2961          {
2962          *errorcodeptr = ERR64;
2963          goto FAILED;
2964          }
2965        goto NORMAL_CHAR;
2966    
2967        case CHAR_LEFT_SQUARE_BRACKET:
2968      previous = code;      previous = code;
2969    
2970      /* 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
2971      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. */
2972    
2973      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2974             ptr[1] == CHAR_EQUALS_SIGN) &&
2975          check_posix_syntax(ptr, &tempptr))          check_posix_syntax(ptr, &tempptr))
2976        {        {
2977        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
2978        goto FAILED;        goto FAILED;
2979        }        }
2980    
# Line 2677  for (;; ptr++) Line 2986  for (;; ptr++)
2986      for (;;)      for (;;)
2987        {        {
2988        c = *(++ptr);        c = *(++ptr);
2989        if (c == '\\')        if (c == CHAR_BACKSLASH)
2990          {          {
2991          if (ptr[1] == 'E') ptr++;          if (ptr[1] == CHAR_E)
2992            else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;            ptr++;
2993              else break;          else if (strncmp((const char *)ptr+1,
2994                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
2995              ptr += 3;
2996            else
2997              break;
2998          }          }
2999        else if (!negate_class && c == '^')        else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
3000          negate_class = TRUE;          negate_class = TRUE;
3001        else break;        else break;
3002        }        }
3003    
3004        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
3005        an initial ']' is taken as a data character -- the code below handles
3006        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3007        [^] must match any character, so generate OP_ALLANY. */
3008    
3009        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3010            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3011          {
3012          *code++ = negate_class? OP_ALLANY : OP_FAIL;
3013          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3014          zerofirstbyte = firstbyte;
3015          break;
3016          }
3017    
3018      /* 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
3019      negation flag at the end, so that support for characters > 255 works      negation flag at the end, so that support for characters > 255 works
3020      correctly (they are all included in the class). */      correctly (they are all included in the class). */
# Line 2745  for (;; ptr++) Line 3072  for (;; ptr++)
3072    
3073        if (inescq)        if (inescq)
3074          {          {
3075          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3076            {            {
3077            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
3078            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2760  for (;; ptr++) Line 3087  for (;; ptr++)
3087        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3088        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3089    
3090        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3091            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3092            check_posix_syntax(ptr, &tempptr))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3093          {          {
3094          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3095          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3096          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3097          uschar pbits[32];          uschar pbits[32];
3098    
3099          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3100            {            {
3101            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3102            goto FAILED;            goto FAILED;
3103            }            }
3104    
3105          ptr += 2;          ptr += 2;
3106          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3107            {            {
3108            local_negate = TRUE;            local_negate = TRUE;
3109            should_flip_negation = TRUE;  /* Note negative special */            should_flip_negation = TRUE;  /* Note negative special */
# Line 2849  for (;; ptr++) Line 3176  for (;; ptr++)
3176        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
3177        character in them, so set class_charcount bigger than one. */        character in them, so set class_charcount bigger than one. */
3178    
3179        if (c == '\\')        if (c == CHAR_BACKSLASH)
3180          {          {
3181          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3182          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3183    
3184          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 */
3185          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 */
3186          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 */
3187          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3188            {            {
3189            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3190              {              {
3191              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3192              }              }
# Line 3085  for (;; ptr++) Line 3412  for (;; ptr++)
3412        entirely. The code for handling \Q and \E is messy. */        entirely. The code for handling \Q and \E is messy. */
3413    
3414        CHECK_RANGE:        CHECK_RANGE:
3415        while (ptr[1] == '\\' && ptr[2] == 'E')        while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3416          {          {
3417          inescq = FALSE;          inescq = FALSE;
3418          ptr += 2;          ptr += 2;
# Line 3095  for (;; ptr++) Line 3422  for (;; ptr++)
3422    
3423        /* Remember \r or \n */        /* Remember \r or \n */
3424    
3425        if (c == '\r' || c == '\n') cd->external_flags |= PCRE_HASCRORLF;        if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3426    
3427        /* Check for range */        /* Check for range */
3428    
3429        if (!inescq && ptr[1] == '-')        if (!inescq && ptr[1] == CHAR_MINUS)
3430          {          {
3431          int d;          int d;
3432          ptr += 2;          ptr += 2;
3433          while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3434    
3435          /* 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
3436          mode. */          mode. */
3437    
3438          while (*ptr == '\\' && ptr[1] == 'Q')          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3439            {            {
3440            ptr += 2;            ptr += 2;
3441            if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }            if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3442                { ptr += 2; continue; }
3443            inescq = TRUE;            inescq = TRUE;
3444            break;            break;
3445            }            }
3446    
3447          if (*ptr == 0 || (!inescq && *ptr == ']'))          if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3448            {            {
3449            ptr = oldptr;            ptr = oldptr;
3450            goto LONE_SINGLE_CHARACTER;            goto LONE_SINGLE_CHARACTER;
# Line 3135  for (;; ptr++) Line 3463  for (;; ptr++)
3463          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
3464          in such circumstances. */          in such circumstances. */
3465    
3466          if (!inescq && d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3467            {            {
3468            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3469            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
# Line 3145  for (;; ptr++) Line 3473  for (;; ptr++)
3473    
3474            if (d < 0)            if (d < 0)
3475              {              {
3476              if (d == -ESC_b) d = '\b';              if (d == -ESC_b) d = CHAR_BS;
3477              else if (d == -ESC_X) d = 'X';              else if (d == -ESC_X) d = CHAR_X;
3478              else if (d == -ESC_R) d = 'R'; else              else if (d == -ESC_R) d = CHAR_R; else
3479                {                {
3480                ptr = oldptr;                ptr = oldptr;
3481                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 3168  for (;; ptr++) Line 3496  for (;; ptr++)
3496    
3497          /* Remember \r or \n */          /* Remember \r or \n */
3498    
3499          if (d == '\r' || d == '\n') cd->external_flags |= PCRE_HASCRORLF;          if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3500    
3501          /* 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
3502          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 3288  for (;; ptr++) Line 3616  for (;; ptr++)
3616          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3617            {            {
3618            unsigned int othercase;            unsigned int othercase;
3619            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)            if ((othercase = UCD_OTHERCASE(c)) != c)
3620              {              {
3621              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3622              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 3315  for (;; ptr++) Line 3643  for (;; ptr++)
3643    
3644      /* Loop until ']' reached. This "while" is the end of the "do" above. */      /* Loop until ']' reached. This "while" is the end of the "do" above. */
3645    
3646      while ((c = *(++ptr)) != 0 && (c != ']' || inescq));      while ((c = *(++ptr)) != 0 && (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
3647    
3648      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)                          /* Missing terminating ']' */
3649        {        {
# Line 3460  we set the flag only if there is a liter Line 3788  we set the flag only if there is a liter
3788      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
3789      has been tested above. */      has been tested above. */
3790    
3791      case '{':      case CHAR_LEFT_CURLY_BRACKET:
3792      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
3793      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
3794      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
3795      goto REPEAT;      goto REPEAT;
3796    
3797      case '*':      case CHAR_ASTERISK:
3798      repeat_min = 0;      repeat_min = 0;
3799      repeat_max = -1;      repeat_max = -1;
3800      goto REPEAT;      goto REPEAT;
3801    
3802      case '+':      case CHAR_PLUS:
3803      repeat_min = 1;      repeat_min = 1;
3804      repeat_max = -1;      repeat_max = -1;
3805      goto REPEAT;      goto REPEAT;
3806    
3807      case '?':      case CHAR_QUESTION_MARK:
3808      repeat_min = 0;      repeat_min = 0;
3809      repeat_max = 1;      repeat_max = 1;
3810    
# Line 3511  we set the flag only if there is a liter Line 3839  we set the flag only if there is a liter
3839      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
3840      repeat type to the non-default. */      repeat type to the non-default. */
3841    
3842      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
3843        {        {
3844        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
3845        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
3846        ptr++;        ptr++;
3847        }        }
3848      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
3849        {        {
3850        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
3851        ptr++;        ptr++;
# Line 3632  we set the flag only if there is a liter Line 3960  we set the flag only if there is a liter
3960    
3961        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
3962    
3963          /*--------------------------------------------------------------------*/
3964          /* This code is obsolete from release 8.00; the restriction was finally
3965          removed: */
3966    
3967        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
3968        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
3969    
3970        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
3971          /*--------------------------------------------------------------------*/
3972    
3973        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
3974    
# Line 3782  we set the flag only if there is a liter Line 4115  we set the flag only if there is a liter
4115          goto END_REPEAT;          goto END_REPEAT;
4116          }          }
4117    
4118          /*--------------------------------------------------------------------*/
4119          /* This code is obsolete from release 8.00; the restriction was finally
4120          removed: */
4121    
4122        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4123        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4124    
4125        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4126          /*--------------------------------------------------------------------*/
4127    
4128        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4129          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 3851  we set the flag only if there is a liter Line 4189  we set the flag only if there is a liter
4189          **   code = previous;          **   code = previous;
4190          **   goto END_REPEAT;          **   goto END_REPEAT;
4191          **   }          **   }
4192    
4193          However, that fails when a group is referenced as a subroutine from          However, that fails when a group is referenced as a subroutine from
4194          elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it          elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it
4195          so that it is skipped on execution. As we don't have a list of which          so that it is skipped on execution. As we don't have a list of which
4196          groups are referenced, we cannot do this selectively.          groups are referenced, we cannot do this selectively.
