/[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 295 by ph10, Mon Dec 31 17:00:24 2007 UTC revision 513 by ph10, Mon May 3 11:13:37 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) 1997-2007 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;                 /* Length of verb name */
192    int   op;    int   op;                  /* Op when no arg, or -1 if arg mandatory */
193      int   op_arg;              /* Op when arg present, or -1 if not allowed */
194  } verbitem;  } verbitem;
195    
196  static const char verbnames[] =  static const char verbnames[] =
197    "ACCEPT\0"    "\0"                       /* Empty name is a shorthand for MARK */
198    "COMMIT\0"    STRING_MARK0
199    "F\0"    STRING_ACCEPT0
200    "FAIL\0"    STRING_COMMIT0
201    "PRUNE\0"    STRING_F0
202    "SKIP\0"    STRING_FAIL0
203    "THEN";    STRING_PRUNE0
204      STRING_SKIP0
205  static verbitem verbs[] = {    STRING_THEN;
206    { 6, OP_ACCEPT },  
207    { 6, OP_COMMIT },  static const verbitem verbs[] = {
208    { 1, OP_FAIL },    { 0, -1,        OP_MARK },
209    { 4, OP_FAIL },    { 4, -1,        OP_MARK },
210    { 5, OP_PRUNE },    { 6, OP_ACCEPT, -1 },
211    { 4, OP_SKIP  },    { 6, OP_COMMIT, -1 },
212    { 4, OP_THEN  }    { 1, OP_FAIL,   -1 },
213      { 4, OP_FAIL,   -1 },
214      { 5, OP_PRUNE,  OP_PRUNE_ARG },
215      { 4, OP_SKIP,   OP_SKIP_ARG  },
216      { 4, OP_THEN,   OP_THEN_ARG  }
217  };  };
218    
219  static int verbcount = sizeof(verbs)/sizeof(verbitem);  static const int verbcount = sizeof(verbs)/sizeof(verbitem);
220    
221    
222  /* Tables of names of POSIX character classes and their lengths. The names are  /* Tables of names of POSIX character classes and their lengths. The names are
# Line 178  length entry. The first three must be al Line 226  length entry. The first three must be al
226  for handling case independence. */  for handling case independence. */
227    
228  static const char posix_names[] =  static const char posix_names[] =
229    "alpha\0"  "lower\0"  "upper\0"  "alnum\0"  "ascii\0"  "blank\0"    STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
230    "cntrl\0"  "digit\0"  "graph\0"  "print\0"  "punct\0"  "space\0"    STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
231    "word\0"   "xdigit";    STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
232      STRING_word0  STRING_xdigit;
233    
234  static const uschar posix_name_lengths[] = {  static const uschar posix_name_lengths[] = {
235    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 273  the number of relocations needed when a
273  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
274  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
275  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
276  because these strings are used only when there is a compilation error. */  because these strings are used only when there is a compilation error.
277    
278    Each substring ends with \0 to insert a null character. This includes the final
279    substring, so that the whole string ends with \0\0, which can be detected when
280    counting through. */
281    
282  static const char error_texts[] =  static const char error_texts[] =
283    "no error\0"    "no error\0"
# Line 295  static const char error_texts[] = Line 348  static const char error_texts[] =
348    /* 55 */    /* 55 */
349    "repeating a DEFINE group is not allowed\0"    "repeating a DEFINE group is not allowed\0"
350    "inconsistent NEWLINE options\0"    "inconsistent NEWLINE options\0"
351    "\\g is not followed by a braced name or an optionally braced non-zero number\0"    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
352    "(?+ or (?- or (?(+ or (?(- must be followed by a non-zero number\0"    "a numbered reference must not be zero\0"
353    "(*VERB) with an argument is not supported\0"    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
354    /* 60 */    /* 60 */
355    "(*VERB) not recognized\0"    "(*VERB) not recognized\0"
356    "number is too big\0"    "number is too big\0"
357    "subpattern name expected\0"    "subpattern name expected\0"
358    "digit expected after (?+";    "digit expected after (?+\0"
359      "] is an invalid data character in JavaScript compatibility mode\0"
360      /* 65 */
361      "different names for subpatterns of the same number are not allowed\0"
362      "(*MARK) must have an argument\0"
363      ;
364    
365  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
366  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 378  For convenience, we use the same bit def
378    
379  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
380    
381  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC
382    
383    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
384    UTF-8 mode. */
385    
386  static const unsigned char digitab[] =  static const unsigned char digitab[] =
387    {    {
388    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 418  static const unsigned char digitab[] =
418    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
419    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
420    
421  #else           /* This is the "abnormal" case, for EBCDIC systems */  #else
422    
423    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
424    
425  static const unsigned char digitab[] =  static const unsigned char digitab[] =
426    {    {
427    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 518  static const char *
518  find_error_text(int n)  find_error_text(int n)
519  {  {
520  const char *s = error_texts;  const char *s = error_texts;
521  for (; n > 0; n--) while (*s++ != 0);  for (; n > 0; n--)
522      {
523      while (*s++ != 0) {};
524      if (*s == 0) return "Error text not found (please report)";
525      }
526  return s;  return s;
527  }  }
528    
# Line 502  if (c == 0) *errorcodeptr = ERR1; Line 570  if (c == 0) *errorcodeptr = ERR1;
570  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.
571  Otherwise further processing may be required. */  Otherwise further processing may be required. */
572    
573  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
574  else if (c < '0' || c > 'z') {}                           /* Not alphanumeric */  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */
575  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0) c = i;
576    
577  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
578  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 591  else
591      /* 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
592      error. */      error. */
593    
594      case 'l':      case CHAR_l:
595      case 'L':      case CHAR_L:
596      case 'N':      case CHAR_N:
597      case 'u':      case CHAR_u:
598      case 'U':      case CHAR_U:
599      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
600      break;      break;
601    
602      /* \g must be followed by a number, either plain or braced. If positive, it      /* \g must be followed by one of a number of specific things:
603      is an absolute backreference. If negative, it is a relative backreference.  
604      This is a Perl 5.10 feature. Perl 5.10 also supports \g{name} as a      (1) A number, either plain or braced. If positive, it is an absolute
605      reference to a named group. This is part of Perl's movement towards a      backreference. If negative, it is a relative backreference. This is a Perl
606      unified syntax for back references. As this is synonymous with \k{name}, we      5.10 feature.
607      fudge it up by pretending it really was \k. */  
608        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
609        is part of Perl's movement towards a unified syntax for back references. As
610        this is synonymous with \k{name}, we fudge it up by pretending it really
611        was \k.
612    
613        (3) For Oniguruma compatibility we also support \g followed by a name or a
614        number either in angle brackets or in single quotes. However, these are
615        (possibly recursive) subroutine calls, _not_ backreferences. Just return
616        the -ESC_g code (cf \k). */
617    
618        case CHAR_g:
619        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
620          {
621          c = -ESC_g;
622          break;
623          }
624    
625        /* Handle the Perl-compatible cases */
626    
627      case 'g':      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
     if (ptr[1] == '{')  
628        {        {
629        const uschar *p;        const uschar *p;
630        for (p = ptr+2; *p != 0 && *p != '}'; p++)        for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
631          if (*p != '-' && (digitab[*p] & ctype_digit) == 0) break;          if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
632        if (*p != 0 && *p != '}')        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
633          {          {
634          c = -ESC_k;          c = -ESC_k;
635          break;          break;
# Line 554  else Line 639  else
639        }        }
640      else braced = FALSE;      else braced = FALSE;
641    
642      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
643        {        {
644        negated = TRUE;        negated = TRUE;
645        ptr++;        ptr++;
# Line 563  else Line 648  else
648    
649      c = 0;      c = 0;
650      while ((digitab[ptr[1]] & ctype_digit) != 0)      while ((digitab[ptr[1]] & ctype_digit) != 0)
651        c = c * 10 + *(++ptr) - '0';        c = c * 10 + *(++ptr) - CHAR_0;
652    
653      if (c < 0)      if (c < 0)   /* Integer overflow */
654        {        {
655        *errorcodeptr = ERR61;        *errorcodeptr = ERR61;
656        break;        break;
657        }        }
658    
659      if (c == 0 || (braced && *(++ptr) != '}'))      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
660        {        {
661        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
662        break;        break;
663        }        }
664    
665        if (c == 0)
666          {
667          *errorcodeptr = ERR58;
668          break;
669          }
670    
671      if (negated)      if (negated)
672        {        {
673        if (c > bracount)        if (c > bracount)
# Line 602  else Line 693  else
693      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
694      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
695    
696      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:
697      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
698    
699      if (!isclass)      if (!isclass)
700        {        {
701        oldptr = ptr;        oldptr = ptr;
702        c -= '0';        c -= CHAR_0;
703        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
704          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - CHAR_0;
705        if (c < 0)        if (c < 0)    /* Integer overflow */
706          {          {
707          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
708          break;          break;
# Line 628  else Line 719  else
719      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.
720      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
721    
722      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
723        {        {
724        ptr--;        ptr--;
725        c = 0;        c = 0;
# Line 641  else Line 732  else
732      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
733      than 3 octal digits. */      than 3 octal digits. */
734    
735      case '0':      case CHAR_0:
736      c -= '0';      c -= CHAR_0;
737      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
738          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
739      if (!utf8 && c > 255) *errorcodeptr = ERR51;      if (!utf8 && c > 255) *errorcodeptr = ERR51;
740      break;      break;
741    
# Line 652  else Line 743  else
743      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
744      treated as a data character. */      treated as a data character. */
745    
746      case 'x':      case CHAR_x:
747      if (ptr[1] == '{')      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
748        {        {
749        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
750        int count = 0;        int count = 0;
# Line 662  else Line 753  else
753        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
754          {          {
755          register int cc = *pt++;          register int cc = *pt++;
756          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
757          count++;          count++;
758    
759  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
760          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
761          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
762  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
763          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
764          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
765  #endif  #endif
766          }          }
767    
768        if (*pt == '}')        if (*pt == CHAR_RIGHT_CURLY_BRACKET)
769          {          {
770          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
771          ptr = pt;          ptr = pt;
# Line 690  else Line 781  else
781      c = 0;      c = 0;
782      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
783        {        {
784        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
785        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
786  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
787        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
788        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
789  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
790        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
791        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
792  #endif  #endif
793        }        }
794      break;      break;
# Line 706  else Line 797  else
797      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
798      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
799    
800      case 'c':      case CHAR_c:
801      c = *(++ptr);      c = *(++ptr);
802      if (c == 0)      if (c == 0)
803        {        {
# Line 714  else Line 805  else
805        break;        break;
806        }        }
807    
808  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
809      if (c >= 'a' && c <= 'z') c -= 32;      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
810      c ^= 0x40;      c ^= 0x40;
811  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
812      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
813      c ^= 0xC0;      c ^= 0xC0;
814  #endif  #endif
815      break;      break;
# Line 780  if (c == 0) goto ERROR_RETURN; Line 871  if (c == 0) goto ERROR_RETURN;
871  /* \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
872  negation. */  negation. */
873    
874  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
875    {    {
876    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
877      {      {
878      *negptr = TRUE;      *negptr = TRUE;
879      ptr++;      ptr++;
# Line 791  if (c == '{') Line 882  if (c == '{')
882      {      {
883      c = *(++ptr);      c = *(++ptr);
884      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
885      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
886      name[i] = c;      name[i] = c;
887      }      }
888    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
889    name[i] = 0;    name[i] = 0;
890    }    }
891    
# Line 859  is_counted_repeat(const uschar *p) Line 950  is_counted_repeat(const uschar *p)
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  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
954    
955  if (*p++ != ',') return FALSE;  if (*p++ != CHAR_COMMA) return FALSE;
956  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
957    
958  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
959  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
960    
961  return (*p == '}');  return (*p == CHAR_RIGHT_CURLY_BRACKET);
962  }  }
963    
964    
# Line 900  int max = -1; Line 991  int max = -1;
991  /* 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
992  an integer overflow. */  an integer overflow. */
993    
994  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;
995  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
996    {    {
997    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 910  if (min < 0 || min > 65535) Line 1001  if (min < 0 || min > 65535)
1001  /* 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.
1002  Also, max must not be less than min. */  Also, max must not be less than min. */
1003    
1004  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1005    {    {
1006    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1007      {      {
1008      max = 0;      max = 0;
1009      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;
1010      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1011        {        {
1012        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 940  return p; Line 1031  return p;
1031    
1032    
1033  /*************************************************  /*************************************************
1034  *       Find forward referenced subpattern       *  *  Subroutine for finding forward reference      *
1035  *************************************************/  *************************************************/
1036    
1037  /* This function scans along a pattern's text looking for capturing  /* This recursive function is called only from find_parens() below. The
1038    top-level call starts at the beginning of the pattern. All other calls must
1039    start at a parenthesis. It scans along a pattern's text looking for capturing
1040  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
1041  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
1042  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
1043  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
1044  be terminated by '>' because that is checked in the first pass.  first pass. Recursion is used to keep track of subpatterns that reset the
1045    capturing group numbers - the (?| feature.
