/[pcre]/code/branches/pcre16/pcre_compile.c
ViewVC logotype

Diff of /code/branches/pcre16/pcre_compile.c

Parent Directory Parent Directory | Revision Log Revision Log | View Patch Patch

revision 369 by ph10, Sun Aug 24 16:53:47 2008 UTC revision 514 by ph10, Mon May 3 12:54:22 2010 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2008 University of Cambridge             Copyright (c) 1997-2010 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 53  supporting internal functions that are n Line 53  supporting internal functions that are n
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56  /* When DEBUG is defined, we need the pcre_printint() function, which is also  /* When PCRE_DEBUG is defined, we need the pcre_printint() function, which is
57  used by pcretest. DEBUG is not defined when building a production library. */  also used by pcretest. PCRE_DEBUG is not defined when building a production
58    library. */
59    
60  #ifdef DEBUG  #ifdef PCRE_DEBUG
61  #include "pcre_printint.src"  #include "pcre_printint.src"
62  #endif  #endif
63    
# Line 91  is 4 there is plenty of room. */ Line 92  is 4 there is plenty of room. */
92    
93  #define COMPILE_WORK_SIZE (4096)  #define COMPILE_WORK_SIZE (4096)
94    
95    /* The overrun tests check for a slightly smaller size so that they detect the
96    overrun before it actually does run off the end of the data block. */
97    
98    #define WORK_SIZE_CHECK (COMPILE_WORK_SIZE - 100)
99    
100    
101  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
102  are simple data values; negative values are for special things like \d and so  are simple data values; negative values are for special things like \d and so
103  on. Zero means further processing is needed (for things like \x), or the escape  on. Zero means further processing is needed (for things like \x), or the escape
104  is invalid. */  is invalid. */
105    
106  #ifndef EBCDIC  /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC
107    
108    /* This is the "normal" table for ASCII systems or for EBCDIC systems running
109    in UTF-8 mode. */
110    
111  static const short int escapes[] = {  static const short int escapes[] = {
112       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,                       0,
113       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,                       0,
114     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */       0,                       0,
115  -ESC_H,      0,      0, -ESC_K,      0,      0,      0,      0,   /* H - O */       0,                       0,
116  -ESC_P, -ESC_Q, -ESC_R, -ESC_S,      0,      0, -ESC_V, -ESC_W,   /* P - W */       0,                       0,
117  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */       CHAR_COLON,              CHAR_SEMICOLON,
118     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
119  -ESC_h,      0,      0, -ESC_k,      0,      0,  ESC_n,      0,   /* h - o */       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
120  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0, -ESC_v, -ESC_w,   /* p - w */       CHAR_COMMERCIAL_AT,      -ESC_A,
121       0,      0, -ESC_z                                            /* x - z */       -ESC_B,                  -ESC_C,
122         -ESC_D,                  -ESC_E,
123         0,                       -ESC_G,
124         -ESC_H,                  0,
125         0,                       -ESC_K,
126         0,                       0,
127         -ESC_N,                  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 130  static const short int escapes[] = { Line 171  static const short int escapes[] = {
171  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
172  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,
173  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
174  /*  D0 */   '}',     0, -ESC_K,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,-ESC_N,      0, -ESC_P,
175  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
176  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
177  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      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      STRING_THEN;
206    
207  static const verbitem verbs[] = {  static const verbitem verbs[] = {
208    { 6, OP_ACCEPT },    { 0, -1,        OP_MARK },
209    { 6, OP_COMMIT },    { 4, -1,        OP_MARK },
210    { 1, OP_FAIL },    { 6, OP_ACCEPT, -1 },
211    { 4, OP_FAIL },    { 6, OP_COMMIT, -1 },
212    { 5, OP_PRUNE },    { 1, OP_FAIL,   -1 },
213    { 4, OP_SKIP  },    { 4, OP_FAIL,   -1 },
214    { 4, OP_THEN  }    { 5, OP_PRUNE,  OP_PRUNE_ARG },
215      { 4, OP_SKIP,   OP_SKIP_ARG  },
216      { 4, OP_THEN,   OP_THEN_ARG  }
217  };  };
218    
219  static const int verbcount = sizeof(verbs)/sizeof(verbitem);  static const int verbcount = sizeof(verbs)/sizeof(verbitem);
# 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 271  static const char error_texts[] = Line 324  static const char error_texts[] =
324    /* 35 */    /* 35 */
325    "invalid condition (?(0)\0"    "invalid condition (?(0)\0"
326    "\\C not allowed in lookbehind assertion\0"    "\\C not allowed in lookbehind assertion\0"
327    "PCRE does not support \\L, \\l, \\N, \\U, or \\u\0"    "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
328    "number after (?C is > 255\0"    "number after (?C is > 255\0"
329    "closing ) for (?C expected\0"    "closing ) for (?C expected\0"
330    /* 40 */    /* 40 */
# Line 297  static const char error_texts[] = Line 350  static const char error_texts[] =
350    "inconsistent NEWLINE options\0"    "inconsistent NEWLINE options\0"
351    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
352    "a numbered reference must not be zero\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 (?+\0"    "digit expected after (?+\0"
359    "] is an invalid data character in JavaScript compatibility mode";    "] 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 322  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 358  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 455  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 503  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 524  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_u:
597      case 'u':      case CHAR_U:
     case 'U':  
598      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
599      break;      break;
600    
# Line 548  else Line 614  else
614      (possibly recursive) subroutine calls, _not_ backreferences. Just return      (possibly recursive) subroutine calls, _not_ backreferences. Just return
615      the -ESC_g code (cf \k). */      the -ESC_g code (cf \k). */
616    
617      case 'g':      case CHAR_g:
618      if (ptr[1] == '<' || ptr[1] == '\'')      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
619        {        {
620        c = -ESC_g;        c = -ESC_g;
621        break;        break;
# Line 557  else Line 623  else
623    
624      /* Handle the Perl-compatible cases */      /* Handle the Perl-compatible cases */
625    
626      if (ptr[1] == '{')      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
627        {        {
628        const uschar *p;        const uschar *p;
629        for (p = ptr+2; *p != 0 && *p != '}'; p++)        for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
630          if (*p != '-' && (digitab[*p] & ctype_digit) == 0) break;          if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
631        if (*p != 0 && *p != '}')        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
632          {          {
633          c = -ESC_k;          c = -ESC_k;
634          break;          break;
# Line 572  else Line 638  else
638        }        }
639      else braced = FALSE;      else braced = FALSE;
640    
641      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
642        {        {
643        negated = TRUE;        negated = TRUE;
644        ptr++;        ptr++;
# Line 581  else Line 647  else
647    
648      c = 0;      c = 0;
649      while ((digitab[ptr[1]] & ctype_digit) != 0)      while ((digitab[ptr[1]] & ctype_digit) != 0)
650        c = c * 10 + *(++ptr) - '0';        c = c * 10 + *(++ptr) - CHAR_0;
651    
652      if (c < 0)   /* Integer overflow */      if (c < 0)   /* Integer overflow */
653        {        {
# Line 589  else Line 655  else
655        break;        break;
656        }        }
657    
658      if (braced && *(++ptr) != '}')      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
659        {        {
660        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
661        break;        break;
# Line 626  else Line 692  else
692      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
693      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
694    
695      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:
696      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
697    
698      if (!isclass)      if (!isclass)
699        {        {
700        oldptr = ptr;        oldptr = ptr;
701        c -= '0';        c -= CHAR_0;
702        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
703          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - CHAR_0;
704        if (c < 0)    /* Integer overflow */        if (c < 0)    /* Integer overflow */
705          {          {
706          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
# Line 652  else Line 718  else
718      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.