4197    
4198          If the maximum is 1 or unlimited, we just have to stick in the BRAZERO          If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4199          and do no more at this point. However, we do need to adjust any          and do no more at this point. However, we do need to adjust any
# Line 3874  we set the flag only if there is a liter Line 4212  we set the flag only if there is a liter
4212              {              {
4213              *previous++ = OP_SKIPZERO;              *previous++ = OP_SKIPZERO;
4214              goto END_REPEAT;              goto END_REPEAT;
4215              }              }
4216            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4217            }            }
4218    
# Line 3920  we set the flag only if there is a liter Line 4258  we set the flag only if there is a liter
4258            {            {
4259            /* 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
4260            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
4261            potential integer overflow. */            potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
4262              integer type when available, otherwise double. */
4263    
4264            if (lengthptr != NULL)            if (lengthptr != NULL)
4265              {              {
4266              int delta = (repeat_min - 1)*length_prevgroup;              int delta = (repeat_min - 1)*length_prevgroup;
4267              if ((double)(repeat_min - 1)*(double)length_prevgroup >              if ((INT64_OR_DOUBLE)(repeat_min - 1)*
4268                                                              (double)INT_MAX ||                    (INT64_OR_DOUBLE)length_prevgroup >
4269                        (INT64_OR_DOUBLE)INT_MAX ||
4270                  OFLOW_MAX - *lengthptr < delta)                  OFLOW_MAX - *lengthptr < delta)
4271                {                {
4272                *errorcodeptr = ERR20;                *errorcodeptr = ERR20;
# Line 3972  we set the flag only if there is a liter Line 4312  we set the flag only if there is a liter
4312          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
4313          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
4314          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
4315          paranoid checks to avoid integer overflow. */          paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
4316            a 64-bit integer type when available, otherwise double. */
4317    
4318          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4319            {            {
4320            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4321                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4322            if ((double)repeat_max *            if ((INT64_OR_DOUBLE)repeat_max *
4323                  (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)                  (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4324                    > (double)INT_MAX ||                    > (INT64_OR_DOUBLE)INT_MAX ||
4325                OFLOW_MAX - *lengthptr < delta)                OFLOW_MAX - *lengthptr < delta)
4326              {              {
4327              *errorcodeptr = ERR20;              *errorcodeptr = ERR20;
# Line 4057  we set the flag only if there is a liter Line 4398  we set the flag only if there is a liter
4398            uschar *scode = bracode;            uschar *scode = bracode;
4399            do            do
4400              {              {
4401              if (could_be_empty_branch(scode, ketcode, utf8))              if (could_be_empty_branch(scode, ketcode, utf8, cd))
4402                {                {
4403                *bracode += OP_SBRA - OP_BRA;                *bracode += OP_SBRA - OP_BRA;
4404                break;                break;
# Line 4069  we set the flag only if there is a liter Line 4410  we set the flag only if there is a liter
4410          }          }
4411        }        }
4412    
4413        /* If previous is OP_FAIL, it was generated by an empty class [] in
4414        JavaScript mode. The other ways in which OP_FAIL can be generated, that is
4415        by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
4416        error above. We can just ignore the repeat in JS case. */
4417    
4418        else if (*previous == OP_FAIL) goto END_REPEAT;
4419    
4420      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
4421    
4422      else      else
# Line 4093  we set the flag only if there is a liter Line 4441  we set the flag only if there is a liter
4441      if (possessive_quantifier)      if (possessive_quantifier)
4442        {        {
4443        int len;        int len;
4444        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||  
4445            *tempcode == OP_NOTEXACT)        if (*tempcode == OP_TYPEEXACT)
4446          tempcode += _pcre_OP_lengths[*tempcode] +          tempcode += _pcre_OP_lengths[*tempcode] +
4447            ((*tempcode == OP_TYPEEXACT &&            ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
4448               (tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP))? 2:0);  
4449          else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
4450            {
4451            tempcode += _pcre_OP_lengths[*tempcode];
4452    #ifdef SUPPORT_UTF8
4453            if (utf8 && tempcode[-1] >= 0xc0)
4454              tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];
4455    #endif
4456            }
4457    
4458        len = code - tempcode;        len = code - tempcode;
4459        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4460          {          {
# Line 4116  we set the flag only if there is a liter Line 4473  we set the flag only if there is a liter
4473          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4474          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4475    
4476            /* Because we are moving code along, we must ensure that any
4477            pending recursive references are updated. */
4478    
4479          default:          default:
4480            *code = OP_END;
4481            adjust_recurse(tempcode, 1 + LINK_SIZE, utf8, cd, save_hwm);
4482          memmove(tempcode + 1+LINK_SIZE, tempcode, len);          memmove(tempcode + 1+LINK_SIZE, tempcode, len);
4483          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
4484          len += 1 + LINK_SIZE;          len += 1 + LINK_SIZE;
# Line 4143  we set the flag only if there is a liter Line 4505  we set the flag only if there is a liter
4505      lookbehind or option setting or condition or all the other extended      lookbehind or option setting or condition or all the other extended
4506      parenthesis forms.  */      parenthesis forms.  */
4507    
4508      case '(':      case CHAR_LEFT_PARENTHESIS:
4509      newoptions = options;      newoptions = options;
4510      skipbytes = 0;      skipbytes = 0;
4511      bravalue = OP_CBRA;      bravalue = OP_CBRA;
4512      save_hwm = cd->hwm;      save_hwm = cd->hwm;
4513      reset_bracount = FALSE;      reset_bracount = FALSE;
4514    
4515      /* First deal with various "verbs" that can be introduced by '*'. */      /* First deal with various "verbs" that can be introduced by '*'. */
4516    
4517      if (*(++ptr) == '*' && (cd->ctypes[ptr[1]] & ctype_letter) != 0)      if (*(++ptr) == CHAR_ASTERISK && (cd->ctypes[ptr[1]] & ctype_letter) != 0)
4518        {        {
4519        int i, namelen;        int i, namelen;
4520        const char *vn = verbnames;        const char *vn = verbnames;
4521        const uschar *name = ++ptr;        const uschar *name = ++ptr;
4522        previous = NULL;        previous = NULL;
4523        while ((cd->ctypes[*++ptr] & ctype_letter) != 0);        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
4524        if (*ptr == ':')        if (*ptr == CHAR_COLON)
4525          {          {
4526          *errorcodeptr = ERR59;   /* Not supported */          *errorcodeptr = ERR59;   /* Not supported */
4527          goto FAILED;          goto FAILED;
4528          }          }
4529        if (*ptr != ')')        if (*ptr != CHAR_RIGHT_PARENTHESIS)
4530          {          {
4531          *errorcodeptr = ERR60;          *errorcodeptr = ERR60;
4532          goto FAILED;          goto FAILED;
# Line 4175  we set the flag only if there is a liter Line 4537  we set the flag only if there is a liter
4537          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
4538              strncmp((char *)name, vn, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
4539            {            {
4540            *code = verbs[i].op;            /* Check for open captures before ACCEPT */
4541            if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;  
4542              if (verbs[i].op == OP_ACCEPT)
4543                {
4544                open_capitem *oc;
4545                cd->had_accept = TRUE;
4546                for (oc = cd->open_caps; oc != NULL; oc = oc->next)
4547                  {
4548                  *code++ = OP_CLOSE;
4549                  PUT2INC(code, 0, oc->number);
4550                  }
4551                }
4552              *code++ = verbs[i].op;
4553            break;            break;
4554            }            }
4555          vn += verbs[i].len + 1;          vn += verbs[i].len + 1;
# Line 4189  we set the flag only if there is a liter Line 4562  we set the flag only if there is a liter
4562      /* Deal with the extended parentheses; all are introduced by '?', and the      /* Deal with the extended parentheses; all are introduced by '?', and the
4563      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. */
4564    
4565      else if (*ptr == '?')      else if (*ptr == CHAR_QUESTION_MARK)
4566        {        {
4567        int i, set, unset, namelen;        int i, set, unset, namelen;
4568        int *optset;        int *optset;
# Line 4198  we set the flag only if there is a liter Line 4571  we set the flag only if there is a liter
4571    
4572        switch (*(++ptr))        switch (*(++ptr))
4573          {          {
4574          case '#':                 /* Comment; skip to ket */          case CHAR_NUMBER_SIGN:                 /* Comment; skip to ket */
4575          ptr++;          ptr++;
4576          while (*ptr != 0 && *ptr != ')') ptr++;          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
4577          if (*ptr == 0)          if (*ptr == 0)
4578            {            {
4579            *errorcodeptr = ERR18;            *errorcodeptr = ERR18;
# Line 4210  we set the flag only if there is a liter Line 4583  we set the flag only if there is a liter
4583    
4584    
4585          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4586          case '|':                 /* Reset capture count for each branch */          case CHAR_VERTICAL_LINE:  /* Reset capture count for each branch */
4587          reset_bracount = TRUE;          reset_bracount = TRUE;
4588          /* Fall through */          /* Fall through */
4589    
4590          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4591          case ':':                 /* Non-capturing bracket */          case CHAR_COLON:          /* Non-capturing bracket */
4592          bravalue = OP_BRA;          bravalue = OP_BRA;
4593          ptr++;          ptr++;
4594          break;          break;
4595    
4596    
4597          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4598          case '(':          case CHAR_LEFT_PARENTHESIS:
4599          bravalue = OP_COND;       /* Conditional group */          bravalue = OP_COND;       /* Conditional group */
4600    
4601          /* 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 4242  we set the flag only if there is a liter Line 4615  we set the flag only if there is a liter
4615          the switch. This will take control down to where bracketed groups,          the switch. This will take control down to where bracketed groups,
4616          including assertions, are processed. */          including assertions, are processed. */
4617    
4618          if (ptr[1] == '?' && (ptr[2] == '=' || ptr[2] == '!' || ptr[2] == '<'))          if (ptr[1] == CHAR_QUESTION_MARK && (ptr[2] == CHAR_EQUALS_SIGN ||
4619                ptr[2] == CHAR_EXCLAMATION_MARK || ptr[2] == CHAR_LESS_THAN_SIGN))
4620            break;            break;
4621    
4622          /* 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 4254  we set the flag only if there is a liter Line 4628  we set the flag only if there is a liter
4628    
4629          /* Check for a test for recursion in a named group. */          /* Check for a test for recursion in a named group. */
4630    
4631          if (ptr[1] == 'R' && ptr[2] == '&')          if (ptr[1] == CHAR_R && ptr[2] == CHAR_AMPERSAND)
4632            {            {
4633            terminator = -1;            terminator = -1;
4634            ptr += 2;            ptr += 2;
# Line 4264  we set the flag only if there is a liter Line 4638  we set the flag only if there is a liter
4638          /* 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
4639          syntax (?(<name>) or (?('name') */          syntax (?(<name>) or (?('name') */
4640    
4641          else if (ptr[1] == '<')          else if (ptr[1] == CHAR_LESS_THAN_SIGN)
4642            {            {
4643            terminator = '>';            terminator = CHAR_GREATER_THAN_SIGN;
4644            ptr++;            ptr++;
4645            }            }
4646          else if (ptr[1] == '\'')          else if (ptr[1] == CHAR_APOSTROPHE)
4647            {            {
4648            terminator = '\'';            terminator = CHAR_APOSTROPHE;
4649            ptr++;            ptr++;
4650            }            }
4651          else          else
4652            {            {
4653            terminator = 0;            terminator = 0;
4654            if (ptr[1] == '-' || ptr[1] == '+') refsign = *(++ptr);            if (ptr[1] == CHAR_MINUS || ptr[1] == CHAR_PLUS) refsign = *(++ptr);
4655            }            }
4656    
4657          /* 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 4297  we set the flag only if there is a liter Line 4671  we set the flag only if there is a liter
4671            {            {
4672            if (recno >= 0)            if (recno >= 0)
4673              recno = ((digitab[*ptr] & ctype_digit) != 0)?              recno = ((digitab[*ptr] & ctype_digit) != 0)?