1046    
1047  Arguments:  Arguments:
1048    ptr          current position in the pattern    ptrptr       address of the current character pointer (updated)
1049    count        current count of capturing parens so far encountered    cd           compile background data
1050    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1051    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1052    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1053      count        pointer to the current capturing subpattern number (updated)
1054    
1055  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1056  */  */
1057    
1058  static int  static int
1059  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,
1060    BOOL xmode)    BOOL xmode, int *count)
1061  {  {
1062  const uschar *thisname;  uschar *ptr = *ptrptr;
1063    int start_count = *count;
1064    int hwm_count = start_count;
1065    BOOL dup_parens = FALSE;
1066    
1067  for (; *ptr != 0; ptr++)  /* If the first character is a parenthesis, check on the type of group we are
1068    dealing with. The very first call may not start with a parenthesis. */
1069    
1070    if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1071    {    {
1072    int term;    if (ptr[1] == CHAR_QUESTION_MARK &&
1073          ptr[2] == CHAR_VERTICAL_LINE)
1074        {
1075        ptr += 3;
1076        dup_parens = TRUE;
1077        }
1078    
1079      /* Handle a normal, unnamed capturing parenthesis */
1080    
1081      else if (ptr[1] != CHAR_QUESTION_MARK && ptr[1] != CHAR_ASTERISK)
1082        {
1083        *count += 1;
1084        if (name == NULL && *count == lorn) return *count;
1085        ptr++;
1086        }
1087    
1088      /* Handle a condition. If it is an assertion, just carry on so that it
1089      is processed as normal. If not, skip to the closing parenthesis of the
1090      condition (there can't be any nested parens. */
1091    
1092      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1093        {
1094        ptr += 2;
1095        if (ptr[1] != CHAR_QUESTION_MARK)
1096          {
1097          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1098          if (*ptr != 0) ptr++;
1099          }
1100        }
1101    
1102      /* We have either (? or (* and not a condition */
1103    
1104      else
1105        {
1106        ptr += 2;
1107        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1108    
1109        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1110    
1111        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1112            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1113          {
1114          int term;
1115          const uschar *thisname;
1116          *count += 1;
1117          if (name == NULL && *count == lorn) return *count;
1118          term = *ptr++;
1119          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1120          thisname = ptr;
1121          while (*ptr != term) ptr++;
1122          if (name != NULL && lorn == ptr - thisname &&
1123              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1124            return *count;
1125          term++;
1126          }
1127        }
1128      }
1129    
1130    /* Past any initial parenthesis handling, scan for parentheses or vertical
1131    bars. */
1132    
1133    for (; *ptr != 0; ptr++)
1134      {
1135    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1136    
1137    if (*ptr == '\\')    if (*ptr == CHAR_BACKSLASH)
1138      {      {
1139      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1140      if (*ptr == 'Q') for (;;)      if (*ptr == CHAR_Q) for (;;)
1141        {        {
1142        while (*(++ptr) != 0 && *ptr != '\\');        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1143        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1144        if (*(++ptr) == 'E') break;        if (*(++ptr) == CHAR_E) break;
1145        }        }
1146      continue;      continue;
1147      }      }
1148    
1149    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1150      are handled for real. If the first character is '^', skip it. Also, if the
1151      first few characters (either before or after ^) are \Q\E or \E we skip them
1152      too. This makes for compatibility with Perl. Note the use of STR macros to
1153      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1154    
1155    if (*ptr == '[')    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1156      {      {
1157      while (*(++ptr) != ']')      BOOL negate_class = FALSE;
1158        for (;;)
1159          {
1160          if (ptr[1] == CHAR_BACKSLASH)
1161            {
1162            if (ptr[2] == CHAR_E)
1163              ptr+= 2;
1164            else if (strncmp((const char *)ptr+2,
1165                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1166              ptr += 4;
1167            else
1168              break;
1169            }
1170          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1171            {
1172            negate_class = TRUE;
1173            ptr++;
1174            }
1175          else break;
1176          }
1177    
1178        /* If the next character is ']', it is a data character that must be
1179        skipped, except in JavaScript compatibility mode. */
1180    
1181        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1182            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1183          ptr++;
1184    
1185        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1186        {        {
1187        if (*ptr == 0) return -1;        if (*ptr == 0) return -1;
1188        if (*ptr == '\\')        if (*ptr == CHAR_BACKSLASH)
1189          {          {
1190          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) goto FAIL_EXIT;
1191          if (*ptr == 'Q') for (;;)          if (*ptr == CHAR_Q) for (;;)
1192            {            {
1193            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1194            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1195            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1196            }            }
1197          continue;          continue;
1198          }          }
# Line 1008  for (; *ptr != 0; ptr++) Line 1202  for (; *ptr != 0; ptr++)
1202    
1203    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1204    
1205    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1206      {      {
1207      while (*(++ptr) != 0 && *ptr != '\n');      while (*(++ptr) != 0 && *ptr != CHAR_NL) {};
1208      if (*ptr == 0) return -1;      if (*ptr == 0) goto FAIL_EXIT;
1209      continue;      continue;
1210      }      }
1211    
1212    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1213    
1214    if (*ptr != '(') continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != '?' && ptr[1] != '*')  
1215      {      {
1216      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, count);
1217      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1218      continue;      if (*ptr == 0) goto FAIL_EXIT;
1219      }      }
1220    
1221    ptr += 2;    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1222    if (*ptr == 'P') ptr++;                      /* Allow optional P */      {
1223        if (dup_parens && *count < hwm_count) *count = hwm_count;
1224        *ptrptr = ptr;
1225        return -1;
1226        }
1227    
1228    /* We have to disambiguate (?<! and (?<= from (?<name> */    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1229        {
1230        if (*count > hwm_count) hwm_count = *count;
1231        *count = start_count;
1232        }
1233      }
1234    
1235    if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  FAIL_EXIT:
1236         *ptr != '\'')  *ptrptr = ptr;
1237      continue;  return -1;
1238    }
1239    
1240    
1241    
1242    
1243    /*************************************************
1244    *       Find forward referenced subpattern       *
1245    *************************************************/
1246    
1247    /* This function scans along a pattern's text looking for capturing
1248    subpatterns, and counting them. If it finds a named pattern that matches the
1249    name it is given, it returns its number. Alternatively, if the name is NULL, it
1250    returns when it reaches a given numbered subpattern. This is used for forward
1251    references to subpatterns. We used to be able to start this scan from the
1252    current compiling point, using the current count value from cd->bracount, and
1253    do it all in a single loop, but the addition of the possibility of duplicate
1254    subpattern numbers means that we have to scan from the very start, in order to
1255    take account of such duplicates, and to use a recursive function to keep track
1256    of the different types of group.
1257    
1258    count++;  Arguments:
1259      cd           compile background data
1260      name         name to seek, or NULL if seeking a numbered subpattern
1261      lorn         name length, or subpattern number if name is NULL
1262      xmode        TRUE if we are in /x mode
1263    
1264    if (name == NULL && count == lorn) return count;  Returns:       the number of the found subpattern, or -1 if not found
1265    term = *ptr++;  */
1266    if (term == '<') term = '>';  
1267    thisname = ptr;  static int
1268    while (*ptr != term) ptr++;  find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode)
1269    if (name != NULL && lorn == ptr - thisname &&  {
1270        strncmp((const char *)name, (const char *)thisname, lorn) == 0)  uschar *ptr = (uschar *)cd->start_pattern;
1271      return count;  int count = 0;
1272    int rc;
1273    
1274    /* If the pattern does not start with an opening parenthesis, the first call
1275    to find_parens_sub() will scan right to the end (if necessary). However, if it
1276    does start with a parenthesis, find_parens_sub() will return when it hits the
1277    matching closing parens. That is why we have to have a loop. */
1278    
1279    for (;;)
1280      {
1281      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, &count);
1282      if (rc > 0 || *ptr++ == 0) break;
1283    }    }
1284    
1285  return -1;  return rc;
1286  }  }
1287    
1288    
1289    
1290    
1291  /*************************************************  /*************************************************
1292  *      Find first significant op code            *  *      Find first significant op code            *
1293  *************************************************/  *************************************************/
# Line 1100  for (;;) Line 1337  for (;;)
1337    
1338      case OP_CALLOUT:      case OP_CALLOUT:
1339      case OP_CREF:      case OP_CREF:
1340        case OP_NCREF:
1341      case OP_RREF:      case OP_RREF:
1342        case OP_NRREF:
1343      case OP_DEF:      case OP_DEF:
1344      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1345      break;      break;
# Line 1116  for (;;) Line 1355  for (;;)
1355    
1356    
1357  /*************************************************  /*************************************************
1358  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1359  *************************************************/  *************************************************/
1360    
1361  /* 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,
1362  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.
1363  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
1364    temporarily terminated with OP_END when this function is called.
1365    
1366    This function is called when a backward assertion is encountered, so that if it
1367    fails, the error message can point to the correct place in the pattern.
1368    However, we cannot do this when the assertion contains subroutine calls,
1369    because they can be forward references. We solve this by remembering this case
1370    and doing the check at the end; a flag specifies which mode we are running in.
1371    
1372  Arguments:  Arguments:
1373    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1374    options  the compiling options    options  the compiling options
1375      atend    TRUE if called when the pattern is complete
1376      cd       the "compile data" structure
1377    
1378  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1379                 or -1 if there is no fixed length,
1380               or -2 if \C was encountered               or -2 if \C was encountered
1381                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1382  */  */
1383    
1384  static int  static int
1385  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)
1386  {  {
1387  int length = -1;  int length = -1;
1388    
# Line 1145  branch, check the length against that of Line 1395  branch, check the length against that of
1395  for (;;)  for (;;)
1396    {    {
1397    int d;    int d;
1398      uschar *ce, *cs;
1399    register int op = *cc;    register int op = *cc;
1400    switch (op)    switch (op)
1401      {      {
# Line 1152  for (;;) Line 1403  for (;;)
1403      case OP_BRA:      case OP_BRA:
1404      case OP_ONCE:      case OP_ONCE:
1405      case OP_COND:      case OP_COND:
1406      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);
1407      if (d < 0) return d;      if (d < 0) return d;
1408      branchlength += d;      branchlength += d;
1409      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1175  for (;;) Line 1426  for (;;)
1426      branchlength = 0;      branchlength = 0;
1427      break;      break;
1428    
1429        /* A true recursion implies not fixed length, but a subroutine call may
1430        be OK. If the subroutine is a forward reference, we can't deal with
1431        it until the end of the pattern, so return -3. */
1432    
1433        case OP_RECURSE:
1434        if (!atend) return -3;
1435        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1436        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1437        if (cc > cs && cc < ce) return -1;                /* Recursion */
1438        d = find_fixedlength(cs + 2, options, atend, cd);
1439        if (d < 0) return d;
1440        branchlength += d;
1441        cc += 1 + LINK_SIZE;
1442        break;
1443    
1444      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1445    
1446      case OP_ASSERT:      case OP_ASSERT:
# Line 1188  for (;;) Line 1454  for (;;)
1454    
1455      case OP_REVERSE:      case OP_REVERSE:
1456      case OP_CREF:      case OP_CREF:
1457        case OP_NCREF:
1458      case OP_RREF:      case OP_RREF:
1459        case OP_NRREF:
1460      case OP_DEF:      case OP_DEF:
1461      case OP_OPT:      case OP_OPT:
1462      case OP_CALLOUT:      case OP_CALLOUT:
1463      case OP_SOD:      case OP_SOD:
1464      case OP_SOM:      case OP_SOM:
1465        case OP_SET_SOM:
1466      case OP_EOD:      case OP_EOD:
1467      case OP_EODN:      case OP_EODN:
1468      case OP_CIRC:      case OP_CIRC:
# Line 1211  for (;;) Line 1480  for (;;)
1480      branchlength++;      branchlength++;
1481      cc += 2;      cc += 2;
1482  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1483      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1484        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1485  #endif  #endif
1486      break;      break;
1487    
# Line 1225  for (;;) Line 1492  for (;;)
1492      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1493      cc += 4;      cc += 4;
1494  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1495      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1496        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1497  #endif  #endif
1498      break;      break;
1499    
# Line 1252  for (;;) Line 1517  for (;;)
1517      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1518      case OP_WORDCHAR:      case OP_WORDCHAR:
1519      case OP_ANY:      case OP_ANY:
1520        case OP_ALLANY:
1521      branchlength++;      branchlength++;
1522      cc++;      cc++;
1523      break;      break;
# Line 1306  for (;;) Line 1572  for (;;)
1572    
1573    
1574  /*************************************************  /*************************************************
1575  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1576  *************************************************/  *************************************************/
1577    
1578  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1579  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1580    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1581    so that it can be called from pcre_study() when finding the minimum matching
1582    length.
1583    
1584  Arguments:  Arguments:
1585    code        points to start of expression    code        points to start of expression
1586    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1587    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1588    
1589  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
1590  */  */
1591    
1592  static const uschar *  const uschar *
1593  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1594  {  {
1595  for (;;)  for (;;)
1596    {    {
# Line 1334  for (;;) Line 1603  for (;;)
1603    
1604    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1605    
1606      /* Handle recursion */
1607    
1608      else if (c == OP_REVERSE)
1609        {
1610        if (number < 0) return (uschar *)code;
1611        code += _pcre_OP_lengths[c];
1612        }
1613    
1614    /* Handle capturing bracket */    /* Handle capturing bracket */
1615    
1616    else if (c == OP_CBRA)    else if (c == OP_CBRA)
# Line 1345  for (;;) Line 1622  for (;;)
1622    
1623    /* Otherwise, we can get the item's length from the table, except that for    /* Otherwise, we can get the item's length from the table, except that for
1624    repeated character types, we have to test for \p and \P, which have an extra    repeated character types, we have to test for \p and \P, which have an extra
1625    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1626      must add in its length. */
1627    
1628    else    else
1629      {      {
# Line 1369  for (;;) Line 1647  for (;;)
1647        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
1648        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1649        break;        break;
1650    
1651          case OP_MARK:
1652          case OP_PRUNE_ARG:
1653          case OP_SKIP_ARG:
1654          case OP_THEN_ARG:
1655          code += code[1];
1656          break;
1657        }        }
1658    
1659      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1400  for (;;) Line 1685  for (;;)
1685        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1686        break;        break;
1687        }        }
1688    #else
1689        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1690  #endif  #endif
1691      }      }
1692    }    }
# Line 1438  for (;;) Line 1725  for (;;)
1725    
1726    /* Otherwise, we can get the item's length from the table, except that for    /* Otherwise, we can get the item's length from the table, except that for
1727    repeated character types, we have to test for \p and \P, which have an extra    repeated character types, we have to test for \p and \P, which have an extra
1728    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1729      must add in its length. */
1730    
1731    else    else
1732      {      {
# Line 1462  for (;;) Line 1750  for (;;)
1750        case OP_TYPEEXACT:        case OP_TYPEEXACT:
1751        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1752        break;        break;
1753    
1754          case OP_MARK:
1755          case OP_PRUNE_ARG:
1756          case OP_SKIP_ARG:
1757          case OP_THEN_ARG:
1758          code += code[1];
1759          break;
1760        }        }
1761    
1762      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1493  for (;;) Line 1788  for (;;)
1788        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1789        break;        break;
1790        }        }
1791    #else
1792        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1793  #endif  #endif
1794      }      }
1795    }    }
# Line 1516  Arguments: Line 1813  Arguments:
1813    code        points to start of search    code        points to start of search
1814    endcode     points to where to stop    endcode     points to where to stop
1815    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
1816      cd          contains pointers to tables etc.
1817    
1818  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
1819  */  */
1820    
1821  static BOOL  static BOOL
1822  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
1823      compile_data *cd)
1824  {  {
1825  register int c;  register int c;
1826  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 1544  for (code = first_significant_code(code Line 1843  for (code = first_significant_code(code
1843    
1844    /* Groups with zero repeats can of course be empty; skip them. */    /* Groups with zero repeats can of course be empty; skip them. */
1845    
1846    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)
1847      {      {
1848      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1849      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
# Line 1552  for (code = first_significant_code(code Line 1851  for (code = first_significant_code(code
1851      continue;      continue;
1852      }      }
1853    
1854      /* For a recursion/subroutine call, if its end has been reached, which
1855      implies a subroutine call, we can scan it. */
1856    
1857      if (c == OP_RECURSE)
1858        {
1859        BOOL empty_branch = FALSE;
1860        const uschar *scode = cd->start_code + GET(code, 1);
1861        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
1862        do
1863          {
1864          if (could_be_empty_branch(scode, endcode, utf8, cd))
1865            {
1866            empty_branch = TRUE;
1867            break;
1868            }
1869          scode += GET(scode, 1);
1870          }
1871        while (*scode == OP_ALT);
1872        if (!empty_branch) return FALSE;  /* All branches are non-empty */
1873        continue;
1874        }
1875    
1876    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
1877    
1878    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 1559  for (code = first_significant_code(code Line 1880  for (code = first_significant_code(code
1880      BOOL empty_branch;      BOOL empty_branch;
1881      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
1882    
1883      /* Scan a closed bracket */      /* If a conditional group has only one branch, there is a second, implied,
1884        empty branch, so just skip over the conditional, because it could be empty.