719      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
720    
721      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
722        {        {
723        ptr--;        ptr--;
724        c = 0;        c = 0;
# Line 665  else Line 731  else
731      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
732      than 3 octal digits. */      than 3 octal digits. */
733    
734      case '0':      case CHAR_0:
735      c -= '0';      c -= CHAR_0;
736      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
737          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
738      if (!utf8 && c > 255) *errorcodeptr = ERR51;      if (!utf8 && c > 255) *errorcodeptr = ERR51;
739      break;      break;
740    
# Line 676  else Line 742  else
742      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
743      treated as a data character. */      treated as a data character. */
744    
745      case 'x':      case CHAR_x:
746      if (ptr[1] == '{')      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
747        {        {
748        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
749        int count = 0;        int count = 0;
# Line 686  else Line 752  else
752        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
753          {          {
754          register int cc = *pt++;          register int cc = *pt++;
755          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
756          count++;          count++;
757    
758  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
759          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
760          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
761  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
762          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
763          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
764  #endif  #endif
765          }          }
766    
767        if (*pt == '}')        if (*pt == CHAR_RIGHT_CURLY_BRACKET)
768          {          {
769          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
770          ptr = pt;          ptr = pt;
# Line 714  else Line 780  else
780      c = 0;      c = 0;
781      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
782        {        {
783        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
784        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
785  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
786        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
787        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
788  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
789        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
790        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
791  #endif  #endif
792        }        }
793      break;      break;
# Line 730  else Line 796  else
796      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
797      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
798    
799      case 'c':      case CHAR_c:
800      c = *(++ptr);      c = *(++ptr);
801      if (c == 0)      if (c == 0)
802        {        {
# Line 738  else Line 804  else
804        break;        break;
805        }        }
806    
807  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
808      if (c >= 'a' && c <= 'z') c -= 32;      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
809      c ^= 0x40;      c ^= 0x40;
810  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
811      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
812      c ^= 0xC0;      c ^= 0xC0;
813  #endif  #endif
814      break;      break;
# Line 763  else Line 829  else
829      break;      break;
830      }      }
831    }    }
832    
833    /* Perl supports \N{name} for character names, as well as plain \N for "not
834    newline". PCRE does not support \N{name}. */
835    
836    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET)
837      *errorcodeptr = ERR37;
838    
839  *ptrptr = ptr;  *ptrptr = ptr;
840  return c;  return c;
# Line 804  if (c == 0) goto ERROR_RETURN; Line 876  if (c == 0) goto ERROR_RETURN;
876  /* \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
877  negation. */  negation. */
878    
879  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
880    {    {
881    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
882      {      {
883      *negptr = TRUE;      *negptr = TRUE;
884      ptr++;      ptr++;
# Line 815  if (c == '{') Line 887  if (c == '{')
887      {      {
888      c = *(++ptr);      c = *(++ptr);
889      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
890      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
891      name[i] = c;      name[i] = c;
892      }      }
893    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
894    name[i] = 0;    name[i] = 0;
895    }    }
896    
# Line 883  is_counted_repeat(const uschar *p) Line 955  is_counted_repeat(const uschar *p)
955  {  {
956  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
957  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
958  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
959    
960  if (*p++ != ',') return FALSE;  if (*p++ != CHAR_COMMA) return FALSE;
961  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
962    
963  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
964  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
965    
966  return (*p == '}');  return (*p == CHAR_RIGHT_CURLY_BRACKET);
967  }  }
968    
969    
# Line 924  int max = -1; Line 996  int max = -1;
996  /* 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
997  an integer overflow. */  an integer overflow. */
998    
999  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;
1000  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1001    {    {
1002    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 934  if (min < 0 || min > 65535) Line 1006  if (min < 0 || min > 65535)
1006  /* 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.
1007  Also, max must not be less than min. */  Also, max must not be less than min. */
1008    
1009  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1010    {    {
1011    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1012      {      {
1013      max = 0;      max = 0;
1014      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;
1015      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1016        {        {
1017        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 964  return p; Line 1036  return p;
1036    
1037    
1038  /*************************************************  /*************************************************
1039  *       Find forward referenced subpattern       *  *  Subroutine for finding forward reference      *
1040  *************************************************/  *************************************************/
1041    
1042  /* This function scans along a pattern's text looking for capturing  /* This recursive function is called only from find_parens() below. The
1043    top-level call starts at the beginning of the pattern. All other calls must
1044    start at a parenthesis. It scans along a pattern's text looking for capturing
1045  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
1046  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
1047  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
1048  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
1049  be terminated by '>' because that is checked in the first pass.  first pass. Recursion is used to keep track of subpatterns that reset the
1050    capturing group numbers - the (?| feature.