4674                recno * 10 + *ptr - '0' : -1;                recno * 10 + *ptr - CHAR_0 : -1;
4675            ptr++;            ptr++;
4676            }            }
4677          namelen = ptr - name;          namelen = ptr - name;
4678    
4679          if ((terminator > 0 && *ptr++ != terminator) || *ptr++ != ')')          if ((terminator > 0 && *ptr++ != terminator) ||
4680                *ptr++ != CHAR_RIGHT_PARENTHESIS)
4681            {            {
4682            ptr--;      /* Error offset */            ptr--;      /* Error offset */
4683            *errorcodeptr = ERR26;            *errorcodeptr = ERR26;
# Line 4324  we set the flag only if there is a liter Line 4699  we set the flag only if there is a liter
4699              *errorcodeptr = ERR58;              *errorcodeptr = ERR58;
4700              goto FAILED;              goto FAILED;
4701              }              }
4702            recno = (refsign == '-')?            recno = (refsign == CHAR_MINUS)?
4703              cd->bracount - recno + 1 : recno +cd->bracount;              cd->bracount - recno + 1 : recno +cd->bracount;
4704            if (recno <= 0 || recno > cd->final_bracount)            if (recno <= 0 || recno > cd->final_bracount)
4705              {              {
# Line 4336  we set the flag only if there is a liter Line 4711  we set the flag only if there is a liter
4711            }            }
4712    
4713          /* Otherwise (did not start with "+" or "-"), start by looking for the          /* Otherwise (did not start with "+" or "-"), start by looking for the
4714          name. */          name. If we find a name, add one to the opcode to change OP_CREF or
4715            OP_RREF into OP_NCREF or OP_NRREF. These behave exactly the same,
4716            except they record that the reference was originally to a name. The
4717            information is used to check duplicate names. */
4718    
4719          slot = cd->name_table;          slot = cd->name_table;
4720          for (i = 0; i < cd->names_found; i++)          for (i = 0; i < cd->names_found; i++)
# Line 4351  we set the flag only if there is a liter Line 4729  we set the flag only if there is a liter
4729            {            {
4730            recno = GET2(slot, 0);            recno = GET2(slot, 0);
4731            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4732              code[1+LINK_SIZE]++;
4733            }            }
4734    
4735          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
4736    
4737          else if ((i = find_parens(ptr, cd->bracount, name, namelen,          else if ((i = find_parens(cd, name, namelen,
4738                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0)) > 0)
4739            {            {
4740            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
4741              code[1+LINK_SIZE]++;
4742            }            }
4743    
4744          /* If terminator == 0 it means that the name followed directly after          /* If terminator == 0 it means that the name followed directly after
# Line 4376  we set the flag only if there is a liter Line 4756  we set the flag only if there is a liter
4756          /* Check for (?(R) for recursion. Allow digits after R to specify a          /* Check for (?(R) for recursion. Allow digits after R to specify a
4757          specific group number. */          specific group number. */
4758    
4759          else if (*name == 'R')          else if (*name == CHAR_R)
4760            {            {
4761            recno = 0;            recno = 0;
4762            for (i = 1; i < namelen; i++)            for (i = 1; i < namelen; i++)
# Line 4386  we set the flag only if there is a liter Line 4766  we set the flag only if there is a liter
4766                *errorcodeptr = ERR15;                *errorcodeptr = ERR15;
4767                goto FAILED;                goto FAILED;
4768                }                }
4769              recno = recno * 10 + name[i] - '0';              recno = recno * 10 + name[i] - CHAR_0;
4770              }              }
4771            if (recno == 0) recno = RREF_ANY;            if (recno == 0) recno = RREF_ANY;
4772            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */
# Line 4396  we set the flag only if there is a liter Line 4776  we set the flag only if there is a liter
4776          /* Similarly, check for the (?(DEFINE) "condition", which is always          /* Similarly, check for the (?(DEFINE) "condition", which is always
4777          false. */          false. */
4778    
4779          else if (namelen == 6 && strncmp((char *)name, "DEFINE", 6) == 0)          else if (namelen == 6 && strncmp((char *)name, STRING_DEFINE, 6) == 0)
4780            {            {
4781            code[1+LINK_SIZE] = OP_DEF;            code[1+LINK_SIZE] = OP_DEF;
4782            skipbytes = 1;            skipbytes = 1;
# Line 4421  we set the flag only if there is a liter Line 4801  we set the flag only if there is a liter
4801    
4802    
4803          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4804          case '=':                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
4805          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
4806          ptr++;          ptr++;
4807          break;          break;
4808    
4809    
4810          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4811          case '!':                 /* Negative lookahead */          case CHAR_EXCLAMATION_MARK:            /* Negative lookahead */
4812          ptr++;          ptr++;
4813          if (*ptr == ')')          /* Optimize (?!) */          if (*ptr == CHAR_RIGHT_PARENTHESIS)    /* Optimize (?!) */
4814            {            {
4815            *code++ = OP_FAIL;            *code++ = OP_FAIL;
4816            previous = NULL;            previous = NULL;
# Line 4441  we set the flag only if there is a liter Line 4821  we set the flag only if there is a liter
4821    
4822    
4823          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4824          case '<':                 /* Lookbehind or named define */          case CHAR_LESS_THAN_SIGN:              /* Lookbehind or named define */
4825          switch (ptr[1])          switch (ptr[1])
4826            {            {
4827            case '=':               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
4828            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
4829            ptr += 2;            ptr += 2;
4830            break;            break;
4831    
4832            case '!':               /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
4833            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
4834            ptr += 2;            ptr += 2;
4835            break;            break;
# Line 4464  we set the flag only if there is a liter Line 4844  we set the flag only if there is a liter
4844    
4845    
4846          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4847          case '>':                 /* One-time brackets */          case CHAR_GREATER_THAN_SIGN:           /* One-time brackets */
4848          bravalue = OP_ONCE;          bravalue = OP_ONCE;
4849          ptr++;          ptr++;
4850          break;          break;
4851    
4852    
4853          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4854          case 'C':                 /* Callout - may be followed by digits; */          case CHAR_C:                 /* Callout - may be followed by digits; */
4855          previous_callout = code;  /* Save for later completion */          previous_callout = code;  /* Save for later completion */
4856          after_manual_callout = 1; /* Skip one item before completing */          after_manual_callout = 1; /* Skip one item before completing */
4857          *code++ = OP_CALLOUT;          *code++ = OP_CALLOUT;
4858            {            {
4859            int n = 0;            int n = 0;
4860            while ((digitab[*(++ptr)] & ctype_digit) != 0)            while ((digitab[*(++ptr)] & ctype_digit) != 0)
4861              n = n * 10 + *ptr - '0';              n = n * 10 + *ptr - CHAR_0;
4862            if (*ptr != ')')            if (*ptr != CHAR_RIGHT_PARENTHESIS)
4863              {              {
4864              *errorcodeptr = ERR39;              *errorcodeptr = ERR39;
4865              goto FAILED;              goto FAILED;
# Line 4499  we set the flag only if there is a liter Line 4879  we set the flag only if there is a liter
4879    
4880    
4881          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4882          case 'P':                 /* Python-style named subpattern handling */          case CHAR_P:              /* Python-style named subpattern handling */
4883          if (*(++ptr) == '=' || *ptr == '>')  /* Reference or recursion */          if (*(++ptr) == CHAR_EQUALS_SIGN ||
4884                *ptr == CHAR_GREATER_THAN_SIGN)  /* Reference or recursion */
4885            {            {
4886            is_recurse = *ptr == '>';            is_recurse = *ptr == CHAR_GREATER_THAN_SIGN;
4887            terminator = ')';            terminator = CHAR_RIGHT_PARENTHESIS;
4888            goto NAMED_REF_OR_RECURSE;            goto NAMED_REF_OR_RECURSE;
4889            }            }
4890          else if (*ptr != '<')    /* Test for Python-style definition */          else if (*ptr != CHAR_LESS_THAN_SIGN)  /* Test for Python-style defn */
4891            {            {
4892            *errorcodeptr = ERR41;            *errorcodeptr = ERR41;
4893            goto FAILED;            goto FAILED;
# Line 4516  we set the flag only if there is a liter Line 4897  we set the flag only if there is a liter
4897    
4898          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4899          DEFINE_NAME:    /* Come here from (?< handling */          DEFINE_NAME:    /* Come here from (?< handling */
4900          case '\'':          case CHAR_APOSTROPHE:
4901            {            {
4902            terminator = (*ptr == '<')? '>' : '\'';            terminator = (*ptr == CHAR_LESS_THAN_SIGN)?