1885        Otherwise, scan the individual branches of the group. */
1886    
1887      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;  
1888        code += GET(code, 1);        code += GET(code, 1);
1889        else
1890          {
1891          empty_branch = FALSE;
1892          do
1893            {
1894            if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
1895              empty_branch = TRUE;
1896            code += GET(code, 1);
1897            }
1898          while (*code == OP_ALT);
1899          if (!empty_branch) return FALSE;   /* All branches are non-empty */
1900        }        }
1901      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
1902      c = *code;      c = *code;
1903      continue;      continue;
1904      }      }
# Line 1630  for (code = first_significant_code(code Line 1959  for (code = first_significant_code(code
1959      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1960      case OP_WORDCHAR:      case OP_WORDCHAR:
1961      case OP_ANY:      case OP_ANY:
1962        case OP_ALLANY:
1963      case OP_ANYBYTE:      case OP_ANYBYTE:
1964      case OP_CHAR:      case OP_CHAR:
1965      case OP_CHARNC:      case OP_CHARNC:
# Line 1686  for (code = first_significant_code(code Line 2016  for (code = first_significant_code(code
2016      case OP_QUERY:      case OP_QUERY:
2017      case OP_MINQUERY:      case OP_MINQUERY:
2018      case OP_POSQUERY:      case OP_POSQUERY:
2019        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2020        break;
2021    
2022      case OP_UPTO:      case OP_UPTO:
2023      case OP_MINUPTO:      case OP_MINUPTO:
2024      case OP_POSUPTO:      case OP_POSUPTO:
2025      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2026      break;      break;
2027  #endif  #endif
2028    
2029        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2030        string. */
2031    
2032        case OP_MARK:
2033        case OP_PRUNE_ARG:
2034        case OP_SKIP_ARG:
2035        case OP_THEN_ARG:
2036        code += code[1];
2037        break;
2038    
2039        /* None of the remaining opcodes are required to match a character. */
2040    
2041        default:
2042        break;
2043      }      }
2044    }    }
2045    
# Line 1714  Arguments: Line 2062  Arguments:
2062    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2063    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2064    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2065      cd          pointers to tables etc
2066    
2067  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2068  */  */
2069    
2070  static BOOL  static BOOL
2071  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2072    BOOL utf8)    BOOL utf8, compile_data *cd)
2073  {  {
2074  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2075    {    {
2076    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2077        return FALSE;
2078    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2079    }    }
2080  return TRUE;  return TRUE;
# Line 1738  return TRUE; Line 2088  return TRUE;
2088    
2089  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2090  encountered in a character class. It checks whether this is followed by a  encountered in a character class. It checks whether this is followed by a
2091  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2092  reach an unescaped ']' without the special preceding character, return FALSE.  reach an unescaped ']' without the special preceding character, return FALSE.
2093    
2094  Originally, this function only recognized a sequence of letters between the  Originally, this function only recognized a sequence of letters between the
2095  terminators, but it seems that Perl recognizes any sequence of characters,  terminators, but it seems that Perl recognizes any sequence of characters,
2096  though of course unknown POSIX names are subsequently rejected. Perl gives an  though of course unknown POSIX names are subsequently rejected. Perl gives an
2097  "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE  "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2098  didn't consider this to be a POSIX class. Likewise for [:1234:].  didn't consider this to be a POSIX class. Likewise for [:1234:].
2099    
2100  The problem in trying to be exactly like Perl is in the handling of escapes. We  The problem in trying to be exactly like Perl is in the handling of escapes. We
2101  have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX  have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2102  class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code  class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2103  below handles the special case of \], but does not try to do any other escape  below handles the special case of \], but does not try to do any other escape
2104  processing. This makes it different from Perl for cases such as [:l\ower:]  processing. This makes it different from Perl for cases such as [:l\ower:]
2105  where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize  where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2106  "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,  "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2107  I think.  I think.
2108    
2109  Arguments:  Arguments:
# Line 1770  int terminator;          /* Don't combin Line 2120  int terminator;          /* Don't combin
2120  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2121  for (++ptr; *ptr != 0; ptr++)  for (++ptr; *ptr != 0; ptr++)
2122    {    {
2123    if (*ptr == '\\' && ptr[1] == ']') ptr++; else    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else
2124      {      {
2125      if (*ptr == ']') return FALSE;      if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2126      if (*ptr == terminator && ptr[1] == ']')      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2127        {        {
2128        *endptr = ptr;        *endptr = ptr;
2129        return TRUE;        return TRUE;
2130        }        }
2131      }      }
2132    }    }
2133  return FALSE;  return FALSE;
2134  }  }
2135    
# Line 1824  return -1; Line 2174  return -1;
2174  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2175  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2176  earlier groups that are outside the current group). However, when a group is  earlier groups that are outside the current group). However, when a group is
2177  optional (i.e. the minimum quantifier is zero), OP_BRAZERO is inserted before  optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is
2178  it, after it has been compiled. This means that any OP_RECURSE items within it  inserted before it, after it has been compiled. This means that any OP_RECURSE
2179  that refer to the group itself or any contained groups have to have their  items within it that refer to the group itself or any contained groups have to
2180  offsets adjusted. That one of the jobs of this function. Before it is called,  have their offsets adjusted. That one of the jobs of this function. Before it
2181  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2182    OP_END.
2183    
2184  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2185  recursions and subroutine calls. It must also check the list of such references  recursions and subroutine calls. It must also check the list of such references
# Line 1963  get_othercase_range(unsigned int *cptr, Line 2314  get_othercase_range(unsigned int *cptr,
2314  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2315    
2316  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2317    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2318    
2319  if (c > d) return FALSE;  if (c > d) return FALSE;
2320    
# Line 1972  next = othercase + 1; Line 2323  next = othercase + 1;
2323    
2324  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2325    {    {
2326    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2327    next++;    next++;
2328    }    }
2329    
# Line 2018  if ((options & PCRE_EXTENDED) != 0) Line 2369  if ((options & PCRE_EXTENDED) != 0)
2369    for (;;)    for (;;)
2370      {      {
2371      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2372      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2373        {        {
2374        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2375          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
# Line 2030  if ((options & PCRE_EXTENDED) != 0) Line 2381  if ((options & PCRE_EXTENDED) != 0)
2381  /* 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
2382  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2383    
2384  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2385    {    {
2386    int temperrorcode = 0;    int temperrorcode = 0;
2387    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
# Line 2055  if ((options & PCRE_EXTENDED) != 0) Line 2406  if ((options & PCRE_EXTENDED) != 0)
2406    for (;;)    for (;;)
2407      {      {
2408      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2409      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2410        {        {
2411        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2412          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
# Line 2066  if ((options & PCRE_EXTENDED) != 0) Line 2417  if ((options & PCRE_EXTENDED) != 0)
2417    
2418  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
2419    
2420  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2421    return FALSE;    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2422        return FALSE;
2423    
2424  /* 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
2425  positive single character match, "item" either contains the character or, if  positive single character match, "item" either contains the character or, if
# Line 2082  if (next >= 0) switch(op_code) Line 2434  if (next >= 0) switch(op_code)
2434    case OP_CHAR:    case OP_CHAR:
2435  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2436    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2437    #else
2438      (void)(utf8_char);  /* Keep compiler happy by referencing function argument */
2439  #endif  #endif
2440    return item != next;    return item != next;
2441    
# Line 2100  if (next >= 0) switch(op_code) Line 2454  if (next >= 0) switch(op_code)
2454      unsigned int othercase;      unsigned int othercase;
2455      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2456  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2457      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
2458  #else  #else
2459      othercase = NOTACHAR;      othercase = NOTACHAR;
2460  #endif  #endif
# Line 2113  if (next >= 0) switch(op_code) Line 2467  if (next >= 0) switch(op_code)
2467    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT, "item" must be a single-byte character. */
2468    
2469    case OP_NOT:    case OP_NOT:
   if (next < 0) return FALSE;  /* Not a character */  
2470    if (item == next) return TRUE;    if (item == next) return TRUE;
2471    if ((options & PCRE_CASELESS) == 0) return FALSE;    if ((options & PCRE_CASELESS) == 0) return FALSE;
2472  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 2122  if (next >= 0) switch(op_code) Line 2475  if (next >= 0) switch(op_code)
2475      unsigned int othercase;      unsigned int othercase;
2476      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2477  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2478      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
2479  #else  #else
2480      othercase = NOTACHAR;      othercase = NOTACHAR;
2481  #endif  #endif
# Line 2376  uschar classbits[32]; Line 2729  uschar classbits[32];
2729  BOOL class_utf8;  BOOL class_utf8;
2730  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
2731  uschar *class_utf8data;  uschar *class_utf8data;
2732    uschar *class_utf8data_base;
2733  uschar utf8_char[6];  uschar utf8_char[6];
2734  #else  #else
2735  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
2736  uschar *utf8_char = NULL;  uschar *utf8_char = NULL;
2737  #endif  #endif
2738    
2739  #ifdef DEBUG  #ifdef PCRE_DEBUG
2740  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2741  #endif  #endif
2742    
# Line 2441  for (;; ptr++) Line 2795  for (;; ptr++)
2795    
2796    if (lengthptr != NULL)    if (lengthptr != NULL)
2797      {      {
2798  #ifdef DEBUG  #ifdef PCRE_DEBUG
2799      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
2800  #endif  #endif
2801      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
2802        {        {
2803        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
2804        goto FAILED;        goto FAILED;
# Line 2493  for (;; ptr++) Line 2847  for (;; ptr++)
2847    /* 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
2848    reference list. */    reference list. */
2849    
2850    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
2851      {      {
2852      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
2853      goto FAILED;      goto FAILED;
# Line 2503  for (;; ptr++) Line 2857  for (;; ptr++)
2857    
2858    if (inescq && c != 0)    if (inescq && c != 0)
2859      {      {
2860      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
2861        {        {
2862        inescq = FALSE;        inescq = FALSE;
2863        ptr++;        ptr++;
# Line 2529  for (;; ptr++) Line 2883  for (;; ptr++)
2883    /* 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
2884    a quantifier. */    a quantifier. */
2885    
2886    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
2887      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
2888        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
2889    
2890    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
2891         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2545  for (;; ptr++) Line 2900  for (;; ptr++)
2900    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
2901      {      {
2902      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
2903      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
2904        {        {
2905        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2906          {          {
# Line 2570  for (;; ptr++) Line 2925  for (;; ptr++)
2925      {      {
2926      /* ===================================================================*/      /* ===================================================================*/
2927      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
2928      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
2929      case ')':      case CHAR_RIGHT_PARENTHESIS:
2930      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
2931      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
2932      *codeptr = code;      *codeptr = code;
# Line 2593  for (;; ptr++) Line 2948  for (;; ptr++)
2948      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
2949      the setting of any following char as a first character. */      the setting of any following char as a first character. */
2950    
2951      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
2952      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
2953        {        {
2954        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
# Line 2602  for (;; ptr++) Line 2957  for (;; ptr++)
2957      *code++ = OP_CIRC;      *code++ = OP_CIRC;
2958      break;      break;
2959    
2960      case '$':      case CHAR_DOLLAR_SIGN:
2961      previous = NULL;      previous = NULL;
2962      *code++ = OP_DOLL;      *code++ = OP_DOLL;
2963      break;      break;
# Line 2610  for (;; ptr++) Line 2965  for (;; ptr++)
2965      /* 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
2966      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
2967    
2968      case '.':      case CHAR_DOT:
2969      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
2970      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
2971      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
2972      previous = code;      previous = code;
2973      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
2974      break;      break;
2975    
2976    
# Line 2630  for (;; ptr++) Line 2985  for (;; ptr++)
2985      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,
2986      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
2987      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.