1051    
1052  Arguments:  Arguments:
1053    ptr          current position in the pattern    ptrptr       address of the current character pointer (updated)
1054    cd           compile background data    cd           compile background data
1055    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1056    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1057    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1058      count        pointer to the current capturing subpattern number (updated)
1059    
1060  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1061  */  */
1062    
1063  static int  static int
1064  find_parens(const uschar *ptr, compile_data *cd, const uschar *name, int lorn,  find_parens_sub(uschar **ptrptr, compile_data *cd, const uschar *name, int lorn,
1065    BOOL xmode)    BOOL xmode, int *count)
1066  {  {
1067  const uschar *thisname;  uschar *ptr = *ptrptr;
1068  int count = cd->bracount;  int start_count = *count;
1069    int hwm_count = start_count;
1070    BOOL dup_parens = FALSE;
1071    
1072  for (; *ptr != 0; ptr++)  /* If the first character is a parenthesis, check on the type of group we are
1073    dealing with. The very first call may not start with a parenthesis. */
1074    
1075    if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1076    {    {
1077    int term;    if (ptr[1] == CHAR_QUESTION_MARK &&
1078          ptr[2] == CHAR_VERTICAL_LINE)
1079        {
1080        ptr += 3;
1081        dup_parens = TRUE;
1082        }
1083    
1084      /* Handle a normal, unnamed capturing parenthesis */
1085    
1086      else if (ptr[1] != CHAR_QUESTION_MARK && ptr[1] != CHAR_ASTERISK)
1087        {
1088        *count += 1;
1089        if (name == NULL && *count == lorn) return *count;
1090        ptr++;
1091        }
1092    
1093      /* Handle a condition. If it is an assertion, just carry on so that it
1094      is processed as normal. If not, skip to the closing parenthesis of the
1095      condition (there can't be any nested parens. */
1096    
1097      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1098        {
1099        ptr += 2;
1100        if (ptr[1] != CHAR_QUESTION_MARK)
1101          {
1102          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1103          if (*ptr != 0) ptr++;
1104          }
1105        }
1106    
1107      /* We have either (? or (* and not a condition */
1108    
1109      else
1110        {
1111        ptr += 2;
1112        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1113    
1114        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1115    
1116        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1117            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1118          {
1119          int term;
1120          const uschar *thisname;
1121          *count += 1;
1122          if (name == NULL && *count == lorn) return *count;
1123          term = *ptr++;
1124          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1125          thisname = ptr;
1126          while (*ptr != term) ptr++;
1127          if (name != NULL && lorn == ptr - thisname &&
1128              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1129            return *count;
1130          term++;
1131          }
1132        }
1133      }
1134    
1135    /* Past any initial parenthesis handling, scan for parentheses or vertical
1136    bars. */
1137    
1138    for (; *ptr != 0; ptr++)
1139      {
1140    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1141    
1142    if (*ptr == '\\')    if (*ptr == CHAR_BACKSLASH)
1143      {      {
1144      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1145      if (*ptr == 'Q') for (;;)      if (*ptr == CHAR_Q) for (;;)
1146        {        {
1147        while (*(++ptr) != 0 && *ptr != '\\') {};        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1148        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1149        if (*(++ptr) == 'E') break;        if (*(++ptr) == CHAR_E) break;
1150        }        }
1151      continue;      continue;
1152      }      }
# Line 1012  for (; *ptr != 0; ptr++) Line 1154  for (; *ptr != 0; ptr++)
1154    /* Skip over character classes; this logic must be similar to the way they    /* Skip over character classes; this logic must be similar to the way they
1155    are handled for real. If the first character is '^', skip it. Also, if the    are handled for real. If the first character is '^', skip it. Also, if the
1156    first few characters (either before or after ^) are \Q\E or \E we skip them    first few characters (either before or after ^) are \Q\E or \E we skip them
1157    too. This makes for compatibility with Perl. */    too. This makes for compatibility with Perl. Note the use of STR macros to
1158      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1159    
1160    if (*ptr == '[')    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1161      {      {
1162      BOOL negate_class = FALSE;      BOOL negate_class = FALSE;
1163      for (;;)      for (;;)
1164        {        {
1165        int c = *(++ptr);        if (ptr[1] == CHAR_BACKSLASH)
       if (c == '\\')  
1166          {          {
1167          if (ptr[1] == 'E') ptr++;          if (ptr[2] == CHAR_E)
1168            else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;            ptr+= 2;
1169              else break;          else if (strncmp((const char *)ptr+2,
1170                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1171              ptr += 4;
1172            else
1173              break;
1174          }          }
1175        else if (!negate_class && c == '^')        else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1176            {
1177          negate_class = TRUE;          negate_class = TRUE;
1178            ptr++;
1179            }
1180        else break;        else break;
1181        }        }
1182    
1183      /* If the next character is ']', it is a data character that must be      /* If the next character is ']', it is a data character that must be
1184      skipped, except in JavaScript compatibility mode. */      skipped, except in JavaScript compatibility mode. */
1185    
1186      if (ptr[1] == ']' && (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)      if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1187            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1188        ptr++;        ptr++;
1189    
1190      while (*(++ptr) != ']')      while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1191        {        {
1192        if (*ptr == 0) return -1;        if (*ptr == 0) return -1;
1193        if (*ptr == '\\')        if (*ptr == CHAR_BACKSLASH)
1194          {          {
1195          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) goto FAIL_EXIT;
1196          if (*ptr == 'Q') for (;;)          if (*ptr == CHAR_Q) for (;;)
1197            {            {
1198            while (*(++ptr) != 0 && *ptr != '\\') {};            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1199            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1200            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1201            }            }
1202          continue;          continue;
1203          }          }
# Line 1057  for (; *ptr != 0; ptr++) Line 1207  for (; *ptr != 0; ptr++)
1207    
1208    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1209    
1210    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1211      {      {
1212      while (*(++ptr) != 0 && *ptr != '\n') {};      while (*(++ptr) != 0 && *ptr != CHAR_NL) {};
1213      if (*ptr == 0) return -1;      if (*ptr == 0) goto FAIL_EXIT;
1214      continue;      continue;
1215      }      }
1216    
1217    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1218    
1219    if (*ptr != '(') continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != '?' && ptr[1] != '*')  
1220      {      {
1221      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, count);
1222      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1223      continue;      if (*ptr == 0) goto FAIL_EXIT;
1224        }
1225    
1226      else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1227        {
1228        if (dup_parens && *count < hwm_count) *count = hwm_count;
1229        *ptrptr = ptr;
1230        return -1;
1231      }      }
1232    
1233    ptr += 2;    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1234    if (*ptr == 'P') ptr++;                      /* Allow optional P */      {
1235        if (*count > hwm_count) hwm_count = *count;
1236        *count = start_count;
1237        }
1238      }
1239    
1240    /* We have to disambiguate (?<! and (?<= from (?<name> */  FAIL_EXIT:
1241    *ptrptr = ptr;
1242    return -1;
1243    }
1244    
   if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  
        *ptr != '\'')  
     continue;  
1245    
   count++;  
1246    
1247    if (name == NULL && count == lorn) return count;  
1248    term = *ptr++;  /*************************************************
1249    if (term == '<') term = '>';  *       Find forward referenced subpattern       *
1250    thisname = ptr;  *************************************************/
1251    while (*ptr != term) ptr++;  
1252    if (name != NULL && lorn == ptr - thisname &&  /* This function scans along a pattern's text looking for capturing
1253        strncmp((const char *)name, (const char *)thisname, lorn) == 0)  subpatterns, and counting them. If it finds a named pattern that matches the
1254      return count;  name it is given, it returns its number. Alternatively, if the name is NULL, it
1255    returns when it reaches a given numbered subpattern. This is used for forward
1256    references to subpatterns. We used to be able to start this scan from the
1257    current compiling point, using the current count value from cd->bracount, and
1258    do it all in a single loop, but the addition of the possibility of duplicate
1259    subpattern numbers means that we have to scan from the very start, in order to
1260    take account of such duplicates, and to use a recursive function to keep track
1261    of the different types of group.