4903                CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
4904            name = ++ptr;            name = ++ptr;
4905    
4906            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
# Line 4549  we set the flag only if there is a liter Line 4931  we set the flag only if there is a liter
4931                }                }
4932              }              }
4933    
4934            /* In the real compile, create the entry in the table */            /* In the real compile, create the entry in the table, maintaining
4935              alphabetical order. Duplicate names for different numbers are
4936              permitted only if PCRE_DUPNAMES is set. Duplicate names for the same
4937              number are always OK. (An existing number can be re-used if (?|
4938              appears in the pattern.) In either event, a duplicate name results in
4939              a duplicate entry in the table, even if the number is the same. This
4940              is because the number of names, and hence the table size, is computed
4941              in the pre-compile, and it affects various numbers and pointers which
4942              would all have to be modified, and the compiled code moved down, if
4943              duplicates with the same number were omitted from the table. This
4944              doesn't seem worth the hassle. However, *different* names for the
4945              same number are not permitted. */
4946    
4947            else            else
4948              {              {
4949                BOOL dupname = FALSE;
4950              slot = cd->name_table;              slot = cd->name_table;
4951    
4952              for (i = 0; i < cd->names_found; i++)              for (i = 0; i < cd->names_found; i++)
4953                {                {
4954                int crc = memcmp(name, slot+2, namelen);                int crc = memcmp(name, slot+2, namelen);
# Line 4561  we set the flag only if there is a liter Line 4956  we set the flag only if there is a liter
4956                  {                  {
4957                  if (slot[2+namelen] == 0)                  if (slot[2+namelen] == 0)
4958                    {                    {
4959                    if ((options & PCRE_DUPNAMES) == 0)                    if (GET2(slot, 0) != cd->bracount + 1 &&
4960                          (options & PCRE_DUPNAMES) == 0)
4961                      {                      {
4962                      *errorcodeptr = ERR43;                      *errorcodeptr = ERR43;
4963                      goto FAILED;                      goto FAILED;
4964                      }                      }
4965                      else dupname = TRUE;
4966                    }                    }
4967                  else crc = -1;      /* Current name is substring */                  else crc = -1;      /* Current name is a substring */
4968                  }                  }
4969    
4970                  /* Make space in the table and break the loop for an earlier
4971                  name. For a duplicate or later name, carry on. We do this for
4972                  duplicates so that in the simple case (when ?(| is not used) they
4973                  are in order of their numbers. */
4974    
4975                if (crc < 0)                if (crc < 0)
4976                  {                  {
4977                  memmove(slot + cd->name_entry_size, slot,                  memmove(slot + cd->name_entry_size, slot,
4978                    (cd->names_found - i) * cd->name_entry_size);                    (cd->names_found - i) * cd->name_entry_size);
4979                  break;                  break;
4980                  }                  }
4981    
4982                  /* Continue the loop for a later or duplicate name */
4983    
4984                slot += cd->name_entry_size;                slot += cd->name_entry_size;
4985                }                }
4986    
4987                /* For non-duplicate names, check for a duplicate number before
4988                adding the new name. */
4989    
4990                if (!dupname)
4991                  {
4992                  uschar *cslot = cd->name_table;
4993                  for (i = 0; i < cd->names_found; i++)
4994                    {
4995                    if (cslot != slot)
4996                      {
4997                      if (GET2(cslot, 0) == cd->bracount + 1)
4998                        {
4999                        *errorcodeptr = ERR65;
5000                        goto FAILED;
5001                        }
5002                      }
5003                    else i--;
5004                    cslot += cd->name_entry_size;
5005                    }
5006                  }
5007    
5008              PUT2(slot, 0, cd->bracount + 1);              PUT2(slot, 0, cd->bracount + 1);
5009              memcpy(slot + 2, name, namelen);              memcpy(slot + 2, name, namelen);
5010              slot[2+namelen] = 0;              slot[2+namelen] = 0;
5011              }              }
5012            }            }
5013    
5014          /* In both cases, count the number of names we've encountered. */          /* In both pre-compile and compile, count the number of names we've
5015            encountered. */
5016    
         ptr++;                    /* Move past > or ' */  
5017          cd->names_found++;          cd->names_found++;
5018            ptr++;                    /* Move past > or ' */
5019          goto NUMBERED_GROUP;          goto NUMBERED_GROUP;
5020    
5021    
5022          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5023          case '&':                 /* Perl recursion/subroutine syntax */          case CHAR_AMPERSAND:            /* Perl recursion/subroutine syntax */
5024          terminator = ')';          terminator = CHAR_RIGHT_PARENTHESIS;
5025          is_recurse = TRUE;          is_recurse = TRUE;
5026          /* Fall through */          /* Fall through */
5027    
# Line 4652  we set the flag only if there is a liter Line 5080  we set the flag only if there is a liter
5080              recno = GET2(slot, 0);              recno = GET2(slot, 0);
5081              }              }
5082            else if ((recno =                /* Forward back reference */            else if ((recno =                /* Forward back reference */
5083                      find_parens(ptr, cd->bracount, name, namelen,                      find_parens(cd, name, namelen,
5084                        (options & PCRE_EXTENDED) != 0)) <= 0)                        (options & PCRE_EXTENDED) != 0)) <= 0)
5085              {              {
5086              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
# Line 4668  we set the flag only if there is a liter Line 5096  we set the flag only if there is a liter
5096    
5097    
5098          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5099          case 'R':                 /* Recursion */          case CHAR_R:              /* Recursion */
5100          ptr++;                    /* Same as (?0)      */          ptr++;                    /* Same as (?0)      */
5101          /* Fall through */          /* Fall through */
5102    
5103    
5104          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5105          case '-': case '+':          case CHAR_MINUS: case CHAR_PLUS:  /* Recursion or subroutine */
5106          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:
5107          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:
5108            {            {
5109            const uschar *called;            const uschar *called;
5110            terminator = ')';            terminator = CHAR_RIGHT_PARENTHESIS;
5111    
5112            /* Come here from the \g<...> and \g'...' code (Oniguruma            /* Come here from the \g<...> and \g'...' code (Oniguruma
5113            compatibility). However, the syntax has been checked to ensure that            compatibility). However, the syntax has been checked to ensure that
5114            the ... are a (signed) number, so that neither ERR63 nor ERR29 will            the ... are a (signed) number, so that neither ERR63 nor ERR29 will
5115            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
5116            ever be taken. */            ever be taken. */
   
           HANDLE_NUMERICAL_RECURSION:  
5117    
5118            if ((refsign = *ptr) == '+')            HANDLE_NUMERICAL_RECURSION:
5119    
5120              if ((refsign = *ptr) == CHAR_PLUS)
5121              {              {
5122              ptr++;              ptr++;
5123              if ((digitab[*ptr] & ctype_digit) == 0)              if ((digitab[*ptr] & ctype_digit) == 0)
# Line 4698  we set the flag only if there is a liter Line 5126  we set the flag only if there is a liter
5126                goto FAILED;                goto FAILED;
5127                }                }
5128              }              }
5129            else if (refsign == '-')            else if (refsign == CHAR_MINUS)