     */  
2988    
2989      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
2990        default (Perl) mode, it is treated as a data character. */
2991    
2992        case CHAR_RIGHT_SQUARE_BRACKET:
2993        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2994          {
2995          *errorcodeptr = ERR64;
2996          goto FAILED;
2997          }
2998        goto NORMAL_CHAR;
2999    
3000        case CHAR_LEFT_SQUARE_BRACKET:
3001      previous = code;      previous = code;
3002    
3003      /* 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
3004      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. */
3005    
3006      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3007             ptr[1] == CHAR_EQUALS_SIGN) &&
3008          check_posix_syntax(ptr, &tempptr))          check_posix_syntax(ptr, &tempptr))
3009        {        {
3010        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
3011        goto FAILED;        goto FAILED;
3012        }        }
3013    
# Line 2653  for (;; ptr++) Line 3019  for (;; ptr++)
3019      for (;;)      for (;;)
3020        {        {
3021        c = *(++ptr);        c = *(++ptr);
3022        if (c == '\\')        if (c == CHAR_BACKSLASH)
3023          {          {
3024          if (ptr[1] == 'E') ptr++;          if (ptr[1] == CHAR_E)
3025            else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;            ptr++;
3026              else break;          else if (strncmp((const char *)ptr+1,
3027                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
3028              ptr += 3;
3029            else
3030              break;
3031          }          }
3032        else if (!negate_class && c == '^')        else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
3033          negate_class = TRUE;          negate_class = TRUE;
3034        else break;        else break;
3035        }        }
3036    
3037        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
3038        an initial ']' is taken as a data character -- the code below handles
3039        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3040        [^] must match any character, so generate OP_ALLANY. */
3041    
3042        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3043            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3044          {
3045          *code++ = negate_class? OP_ALLANY : OP_FAIL;
3046          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3047          zerofirstbyte = firstbyte;
3048          break;
3049          }
3050    
3051      /* 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
3052      negation flag at the end, so that support for characters > 255 works      negation flag at the end, so that support for characters > 255 works
3053      correctly (they are all included in the class). */      correctly (they are all included in the class). */
# Line 2687  for (;; ptr++) Line 3071  for (;; ptr++)
3071  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3072      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
3073      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
3074        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
3075  #endif  #endif
3076    
3077      /* Process characters until ] is reached. By writing this as a "do" it      /* Process characters until ] is reached. By writing this as a "do" it
# Line 2702  for (;; ptr++) Line 3087  for (;; ptr++)
3087          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3088          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3089          }          }
3090    
3091          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
3092          data and reset the pointer. This is so that very large classes that
3093          contain a zillion UTF-8 characters no longer overwrite the work space
3094          (which is on the stack). */
3095    
3096          if (lengthptr != NULL)
3097            {
3098            *lengthptr += class_utf8data - class_utf8data_base;
3099            class_utf8data = class_utf8data_base;
3100            }
3101    
3102  #endif  #endif
3103    
3104        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
3105    
3106        if (inescq)        if (inescq)
3107          {          {
3108          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3109            {            {
3110            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
3111            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2723  for (;; ptr++) Line 3120  for (;; ptr++)
3120        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3121        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3122    
3123        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3124            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3125            check_posix_syntax(ptr, &tempptr))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3126          {          {
3127          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3128          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3129          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3130          uschar pbits[32];          uschar pbits[32];
3131    
3132          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3133            {            {
3134            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3135            goto FAILED;            goto FAILED;
3136            }            }
3137    
3138          ptr += 2;          ptr += 2;
3139          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3140            {            {
3141            local_negate = TRUE;            local_negate = TRUE;
3142            should_flip_negation = TRUE;  /* Note negative special */            should_flip_negation = TRUE;  /* Note negative special */
# Line 2807  for (;; ptr++) Line 3204  for (;; ptr++)
3204    
3205        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3206        of the specials, which just set a flag. The sequence \b is a special        of the specials, which just set a flag. The sequence \b is a special
3207        case. Inside a class (and only there) it is treated as backspace.        case. Inside a class (and only there) it is treated as backspace. We
3208        Elsewhere it marks a word boundary. Other escapes have preset maps ready        assume that other escapes have more than one character in them, so set
3209        to 'or' into the one we are building. We assume they have more than one        class_charcount bigger than one. Unrecognized escapes fall through and
3210        character in them, so set class_charcount bigger than one. */        are either treated as literal characters (by default), or are faulted if
3211          PCRE_EXTRA is set. */
3212    
3213        if (c == '\\')        if (c == CHAR_BACKSLASH)
3214          {          {
3215          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3216          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3217    
3218          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 */
         else if (-c == ESC_X) c = 'X';   /* \X is literal X in a class */  
         else if (-c == ESC_R) c = 'R';   /* \R is literal R in a class */  
3219          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3220            {            {
3221            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3222              {              {
3223              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3224              }              }
# Line 3048  for (;; ptr++) Line 3444  for (;; ptr++)
3444        entirely. The code for handling \Q and \E is messy. */        entirely. The code for handling \Q and \E is messy. */
3445    
3446        CHECK_RANGE:        CHECK_RANGE:
3447        while (ptr[1] == '\\' && ptr[2] == 'E')        while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3448          {          {
3449          inescq = FALSE;          inescq = FALSE;
3450          ptr += 2;          ptr += 2;
# Line 3058  for (;; ptr++) Line 3454  for (;; ptr++)
3454    
3455        /* Remember \r or \n */        /* Remember \r or \n */
3456    
3457        if (c == '\r' || c == '\n') cd->external_flags |= PCRE_HASCRORLF;        if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3458    
3459        /* Check for range */        /* Check for range */
3460    
3461        if (!inescq && ptr[1] == '-')        if (!inescq && ptr[1] == CHAR_MINUS)
3462          {          {
3463          int d;          int d;
3464          ptr += 2;          ptr += 2;
3465          while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3466    
3467          /* 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
3468          mode. */          mode. */
3469    
3470          while (*ptr == '\\' && ptr[1] == 'Q')          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3471            {            {
3472            ptr += 2;            ptr += 2;
3473            if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }            if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3474                { ptr += 2; continue; }
3475            inescq = TRUE;            inescq = TRUE;
3476            break;            break;
3477            }            }
3478    
3479          if (*ptr == 0 || (!inescq && *ptr == ']'))          if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3480            {            {
3481            ptr = oldptr;            ptr = oldptr;
3482            goto LONE_SINGLE_CHARACTER;            goto LONE_SINGLE_CHARACTER;
# Line 3098  for (;; ptr++) Line 3495  for (;; ptr++)
3495          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
3496          in such circumstances. */          in such circumstances. */
3497    
3498          if (!inescq && d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3499            {            {
3500            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3501            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
# Line 3108  for (;; ptr++) Line 3505  for (;; ptr++)
3505    
3506            if (d < 0)            if (d < 0)
3507              {              {
3508              if (d == -ESC_b) d = '\b';              if (d == -ESC_b) d = CHAR_BS;
3509              else if (d == -ESC_X) d = 'X';              else if (d == -ESC_X) d = CHAR_X;
3510              else if (d == -ESC_R) d = 'R'; else              else if (d == -ESC_R) d = CHAR_R; else
3511                {                {
3512                ptr = oldptr;                ptr = oldptr;
3513                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 3131  for (;; ptr++) Line 3528  for (;; ptr++)
3528    
3529          /* Remember \r or \n */          /* Remember \r or \n */
3530    
3531          if (d == '\r' || d == '\n') cd->external_flags |= PCRE_HASCRORLF;          if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3532    
3533          /* 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
3534          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 3251  for (;; ptr++) Line 3648  for (;; ptr++)
3648          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3649            {            {
3650            unsigned int othercase;            unsigned int othercase;
3651            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)            if ((othercase = UCD_OTHERCASE(c)) != c)
3652              {              {
3653              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3654              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 3278  for (;; ptr++) Line 3675  for (;; ptr++)
3675    
3676      /* Loop until ']' reached. This "while" is the end of the "do" above. */      /* Loop until ']' reached. This "while" is the end of the "do" above. */
3677    
3678      while ((c = *(++ptr)) != 0 && (c != ']' || inescq));      while ((c = *(++ptr)) != 0 && (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
3679    
3680      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)                          /* Missing terminating ']' */
3681        {        {
# Line 3423  we set the flag only if there is a liter Line 3820  we set the flag only if there is a liter
3820      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
3821      has been tested above. */      has been tested above. */
3822    
3823      case '{':      case CHAR_LEFT_CURLY_BRACKET:
3824      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
3825      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
3826      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
3827      goto REPEAT;      goto REPEAT;
3828    
3829      case '*':      case CHAR_ASTERISK:
3830      repeat_min = 0;      repeat_min = 0;
3831      repeat_max = -1;      repeat_max = -1;
3832      goto REPEAT;      goto REPEAT;
3833    
3834      case '+':      case CHAR_PLUS:
3835      repeat_min = 1;      repeat_min = 1;
3836      repeat_max = -1;      repeat_max = -1;
3837      goto REPEAT;      goto REPEAT;
3838    
3839      case '?':      case CHAR_QUESTION_MARK:
3840      repeat_min = 0;      repeat_min = 0;
3841      repeat_max = 1;      repeat_max = 1;
3842    
# Line 3474  we set the flag only if there is a liter Line 3871  we set the flag only if there is a liter
3871      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
3872      repeat type to the non-default. */      repeat type to the non-default. */
3873    
3874      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
3875        {        {
3876        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
3877        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
3878        ptr++;        ptr++;
3879        }        }
3880      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
3881        {        {
3882        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
3883        ptr++;        ptr++;
# Line 3595  we set the flag only if there is a liter Line 3992  we set the flag only if there is a liter
3992    
3993        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
3994    
3995          /*--------------------------------------------------------------------*/
3996          /* This code is obsolete from release 8.00; the restriction was finally
3997          removed: */
3998    
3999        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4000        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4001    
4002        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4003          /*--------------------------------------------------------------------*/
4004    
4005        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
4006    
# Line 3745  we set the flag only if there is a liter Line 4147  we set the flag only if there is a liter
4147          goto END_REPEAT;          goto END_REPEAT;
4148          }          }
4149    
4150          /*--------------------------------------------------------------------*/
4151          /* This code is obsolete from release 8.00; the restriction was finally
4152          removed: */
4153    
4154        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4155        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4156    
4157        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4158          /*--------------------------------------------------------------------*/
4159    
4160        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4161          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 3806  we set the flag only if there is a liter Line 4213  we set the flag only if there is a liter
4213    
4214        if (repeat_min == 0)        if (repeat_min == 0)
4215          {          {
4216          /* If the maximum is also zero, we just omit the group from the output          /* If the maximum is also zero, we used to just omit the group from the
4217          altogether. */          output altogether, like this:
4218    
4219          if (repeat_max == 0)          ** if (repeat_max == 0)
4220            {          **   {
4221            code = previous;          **   code = previous;
4222            goto END_REPEAT;          **   goto END_REPEAT;
4223            }          **   }
4224    
4225            However, that fails when a group is referenced as a subroutine from
4226            elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it
4227            so that it is skipped on execution. As we don't have a list of which
4228            groups are referenced, we cannot do this selectively.
4229    
4230            If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4231            and do no more at this point. However, we do need to adjust any
4232            OP_RECURSE calls inside the group that refer to the group itself or any
4233            internal or forward referenced group, because the offset is from the
4234            start of the whole regex. Temporarily terminate the pattern while doing
4235            this. */
4236    
4237          /* If the maximum is 1 or unlimited, we just have to stick in the          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
         BRAZERO and do no more at this point. However, we do need to adjust  
         any OP_RECURSE calls inside the group that refer to the group itself or  
         any internal or forward referenced group, because the offset is from  
         the start of the whole regex. Temporarily terminate the pattern while  
         doing this. */  
   
         if (repeat_max <= 1)  
4238            {            {
4239            *code = OP_END;            *code = OP_END;
4240            adjust_recurse(previous, 1, utf8, cd, save_hwm);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
4241            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
4242            code++;            code++;
4243              if (repeat_max == 0)
4244                {
4245                *previous++ = OP_SKIPZERO;
4246                goto END_REPEAT;
4247                }
4248            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4249            }            }
4250    
# Line 3873  we set the flag only if there is a liter Line 4290  we set the flag only if there is a liter
4290            {            {
4291            /* 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
4292            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
4293            potential integer overflow. */            potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
4294              integer type when available, otherwise double. */
4295    
4296            if (lengthptr != NULL)            if (lengthptr != NULL)
4297              {              {
4298              int delta = (repeat_min - 1)*length_prevgroup;              int delta = (repeat_min - 1)*length_prevgroup;
4299              if ((double)(repeat_min - 1)*(double)length_prevgroup >              if ((INT64_OR_DOUBLE)(repeat_min - 1)*
4300                                                              (double)INT_MAX ||                    (INT64_OR_DOUBLE)length_prevgroup >
4301                        (INT64_OR_DOUBLE)INT_MAX ||
4302                  OFLOW_MAX - *lengthptr < delta)                  OFLOW_MAX - *lengthptr < delta)
4303                {                {
4304                *errorcodeptr = ERR20;                *errorcodeptr = ERR20;
# Line 3925  we set the flag only if there is a liter Line 4344  we set the flag only if there is a liter
4344          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
4345          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
4346          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
4347          paranoid checks to avoid integer overflow. */          paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
4348            a 64-bit integer type when available, otherwise double. */
4349    
4350          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4351            {            {
4352            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4353                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4354            if ((double)repeat_max *            if ((INT64_OR_DOUBLE)repeat_max *
4355                  (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)                  (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4356                    > (double)INT_MAX ||                    > (INT64_OR_DOUBLE)INT_MAX ||
4357                OFLOW_MAX - *lengthptr < delta)                OFLOW_MAX - *lengthptr < delta)
4358              {              {
4359              *errorcodeptr = ERR20;              *errorcodeptr = ERR20;
# Line 4010  we set the flag only if there is a liter Line 4430  we set the flag only if there is a liter
4430            uschar *scode = bracode;            uschar *scode = bracode;
4431            do            do
4432              {              {
4433              if (could_be_empty_branch(scode, ketcode, utf8))              if (could_be_empty_branch(scode, ketcode, utf8, cd))
4434                {                {
4435                *bracode += OP_SBRA - OP_BRA;                *bracode += OP_SBRA - OP_BRA;
4436                break;                break;
# Line 4022  we set the flag only if there is a liter Line 4442  we set the flag only if there is a liter
4442          }          }
4443        }        }
4444    
4445        /* If previous is OP_FAIL, it was generated by an empty class [] in
4446        JavaScript mode. The other ways in which OP_FAIL can be generated, that is
4447        by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
4448        error above. We can just ignore the repeat in JS case. */
4449    