1262    
1263    Arguments:
1264      cd           compile background data
1265      name         name to seek, or NULL if seeking a numbered subpattern
1266      lorn         name length, or subpattern number if name is NULL
1267      xmode        TRUE if we are in /x mode
1268    
1269    Returns:       the number of the found subpattern, or -1 if not found
1270    */
1271    
1272    static int
1273    find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode)
1274    {
1275    uschar *ptr = (uschar *)cd->start_pattern;
1276    int count = 0;
1277    int rc;
1278    
1279    /* If the pattern does not start with an opening parenthesis, the first call
1280    to find_parens_sub() will scan right to the end (if necessary). However, if it
1281    does start with a parenthesis, find_parens_sub() will return when it hits the
1282    matching closing parens. That is why we have to have a loop. */
1283    
1284    for (;;)
1285      {
1286      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, &count);
1287      if (rc > 0 || *ptr++ == 0) break;
1288    }    }
1289    
1290  return -1;  return rc;
1291  }  }
1292    
1293    
1294    
1295    
1296  /*************************************************  /*************************************************
1297  *      Find first significant op code            *  *      Find first significant op code            *
1298  *************************************************/  *************************************************/
# Line 1149  for (;;) Line 1342  for (;;)
1342    
1343      case OP_CALLOUT:      case OP_CALLOUT:
1344      case OP_CREF:      case OP_CREF:
1345        case OP_NCREF:
1346      case OP_RREF:      case OP_RREF:
1347        case OP_NRREF:
1348      case OP_DEF:      case OP_DEF:
1349      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1350      break;      break;
# Line 1165  for (;;) Line 1360  for (;;)
1360    
1361    
1362  /*************************************************  /*************************************************
1363  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1364  *************************************************/  *************************************************/
1365    
1366  /* 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,
1367  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.
1368  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
1369    temporarily terminated with OP_END when this function is called.
1370    
1371    This function is called when a backward assertion is encountered, so that if it
1372    fails, the error message can point to the correct place in the pattern.
1373    However, we cannot do this when the assertion contains subroutine calls,
1374    because they can be forward references. We solve this by remembering this case
1375    and doing the check at the end; a flag specifies which mode we are running in.
1376    
1377  Arguments:  Arguments:
1378    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1379    options  the compiling options    options  the compiling options
1380      atend    TRUE if called when the pattern is complete
1381      cd       the "compile data" structure
1382    
1383  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1384                 or -1 if there is no fixed length,
1385               or -2 if \C was encountered               or -2 if \C was encountered
1386                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1387  */  */
1388    
1389  static int  static int
1390  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)
1391  {  {
1392  int length = -1;  int length = -1;
1393    
# Line 1194  branch, check the length against that of Line 1400  branch, check the length against that of
1400  for (;;)  for (;;)
1401    {    {
1402    int d;    int d;
1403      uschar *ce, *cs;
1404    register int op = *cc;    register int op = *cc;
1405    switch (op)    switch (op)
1406      {      {
# Line 1201  for (;;) Line 1408  for (;;)
1408      case OP_BRA:      case OP_BRA:
1409      case OP_ONCE:      case OP_ONCE:
1410      case OP_COND:      case OP_COND:
1411      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);
1412      if (d < 0) return d;      if (d < 0) return d;
1413      branchlength += d;      branchlength += d;
1414      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1224  for (;;) Line 1431  for (;;)
1431      branchlength = 0;      branchlength = 0;
1432      break;      break;
1433    
1434        /* A true recursion implies not fixed length, but a subroutine call may
1435        be OK. If the subroutine is a forward reference, we can't deal with
1436        it until the end of the pattern, so return -3. */
1437    
1438        case OP_RECURSE:
1439        if (!atend) return -3;
1440        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1441        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1442        if (cc > cs && cc < ce) return -1;                /* Recursion */
1443        d = find_fixedlength(cs + 2, options, atend, cd);
1444        if (d < 0) return d;
1445        branchlength += d;
1446        cc += 1 + LINK_SIZE;
1447        break;
1448    
1449      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1450    
1451      case OP_ASSERT:      case OP_ASSERT:
# Line 1237  for (;;) Line 1459  for (;;)
1459    
1460      case OP_REVERSE:      case OP_REVERSE:
1461      case OP_CREF:      case OP_CREF:
1462        case OP_NCREF:
1463      case OP_RREF:      case OP_RREF:
1464        case OP_NRREF:
1465      case OP_DEF:      case OP_DEF:
1466      case OP_OPT:      case OP_OPT:
1467      case OP_CALLOUT:      case OP_CALLOUT:
1468      case OP_SOD:      case OP_SOD:
1469      case OP_SOM:      case OP_SOM:
1470        case OP_SET_SOM:
1471      case OP_EOD:      case OP_EOD:
1472      case OP_EODN:      case OP_EODN:
1473      case OP_CIRC:      case OP_CIRC:
# Line 1260  for (;;) Line 1485  for (;;)
1485      branchlength++;      branchlength++;
1486      cc += 2;      cc += 2;
1487  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1488      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1489        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1490  #endif  #endif
1491      break;      break;
1492    
# Line 1274  for (;;) Line 1497  for (;;)
1497      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1498      cc += 4;      cc += 4;
1499  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1500      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1501        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1502  #endif  #endif
1503      break;      break;
1504    
# Line 1356  for (;;) Line 1577  for (;;)
1577    
1578    
1579  /*************************************************  /*************************************************
1580  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1581  *************************************************/  *************************************************/
1582    
1583  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1584  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1585    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1586    so that it can be called from pcre_study() when finding the minimum matching
1587    length.