5130              {              {
5131              if ((digitab[ptr[1]] & ctype_digit) == 0)              if ((digitab[ptr[1]] & ctype_digit) == 0)
5132                goto OTHER_CHAR_AFTER_QUERY;                goto OTHER_CHAR_AFTER_QUERY;
# Line 4707  we set the flag only if there is a liter Line 5135  we set the flag only if there is a liter
5135    
5136            recno = 0;            recno = 0;
5137            while((digitab[*ptr] & ctype_digit) != 0)            while((digitab[*ptr] & ctype_digit) != 0)
5138              recno = recno * 10 + *ptr++ - '0';              recno = recno * 10 + *ptr++ - CHAR_0;
5139    
5140            if (*ptr != terminator)            if (*ptr != terminator)
5141              {              {
# Line 4715  we set the flag only if there is a liter Line 5143  we set the flag only if there is a liter
5143              goto FAILED;              goto FAILED;
5144              }              }
5145    
5146            if (refsign == '-')            if (refsign == CHAR_MINUS)
5147              {              {
5148              if (recno == 0)              if (recno == 0)
5149                {                {
# Line 4729  we set the flag only if there is a liter Line 5157  we set the flag only if there is a liter
5157                goto FAILED;                goto FAILED;
5158                }                }
5159              }              }
5160            else if (refsign == '+')            else if (refsign == CHAR_PLUS)
5161              {              {
5162              if (recno == 0)              if (recno == 0)
5163                {                {
# Line 4756  we set the flag only if there is a liter Line 5184  we set the flag only if there is a liter
5184            if (lengthptr == NULL)            if (lengthptr == NULL)
5185              {              {
5186              *code = OP_END;              *code = OP_END;
5187              if (recno != 0) called = find_bracket(cd->start_code, utf8, recno);              if (recno != 0)
5188                  called = _pcre_find_bracket(cd->start_code, utf8, recno);
5189    
5190              /* Forward reference */              /* Forward reference */
5191    
5192              if (called == NULL)              if (called == NULL)
5193                {                {
5194                if (find_parens(ptr, cd->bracount, NULL, recno,                if (find_parens(cd, NULL, recno,
5195                     (options & PCRE_EXTENDED) != 0) < 0)                      (options & PCRE_EXTENDED) != 0) < 0)
5196                  {                  {
5197                  *errorcodeptr = ERR15;                  *errorcodeptr = ERR15;
5198                  goto FAILED;                  goto FAILED;
5199                  }                  }
5200    
5201                  /* Fudge the value of "called" so that when it is inserted as an
5202                  offset below, what it actually inserted is the reference number
5203                  of the group. */
5204    
5205                called = cd->start_code + recno;                called = cd->start_code + recno;
5206                PUTINC(cd->hwm, 0, code + 2 + LINK_SIZE - cd->start_code);                PUTINC(cd->hwm, 0, code + 2 + LINK_SIZE - cd->start_code);
5207                }                }
# Line 4777  we set the flag only if there is a liter Line 5211  we set the flag only if there is a liter
5211              recursion that could loop for ever, and diagnose that case. */              recursion that could loop for ever, and diagnose that case. */
5212    
5213              else if (GET(called, 1) == 0 &&              else if (GET(called, 1) == 0 &&
5214                       could_be_empty(called, code, bcptr, utf8))                       could_be_empty(called, code, bcptr, utf8, cd))
5215                {                {
5216                *errorcodeptr = ERR40;                *errorcodeptr = ERR40;
5217                goto FAILED;                goto FAILED;
# Line 4815  we set the flag only if there is a liter Line 5249  we set the flag only if there is a liter
5249          set = unset = 0;          set = unset = 0;
5250          optset = &set;          optset = &set;
5251    
5252          while (*ptr != ')' && *ptr != ':')          while (*ptr != CHAR_RIGHT_PARENTHESIS && *ptr != CHAR_COLON)
5253            {            {
5254            switch (*ptr++)            switch (*ptr++)
5255              {              {
5256              case '-': optset = &unset; break;              case CHAR_MINUS: optset = &unset; break;
5257    
5258              case 'J':    /* Record that it changed in the external options */              case CHAR_J:    /* Record that it changed in the external options */
5259              *optset |= PCRE_DUPNAMES;              *optset |= PCRE_DUPNAMES;
5260              cd->external_flags |= PCRE_JCHANGED;              cd->external_flags |= PCRE_JCHANGED;
5261              break;              break;
5262    
5263              case 'i': *optset |= PCRE_CASELESS; break;              case CHAR_i: *optset |= PCRE_CASELESS; break;
5264              case 'm': *optset |= PCRE_MULTILINE; break;              case CHAR_m: *optset |= PCRE_MULTILINE; break;
5265              case 's': *optset |= PCRE_DOTALL; break;              case CHAR_s: *optset |= PCRE_DOTALL; break;
5266              case 'x': *optset |= PCRE_EXTENDED; break;              case CHAR_x: *optset |= PCRE_EXTENDED; break;
5267              case 'U': *optset |= PCRE_UNGREEDY; break;              case CHAR_U: *optset |= PCRE_UNGREEDY; break;
5268              case 'X': *optset |= PCRE_EXTRA; break;              case CHAR_X: *optset |= PCRE_EXTRA; break;
5269    
5270              default:  *errorcodeptr = ERR12;              default:  *errorcodeptr = ERR12;
5271                        ptr--;    /* Correct the offset */                        ptr--;    /* Correct the offset */
# Line 4862  we set the flag only if there is a liter Line 5296  we set the flag only if there is a liter
5296          both phases.          both phases.
5297    
5298          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
5299          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
5300          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. */  
5301    
5302          if (*ptr == ')')          if (*ptr == CHAR_RIGHT_PARENTHESIS)
5303            {            {
5304            if (code == cd->start_code + 1 + LINK_SIZE &&            if (code == cd->start_code + 1 + LINK_SIZE &&
5305                 (lengthptr == NULL || *lengthptr == 2 + 2*LINK_SIZE))                 (lengthptr == NULL || *lengthptr == 2 + 2*LINK_SIZE))
5306              {              {
5307              cd->external_options = newoptions;              cd->external_options = newoptions;
             options = newoptions;  
5308              }              }
5309           else            else
5310              {              {
5311              if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))              if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))
5312                {                {
5313                *code++ = OP_OPT;                *code++ = OP_OPT;
5314                *code++ = newoptions & PCRE_IMS;                *code++ = newoptions & PCRE_IMS;
5315                }                }
   
             /* 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;  
5316              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);
5317              greedy_non_default = greedy_default ^ 1;              greedy_non_default = greedy_default ^ 1;
5318              req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
5319              }              }
5320    
5321              /* Change options at this level, and pass them back for use
5322              in subsequent branches. When not at the start of the pattern, this
5323              information is also necessary so that a resetting item can be
5324              compiled at the end of a group (if we are in a group). */
5325    
5326              *optionsptr = options = newoptions;
5327            previous = NULL;       /* This item can't be repeated */            previous = NULL;       /* This item can't be repeated */
5328            continue;              /* It is complete */            continue;              /* It is complete */
5329            }            }
# Line 5008  we set the flag only if there is a liter Line 5439  we set the flag only if there is a liter
5439    
5440      /* Error if hit end of pattern */      /* Error if hit end of pattern */
5441    
5442      if (*ptr != ')')      if (*ptr != CHAR_RIGHT_PARENTHESIS)
5443        {        {
5444        *errorcodeptr = ERR14;        *errorcodeptr = ERR14;
5445        goto FAILED;        goto FAILED;
# Line 5105  we set the flag only if there is a liter Line 5536  we set the flag only if there is a liter
5536      back references and those types that consume a character may be repeated.      back references and those types that consume a character may be repeated.