4450        else if (*previous == OP_FAIL) goto END_REPEAT;
4451    
4452      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
4453    
4454      else      else
# Line 4046  we set the flag only if there is a liter Line 4473  we set the flag only if there is a liter
4473      if (possessive_quantifier)      if (possessive_quantifier)
4474        {        {
4475        int len;        int len;
4476        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||  
4477            *tempcode == OP_NOTEXACT)        if (*tempcode == OP_TYPEEXACT)
4478          tempcode += _pcre_OP_lengths[*tempcode] +          tempcode += _pcre_OP_lengths[*tempcode] +
4479            ((*tempcode == OP_TYPEEXACT &&            ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
4480               (tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP))? 2:0);  
4481          else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
4482            {
4483            tempcode += _pcre_OP_lengths[*tempcode];
4484    #ifdef SUPPORT_UTF8
4485            if (utf8 && tempcode[-1] >= 0xc0)
4486              tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];
4487    #endif
4488            }
4489    
4490        len = code - tempcode;        len = code - tempcode;
4491        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4492          {          {
# Line 4069  we set the flag only if there is a liter Line 4505  we set the flag only if there is a liter
4505          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4506          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4507    
4508            /* Because we are moving code along, we must ensure that any
4509            pending recursive references are updated. */
4510    
4511          default:          default:
4512            *code = OP_END;
4513            adjust_recurse(tempcode, 1 + LINK_SIZE, utf8, cd, save_hwm);
4514          memmove(tempcode + 1+LINK_SIZE, tempcode, len);          memmove(tempcode + 1+LINK_SIZE, tempcode, len);
4515          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
4516          len += 1 + LINK_SIZE;          len += 1 + LINK_SIZE;
# Line 4096  we set the flag only if there is a liter Line 4537  we set the flag only if there is a liter
4537      lookbehind or option setting or condition or all the other extended      lookbehind or option setting or condition or all the other extended
4538      parenthesis forms.  */      parenthesis forms.  */
4539    
4540      case '(':      case CHAR_LEFT_PARENTHESIS:
4541      newoptions = options;      newoptions = options;
4542      skipbytes = 0;      skipbytes = 0;
4543      bravalue = OP_CBRA;      bravalue = OP_CBRA;
# Line 4105  we set the flag only if there is a liter Line 4546  we set the flag only if there is a liter
4546    
4547      /* First deal with various "verbs" that can be introduced by '*'. */      /* First deal with various "verbs" that can be introduced by '*'. */
4548    
4549      if (*(++ptr) == '*' && (cd->ctypes[ptr[1]] & ctype_letter) != 0)      if (*(++ptr) == CHAR_ASTERISK &&
4550             ((cd->ctypes[ptr[1]] & ctype_letter) != 0 || ptr[1] == ':'))
4551        {        {
4552        int i, namelen;        int i, namelen;
4553          int arglen = 0;
4554        const char *vn = verbnames;        const char *vn = verbnames;
4555        const uschar *name = ++ptr;        const uschar *name = ptr + 1;
4556          const uschar *arg = NULL;
4557        previous = NULL;        previous = NULL;
4558        while ((cd->ctypes[*++ptr] & ctype_letter) != 0);        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
4559        if (*ptr == ':')        namelen = ptr - name;
4560    
4561          if (*ptr == CHAR_COLON)
4562          {          {
4563          *errorcodeptr = ERR59;   /* Not supported */          arg = ++ptr;
4564          goto FAILED;          while ((cd->ctypes[*ptr] & (ctype_letter|ctype_digit)) != 0
4565              || *ptr == '_') ptr++;
4566            arglen = ptr - arg;
4567          }          }
4568        if (*ptr != ')')  
4569          if (*ptr != CHAR_RIGHT_PARENTHESIS)
4570          {          {
4571          *errorcodeptr = ERR60;          *errorcodeptr = ERR60;
4572          goto FAILED;          goto FAILED;
4573          }          }
4574        namelen = ptr - name;  
4575          /* Scan the table of verb names */
4576    
4577        for (i = 0; i < verbcount; i++)        for (i = 0; i < verbcount; i++)
4578          {          {
4579          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
4580              strncmp((char *)name, vn, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
4581            {            {
4582            *code = verbs[i].op;            /* Check for open captures before ACCEPT */
4583            if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;  
4584            break;            if (verbs[i].op == OP_ACCEPT)
4585                {
4586                open_capitem *oc;
4587                cd->had_accept = TRUE;
4588                for (oc = cd->open_caps; oc != NULL; oc = oc->next)
4589                  {
4590                  *code++ = OP_CLOSE;
4591                  PUT2INC(code, 0, oc->number);
4592                  }
4593                }
4594    
4595              /* Handle the cases with/without an argument */
4596    
4597              if (arglen == 0)
4598                {
4599                if (verbs[i].op < 0)   /* Argument is mandatory */
4600                  {
4601                  *errorcodeptr = ERR66;
4602                  goto FAILED;
4603                  }
4604                *code++ = verbs[i].op;
4605                }
4606    
4607              else
4608                {
4609                if (verbs[i].op_arg < 0)   /* Argument is forbidden */
4610                  {
4611                  *errorcodeptr = ERR59;
4612                  goto FAILED;
4613                  }
4614                *code++ = verbs[i].op_arg;
4615                *code++ = arglen;
4616                memcpy(code, arg, arglen);
4617                code += arglen;
4618                *code++ = 0;
4619                }
4620    
4621              break;  /* Found verb, exit loop */
4622            }            }
4623    
4624          vn += verbs[i].len + 1;          vn += verbs[i].len + 1;
4625          }          }
4626        if (i < verbcount) continue;  
4627        *errorcodeptr = ERR60;        if (i < verbcount) continue;    /* Successfully handled a verb */
4628          *errorcodeptr = ERR60;          /* Verb not recognized */
4629        goto FAILED;        goto FAILED;
4630        }        }
4631    
4632      /* Deal with the extended parentheses; all are introduced by '?', and the      /* Deal with the extended parentheses; all are introduced by '?', and the
4633      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. */
4634    
4635      else if (*ptr == '?')      else if (*ptr == CHAR_QUESTION_MARK)
4636        {        {
4637        int i, set, unset, namelen;        int i, set, unset, namelen;
4638        int *optset;        int *optset;
# Line 4151  we set the flag only if there is a liter Line 4641  we set the flag only if there is a liter
4641    
4642        switch (*(++ptr))        switch (*(++ptr))
4643          {          {
4644          case '#':                 /* Comment; skip to ket */          case CHAR_NUMBER_SIGN:                 /* Comment; skip to ket */
4645          ptr++;          ptr++;
4646          while (*ptr != 0 && *ptr != ')') ptr++;          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
4647          if (*ptr == 0)          if (*ptr == 0)
4648            {            {
4649            *errorcodeptr = ERR18;            *errorcodeptr = ERR18;
# Line 4163  we set the flag only if there is a liter Line 4653  we set the flag only if there is a liter
4653    
4654    
4655          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4656          case '|':                 /* Reset capture count for each branch */          case CHAR_VERTICAL_LINE:  /* Reset capture count for each branch */
4657          reset_bracount = TRUE;          reset_bracount = TRUE;
4658          /* Fall through */          /* Fall through */
4659    
4660          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4661          case ':':                 /* Non-capturing bracket */          case CHAR_COLON:          /* Non-capturing bracket */
4662          bravalue = OP_BRA;          bravalue = OP_BRA;
4663          ptr++;          ptr++;
4664          break;          break;
4665    
4666    
4667          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4668          case '(':          case CHAR_LEFT_PARENTHESIS:
4669          bravalue = OP_COND;       /* Conditional group */          bravalue = OP_COND;       /* Conditional group */
4670    
4671          /* 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 4195  we set the flag only if there is a liter Line 4685  we set the flag only if there is a liter
4685          the switch. This will take control down to where bracketed groups,          the switch. This will take control down to where bracketed groups,
4686          including assertions, are processed. */          including assertions, are processed. */
4687    
4688          if (ptr[1] == '?' && (ptr[2] == '=' || ptr[2] == '!' || ptr[2] == '<'))          if (ptr[1] == CHAR_QUESTION_MARK && (ptr[2] == CHAR_EQUALS_SIGN ||
4689                ptr[2] == CHAR_EXCLAMATION_MARK || ptr[2] == CHAR_LESS_THAN_SIGN))
4690            break;            break;
4691    
4692          /* 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 4207  we set the flag only if there is a liter Line 4698  we set the flag only if there is a liter
4698    
4699          /* Check for a test for recursion in a named group. */          /* Check for a test for recursion in a named group. */
4700    
4701          if (ptr[1] == 'R' && ptr[2] == '&')          if (ptr[1] == CHAR_R && ptr[2] == CHAR_AMPERSAND)
4702            {            {
4703            terminator = -1;            terminator = -1;
4704            ptr += 2;            ptr += 2;
# Line 4217  we set the flag only if there is a liter Line 4708  we set the flag only if there is a liter
4708          /* 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
4709          syntax (?(<name>) or (?('name') */          syntax (?(<name>) or (?('name') */
4710    
4711          else if (ptr[1] == '<')          else if (ptr[1] == CHAR_LESS_THAN_SIGN)
4712            {            {
4713            terminator = '>';            terminator = CHAR_GREATER_THAN_SIGN;
4714            ptr++;            ptr++;
4715            }            }
4716          else if (ptr[1] == '\'')          else if (ptr[1] == CHAR_APOSTROPHE)
4717            {            {
4718            terminator = '\'';            terminator = CHAR_APOSTROPHE;
4719            ptr++;            ptr++;
4720            }            }
4721          else          else
4722            {            {
4723            terminator = 0;            terminator = 0;
4724            if (ptr[1] == '-' || ptr[1] == '+') refsign = *(++ptr);            if (ptr[1] == CHAR_MINUS || ptr[1] == CHAR_PLUS) refsign = *(++ptr);
4725            }            }
4726    
4727          /* 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 4250  we set the flag only if there is a liter Line 4741  we set the flag only if there is a liter
4741            {            {
4742            if (recno >= 0)            if (recno >= 0)
4743              recno = ((digitab[*ptr] & ctype_digit) != 0)?              recno = ((digitab[*ptr] & ctype_digit) != 0)?
4744                recno * 10 + *ptr - '0' : -1;                recno * 10 + *ptr - CHAR_0 : -1;
4745            ptr++;            ptr++;
4746            }            }
4747          namelen = ptr - name;          namelen = ptr - name;
4748    
4749          if ((terminator > 0 && *ptr++ != terminator) || *ptr++ != ')')          if ((terminator > 0 && *ptr++ != terminator) ||
4750                *ptr++ != CHAR_RIGHT_PARENTHESIS)
4751            {            {
4752            ptr--;      /* Error offset */            ptr--;      /* Error offset */
4753            *errorcodeptr = ERR26;            *errorcodeptr = ERR26;
# Line 4277  we set the flag only if there is a liter Line 4769  we set the flag only if there is a liter
4769              *errorcodeptr = ERR58;              *errorcodeptr = ERR58;
4770              goto FAILED;              goto FAILED;
4771              }              }
4772            recno = (refsign == '-')?            recno = (refsign == CHAR_MINUS)?
4773              cd->bracount - recno + 1 : recno +cd->bracount;              cd->bracount - recno + 1 : recno +cd->bracount;
4774            if (recno <= 0 || recno > cd->final_bracount)            if (recno <= 0 || recno > cd->final_bracount)
4775              {              {
# Line 4289  we set the flag only if there is a liter Line 4781  we set the flag only if there is a liter
4781            }            }
4782    
4783          /* Otherwise (did not start with "+" or "-"), start by looking for the          /* Otherwise (did not start with "+" or "-"), start by looking for the
4784          name. */          name. If we find a name, add one to the opcode to change OP_CREF or
4785            OP_RREF into OP_NCREF or OP_NRREF. These behave exactly the same,
4786            except they record that the reference was originally to a name. The
4787            information is used to check duplicate names. */
4788    
4789          slot = cd->name_table;          slot = cd->name_table;
4790          for (i = 0; i < cd->names_found; i++)          for (i = 0; i < cd->names_found; i++)
# Line 4304  we set the flag only if there is a liter Line 4799  we set the flag only if there is a liter
4799            {            {
4800            recno = GET2(slot, 0);            recno = GET2(slot, 0);
4801            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4802              code[1+LINK_SIZE]++;
4803            }            }
4804    
4805          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
4806    
4807          else if ((i = find_parens(ptr, cd->bracount, name, namelen,          else if ((i = find_parens(cd, name, namelen,
4808                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0)) > 0)
4809            {            {
4810            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
4811              code[1+LINK_SIZE]++;
4812            }            }
4813    
4814          /* If terminator == 0 it means that the name followed directly after          /* If terminator == 0 it means that the name followed directly after
# Line 4329  we set the flag only if there is a liter Line 4826  we set the flag only if there is a liter
4826          /* Check for (?(R) for recursion. Allow digits after R to specify a          /* Check for (?(R) for recursion. Allow digits after R to specify a
4827          specific group number. */          specific group number. */
4828    
4829          else if (*name == 'R')          else if (*name == CHAR_R)
4830            {            {
4831            recno = 0;            recno = 0;
4832            for (i = 1; i < namelen; i++)            for (i = 1; i < namelen; i++)
# Line 4339  we set the flag only if there is a liter Line 4836  we set the flag only if there is a liter
4836                *errorcodeptr = ERR15;                *errorcodeptr = ERR15;
4837                goto FAILED;                goto FAILED;
4838                }                }
4839              recno = recno * 10 + name[i] - '0';              recno = recno * 10 + name[i] - CHAR_0;
4840              }              }
4841            if (recno == 0) recno = RREF_ANY;            if (recno == 0) recno = RREF_ANY;
4842            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */
# Line 4349  we set the flag only if there is a liter Line 4846  we set the flag only if there is a liter
4846          /* Similarly, check for the (?(DEFINE) "condition", which is always          /* Similarly, check for the (?(DEFINE) "condition", which is always
4847          false. */          false. */
4848    
4849          else if (namelen == 6 && strncmp((char *)name, "DEFINE", 6) == 0)          else if (namelen == 6 && strncmp((char *)name, STRING_DEFINE, 6) == 0)
4850            {            {
4851            code[1+LINK_SIZE] = OP_DEF;            code[1+LINK_SIZE] = OP_DEF;
4852            skipbytes = 1;            skipbytes = 1;
# Line 4374  we set the flag only if there is a liter Line 4871  we set the flag only if there is a liter
4871    
4872    
4873          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4874          case '=':                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
4875          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
4876          ptr++;          ptr++;
4877          break;          break;
4878    
4879    
4880          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4881          case '!':                 /* Negative lookahead */          case CHAR_EXCLAMATION_MARK:            /* Negative lookahead */
4882          ptr++;          ptr++;
4883          if (*ptr == ')')          /* Optimize (?!) */          if (*ptr == CHAR_RIGHT_PARENTHESIS)    /* Optimize (?!) */
4884            {            {
4885            *code++ = OP_FAIL;            *code++ = OP_FAIL;
4886            previous = NULL;            previous = NULL;
# Line 4394  we set the flag only if there is a liter Line 4891  we set the flag only if there is a liter
4891    
4892    
4893          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4894          case '<':                 /* Lookbehind or named define */          case CHAR_LESS_THAN_SIGN:              /* Lookbehind or named define */
4895          switch (ptr[1])          switch (ptr[1])
4896            {            {
4897            case '=':               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
4898            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
4899            ptr += 2;            ptr += 2;
4900            break;            break;
4901    
4902            case '!':               /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
4903            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
4904            ptr += 2;            ptr += 2;
4905            break;            break;
# Line 4417  we set the flag only if there is a liter Line 4914  we set the flag only if there is a liter
4914    
4915    
4916          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4917          case '>':                 /* One-time brackets */          case CHAR_GREATER_THAN_SIGN:           /* One-time brackets */
4918          bravalue = OP_ONCE;          bravalue = OP_ONCE;
4919          ptr++;          ptr++;
4920          break;          break;
4921    
4922    
4923          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4924          case 'C':                 /* Callout - may be followed by digits; */          case CHAR_C:                 /* Callout - may be followed by digits; */
4925          previous_callout = code;  /* Save for later completion */          previous_callout = code;  /* Save for later completion */
4926          after_manual_callout = 1; /* Skip one item before completing */          after_manual_callout = 1; /* Skip one item before completing */
4927          *code++ = OP_CALLOUT;          *code++ = OP_CALLOUT;
4928            {            {
4929            int n = 0;            int n = 0;
4930            while ((digitab[*(++ptr)] & ctype_digit) != 0)            while ((digitab[*(++ptr)] & ctype_digit) != 0)
4931              n = n * 10 + *ptr - '0';              n = n * 10 + *ptr - CHAR_0;
4932            if (*ptr != ')')            if (*ptr != CHAR_RIGHT_PARENTHESIS)
4933              {              {
4934              *errorcodeptr = ERR39;              *errorcodeptr = ERR39;
4935              goto FAILED;              goto FAILED;
# Line 4452  we set the flag only if there is a liter Line 4949  we set the flag only if there is a liter
4949    
4950    
4951          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4952          case 'P':                 /* Python-style named subpattern handling */          case CHAR_P:              /* Python-style named subpattern handling */
4953          if (*(++ptr) == '=' || *ptr == '>')  /* Reference or recursion */          if (*(++ptr) == CHAR_EQUALS_SIGN ||
4954                *ptr == CHAR_GREATER_THAN_SIGN)  /* Reference or recursion */
4955            {            {
4956            is_recurse = *ptr == '>';            is_recurse = *ptr == CHAR_GREATER_THAN_SIGN;
4957            terminator = ')';            terminator = CHAR_RIGHT_PARENTHESIS;
4958            goto NAMED_REF_OR_RECURSE;            goto NAMED_REF_OR_RECURSE;
4959            }            }
4960          else if (*ptr != '<')    /* Test for Python-style definition */          else if (*ptr != CHAR_LESS_THAN_SIGN)  /* Test for Python-style defn */
4961            {            {
4962            *errorcodeptr = ERR41;            *errorcodeptr = ERR41;
4963            goto FAILED;            goto FAILED;
# Line 4469  we set the flag only if there is a liter Line 4967  we set the flag only if there is a liter
4967    
4968          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4969          DEFINE_NAME:    /* Come here from (?< handling */          DEFINE_NAME:    /* Come here from (?< handling */
4970          case '\'':          case CHAR_APOSTROPHE:
4971            {            {
4972            terminator = (*ptr == '<')? '>' : '\'';            terminator = (*ptr == CHAR_LESS_THAN_SIGN)?