1588    
1589  Arguments:  Arguments:
1590    code        points to start of expression    code        points to start of expression
1591    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1592    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1593    
1594  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
1595  */  */
1596    
1597  static const uschar *  const uschar *
1598  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1599  {  {
1600  for (;;)  for (;;)
1601    {    {
# Line 1384  for (;;) Line 1608  for (;;)
1608    
1609    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1610    
1611      /* Handle recursion */
1612    
1613      else if (c == OP_REVERSE)
1614        {
1615        if (number < 0) return (uschar *)code;
1616        code += _pcre_OP_lengths[c];
1617        }
1618    
1619    /* Handle capturing bracket */    /* Handle capturing bracket */
1620    
1621    else if (c == OP_CBRA)    else if (c == OP_CBRA)
# Line 1395  for (;;) Line 1627  for (;;)
1627    
1628    /* 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
1629    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
1630    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1631      must add in its length. */
1632    
1633    else    else
1634      {      {
# Line 1419  for (;;) Line 1652  for (;;)
1652        case OP_TYPEPOSUPTO:        case OP_TYPEPOSUPTO:
1653        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1654        break;        break;
1655    
1656          case OP_MARK:
1657          case OP_PRUNE_ARG:
1658          case OP_SKIP_ARG:
1659          case OP_THEN_ARG:
1660          code += code[1];
1661          break;
1662        }        }
1663    
1664      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1490  for (;;) Line 1730  for (;;)
1730    
1731    /* 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
1732    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
1733    two bytes of parameters. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1734      must add in its length. */
1735    
1736    else    else
1737      {      {
# Line 1514  for (;;) Line 1755  for (;;)
1755        case OP_TYPEEXACT:        case OP_TYPEEXACT:
1756        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1757        break;        break;
1758    
1759          case OP_MARK:
1760          case OP_PRUNE_ARG:
1761          case OP_SKIP_ARG:
1762          case OP_THEN_ARG:
1763          code += code[1];
1764          break;
1765        }        }
1766    
1767      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1570  Arguments: Line 1818  Arguments:
1818    code        points to start of search    code        points to start of search
1819    endcode     points to where to stop    endcode     points to where to stop
1820    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
1821      cd          contains pointers to tables etc.
1822    
1823  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
1824  */  */
1825    
1826  static BOOL  static BOOL
1827  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
1828      compile_data *cd)
1829  {  {
1830  register int c;  register int c;
1831  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 1606  for (code = first_significant_code(code Line 1856  for (code = first_significant_code(code
1856      continue;      continue;
1857      }      }
1858    
1859      /* For a recursion/subroutine call, if its end has been reached, which
1860      implies a subroutine call, we can scan it. */
1861    
1862      if (c == OP_RECURSE)
1863        {
1864        BOOL empty_branch = FALSE;
1865        const uschar *scode = cd->start_code + GET(code, 1);
1866        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
1867        do
1868          {
1869          if (could_be_empty_branch(scode, endcode, utf8, cd))
1870            {
1871            empty_branch = TRUE;
1872            break;
1873            }
1874          scode += GET(scode, 1);
1875          }
1876        while (*scode == OP_ALT);
1877        if (!empty_branch) return FALSE;  /* All branches are non-empty */
1878        continue;
1879        }
1880    
1881    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
1882    
1883    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 1613  for (code = first_significant_code(code Line 1885  for (code = first_significant_code(code
1885      BOOL empty_branch;      BOOL empty_branch;
1886      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
1887    
1888      /* Scan a closed bracket */      /* If a conditional group has only one branch, there is a second, implied,
1889        empty branch, so just skip over the conditional, because it could be empty.
1890        Otherwise, scan the individual branches of the group. */
1891    
1892      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;  
1893        code += GET(code, 1);        code += GET(code, 1);
1894        else
1895          {
1896          empty_branch = FALSE;
1897          do
1898            {
1899            if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
1900              empty_branch = TRUE;
1901            code += GET(code, 1);
1902            }
1903          while (*code == OP_ALT);
1904          if (!empty_branch) return FALSE;   /* All branches are non-empty */
1905        }        }
1906      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
1907      c = *code;      c = *code;
1908      continue;      continue;
1909      }      }
# Line 1741  for (code = first_significant_code(code Line 2021  for (code = first_significant_code(code
2021      case OP_QUERY:      case OP_QUERY:
2022      case OP_MINQUERY:      case OP_MINQUERY:
2023      case OP_POSQUERY:      case OP_POSQUERY:
2024        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2025        break;
2026    
2027      case OP_UPTO:      case OP_UPTO:
2028      case OP_MINUPTO:      case OP_MINUPTO:
2029      case OP_POSUPTO:      case OP_POSUPTO:
2030      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2031      break;      break;
2032  #endif  #endif
2033    
2034        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2035        string. */
2036    
2037        case OP_MARK:
2038        case OP_PRUNE_ARG:
2039        case OP_SKIP_ARG:
2040        case OP_THEN_ARG:
2041        code += code[1];
2042        break;
2043    
2044        /* None of the remaining opcodes are required to match a character. */
2045    
2046        default:
2047        break;
2048      }      }
2049    }    }
2050    
# Line 1769  Arguments: Line 2067  Arguments:
2067    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2068    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2069    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2070      cd          pointers to tables etc
2071    
2072  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2073  */  */
2074    
2075  static BOOL  static BOOL
2076  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2077    BOOL utf8)    BOOL utf8, compile_data *cd)
2078  {  {
2079  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2080    {    {
2081    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2082        return FALSE;
2083    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2084    }    }
2085  return TRUE;  return TRUE;
# Line 1825  int terminator; /* Don't combin Line 2125  int terminator; /* Don't combin
2125  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2126  for (++ptr; *ptr != 0; ptr++)  for (++ptr; *ptr != 0; ptr++)
2127    {    {
2128    if (*ptr == '\\' && ptr[1] == ']') ptr++; else    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else
2129      {      {
2130      if (*ptr == ']') return FALSE;      if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2131      if (*ptr == terminator && ptr[1] == ']')      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2132        {        {
2133        *endptr = ptr;        *endptr = ptr;
2134        return TRUE;        return TRUE;
# Line 2074  if ((options & PCRE_EXTENDED) != 0) Line 2374  if ((options & PCRE_EXTENDED) != 0)
2374    for (;;)    for (;;)
2375      {      {
2376      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2377      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2378        {        {
2379        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2380          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
# Line 2086  if ((options & PCRE_EXTENDED) != 0) Line 2386  if ((options & PCRE_EXTENDED) != 0)
2386  /* 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