5537      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
5538      have to change if any new ones are ever created. */      have to change if any new ones are ever created. */
5539    
5540      case '\\':      case CHAR_BACKSLASH:
5541      tempptr = ptr;      tempptr = ptr;
5542      c = check_escape(&ptr, errorcodeptr, cd->bracount, options, FALSE);      c = check_escape(&ptr, errorcodeptr, cd->bracount, options, FALSE);
5543      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
# Line 5115  we set the flag only if there is a liter Line 5546  we set the flag only if there is a liter
5546        {        {
5547        if (-c == ESC_Q)            /* Handle start of quoted string */        if (-c == ESC_Q)            /* Handle start of quoted string */
5548          {          {
5549          if (ptr[1] == '\\' && ptr[2] == 'E') ptr += 2; /* avoid empty string */          if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
5550            else inescq = TRUE;            ptr += 2;               /* avoid empty string */
5551                else inescq = TRUE;
5552          continue;          continue;
5553          }          }
5554    
# Line 5132  we set the flag only if there is a liter Line 5564  we set the flag only if there is a liter
5564    
5565        zerofirstbyte = firstbyte;        zerofirstbyte = firstbyte;
5566        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
5567    
5568        /* \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'
5569        is a subroutine call by number (Oniguruma syntax). In fact, the value        is a subroutine call by number (Oniguruma syntax). In fact, the value
5570        -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 <
5571        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
5572        -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
5573        that is a synonym for a named back reference). */        that is a synonym for a named back reference). */
5574    
5575        if (-c == ESC_g)        if (-c == ESC_g)
5576          {          {
5577          const uschar *p;          const uschar *p;
5578          save_hwm = cd->hwm;   /* Normally this is set when '(' is read */          save_hwm = cd->hwm;   /* Normally this is set when '(' is read */
5579          terminator = (*(++ptr) == '<')? '>' : '\'';          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
5580              CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
5581    
5582          /* These two statements stop the compiler for warning about possibly          /* These two statements stop the compiler for warning about possibly
5583          unset variables caused by the jump to HANDLE_NUMERICAL_RECURSION. In          unset variables caused by the jump to HANDLE_NUMERICAL_RECURSION. In
5584          fact, because we actually check for a number below, the paths that          fact, because we actually check for a number below, the paths that
5585          would actually be in error are never taken. */          would actually be in error are never taken. */
5586    
5587          skipbytes = 0;          skipbytes = 0;
5588          reset_bracount = FALSE;          reset_bracount = FALSE;
5589    
5590          /* Test for a name */          /* Test for a name */
5591    
5592          if (ptr[1] != '+' && ptr[1] != '-')          if (ptr[1] != CHAR_PLUS && ptr[1] != CHAR_MINUS)
5593            {            {
5594            BOOL isnumber = TRUE;            BOOL isnumber = TRUE;
5595            for (p = ptr + 1; *p != 0 && *p != terminator; p++)            for (p = ptr + 1; *p != 0 && *p != terminator; p++)
5596              {              {
5597              if ((cd->ctypes[*p] & ctype_digit) == 0) isnumber = FALSE;              if ((cd->ctypes[*p] & ctype_digit) == 0) isnumber = FALSE;
5598              if ((cd->ctypes[*p] & ctype_word) == 0) break;              if ((cd->ctypes[*p] & ctype_word) == 0) break;
5599              }              }
5600            if (*p != terminator)            if (*p != terminator)
5601              {              {
5602              *errorcodeptr = ERR57;              *errorcodeptr = ERR57;
5603              break;              break;
5604              }              }
5605            if (isnumber)            if (isnumber)
5606              {              {
5607              ptr++;              ptr++;
5608              goto HANDLE_NUMERICAL_RECURSION;              goto HANDLE_NUMERICAL_RECURSION;
5609              }              }
5610            is_recurse = TRUE;            is_recurse = TRUE;
5611            goto NAMED_REF_OR_RECURSE;            goto NAMED_REF_OR_RECURSE;
5612            }            }
5613    
5614          /* Test a signed number in angle brackets or quotes. */          /* Test a signed number in angle brackets or quotes. */
5615    
5616          p = ptr + 2;          p = ptr + 2;
5617          while ((digitab[*p] & ctype_digit) != 0) p++;          while ((digitab[*p] & ctype_digit) != 0) p++;
5618          if (*p != terminator)          if (*p != terminator)
# Line 5187  we set the flag only if there is a liter Line 5620  we set the flag only if there is a liter
5620            *errorcodeptr = ERR57;            *errorcodeptr = ERR57;
5621            break;            break;
5622            }            }
5623          ptr++;          ptr++;
5624          goto HANDLE_NUMERICAL_RECURSION;          goto HANDLE_NUMERICAL_RECURSION;
5625          }          }
5626    
5627        /* \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).
5628        We also support \k{name} (.NET syntax) */        We also support \k{name} (.NET syntax) */
5629    
5630        if (-c == ESC_k && (ptr[1] == '<' || ptr[1] == '\'' || ptr[1] == '{'))        if (-c == ESC_k && (ptr[1] == CHAR_LESS_THAN_SIGN ||
5631              ptr[1] == CHAR_APOSTROPHE || ptr[1] == CHAR_LEFT_CURLY_BRACKET))
5632          {          {
5633          is_recurse = FALSE;          is_recurse = FALSE;
5634          terminator = (*(++ptr) == '<')? '>' : (*ptr == '\'')? '\'' : '}';          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
5635              CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?
5636              CHAR_APOSTROPHE : CHAR_RIGHT_CURLY_BRACKET;
5637          goto NAMED_REF_OR_RECURSE;          goto NAMED_REF_OR_RECURSE;
5638          }          }
5639    
# Line 5207  we set the flag only if there is a liter Line 5643  we set the flag only if there is a liter
5643    
5644        if (-c >= ESC_REF)        if (-c >= ESC_REF)
5645          {          {
5646            open_capitem *oc;
5647          recno = -c - ESC_REF;          recno = -c - ESC_REF;
5648    
5649          HANDLE_REFERENCE:    /* Come here from named backref handling */          HANDLE_REFERENCE:    /* Come here from named backref handling */
# Line 5216  we set the flag only if there is a liter Line 5653  we set the flag only if there is a liter
5653          PUT2INC(code, 0, recno);          PUT2INC(code, 0, recno);
5654          cd->backref_map |= (recno < 32)? (1 << recno) : 1;          cd->backref_map |= (recno < 32)? (1 << recno) : 1;
5655          if (recno > cd->top_backref) cd->top_backref = recno;          if (recno > cd->top_backref) cd->top_backref = recno;
5656    
5657            /* Check to see if this back reference is recursive, that it, it
5658            is inside the group that it references. A flag is set so that the
5659            group can be made atomic. */
5660    
5661            for (oc = cd->open_caps; oc != NULL; oc = oc->next)
5662              {
5663              if (oc->number == recno)
5664                {
5665                oc->flag = TRUE;
5666                break;
5667                }
5668              }
5669          }          }
5670    
5671        /* So are Unicode property matches, if supported. */        /* So are Unicode property matches, if supported. */
# Line 5300  we set the flag only if there is a liter Line 5750  we set the flag only if there is a liter
5750    
5751      /* Remember if \r or \n were seen */      /* Remember if \r or \n were seen */
5752    
5753      if (mcbuffer[0] == '\r' || mcbuffer[0] == '\n')      if (mcbuffer[0] == CHAR_CR || mcbuffer[0] == CHAR_NL)
5754        cd->external_flags |= PCRE_HASCRORLF;        cd->external_flags |= PCRE_HASCRORLF;
5755    
5756      /* Set the first and required bytes appropriately. If no previous first      /* Set the first and required bytes appropriately. If no previous first
# Line 5398  uschar *code = *codeptr; Line 5848  uschar *code = *codeptr;
5848  uschar *last_branch = code;  uschar *last_branch = code;
5849  uschar *start_bracket = code;  uschar *start_bracket = code;
5850  uschar *reverse_count = NULL;  uschar *reverse_count = NULL;
5851    open_capitem capitem;
5852    int capnumber = 0;
5853  int firstbyte, reqbyte;  int firstbyte, reqbyte;
5854  int branchfirstbyte, branchreqbyte;  int branchfirstbyte, branchreqbyte;
5855  int length;  int length;
5856  int orig_bracount;  int orig_bracount;
5857  int max_bracount;  int max_bracount;
5858    int old_external_options = cd->external_options;
5859  branch_chain bc;  branch_chain bc;
5860    
5861  bc.outer = bcptr;  bc.outer = bcptr;
5862  bc.current = code;  bc.current_branch = code;
5863    
5864  firstbyte = reqbyte = REQ_UNSET;  firstbyte = reqbyte = REQ_UNSET;
5865    
# Line 5424  the code that abstracts option settings Line 5877  the code that abstracts option settings
5877  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
5878  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. */
5879    
5880    /* If this is a capturing subpattern, add to the chain of open capturing items
5881    so that we can detect them if (*ACCEPT) is encountered. This is also used to
5882    detect groups that contain recursive back references to themselves. */
5883    
5884    if (*code == OP_CBRA)
5885      {
5886      capnumber = GET2(code, 1 + LINK_SIZE);
5887      capitem.number = capnumber;
5888      capitem.next = cd->open_caps;
5889      capitem.flag = FALSE;
5890      cd->open_caps = &capitem;
5891      }
5892    
5893  /* Offset is set zero to mark that this bracket is still open */  /* Offset is set zero to mark that this bracket is still open */
5894    
5895  PUT(code, 1, 0);  PUT(code, 1, 0);
# Line 5468  for (;;) Line 5934  for (;;)
5934      return FALSE;      return FALSE;
5935      }      }
5936    
5937      /* If the external options have changed during this branch, it means that we
5938      are at the top level, and a leading option setting has been encountered. We
5939      need to re-set the original option values to take account of this so that,
5940      during the pre-compile phase, we know to allow for a re-set at the start of
5941      subsequent branches. */
5942    
5943      if (old_external_options != cd->external_options)
5944        oldims = cd->external_options & PCRE_IMS;
5945    
5946    /* Keep the highest bracket count in case (?| was used and some branch    /* Keep the highest bracket count in case (?| was used and some branch
5947    has fewer than the rest. */    has fewer than the rest. */
5948    
# Line 5518  for (;;) Line 5993  for (;;)
5993    
5994      /* If lookbehind, check that this branch matches a fixed-length string, and      /* If lookbehind, check that this branch matches a fixed-length string, and
5995      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
5996      branch with OP_END. */      branch with OP_END. If the branch contains OP_RECURSE, the result is -3
5997        because there may be forward references that we can't check here. Set a
5998        flag to cause another lookbehind check at the end. Why not do it all at the
5999        end? Because common, erroneous checks are picked up here and the offset of
6000        the problem can be shown. */
6001    
6002      if (lookbehind)      if (lookbehind)
6003        {        {
6004        int fixed_length;        int fixed_length;
6005        *code = OP_END;        *code = OP_END;
6006        fixed_length = find_fixedlength(last_branch, options);        fixed_length = find_fixedlength(last_branch, options, FALSE, cd);
6007        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
6008        if (fixed_length < 0)        if (fixed_length == -3)
6009            {
6010            cd->check_lookbehind = TRUE;
6011            }
6012          else if (fixed_length < 0)
6013          {          {
6014          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;
6015          *ptrptr = ptr;          *ptrptr = ptr;
6016          return FALSE;          return FALSE;
6017          }          }
6018        PUT(reverse_count, 0, fixed_length);        else { PUT(reverse_count, 0, fixed_length); }
6019        }        }
6020      }      }
6021    
# Line 5545  for (;;) Line 6028  for (;;)
6028    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.