4973                CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
4974            name = ++ptr;            name = ++ptr;
4975    
4976            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
# Line 4502  we set the flag only if there is a liter Line 5001  we set the flag only if there is a liter
5001                }                }
5002              }              }
5003    
5004            /* In the real compile, create the entry in the table */            /* In the real compile, create the entry in the table, maintaining
5005              alphabetical order. Duplicate names for different numbers are
5006              permitted only if PCRE_DUPNAMES is set. Duplicate names for the same
5007              number are always OK. (An existing number can be re-used if (?|
5008              appears in the pattern.) In either event, a duplicate name results in
5009              a duplicate entry in the table, even if the number is the same. This
5010              is because the number of names, and hence the table size, is computed
5011              in the pre-compile, and it affects various numbers and pointers which
5012              would all have to be modified, and the compiled code moved down, if
5013              duplicates with the same number were omitted from the table. This
5014              doesn't seem worth the hassle. However, *different* names for the
5015              same number are not permitted. */
5016    
5017            else            else
5018              {              {
5019                BOOL dupname = FALSE;
5020              slot = cd->name_table;              slot = cd->name_table;
5021    
5022              for (i = 0; i < cd->names_found; i++)              for (i = 0; i < cd->names_found; i++)
5023                {                {
5024                int crc = memcmp(name, slot+2, namelen);                int crc = memcmp(name, slot+2, namelen);
# Line 4514  we set the flag only if there is a liter Line 5026  we set the flag only if there is a liter
5026                  {                  {
5027                  if (slot[2+namelen] == 0)                  if (slot[2+namelen] == 0)
5028                    {                    {
5029                    if ((options & PCRE_DUPNAMES) == 0)                    if (GET2(slot, 0) != cd->bracount + 1 &&
5030                          (options & PCRE_DUPNAMES) == 0)
5031                      {                      {
5032                      *errorcodeptr = ERR43;                      *errorcodeptr = ERR43;
5033                      goto FAILED;                      goto FAILED;
5034                      }                      }
5035                      else dupname = TRUE;
5036                    }                    }
5037                  else crc = -1;      /* Current name is substring */                  else crc = -1;      /* Current name is a substring */
5038                  }                  }
5039    
5040                  /* Make space in the table and break the loop for an earlier
5041                  name. For a duplicate or later name, carry on. We do this for
5042                  duplicates so that in the simple case (when ?(| is not used) they
5043                  are in order of their numbers. */
5044    
5045                if (crc < 0)                if (crc < 0)
5046                  {                  {
5047                  memmove(slot + cd->name_entry_size, slot,                  memmove(slot + cd->name_entry_size, slot,
5048                    (cd->names_found - i) * cd->name_entry_size);                    (cd->names_found - i) * cd->name_entry_size);
5049                  break;                  break;
5050                  }                  }
5051    
5052                  /* Continue the loop for a later or duplicate name */
5053    
5054                slot += cd->name_entry_size;                slot += cd->name_entry_size;
5055                }                }
5056    
5057                /* For non-duplicate names, check for a duplicate number before
5058                adding the new name. */
5059    
5060                if (!dupname)
5061                  {
5062                  uschar *cslot = cd->name_table;
5063                  for (i = 0; i < cd->names_found; i++)
5064                    {
5065                    if (cslot != slot)
5066                      {
5067                      if (GET2(cslot, 0) == cd->bracount + 1)
5068                        {
5069                        *errorcodeptr = ERR65;
5070                        goto FAILED;
5071                        }
5072                      }
5073                    else i--;
5074                    cslot += cd->name_entry_size;
5075                    }
5076                  }
5077    
5078              PUT2(slot, 0, cd->bracount + 1);              PUT2(slot, 0, cd->bracount + 1);
5079              memcpy(slot + 2, name, namelen);              memcpy(slot + 2, name, namelen);
5080              slot[2+namelen] = 0;              slot[2+namelen] = 0;
5081              }              }
5082            }            }
5083    
5084          /* In both cases, count the number of names we've encountered. */          /* In both pre-compile and compile, count the number of names we've
5085            encountered. */
5086    
         ptr++;                    /* Move past > or ' */  
5087          cd->names_found++;          cd->names_found++;
5088            ptr++;                    /* Move past > or ' */
5089          goto NUMBERED_GROUP;          goto NUMBERED_GROUP;
5090    
5091    
5092          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5093          case '&':                 /* Perl recursion/subroutine syntax */          case CHAR_AMPERSAND:            /* Perl recursion/subroutine syntax */
5094          terminator = ')';          terminator = CHAR_RIGHT_PARENTHESIS;
5095          is_recurse = TRUE;          is_recurse = TRUE;
5096          /* Fall through */          /* Fall through */
5097    
# Line 4554  we set the flag only if there is a liter Line 5099  we set the flag only if there is a liter
5099          references (?P=name) and recursion (?P>name), as well as falling          references (?P=name) and recursion (?P>name), as well as falling
5100          through from the Perl recursion syntax (?&name). We also come here from          through from the Perl recursion syntax (?&name). We also come here from
5101          the Perl \k<name> or \k'name' back reference syntax and the \k{name}          the Perl \k<name> or \k'name' back reference syntax and the \k{name}
5102          .NET syntax. */          .NET syntax, and the Oniguruma \g<...> and \g'...' subroutine syntax. */
5103    
5104          NAMED_REF_OR_RECURSE:          NAMED_REF_OR_RECURSE:
5105          name = ++ptr;          name = ++ptr;
# Line 4605  we set the flag only if there is a liter Line 5150  we set the flag only if there is a liter
5150              recno = GET2(slot, 0);              recno = GET2(slot, 0);
5151              }              }
5152            else if ((recno =                /* Forward back reference */            else if ((recno =                /* Forward back reference */
5153                      find_parens(ptr, cd->bracount, name, namelen,                      find_parens(cd, name, namelen,
5154                        (options & PCRE_EXTENDED) != 0)) <= 0)                        (options & PCRE_EXTENDED) != 0)) <= 0)
5155              {              {
5156              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
# Line 4621  we set the flag only if there is a liter Line 5166  we set the flag only if there is a liter
5166    
5167    
5168          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5169          case 'R':                 /* Recursion */          case CHAR_R:              /* Recursion */
5170          ptr++;                    /* Same as (?0)      */          ptr++;                    /* Same as (?0)      */
5171          /* Fall through */          /* Fall through */
5172    
5173    
5174          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5175          case '-': case '+':          case CHAR_MINUS: case CHAR_PLUS:  /* Recursion or subroutine */
5176          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:
5177          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:
5178            {            {
5179            const uschar *called;            const uschar *called;
5180              terminator = CHAR_RIGHT_PARENTHESIS;
5181    
5182              /* Come here from the \g<...> and \g'...' code (Oniguruma
5183              compatibility). However, the syntax has been checked to ensure that
5184              the ... are a (signed) number, so that neither ERR63 nor ERR29 will
5185              be called on this path, nor with the jump to OTHER_CHAR_AFTER_QUERY
5186              ever be taken. */
5187    
5188              HANDLE_NUMERICAL_RECURSION:
5189    
5190            if ((refsign = *ptr) == '+')            if ((refsign = *ptr) == CHAR_PLUS)
5191              {              {
5192              ptr++;              ptr++;
5193              if ((digitab[*ptr] & ctype_digit) == 0)              if ((digitab[*ptr] & ctype_digit) == 0)
# Line 4642  we set the flag only if there is a liter Line 5196  we set the flag only if there is a liter
5196                goto FAILED;                goto FAILED;
5197                }                }
5198              }              }
5199            else if (refsign == '-')            else if (refsign == CHAR_MINUS)
5200              {              {
5201              if ((digitab[ptr[1]] & ctype_digit) == 0)              if ((digitab[ptr[1]] & ctype_digit) == 0)
5202                goto OTHER_CHAR_AFTER_QUERY;                goto OTHER_CHAR_AFTER_QUERY;
# Line 4651  we set the flag only if there is a liter Line 5205  we set the flag only if there is a liter
5205    
5206            recno = 0;            recno = 0;
5207            while((digitab[*ptr] & ctype_digit) != 0)            while((digitab[*ptr] & ctype_digit) != 0)
5208              recno = recno * 10 + *ptr++ - '0';              recno = recno * 10 + *ptr++ - CHAR_0;
5209    
5210            if (*ptr != ')')            if (*ptr != terminator)
5211              {              {
5212              *errorcodeptr = ERR29;              *errorcodeptr = ERR29;
5213              goto FAILED;              goto FAILED;
5214              }              }
5215    
5216            if (refsign == '-')            if (refsign == CHAR_MINUS)
5217              {              {
5218              if (recno == 0)              if (recno == 0)
5219                {                {
# Line 4673  we set the flag only if there is a liter Line 5227  we set the flag only if there is a liter
5227                goto FAILED;                goto FAILED;
5228                }                }
5229              }              }
5230            else if (refsign == '+')            else if (refsign == CHAR_PLUS)
5231              {              {
5232              if (recno == 0)              if (recno == 0)
5233                {                {
# Line 4700  we set the flag only if there is a liter Line 5254  we set the flag only if there is a liter
5254            if (lengthptr == NULL)            if (lengthptr == NULL)
5255              {              {
5256              *code = OP_END;              *code = OP_END;
5257              if (recno != 0) called = find_bracket(cd->start_code, utf8, recno);              if (recno != 0)
5258                  called = _pcre_find_bracket(cd->start_code, utf8, recno);
5259    
5260              /* Forward reference */              /* Forward reference */
5261    
5262              if (called == NULL)              if (called == NULL)
5263                {                {
5264                if (find_parens(ptr, cd->bracount, NULL, recno,                if (find_parens(cd, NULL, recno,
5265                     (options & PCRE_EXTENDED) != 0) < 0)                      (options & PCRE_EXTENDED) != 0) < 0)
5266                  {                  {
5267                  *errorcodeptr = ERR15;                  *errorcodeptr = ERR15;
5268                  goto FAILED;                  goto FAILED;
5269                  }                  }
5270    
5271                  /* Fudge the value of "called" so that when it is inserted as an
5272                  offset below, what it actually inserted is the reference number
5273                  of the group. */
5274    
5275                called = cd->start_code + recno;                called = cd->start_code + recno;
5276                PUTINC(cd->hwm, 0, code + 2 + LINK_SIZE - cd->start_code);                PUTINC(cd->hwm, 0, code + 2 + LINK_SIZE - cd->start_code);
5277                }                }
# Line 4721  we set the flag only if there is a liter Line 5281  we set the flag only if there is a liter
5281              recursion that could loop for ever, and diagnose that case. */              recursion that could loop for ever, and diagnose that case. */
5282    
5283              else if (GET(called, 1) == 0 &&              else if (GET(called, 1) == 0 &&
5284                       could_be_empty(called, code, bcptr, utf8))                       could_be_empty(called, code, bcptr, utf8, cd))
5285                {                {
5286                *errorcodeptr = ERR40;                *errorcodeptr = ERR40;
5287                goto FAILED;                goto FAILED;
# Line 4759  we set the flag only if there is a liter Line 5319  we set the flag only if there is a liter
5319          set = unset = 0;          set = unset = 0;
5320          optset = &set;          optset = &set;
5321    
5322          while (*ptr != ')' && *ptr != ':')          while (*ptr != CHAR_RIGHT_PARENTHESIS && *ptr != CHAR_COLON)
5323            {            {
5324            switch (*ptr++)            switch (*ptr++)
5325              {              {
5326              case '-': optset = &unset; break;              case CHAR_MINUS: optset = &unset; break;
5327    
5328              case 'J':    /* Record that it changed in the external options */              case CHAR_J:    /* Record that it changed in the external options */
5329              *optset |= PCRE_DUPNAMES;              *optset |= PCRE_DUPNAMES;
5330              cd->external_flags |= PCRE_JCHANGED;              cd->external_flags |= PCRE_JCHANGED;
5331              break;              break;
5332    
5333              case 'i': *optset |= PCRE_CASELESS; break;              case CHAR_i: *optset |= PCRE_CASELESS; break;
5334              case 'm': *optset |= PCRE_MULTILINE; break;              case CHAR_m: *optset |= PCRE_MULTILINE; break;
5335              case 's': *optset |= PCRE_DOTALL; break;              case CHAR_s: *optset |= PCRE_DOTALL; break;
5336              case 'x': *optset |= PCRE_EXTENDED; break;              case CHAR_x: *optset |= PCRE_EXTENDED; break;
5337              case 'U': *optset |= PCRE_UNGREEDY; break;              case CHAR_U: *optset |= PCRE_UNGREEDY; break;
5338              case 'X': *optset |= PCRE_EXTRA; break;              case CHAR_X: *optset |= PCRE_EXTRA; break;
5339    
5340              default:  *errorcodeptr = ERR12;              default:  *errorcodeptr = ERR12;
5341                        ptr--;    /* Correct the offset */                        ptr--;    /* Correct the offset */
# Line 4806  we set the flag only if there is a liter Line 5366  we set the flag only if there is a liter
5366          both phases.          both phases.
5367    
5368          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
5369          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
5370          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. */  
5371    
5372          if (*ptr == ')')          if (*ptr == CHAR_RIGHT_PARENTHESIS)
5373            {            {
5374            if (code == cd->start_code + 1 + LINK_SIZE &&            if (code == cd->start_code + 1 + LINK_SIZE &&
5375                 (lengthptr == NULL || *lengthptr == 2 + 2*LINK_SIZE))                 (lengthptr == NULL || *lengthptr == 2 + 2*LINK_SIZE))
5376              {              {
5377              cd->external_options = newoptions;              cd->external_options = newoptions;
             options = newoptions;  
5378              }              }
5379           else            else
5380              {              {
5381              if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))              if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))
5382                {                {
5383                *code++ = OP_OPT;                *code++ = OP_OPT;
5384                *code++ = newoptions & PCRE_IMS;                *code++ = newoptions & PCRE_IMS;
5385                }                }
   
             /* 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;  
5386              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);
5387              greedy_non_default = greedy_default ^ 1;              greedy_non_default = greedy_default ^ 1;
5388              req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
5389              }              }
5390    
5391              /* Change options at this level, and pass them back for use
5392              in subsequent branches. When not at the start of the pattern, this
5393              information is also necessary so that a resetting item can be
5394              compiled at the end of a group (if we are in a group). */
5395    
5396              *optionsptr = options = newoptions;
5397            previous = NULL;       /* This item can't be repeated */            previous = NULL;       /* This item can't be repeated */
5398            continue;              /* It is complete */            continue;              /* It is complete */
5399            }            }
# Line 4851  we set the flag only if there is a liter Line 5408  we set the flag only if there is a liter
5408          }     /* End of switch for character following (? */          }     /* End of switch for character following (? */
5409        }       /* End of (? handling */        }       /* End of (? handling */
5410    
5411      /* Opening parenthesis not followed by '?'. If PCRE_NO_AUTO_CAPTURE is set,      /* Opening parenthesis not followed by '*' or '?'. If PCRE_NO_AUTO_CAPTURE
5412      all unadorned brackets become non-capturing and behave like (?:...)      is set, all unadorned brackets become non-capturing and behave like (?:...)