2387  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2388    
2389  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2390    {    {
2391    int temperrorcode = 0;    int temperrorcode = 0;
2392    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
# Line 2111  if ((options & PCRE_EXTENDED) != 0) Line 2411  if ((options & PCRE_EXTENDED) != 0)
2411    for (;;)    for (;;)
2412      {      {
2413      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2414      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2415        {        {
2416        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2417          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
# Line 2122  if ((options & PCRE_EXTENDED) != 0) Line 2422  if ((options & PCRE_EXTENDED) != 0)
2422    
2423  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
2424    
2425  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2426    return FALSE;    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2427        return FALSE;
2428    
2429  /* 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
2430  positive single character match, "item" either contains the character or, if  positive single character match, "item" either contains the character or, if
# Line 2440  BOOL utf8 = FALSE; Line 2741  BOOL utf8 = FALSE;
2741  uschar *utf8_char = NULL;  uschar *utf8_char = NULL;
2742  #endif  #endif
2743    
2744  #ifdef DEBUG  #ifdef PCRE_DEBUG
2745  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2746  #endif  #endif
2747    
# Line 2499  for (;; ptr++) Line 2800  for (;; ptr++)
2800    
2801    if (lengthptr != NULL)    if (lengthptr != NULL)
2802      {      {
2803  #ifdef DEBUG  #ifdef PCRE_DEBUG
2804      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
2805  #endif  #endif
2806      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
2807        {        {
2808        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
2809        goto FAILED;        goto FAILED;
# Line 2551  for (;; ptr++) Line 2852  for (;; ptr++)
2852    /* 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
2853    reference list. */    reference list. */
2854    
2855    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
2856      {      {
2857      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
2858      goto FAILED;      goto FAILED;
# Line 2561  for (;; ptr++) Line 2862  for (;; ptr++)
2862    
2863    if (inescq && c != 0)    if (inescq && c != 0)
2864      {      {
2865      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
2866        {        {
2867        inescq = FALSE;        inescq = FALSE;
2868        ptr++;        ptr++;
# Line 2587  for (;; ptr++) Line 2888  for (;; ptr++)
2888    /* 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
2889    a quantifier. */    a quantifier. */
2890    
2891    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
2892      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
2893        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
2894    
2895    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
2896         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2603  for (;; ptr++) Line 2905  for (;; ptr++)
2905    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
2906      {      {
2907      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
2908      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
2909        {        {
2910        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2911          {          {
# Line 2628  for (;; ptr++) Line 2930  for (;; ptr++)
2930      {      {
2931      /* ===================================================================*/      /* ===================================================================*/
2932      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
2933      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
2934      case ')':      case CHAR_RIGHT_PARENTHESIS:
2935      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
2936      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
2937      *codeptr = code;      *codeptr = code;
# Line 2651  for (;; ptr++) Line 2953  for (;; ptr++)
2953      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
2954      the setting of any following char as a first character. */      the setting of any following char as a first character. */
2955    
2956      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
2957      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
2958        {        {
2959        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
# Line 2660  for (;; ptr++) Line 2962  for (;; ptr++)
2962      *code++ = OP_CIRC;      *code++ = OP_CIRC;
2963      break;      break;
2964    
2965      case '$':      case CHAR_DOLLAR_SIGN:
2966      previous = NULL;      previous = NULL;
2967      *code++ = OP_DOLL;      *code++ = OP_DOLL;
2968      break;      break;
# Line 2668  for (;; ptr++) Line 2970  for (;; ptr++)
2970      /* 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
2971      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
2972    
2973      case '.':      case CHAR_DOT:
2974      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
2975      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
2976      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
# Line 2692  for (;; ptr++) Line 2994  for (;; ptr++)
2994      In JavaScript compatibility mode, an isolated ']' causes an error. In      In JavaScript compatibility mode, an isolated ']' causes an error. In
2995      default (Perl) mode, it is treated as a data character. */      default (Perl) mode, it is treated as a data character. */
2996    
2997      case ']':      case CHAR_RIGHT_SQUARE_BRACKET:
2998      if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)      if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2999        {        {
3000        *errorcodeptr = ERR64;        *errorcodeptr = ERR64;
# Line 2700  for (;; ptr++) Line 3002  for (;; ptr++)
3002        }        }
3003      goto NORMAL_CHAR;      goto NORMAL_CHAR;
3004    
3005      case '[':      case CHAR_LEFT_SQUARE_BRACKET:
3006      previous = code;      previous = code;
3007    
3008      /* 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
3009      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. */
3010    
3011      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3012             ptr[1] == CHAR_EQUALS_SIGN) &&
3013          check_posix_syntax(ptr, &tempptr))          check_posix_syntax(ptr, &tempptr))
3014        {        {
3015        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
3016        goto FAILED;        goto FAILED;
3017        }        }
3018    
# Line 2721  for (;; ptr++) Line 3024  for (;; ptr++)
3024      for (;;)      for (;;)
3025        {        {
3026        c = *(++ptr);        c = *(++ptr);
3027        if (c == '\\')        if (c == CHAR_BACKSLASH)
3028          {          {
3029          if (ptr[1] == 'E') ptr++;          if (ptr[1] == CHAR_E)
3030            else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;            ptr++;
3031              else break;          else if (strncmp((const char *)ptr+1,
3032                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
3033              ptr += 3;
3034            else
3035              break;
3036          }          }
3037        else if (!negate_class && c == '^')        else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
3038          negate_class = TRUE;          negate_class = TRUE;
3039        else break;        else break;
3040        }        }
# Line 2737  for (;; ptr++) Line 3044  for (;; ptr++)
3044      that. In JS mode, [] must always fail, so generate OP_FAIL, whereas      that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3045      [^] must match any character, so generate OP_ALLANY. */      [^] must match any character, so generate OP_ALLANY. */
3046    
3047      if (c ==']' && (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)      if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3048            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3049        {        {
3050        *code++ = negate_class? OP_ALLANY : OP_FAIL;        *code++ = negate_class? OP_ALLANY : OP_FAIL;
3051        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
# Line 2802  for (;; ptr++) Line 3110  for (;; ptr++)
3110    
3111        if (inescq)        if (inescq)
3112          {          {