6029    Return leaving the pointer at the terminating char. */    Return leaving the pointer at the terminating char. */
6030    
6031    if (*ptr != '|')    if (*ptr != CHAR_VERTICAL_LINE)
6032      {      {
6033      if (lengthptr == NULL)      if (lengthptr == NULL)
6034        {        {
# Line 5566  for (;;) Line 6049  for (;;)
6049      PUT(code, 1, code - start_bracket);      PUT(code, 1, code - start_bracket);
6050      code += 1 + LINK_SIZE;      code += 1 + LINK_SIZE;
6051    
6052      /* Resetting option if needed */      /* If it was a capturing subpattern, check to see if it contained any
6053        recursive back references. If so, we must wrap it in atomic brackets.
6054        In any event, remove the block from the chain. */
6055    
6056      if ((options & PCRE_IMS) != oldims && *ptr == ')')      if (capnumber > 0)
6057          {
6058          if (cd->open_caps->flag)
6059            {
6060            memmove(start_bracket + 1 + LINK_SIZE, start_bracket,
6061              code - start_bracket);
6062            *start_bracket = OP_ONCE;
6063            code += 1 + LINK_SIZE;
6064            PUT(start_bracket, 1, code - start_bracket);
6065            *code = OP_KET;
6066            PUT(code, 1, code - start_bracket);
6067            code += 1 + LINK_SIZE;
6068            length += 2 + 2*LINK_SIZE;
6069            }
6070          cd->open_caps = cd->open_caps->next;
6071          }
6072    
6073        /* Reset options if needed. */
6074    
6075        if ((options & PCRE_IMS) != oldims && *ptr == CHAR_RIGHT_PARENTHESIS)
6076        {        {
6077        *code++ = OP_OPT;        *code++ = OP_OPT;
6078        *code++ = oldims;        *code++ = oldims;
# Line 5615  for (;;) Line 6119  for (;;)
6119      {      {
6120      *code = OP_ALT;      *code = OP_ALT;
6121      PUT(code, 1, code - last_branch);      PUT(code, 1, code - last_branch);
6122      bc.current = last_branch = code;      bc.current_branch = last_branch = code;
6123      code += 1 + LINK_SIZE;      code += 1 + LINK_SIZE;
6124      }      }
6125    
# Line 5697  do { Line 6201  do {
6201       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;
6202       }       }
6203    
6204     /* .* 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
6205     are or may be referenced. */     it isn't in brackets that are or may be referenced. */
6206    
6207     else if ((op == OP_TYPESTAR || op == OP_TYPEMINSTAR ||     else if ((op == OP_TYPESTAR || op == OP_TYPEMINSTAR ||
6208               op == OP_TYPEPOSSTAR) &&               op == OP_TYPEPOSSTAR))
             (*options & PCRE_DOTALL) != 0)  
6209       {       {
6210       if (scode[1] != OP_ANY || (bracket_map & backref_map) != 0) return FALSE;       if (scode[1] != OP_ALLANY || (bracket_map & backref_map) != 0)
6211           return FALSE;
6212       }       }
6213    
6214     /* Check for explicit anchoring */     /* Check for explicit anchoring */
# Line 5750  do { Line 6254  do {
6254       NULL, 0, FALSE);       NULL, 0, FALSE);
6255     register int op = *scode;     register int op = *scode;
6256    
6257       /* If we are at the start of a conditional assertion group, *both* the
6258       conditional assertion *and* what follows the condition must satisfy the test
6259       for start of line. Other kinds of condition fail. Note that there may be an
6260       auto-callout at the start of a condition. */
6261    
6262       if (op == OP_COND)
6263         {
6264         scode += 1 + LINK_SIZE;
6265         if (*scode == OP_CALLOUT) scode += _pcre_OP_lengths[OP_CALLOUT];
6266         switch (*scode)
6267           {
6268           case OP_CREF:
6269           case OP_NCREF:
6270           case OP_RREF:
6271           case OP_NRREF:
6272           case OP_DEF:
6273           return FALSE;
6274    
6275           default:     /* Assertion */
6276           if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
6277           do scode += GET(scode, 1); while (*scode == OP_ALT);
6278           scode += 1 + LINK_SIZE;
6279           break;
6280           }
6281         scode = first_significant_code(scode, NULL, 0, FALSE);
6282         op = *scode;
6283         }
6284    
6285     /* Non-capturing brackets */     /* Non-capturing brackets */
6286    
6287     if (op == OP_BRA)     if (op == OP_BRA)
# Line 5768  do { Line 6300  do {
6300    
6301     /* Other brackets */     /* Other brackets */
6302    
6303     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_COND)     else if (op == OP_ASSERT || op == OP_ONCE)
6304       { if (!is_startline(scode, bracket_map, backref_map)) return FALSE; }       {
6305         if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
6306         }
6307    
6308     /* .* 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
6309     may be referenced. */     may be referenced. */
# Line 5886  Returns:        pointer to compiled data Line 6420  Returns:        pointer to compiled data
6420                  with errorptr and erroroffset set                  with errorptr and erroroffset set
6421  */  */
6422    
6423  PCRE_EXP_DEFN pcre *  PCRE_EXP_DEFN pcre * PCRE_CALL_CONVENTION
6424  pcre_compile(const char *pattern, int options, const char **errorptr,  pcre_compile(const char *pattern, int options, const char **errorptr,
6425    int *erroroffset, const unsigned char *tables)    int *erroroffset, const unsigned char *tables)
6426  {  {
# Line 5894  return pcre_compile2(pattern, options, N Line 6428  return pcre_compile2(pattern, options, N
6428  }  }
6429    
6430    
6431  PCRE_EXP_DEFN pcre *  PCRE_EXP_DEFN pcre * PCRE_CALL_CONVENTION
6432  pcre_compile2(const char *pattern, int options, int *errorcodeptr,  pcre_compile2(const char *pattern, int options, int *errorcodeptr,
6433    const char **errorptr, int *erroroffset, const unsigned char *tables)    const char **errorptr, int *erroroffset, const unsigned char *tables)
6434  {  {
# Line 5903  int length = 1;  /* For final END opcode Line 6437  int length = 1;  /* For final END opcode
6437  int firstbyte, reqbyte, newline;  int firstbyte, reqbyte, newline;
6438  int errorcode = 0;  int errorcode = 0;
6439  int skipatstart = 0;  int skipatstart = 0;
6440  #ifdef SUPPORT_UTF8  BOOL utf8 = (options & PCRE_UTF8) != 0;
 BOOL utf8;  
 #endif  
6441  size_t size;  size_t size;
6442  uschar *code;  uschar *code;
6443  const uschar *codestart;  const uschar *codestart;
# Line 5948  if (erroroffset == NULL) Line 6480  if (erroroffset == NULL)
6480    
6481  *erroroffset = 0;  *erroroffset = 0;
6482    
 /* 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;  
   }  
   
6483  /* Set up pointers to the individual character tables */  /* Set up pointers to the individual character tables */
6484    
6485  if (tables == NULL) tables = _pcre_default_tables;  if (tables == NULL) tables = _pcre_default_tables;
# Line 5980  cd->fcc = tables + fcc_offset; Line 6488  cd->fcc = tables + fcc_offset;
6488  cd->cbits = tables + cbits_offset;  cd->cbits = tables + cbits_offset;
6489  cd->ctypes = tables + ctypes_offset;  cd->ctypes = tables + ctypes_offset;
6490    
6491    /* Check that all undefined public option bits are zero */
6492    
6493    if ((options & ~PUBLIC_COMPILE_OPTIONS) != 0)
6494      {
6495      errorcode = ERR17;
6496      goto PCRE_EARLY_ERROR_RETURN;
6497      }
6498    
6499  /* 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
6500  the offset for later. */  the offset for later. */
6501    
6502  while (ptr[skipatstart] == '(' && ptr[skipatstart+1] == '*')  while (ptr[skipatstart] == CHAR_LEFT_PARENTHESIS &&
6503           ptr[skipatstart+1] == CHAR_ASTERISK)
6504    {    {
6505    int newnl = 0;    int newnl = 0;
6506    int newbsr = 0;    int newbsr = 0;
6507    
6508    if (strncmp((char *)(ptr+skipatstart+2), "CR)", 3) == 0)    if (strncmp((char *)(ptr+skipatstart+2), STRING_UTF8_RIGHTPAR, 5) == 0)
6509        { skipatstart += 7; options |= PCRE_UTF8; continue; }
6510    
6511      if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)
6512      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }
6513    else if (strncmp((char *)(ptr+skipatstart+2), "LF)", 3)  == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_LF_RIGHTPAR, 3)  == 0)
6514      { skipatstart += 5; newnl = PCRE_NEWLINE_LF; }      { skipatstart += 5; newnl = PCRE_NEWLINE_LF; }
6515    else if (strncmp((char *)(ptr+skipatstart+2), "CRLF)", 5)  == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_CRLF_RIGHTPAR, 5)  == 0)