5413      brackets. */      brackets. */
5414    
5415      else if ((options & PCRE_NO_AUTO_CAPTURE) != 0)      else if ((options & PCRE_NO_AUTO_CAPTURE) != 0)
# Line 4952  we set the flag only if there is a liter Line 5509  we set the flag only if there is a liter
5509    
5510      /* Error if hit end of pattern */      /* Error if hit end of pattern */
5511    
5512      if (*ptr != ')')      if (*ptr != CHAR_RIGHT_PARENTHESIS)
5513        {        {
5514        *errorcodeptr = ERR14;        *errorcodeptr = ERR14;
5515        goto FAILED;        goto FAILED;
# Line 5050  we set the flag only if there is a liter Line 5607  we set the flag only if there is a liter
5607      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
5608      have to change if any new ones are ever created. */      have to change if any new ones are ever created. */
5609    
5610      case '\\':      case CHAR_BACKSLASH:
5611      tempptr = ptr;      tempptr = ptr;
5612      c = check_escape(&ptr, errorcodeptr, cd->bracount, options, FALSE);      c = check_escape(&ptr, errorcodeptr, cd->bracount, options, FALSE);
5613      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
# Line 5059  we set the flag only if there is a liter Line 5616  we set the flag only if there is a liter
5616        {        {
5617        if (-c == ESC_Q)            /* Handle start of quoted string */        if (-c == ESC_Q)            /* Handle start of quoted string */
5618          {          {
5619          if (ptr[1] == '\\' && ptr[2] == 'E') ptr += 2; /* avoid empty string */          if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
5620            else inescq = TRUE;            ptr += 2;               /* avoid empty string */
5621                else inescq = TRUE;
5622          continue;          continue;
5623          }          }
5624    
# Line 5077  we set the flag only if there is a liter Line 5635  we set the flag only if there is a liter
5635        zerofirstbyte = firstbyte;        zerofirstbyte = firstbyte;
5636        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
5637    
5638          /* \g<name> or \g'name' is a subroutine call by name and \g<n> or \g'n'
5639          is a subroutine call by number (Oniguruma syntax). In fact, the value
5640          -ESC_g is returned only for these cases. So we don't need to check for <
5641          or ' if the value is -ESC_g. For the Perl syntax \g{n} the value is
5642          -ESC_REF+n, and for the Perl syntax \g{name} the result is -ESC_k (as
5643          that is a synonym for a named back reference). */
5644    
5645          if (-c == ESC_g)
5646            {
5647            const uschar *p;
5648            save_hwm = cd->hwm;   /* Normally this is set when '(' is read */
5649            terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
5650              CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
5651    
5652            /* These two statements stop the compiler for warning about possibly
5653            unset variables caused by the jump to HANDLE_NUMERICAL_RECURSION. In
5654            fact, because we actually check for a number below, the paths that
5655            would actually be in error are never taken. */
5656    
5657            skipbytes = 0;
5658            reset_bracount = FALSE;
5659    
5660            /* Test for a name */
5661    
5662            if (ptr[1] != CHAR_PLUS && ptr[1] != CHAR_MINUS)
5663              {
5664              BOOL isnumber = TRUE;
5665              for (p = ptr + 1; *p != 0 && *p != terminator; p++)
5666                {
5667                if ((cd->ctypes[*p] & ctype_digit) == 0) isnumber = FALSE;
5668                if ((cd->ctypes[*p] & ctype_word) == 0) break;
5669                }
5670              if (*p != terminator)
5671                {
5672                *errorcodeptr = ERR57;
5673                break;
5674                }
5675              if (isnumber)
5676                {
5677                ptr++;
5678                goto HANDLE_NUMERICAL_RECURSION;
5679                }
5680              is_recurse = TRUE;
5681              goto NAMED_REF_OR_RECURSE;
5682              }
5683    
5684            /* Test a signed number in angle brackets or quotes. */
5685    
5686            p = ptr + 2;
5687            while ((digitab[*p] & ctype_digit) != 0) p++;
5688            if (*p != terminator)
5689              {
5690              *errorcodeptr = ERR57;
5691              break;
5692              }
5693            ptr++;
5694            goto HANDLE_NUMERICAL_RECURSION;
5695            }
5696    
5697        /* \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).
5698        We also support \k{name} (.NET syntax) */        We also support \k{name} (.NET syntax) */
5699    
5700        if (-c == ESC_k && (ptr[1] == '<' || ptr[1] == '\'' || ptr[1] == '{'))        if (-c == ESC_k && (ptr[1] == CHAR_LESS_THAN_SIGN ||
5701              ptr[1] == CHAR_APOSTROPHE || ptr[1] == CHAR_LEFT_CURLY_BRACKET))
5702          {          {
5703          is_recurse = FALSE;          is_recurse = FALSE;
5704          terminator = (*(++ptr) == '<')? '>' : (*ptr == '\'')? '\'' : '}';          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
5705              CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?
5706              CHAR_APOSTROPHE : CHAR_RIGHT_CURLY_BRACKET;
5707          goto NAMED_REF_OR_RECURSE;          goto NAMED_REF_OR_RECURSE;
5708          }          }
5709    
# Line 5093  we set the flag only if there is a liter Line 5713  we set the flag only if there is a liter
5713    
5714        if (-c >= ESC_REF)        if (-c >= ESC_REF)
5715          {          {
5716            open_capitem *oc;
5717          recno = -c - ESC_REF;          recno = -c - ESC_REF;
5718    
5719          HANDLE_REFERENCE:    /* Come here from named backref handling */          HANDLE_REFERENCE:    /* Come here from named backref handling */
# Line 5102  we set the flag only if there is a liter Line 5723  we set the flag only if there is a liter
5723          PUT2INC(code, 0, recno);          PUT2INC(code, 0, recno);
5724          cd->backref_map |= (recno < 32)? (1 << recno) : 1;          cd->backref_map |= (recno < 32)? (1 << recno) : 1;
5725          if (recno > cd->top_backref) cd->top_backref = recno;          if (recno > cd->top_backref) cd->top_backref = recno;
5726    
5727            /* Check to see if this back reference is recursive, that it, it
5728            is inside the group that it references. A flag is set so that the
5729            group can be made atomic. */
5730    
5731            for (oc = cd->open_caps; oc != NULL; oc = oc->next)
5732              {
5733              if (oc->number == recno)
5734                {
5735                oc->flag = TRUE;
5736                break;
5737                }
5738              }
5739          }          }
5740    
5741        /* So are Unicode property matches, if supported. */        /* So are Unicode property matches, if supported. */
# Line 5186  we set the flag only if there is a liter Line 5820  we set the flag only if there is a liter
5820    
5821      /* Remember if \r or \n were seen */      /* Remember if \r or \n were seen */
5822    
5823      if (mcbuffer[0] == '\r' || mcbuffer[0] == '\n')      if (mcbuffer[0] == CHAR_CR || mcbuffer[0] == CHAR_NL)
5824        cd->external_flags |= PCRE_HASCRORLF;        cd->external_flags |= PCRE_HASCRORLF;
5825    
5826      /* Set the first and required bytes appropriately. If no previous first      /* Set the first and required bytes appropriately. If no previous first
# Line 5284  uschar *code = *codeptr; Line 5918  uschar *code = *codeptr;
5918  uschar *last_branch = code;  uschar *last_branch = code;
5919  uschar *start_bracket = code;  uschar *start_bracket = code;
5920  uschar *reverse_count = NULL;  uschar *reverse_count = NULL;
5921    open_capitem capitem;
5922    int capnumber = 0;
5923  int firstbyte, reqbyte;  int firstbyte, reqbyte;
5924  int branchfirstbyte, branchreqbyte;  int branchfirstbyte, branchreqbyte;
5925  int length;  int length;
5926  int orig_bracount;  int orig_bracount;
5927  int max_bracount;  int max_bracount;
5928    int old_external_options = cd->external_options;
5929  branch_chain bc;  branch_chain bc;
5930    
5931  bc.outer = bcptr;  bc.outer = bcptr;
5932  bc.current = code;  bc.current_branch = code;
5933    
5934  firstbyte = reqbyte = REQ_UNSET;  firstbyte = reqbyte = REQ_UNSET;
5935    
# Line 5310  the code that abstracts option settings Line 5947  the code that abstracts option settings
5947  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
5948  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. */
5949    
5950    /* If this is a capturing subpattern, add to the chain of open capturing items
5951    so that we can detect them if (*ACCEPT) is encountered. This is also used to
5952    detect groups that contain recursive back references to themselves. */
5953    
5954    if (*code == OP_CBRA)
5955      {
5956      capnumber = GET2(code, 1 + LINK_SIZE);
5957      capitem.number = capnumber;
5958      capitem.next = cd->open_caps;
5959      capitem.flag = FALSE;
5960      cd->open_caps = &capitem;
5961      }
5962    
5963  /* Offset is set zero to mark that this bracket is still open */  /* Offset is set zero to mark that this bracket is still open */
5964    
5965  PUT(code, 1, 0);  PUT(code, 1, 0);
# Line 5354  for (;;) Line 6004  for (;;)
6004      return FALSE;      return FALSE;
6005      }      }
6006    
6007      /* If the external options have changed during this branch, it means that we
6008      are at the top level, and a leading option setting has been encountered. We
6009      need to re-set the original option values to take account of this so that,
6010      during the pre-compile phase, we know to allow for a re-set at the start of
6011      subsequent branches. */
6012    
6013      if (old_external_options != cd->external_options)
6014        oldims = cd->external_options & PCRE_IMS;
6015    
6016    /* Keep the highest bracket count in case (?| was used and some branch    /* Keep the highest bracket count in case (?| was used and some branch
6017    has fewer than the rest. */    has fewer than the rest. */
6018    
# Line 5404  for (;;) Line 6063  for (;;)
6063    
6064      /* If lookbehind, check that this branch matches a fixed-length string, and      /* If lookbehind, check that this branch matches a fixed-length string, and
6065      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
6066      branch with OP_END. */      branch with OP_END. If the branch contains OP_RECURSE, the result is -3
6067        because there may be forward references that we can't check here. Set a
6068        flag to cause another lookbehind check at the end. Why not do it all at the
6069        end? Because common, erroneous checks are picked up here and the offset of
6070        the problem can be shown. */
6071    
6072      if (lookbehind)      if (lookbehind)
6073        {        {
6074        int fixed_length;        int fixed_length;
6075        *code = OP_END;        *code = OP_END;
6076        fixed_length = find_fixedlength(last_branch, options);        fixed_length = find_fixedlength(last_branch, options, FALSE, cd);
6077        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
6078        if (fixed_length < 0)        if (fixed_length == -3)
6079            {
6080            cd->check_lookbehind = TRUE;
6081            }
6082          else if (fixed_length < 0)
6083          {          {
6084          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;
6085          *ptrptr = ptr;          *ptrptr = ptr;
6086          return FALSE;          return FALSE;
6087          }          }
6088        PUT(reverse_count, 0, fixed_length);        else { PUT(reverse_count, 0, fixed_length); }
6089        }        }
6090      }      }
6091    
# Line 5431  for (;;) Line 6098  for (;;)
6098    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.
6099    Return leaving the pointer at the terminating char. */    Return leaving the pointer at the terminating char. */
6100    
6101    if (*ptr != '|')    if (*ptr != CHAR_VERTICAL_LINE)
6102      {      {
6103      if (lengthptr == NULL)      if (lengthptr == NULL)
6104        {        {
# Line 5452  for (;;) Line 6119  for (;;)
6119      PUT(code, 1, code - start_bracket);      PUT(code, 1, code - start_bracket);
6120      code += 1 + LINK_SIZE;      code += 1 + LINK_SIZE;
6121    
6122      /* Resetting option if needed */      /* If it was a capturing subpattern, check to see if it contained any
6123        recursive back references. If so, we must wrap it in atomic brackets.