3113          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3114            {            {
3115            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
3116            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2817  for (;; ptr++) Line 3125  for (;; ptr++)
3125        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3126        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3127    
3128        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3129            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3130            check_posix_syntax(ptr, &tempptr))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3131          {          {
3132          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3133          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3134          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3135          uschar pbits[32];          uschar pbits[32];
3136    
3137          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3138            {            {
3139            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3140            goto FAILED;            goto FAILED;
3141            }            }
3142    
3143          ptr += 2;          ptr += 2;
3144          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3145            {            {
3146            local_negate = TRUE;            local_negate = TRUE;
3147            should_flip_negation = TRUE;  /* Note negative special */            should_flip_negation = TRUE;  /* Note negative special */
# Line 2901  for (;; ptr++) Line 3209  for (;; ptr++)
3209    
3210        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3211        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
3212        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
3213        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
3214        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
3215        character in them, so set class_charcount bigger than one. */        are either treated as literal characters (by default), or are faulted if
3216          PCRE_EXTRA is set. */
3217    
3218        if (c == '\\')        if (c == CHAR_BACKSLASH)
3219          {          {
3220          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3221          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3222    
3223          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 */  
3224          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3225            {            {
3226            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3227              {              {
3228              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3229              }              }
# Line 3142  for (;; ptr++) Line 3449  for (;; ptr++)
3449        entirely. The code for handling \Q and \E is messy. */        entirely. The code for handling \Q and \E is messy. */
3450    
3451        CHECK_RANGE:        CHECK_RANGE:
3452        while (ptr[1] == '\\' && ptr[2] == 'E')        while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3453          {          {
3454          inescq = FALSE;          inescq = FALSE;
3455          ptr += 2;          ptr += 2;
# Line 3152  for (;; ptr++) Line 3459  for (;; ptr++)
3459    
3460        /* Remember \r or \n */        /* Remember \r or \n */
3461    
3462        if (c == '\r' || c == '\n') cd->external_flags |= PCRE_HASCRORLF;        if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3463    
3464        /* Check for range */        /* Check for range */
3465    
3466        if (!inescq && ptr[1] == '-')        if (!inescq && ptr[1] == CHAR_MINUS)
3467          {          {
3468          int d;          int d;
3469          ptr += 2;          ptr += 2;
3470          while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3471    
3472          /* 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
3473          mode. */          mode. */
3474    
3475          while (*ptr == '\\' && ptr[1] == 'Q')          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3476            {            {
3477            ptr += 2;            ptr += 2;
3478            if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }            if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3479                { ptr += 2; continue; }
3480            inescq = TRUE;            inescq = TRUE;
3481            break;            break;
3482            }            }
3483    
3484          if (*ptr == 0 || (!inescq && *ptr == ']'))          if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3485            {            {
3486            ptr = oldptr;            ptr = oldptr;
3487            goto LONE_SINGLE_CHARACTER;            goto LONE_SINGLE_CHARACTER;
# Line 3192  for (;; ptr++) Line 3500  for (;; ptr++)
3500          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
3501          in such circumstances. */          in such circumstances. */
3502    
3503          if (!inescq && d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3504            {            {
3505            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3506            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3507    
3508            /* \b is backspace; \X is literal X; \R is literal R; any other            /* \b is backspace; any other special means the '-' was literal */
           special means the '-' was literal */  
3509    
3510            if (d < 0)            if (d < 0)
3511              {              {
3512              if (d == -ESC_b) d = '\b';              if (d == -ESC_b) d = CHAR_BS; else
             else if (d == -ESC_X) d = 'X';  
             else if (d == -ESC_R) d = 'R'; else  
3513                {                {
3514                ptr = oldptr;                ptr = oldptr;
3515                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 3225  for (;; ptr++) Line 3530  for (;; ptr++)
3530    
3531          /* Remember \r or \n */          /* Remember \r or \n */
3532    
3533          if (d == '\r' || d == '\n') cd->external_flags |= PCRE_HASCRORLF;          if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3534    
3535          /* 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
3536          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 3372  for (;; ptr++) Line 3677  for (;; ptr++)
3677    
3678      /* Loop until ']' reached. This "while" is the end of the "do" above. */      /* Loop until ']' reached. This "while" is the end of the "do" above. */
3679    
3680      while ((c = *(++ptr)) != 0 && (c != ']' || inescq));      while ((c = *(++ptr)) != 0 && (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
3681    
3682      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)                          /* Missing terminating ']' */
3683        {        {
# Line 3517  we set the flag only if there is a liter Line 3822  we set the flag only if there is a liter
3822      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
3823      has been tested above. */      has been tested above. */
3824    
3825      case '{':      case CHAR_LEFT_CURLY_BRACKET:
3826      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
3827      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
3828      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
3829      goto REPEAT;      goto REPEAT;
3830    
3831      case '*':      case CHAR_ASTERISK:
3832      repeat_min = 0;      repeat_min = 0;
3833      repeat_max = -1;      repeat_max = -1;
3834      goto REPEAT;      goto REPEAT;
3835    
3836      case '+':      case CHAR_PLUS:
3837      repeat_min = 1;      repeat_min = 1;
3838      repeat_max = -1;      repeat_max = -1;
3839      goto REPEAT;      goto REPEAT;
3840    
3841      case '?':      case CHAR_QUESTION_MARK:
3842      repeat_min = 0;      repeat_min = 0;
3843      repeat_max = 1;      repeat_max = 1;
3844    
# Line 3568  we set the flag only if there is a liter Line 3873  we set the flag only if there is a liter
3873      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
3874      repeat type to the non-default. */      repeat type to the non-default. */
3875    
3876      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
3877        {        {
3878        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
3879        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
3880        ptr++;        ptr++;
3881        }        }
3882      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