6516      { skipatstart += 7; newnl = PCRE_NEWLINE_CR + PCRE_NEWLINE_LF; }      { skipatstart += 7; newnl = PCRE_NEWLINE_CR + PCRE_NEWLINE_LF; }
6517    else if (strncmp((char *)(ptr+skipatstart+2), "ANY)", 4) == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_ANY_RIGHTPAR, 4) == 0)
6518      { skipatstart += 6; newnl = PCRE_NEWLINE_ANY; }      { skipatstart += 6; newnl = PCRE_NEWLINE_ANY; }
6519    else if (strncmp((char *)(ptr+skipatstart+2), "ANYCRLF)", 8)  == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_ANYCRLF_RIGHTPAR, 8) == 0)
6520      { skipatstart += 10; newnl = PCRE_NEWLINE_ANYCRLF; }      { skipatstart += 10; newnl = PCRE_NEWLINE_ANYCRLF; }
6521    
6522    else if (strncmp((char *)(ptr+skipatstart+2), "BSR_ANYCRLF)", 12) == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_BSR_ANYCRLF_RIGHTPAR, 12) == 0)
6523      { skipatstart += 14; newbsr = PCRE_BSR_ANYCRLF; }      { skipatstart += 14; newbsr = PCRE_BSR_ANYCRLF; }
6524    else if (strncmp((char *)(ptr+skipatstart+2), "BSR_UNICODE)", 12) == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_BSR_UNICODE_RIGHTPAR, 12) == 0)
6525      { skipatstart += 14; newbsr = PCRE_BSR_UNICODE; }      { skipatstart += 14; newbsr = PCRE_BSR_UNICODE; }
6526    
6527    if (newnl != 0)    if (newnl != 0)
# Line 6011  while (ptr[skipatstart] == '(' && ptr[sk Line 6531  while (ptr[skipatstart] == '(' && ptr[sk
6531    else break;    else break;
6532    }    }
6533    
6534    /* Can't support UTF8 unless PCRE has been compiled to include the code. */
6535    
6536    #ifdef SUPPORT_UTF8
6537    if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&
6538         (*erroroffset = _pcre_valid_utf8((USPTR)pattern, -1)) >= 0)
6539      {
6540      errorcode = ERR44;
6541      goto PCRE_EARLY_ERROR_RETURN2;
6542      }
6543    #else
6544    if (utf8)
6545      {
6546      errorcode = ERR32;
6547      goto PCRE_EARLY_ERROR_RETURN;
6548      }
6549    #endif
6550    
6551  /* Check validity of \R options. */  /* Check validity of \R options. */
6552    
6553  switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))  switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))
# Line 6029  current code allows for fixed one- or tw Line 6566  current code allows for fixed one- or tw
6566  switch (options & PCRE_NEWLINE_BITS)  switch (options & PCRE_NEWLINE_BITS)
6567    {    {
6568    case 0: newline = NEWLINE; break;   /* Build-time default */    case 0: newline = NEWLINE; break;   /* Build-time default */
6569    case PCRE_NEWLINE_CR: newline = '\r'; break;    case PCRE_NEWLINE_CR: newline = CHAR_CR; break;
6570    case PCRE_NEWLINE_LF: newline = '\n'; break;    case PCRE_NEWLINE_LF: newline = CHAR_NL; break;
6571    case PCRE_NEWLINE_CR+    case PCRE_NEWLINE_CR+
6572         PCRE_NEWLINE_LF: newline = ('\r' << 8) | '\n'; break;         PCRE_NEWLINE_LF: newline = (CHAR_CR << 8) | CHAR_NL; break;
6573    case PCRE_NEWLINE_ANY: newline = -1; break;    case PCRE_NEWLINE_ANY: newline = -1; break;
6574    case PCRE_NEWLINE_ANYCRLF: newline = -2; break;    case PCRE_NEWLINE_ANYCRLF: newline = -2; break;
6575    default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;    default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;
# Line 6093  cd->end_pattern = (const uschar *)(patte Line 6630  cd->end_pattern = (const uschar *)(patte
6630  cd->req_varyopt = 0;  cd->req_varyopt = 0;
6631  cd->external_options = options;  cd->external_options = options;
6632  cd->external_flags = 0;  cd->external_flags = 0;
6633    cd->open_caps = NULL;
6634    
6635  /* 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
6636  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 6167  cd->start_code = codestart; Line 6705  cd->start_code = codestart;
6705  cd->hwm = cworkspace;  cd->hwm = cworkspace;
6706  cd->req_varyopt = 0;  cd->req_varyopt = 0;
6707  cd->had_accept = FALSE;  cd->had_accept = FALSE;
6708    cd->check_lookbehind = FALSE;
6709    cd->open_caps = NULL;
6710    
6711  /* Set up a starting, non-extracting bracket, then compile the expression. On  /* Set up a starting, non-extracting bracket, then compile the expression. On
6712  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 6192  if debugging, leave the test till after Line 6732  if debugging, leave the test till after
6732    
6733  *code++ = OP_END;  *code++ = OP_END;
6734    
6735  #ifndef DEBUG  #ifndef PCRE_DEBUG
6736  if (code - codestart > length) errorcode = ERR23;  if (code - codestart > length) errorcode = ERR23;
6737  #endif  #endif
6738    
# Line 6205  while (errorcode == 0 && cd->hwm > cwork Line 6745  while (errorcode == 0 && cd->hwm > cwork
6745    cd->hwm -= LINK_SIZE;    cd->hwm -= LINK_SIZE;
6746    offset = GET(cd->hwm, 0);    offset = GET(cd->hwm, 0);
6747    recno = GET(codestart, offset);    recno = GET(codestart, offset);
6748    groupptr = find_bracket(codestart, (re->options & PCRE_UTF8) != 0, recno);    groupptr = _pcre_find_bracket(codestart, utf8, recno);
6749    if (groupptr == NULL) errorcode = ERR53;    if (groupptr == NULL) errorcode = ERR53;
6750      else PUT(((uschar *)codestart), offset, groupptr - codestart);      else PUT(((uschar *)codestart), offset, groupptr - codestart);
6751    }    }
6752    
6753  /* Give an error if there's back reference to a non-existent capturing  /* Give an error if there's back reference to a non-existent capturing
6754  subpattern. */  subpattern. */
6755    
6756  if (errorcode == 0 && re->top_backref > re->top_bracket) errorcode = ERR15;  if (errorcode == 0 && re->top_backref > re->top_bracket) errorcode = ERR15;
6757    
6758    /* If there were any lookbehind assertions that contained OP_RECURSE
6759    (recursions or subroutine calls), a flag is set for them to be checked here,
6760    because they may contain forward references. Actual recursions can't be fixed
6761    length, but subroutine calls can. It is done like this so that those without
6762    OP_RECURSE that are not fixed length get a diagnosic with a useful offset. The
6763    exceptional ones forgo this. We scan the pattern to check that they are fixed
6764    length, and set their lengths. */
6765    
6766    if (cd->check_lookbehind)
6767      {
6768      uschar *cc = (uschar *)codestart;
6769    
6770      /* Loop, searching for OP_REVERSE items, and process those that do not have
6771      their length set. (Actually, it will also re-process any that have a length
6772      of zero, but that is a pathological case, and it does no harm.) When we find
6773      one, we temporarily terminate the branch it is in while we scan it. */
6774    
6775      for (cc = (uschar *)_pcre_find_bracket(codestart, utf8, -1);
6776           cc != NULL;
6777           cc = (uschar *)_pcre_find_bracket(cc, utf8, -1))
6778        {
6779        if (GET(cc, 1) == 0)
6780          {
6781          int fixed_length;
6782          uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);
6783          int end_op = *be;
6784          *be = OP_END;
6785          fixed_length = find_fixedlength(cc, re->options, TRUE, cd);
6786          *be = end_op;
6787          DPRINTF(("fixed length = %d\n", fixed_length));
6788          if (fixed_length < 0)
6789            {
6790            errorcode = (fixed_length == -2)? ERR36 : ERR25;
6791            break;
6792            }
6793          PUT(cc, 1, fixed_length);
6794          }
6795        cc += 1 + LINK_SIZE;
6796        }
6797      }
6798    
6799  /* Failed to compile, or error while post-processing */  /* Failed to compile, or error while post-processing */
6800    
6801  if (errorcode != 0)  if (errorcode != 0)
# Line 6275  if (reqbyte >= 0 && Line 6856  if (reqbyte >= 0 &&
6856  /* 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
6857  case when building a production library. */  case when building a production library. */
6858    
6859  #ifdef DEBUG  #ifdef PCRE_DEBUG
   
6860  printf("Length = %d top_bracket = %d top_backref = %d\n",  printf("Length = %d top_bracket = %d top_backref = %d\n",
6861    length, re->top_bracket, re->top_backref);    length, re->top_bracket, re->top_backref);
6862    
# Line 6313  if (code - codestart > length) Line 6893  if (code - codestart > length)
6893    if (errorcodeptr != NULL) *errorcodeptr = ERR23;    if (errorcodeptr != NULL) *errorcodeptr = ERR23;
6894    return NULL;    return NULL;
6895    }    }
6896  #endif   /* DEBUG */  #endif   /* PCRE_DEBUG */
6897    
6898  return (pcre *)re;  return (pcre *)re;
6899  }  }

Legend:
Removed from v.335  
changed lines
  Added in v.507

  ViewVC Help
Powered by ViewVC 1.1.5