6124        In any event, remove the block from the chain. */
6125    
6126        if (capnumber > 0)
6127          {
6128          if (cd->open_caps->flag)
6129            {
6130            memmove(start_bracket + 1 + LINK_SIZE, start_bracket,
6131              code - start_bracket);
6132            *start_bracket = OP_ONCE;
6133            code += 1 + LINK_SIZE;
6134            PUT(start_bracket, 1, code - start_bracket);
6135            *code = OP_KET;
6136            PUT(code, 1, code - start_bracket);
6137            code += 1 + LINK_SIZE;
6138            length += 2 + 2*LINK_SIZE;
6139            }
6140          cd->open_caps = cd->open_caps->next;
6141          }
6142    
6143        /* Reset options if needed. */
6144    
6145      if ((options & PCRE_IMS) != oldims && *ptr == ')')      if ((options & PCRE_IMS) != oldims && *ptr == CHAR_RIGHT_PARENTHESIS)
6146        {        {
6147        *code++ = OP_OPT;        *code++ = OP_OPT;
6148        *code++ = oldims;        *code++ = oldims;
# Line 5501  for (;;) Line 6189  for (;;)
6189      {      {
6190      *code = OP_ALT;      *code = OP_ALT;
6191      PUT(code, 1, code - last_branch);      PUT(code, 1, code - last_branch);
6192      bc.current = last_branch = code;      bc.current_branch = last_branch = code;
6193      code += 1 + LINK_SIZE;      code += 1 + LINK_SIZE;
6194      }      }
6195    
# Line 5583  do { Line 6271  do {
6271       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;
6272       }       }
6273    
6274     /* .* 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
6275     are or may be referenced. */     it isn't in brackets that are or may be referenced. */
6276    
6277     else if ((op == OP_TYPESTAR || op == OP_TYPEMINSTAR ||     else if ((op == OP_TYPESTAR || op == OP_TYPEMINSTAR ||
6278               op == OP_TYPEPOSSTAR) &&               op == OP_TYPEPOSSTAR))
             (*options & PCRE_DOTALL) != 0)  
6279       {       {
6280       if (scode[1] != OP_ANY || (bracket_map & backref_map) != 0) return FALSE;       if (scode[1] != OP_ALLANY || (bracket_map & backref_map) != 0)
6281           return FALSE;
6282       }       }
6283    
6284     /* Check for explicit anchoring */     /* Check for explicit anchoring */
# Line 5636  do { Line 6324  do {
6324       NULL, 0, FALSE);       NULL, 0, FALSE);
6325     register int op = *scode;     register int op = *scode;
6326    
6327       /* If we are at the start of a conditional assertion group, *both* the
6328       conditional assertion *and* what follows the condition must satisfy the test
6329       for start of line. Other kinds of condition fail. Note that there may be an
6330       auto-callout at the start of a condition. */
6331    
6332       if (op == OP_COND)
6333         {
6334         scode += 1 + LINK_SIZE;
6335         if (*scode == OP_CALLOUT) scode += _pcre_OP_lengths[OP_CALLOUT];
6336         switch (*scode)
6337           {
6338           case OP_CREF:
6339           case OP_NCREF:
6340           case OP_RREF:
6341           case OP_NRREF:
6342           case OP_DEF:
6343           return FALSE;
6344    
6345           default:     /* Assertion */
6346           if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
6347           do scode += GET(scode, 1); while (*scode == OP_ALT);
6348           scode += 1 + LINK_SIZE;
6349           break;
6350           }
6351         scode = first_significant_code(scode, NULL, 0, FALSE);
6352         op = *scode;
6353         }
6354    
6355     /* Non-capturing brackets */     /* Non-capturing brackets */
6356    
6357     if (op == OP_BRA)     if (op == OP_BRA)
# Line 5654  do { Line 6370  do {
6370    
6371     /* Other brackets */     /* Other brackets */
6372    
6373     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_COND)     else if (op == OP_ASSERT || op == OP_ONCE)
6374       { if (!is_startline(scode, bracket_map, backref_map)) return FALSE; }       {
6375         if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
6376         }
6377    
6378     /* .* 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
6379     may be referenced. */     may be referenced. */
# Line 5772  Returns:        pointer to compiled data Line 6490  Returns:        pointer to compiled data
6490                  with errorptr and erroroffset set                  with errorptr and erroroffset set
6491  */  */
6492    
6493  PCRE_EXP_DEFN pcre *  PCRE_EXP_DEFN pcre * PCRE_CALL_CONVENTION
6494  pcre_compile(const char *pattern, int options, const char **errorptr,  pcre_compile(const char *pattern, int options, const char **errorptr,
6495    int *erroroffset, const unsigned char *tables)    int *erroroffset, const unsigned char *tables)
6496  {  {
# Line 5780  return pcre_compile2(pattern, options, N Line 6498  return pcre_compile2(pattern, options, N
6498  }  }
6499    
6500    
6501  PCRE_EXP_DEFN pcre *  PCRE_EXP_DEFN pcre * PCRE_CALL_CONVENTION
6502  pcre_compile2(const char *pattern, int options, int *errorcodeptr,  pcre_compile2(const char *pattern, int options, int *errorcodeptr,
6503    const char **errorptr, int *erroroffset, const unsigned char *tables)    const char **errorptr, int *erroroffset, const unsigned char *tables)
6504  {  {
# Line 5789  int length = 1;  /* For final END opcode Line 6507  int length = 1;  /* For final END opcode
6507  int firstbyte, reqbyte, newline;  int firstbyte, reqbyte, newline;
6508  int errorcode = 0;  int errorcode = 0;
6509  int skipatstart = 0;  int skipatstart = 0;
6510  #ifdef SUPPORT_UTF8  BOOL utf8 = (options & PCRE_UTF8) != 0;
 BOOL utf8;  
 #endif  
6511  size_t size;  size_t size;
6512  uschar *code;  uschar *code;
6513  const uschar *codestart;  const uschar *codestart;
# Line 5807  to fill in forward references to subpatt Line 6523  to fill in forward references to subpatt
6523    
6524  uschar cworkspace[COMPILE_WORK_SIZE];  uschar cworkspace[COMPILE_WORK_SIZE];
6525    
   
6526  /* Set this early so that early errors get offset 0. */  /* Set this early so that early errors get offset 0. */
6527    
6528  ptr = (const uschar *)pattern;  ptr = (const uschar *)pattern;
# Line 5835  if (erroroffset == NULL) Line 6550  if (erroroffset == NULL)
6550    
6551  *erroroffset = 0;  *erroroffset = 0;
6552    
 /* 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;  
   }  
   
6553  /* Set up pointers to the individual character tables */  /* Set up pointers to the individual character tables */
6554    
6555  if (tables == NULL) tables = _pcre_default_tables;  if (tables == NULL) tables = _pcre_default_tables;
# Line 5867  cd->fcc = tables + fcc_offset; Line 6558  cd->fcc = tables + fcc_offset;
6558  cd->cbits = tables + cbits_offset;  cd->cbits = tables + cbits_offset;
6559  cd->ctypes = tables + ctypes_offset;  cd->ctypes = tables + ctypes_offset;
6560    
6561    /* Check that all undefined public option bits are zero */
6562    
6563    if ((options & ~PUBLIC_COMPILE_OPTIONS) != 0)
6564      {
6565      errorcode = ERR17;
6566      goto PCRE_EARLY_ERROR_RETURN;
6567      }
6568    
6569  /* 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
6570  the offset for later. */  the offset for later. */
6571    
6572  while (ptr[skipatstart] == '(' && ptr[skipatstart+1] == '*')  while (ptr[skipatstart] == CHAR_LEFT_PARENTHESIS &&
6573           ptr[skipatstart+1] == CHAR_ASTERISK)
6574    {    {
6575    int newnl = 0;    int newnl = 0;
6576    int newbsr = 0;    int newbsr = 0;
6577    
6578    if (strncmp((char *)(ptr+skipatstart+2), "CR)", 3) == 0)    if (strncmp((char *)(ptr+skipatstart+2), STRING_UTF8_RIGHTPAR, 5) == 0)
6579        { skipatstart += 7; options |= PCRE_UTF8; continue; }
6580    
6581      if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)
6582      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }
6583    else if (strncmp((char *)(ptr+skipatstart+2), "LF)", 3)  == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_LF_RIGHTPAR, 3)  == 0)
6584      { skipatstart += 5; newnl = PCRE_NEWLINE_LF; }      { skipatstart += 5; newnl = PCRE_NEWLINE_LF; }
6585    else if (strncmp((char *)(ptr+skipatstart+2), "CRLF)", 5)  == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_CRLF_RIGHTPAR, 5)  == 0)
6586      { skipatstart += 7; newnl = PCRE_NEWLINE_CR + PCRE_NEWLINE_LF; }      { skipatstart += 7; newnl = PCRE_NEWLINE_CR + PCRE_NEWLINE_LF; }
6587    else if (strncmp((char *)(ptr+skipatstart+2), "ANY)", 4) == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_ANY_RIGHTPAR, 4) == 0)
6588      { skipatstart += 6; newnl = PCRE_NEWLINE_ANY; }      { skipatstart += 6; newnl = PCRE_NEWLINE_ANY; }
6589    else if (strncmp((char *)(ptr+skipatstart+2), "ANYCRLF)", 8)  == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_ANYCRLF_RIGHTPAR, 8) == 0)
6590      { skipatstart += 10; newnl = PCRE_NEWLINE_ANYCRLF; }      { skipatstart += 10; newnl = PCRE_NEWLINE_ANYCRLF; }
6591    
6592    else if (strncmp((char *)(ptr+skipatstart+2), "BSR_ANYCRLF)", 12) == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_BSR_ANYCRLF_RIGHTPAR, 12) == 0)
6593      { skipatstart += 14; newbsr = PCRE_BSR_ANYCRLF; }      { skipatstart += 14; newbsr = PCRE_BSR_ANYCRLF; }
6594    else if (strncmp((char *)(ptr+skipatstart+2), "BSR_UNICODE)", 12) == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_BSR_UNICODE_RIGHTPAR, 12) == 0)
6595      { skipatstart += 14; newbsr = PCRE_BSR_UNICODE; }      { skipatstart += 14; newbsr = PCRE_BSR_UNICODE; }
6596    
6597    if (newnl != 0)    if (newnl != 0)
# Line 5898  while (ptr[skipatstart] == '(' && ptr[sk Line 6601  while (ptr[skipatstart] == '(' && ptr[sk
6601    else break;    else break;
6602    }    }
6603    
6604    /* Can't support UTF8 unless PCRE has been compiled to include the code. */
6605    
6606    #ifdef SUPPORT_UTF8
6607    if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&
6608         (*erroroffset = _pcre_valid_utf8((USPTR)pattern, -1)) >= 0)
6609      {
6610      errorcode = ERR44;
6611      goto PCRE_EARLY_ERROR_RETURN2;
6612      }
6613    #else
6614    if (utf8)
6615      {
6616      errorcode = ERR32;
6617      goto PCRE_EARLY_ERROR_RETURN;
6618      }
6619    #endif
6620    
6621  /* Check validity of \R options. */  /* Check validity of \R options. */
6622    
6623  switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))  switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))
# Line 5916  current code allows for fixed one- or tw Line 6636  current code allows for fixed one- or tw
6636  switch (options & PCRE_NEWLINE_BITS)  switch (options & PCRE_NEWLINE_BITS)
6637    {    {
6638    case 0: newline = NEWLINE; break;   /* Build-time default */    case 0: newline = NEWLINE; break;   /* Build-time default */
6639    case PCRE_NEWLINE_CR: newline = '\r'; break;    case PCRE_NEWLINE_CR: newline = CHAR_CR; break;
6640    case PCRE_NEWLINE_LF: newline = '\n'; break;    case PCRE_NEWLINE_LF: newline = CHAR_NL; break;
6641    case PCRE_NEWLINE_CR+    case PCRE_NEWLINE_CR+
6642         PCRE_NEWLINE_LF: newline = ('\r' << 8) | '\n'; break;         PCRE_NEWLINE_LF: newline = (CHAR_CR << 8) | CHAR_NL; break;
6643    case PCRE_NEWLINE_ANY: newline = -1; break;    case PCRE_NEWLINE_ANY: newline = -1; break;
6644    case PCRE_NEWLINE_ANYCRLF: newline = -2; break;    case PCRE_NEWLINE_ANYCRLF: newline = -2; break;
6645    default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;    default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;
# Line 5980  cd->end_pattern = (const uschar *)(patte Line 6700  cd->end_pattern = (const uschar *)(patte
6700  cd->req_varyopt = 0;  cd->req_varyopt = 0;
6701  cd->external_options = options;  cd->external_options = options;
6702  cd->external_flags = 0;  cd->external_flags = 0;
6703    cd->open_caps = NULL;
6704    
6705  /* 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
6706  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 6054  cd->start_code = codestart; Line 6775  cd->start_code = codestart;
6775  cd->hwm = cworkspace;  cd->hwm = cworkspace;
6776  cd->req_varyopt = 0;  cd->req_varyopt = 0;
6777  cd->had_accept = FALSE;  cd->had_accept = FALSE;
6778    cd->check_lookbehind = FALSE;
6779    cd->open_caps = NULL;
6780    
6781  /* Set up a starting, non-extracting bracket, then compile the expression. On  /* Set up a starting, non-extracting bracket, then compile the expression. On
6782  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 6079  if debugging, leave the test till after Line 6802  if debugging, leave the test till after
6802    
6803  *code++ = OP_END;  *code++ = OP_END;
6804    
6805  #ifndef DEBUG  #ifndef PCRE_DEBUG
6806  if (code - codestart > length) errorcode = ERR23;  if (code - codestart > length) errorcode = ERR23;
6807  #endif  #endif
6808    
# Line 6092  while (errorcode == 0 && cd->hwm > cwork Line 6815  while (errorcode == 0 && cd->hwm > cwork
6815    cd->hwm -= LINK_SIZE;    cd->hwm -= LINK_SIZE;
6816    offset = GET(cd->hwm, 0);    offset = GET(cd->hwm, 0);
6817    recno = GET(codestart, offset);    recno = GET(codestart, offset);
6818    groupptr = find_bracket(codestart, (re->options & PCRE_UTF8) != 0, recno);    groupptr = _pcre_find_bracket(codestart, utf8, recno);
6819    if (groupptr == NULL) errorcode = ERR53;    if (groupptr == NULL) errorcode = ERR53;
6820      else PUT(((uschar *)codestart), offset, groupptr - codestart);      else PUT(((uschar *)codestart), offset, groupptr - codestart);
6821    }    }
# Line 6102  subpattern. */ Line 6825  subpattern. */
6825    
6826  if (errorcode == 0 && re->top_backref > re->top_bracket) errorcode = ERR15;  if (errorcode == 0 && re->top_backref > re->top_bracket) errorcode = ERR15;
6827    
6828    /* If there were any lookbehind assertions that contained OP_RECURSE
6829    (recursions or subroutine calls), a flag is set for them to be checked here,
6830    because they may contain forward references. Actual recursions can't be fixed
6831    length, but subroutine calls can. It is done like this so that those without
6832    OP_RECURSE that are not fixed length get a diagnosic with a useful offset. The
6833    exceptional ones forgo this. We scan the pattern to check that they are fixed
6834    length, and set their lengths. */
6835    
6836    if (cd->check_lookbehind)
6837      {
6838      uschar *cc = (uschar *)codestart;
6839    
6840      /* Loop, searching for OP_REVERSE items, and process those that do not have
6841      their length set. (Actually, it will also re-process any that have a length
6842      of zero, but that is a pathological case, and it does no harm.) When we find
6843      one, we temporarily terminate the branch it is in while we scan it. */
6844    
6845      for (cc = (uschar *)_pcre_find_bracket(codestart, utf8, -1);
6846           cc != NULL;
6847           cc = (uschar *)_pcre_find_bracket(cc, utf8, -1))
6848        {
6849        if (GET(cc, 1) == 0)
6850          {
6851          int fixed_length;
6852          uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);
6853          int end_op = *be;
6854          *be = OP_END;
6855          fixed_length = find_fixedlength(cc, re->options, TRUE, cd);
6856          *be = end_op;
6857          DPRINTF(("fixed length = %d\n", fixed_length));
6858          if (fixed_length < 0)
6859            {
6860            errorcode = (fixed_length == -2)? ERR36 : ERR25;
6861            break;
6862            }
6863          PUT(cc, 1, fixed_length);
6864          }
6865        cc += 1 + LINK_SIZE;
6866        }
6867      }
6868    
6869  /* Failed to compile, or error while post-processing */  /* Failed to compile, or error while post-processing */
6870    
6871  if (errorcode != 0)  if (errorcode != 0)
# Line 6162  if (reqbyte >= 0 && Line 6926  if (reqbyte >= 0 &&
6926  /* 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
6927  case when building a production library. */  case when building a production library. */
6928    
6929  #ifdef DEBUG  #ifdef PCRE_DEBUG
   
6930  printf("Length = %d top_bracket = %d top_backref = %d\n",  printf("Length = %d top_bracket = %d top_backref = %d\n",
6931    length, re->top_bracket, re->top_backref);    length, re->top_bracket, re->top_backref);
6932    
# Line 6200  if (code - codestart > length) Line 6963  if (code - codestart > length)
6963    if (errorcodeptr != NULL) *errorcodeptr = ERR23;    if (errorcodeptr != NULL) *errorcodeptr = ERR23;
6964    return NULL;    return NULL;
6965    }    }
6966  #endif   /* DEBUG */  #endif   /* PCRE_DEBUG */
6967    
6968  return (pcre *)re;  return (pcre *)re;
6969  }  }

Legend:
Removed from v.295  
changed lines
  Added in v.513

  ViewVC Help
Powered by ViewVC 1.1.5