3883        {        {
3884        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
3885        ptr++;        ptr++;
# Line 3689  we set the flag only if there is a liter Line 3994  we set the flag only if there is a liter
3994    
3995        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
3996    
3997          /*--------------------------------------------------------------------*/
3998          /* This code is obsolete from release 8.00; the restriction was finally
3999          removed: */
4000    
4001        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4002        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4003    
4004        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4005          /*--------------------------------------------------------------------*/
4006    
4007        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
4008    
# Line 3839  we set the flag only if there is a liter Line 4149  we set the flag only if there is a liter
4149          goto END_REPEAT;          goto END_REPEAT;
4150          }          }
4151    
4152          /*--------------------------------------------------------------------*/
4153          /* This code is obsolete from release 8.00; the restriction was finally
4154          removed: */
4155    
4156        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4157        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4158    
4159        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4160          /*--------------------------------------------------------------------*/
4161    
4162        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4163          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 3977  we set the flag only if there is a liter Line 4292  we set the flag only if there is a liter
4292            {            {
4293            /* 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
4294            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
4295            potential integer overflow. */            potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
4296              integer type when available, otherwise double. */
4297    
4298            if (lengthptr != NULL)            if (lengthptr != NULL)
4299              {              {
4300              int delta = (repeat_min - 1)*length_prevgroup;              int delta = (repeat_min - 1)*length_prevgroup;
4301              if ((double)(repeat_min - 1)*(double)length_prevgroup >              if ((INT64_OR_DOUBLE)(repeat_min - 1)*
4302                                                              (double)INT_MAX ||                    (INT64_OR_DOUBLE)length_prevgroup >
4303                        (INT64_OR_DOUBLE)INT_MAX ||
4304                  OFLOW_MAX - *lengthptr < delta)                  OFLOW_MAX - *lengthptr < delta)
4305                {                {
4306                *errorcodeptr = ERR20;                *errorcodeptr = ERR20;
# Line 4029  we set the flag only if there is a liter Line 4346  we set the flag only if there is a liter
4346          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
4347          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
4348          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
4349          paranoid checks to avoid integer overflow. */          paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
4350            a 64-bit integer type when available, otherwise double. */
4351    
4352          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4353            {            {
4354            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4355                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4356            if ((double)repeat_max *            if ((INT64_OR_DOUBLE)repeat_max *
4357                  (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)                  (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4358                    > (double)INT_MAX ||                    > (INT64_OR_DOUBLE)INT_MAX ||
4359                OFLOW_MAX - *lengthptr < delta)                OFLOW_MAX - *lengthptr < delta)
4360              {              {
4361              *errorcodeptr = ERR20;              *errorcodeptr = ERR20;
# Line 4114  we set the flag only if there is a liter Line 4432  we set the flag only if there is a liter
4432            uschar *scode = bracode;            uschar *scode = bracode;
4433            do            do
4434              {              {
4435              if (could_be_empty_branch(scode, ketcode, utf8))              if (could_be_empty_branch(scode, ketcode, utf8, cd))
4436                {                {
4437                *bracode += OP_SBRA - OP_BRA;                *bracode += OP_SBRA - OP_BRA;
4438                break;                break;
# Line 4157  we set the flag only if there is a liter Line 4475  we set the flag only if there is a liter
4475      if (possessive_quantifier)      if (possessive_quantifier)
4476        {        {
4477        int len;        int len;
4478        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||  
4479            *tempcode == OP_NOTEXACT)        if (*tempcode == OP_TYPEEXACT)
4480          tempcode += _pcre_OP_lengths[*tempcode] +          tempcode += _pcre_OP_lengths[*tempcode] +
4481            ((*tempcode == OP_TYPEEXACT &&            ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
4482               (tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP))? 2:0);  
4483          else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
4484            {
4485            tempcode += _pcre_OP_lengths[*tempcode];
4486    #ifdef SUPPORT_UTF8
4487            if (utf8 && tempcode[-1] >= 0xc0)
4488              tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];
4489    #endif
4490            }
4491    
4492        len = code - tempcode;        len = code - tempcode;
4493        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4494          {          {
# Line 4180  we set the flag only if there is a liter Line 4507  we set the flag only if there is a liter
4507          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4508          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4509    
4510            /* Because we are moving code along, we must ensure that any
4511            pending recursive references are updated. */
4512    
4513          default:          default:
4514            *code = OP_END;
4515            adjust_recurse(tempcode, 1 + LINK_SIZE, utf8, cd, save_hwm);
4516          memmove(tempcode + 1+LINK_SIZE, tempcode, len);          memmove(tempcode + 1+LINK_SIZE, tempcode, len);
4517          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
4518          len += 1 + LINK_SIZE;          len += 1 + LINK_SIZE;
# Line 4207  we set the flag only if there is a liter Line 4539  we set the flag only if there is a liter
4539      lookbehind or option setting or condition or all the other extended      lookbehind or option setting or condition or all the other extended
4540      parenthesis forms.  */      parenthesis forms.  */
4541    
4542      case '(':      case CHAR_LEFT_PARENTHESIS:
4543      newoptions = options;      newoptions = options;
4544      skipbytes = 0;      skipbytes = 0;
4545      bravalue = OP_CBRA;      bravalue = OP_CBRA;
# Line 4216  we set the flag only if there is a liter Line 4548  we set the flag only if there is a liter
4548    
4549      /* First deal with various "verbs" that can be introduced by '*'. */      /* First deal with various "verbs" that can be introduced by '*'. */
4550    
4551      if (*(++ptr) == '*' && (cd->ctypes[ptr[1]] & ctype_letter) != 0)      if (*(++ptr) == CHAR_ASTERISK &&
4552             ((cd->ctypes[ptr[1]] & ctype_letter) != 0 || ptr[1] == ':'))
4553        {        {
4554        int i, namelen;        int i, namelen;
4555          int arglen = 0;
4556        const char *vn = verbnames;        const char *vn = verbnames;
4557        const uschar *name = ++ptr;        const uschar *name = ptr + 1;
4558          const uschar *arg = NULL;
4559        previous = NULL;        previous = NULL;
4560        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
4561        if (*ptr == ':')        namelen = ptr - name;
4562    
4563          if (*ptr == CHAR_COLON)
4564          {          {
4565          *errorcodeptr = ERR59;   /* Not supported */          arg = ++ptr;
4566          goto FAILED;          while ((cd->ctypes[*ptr] & (ctype_letter|ctype_digit)) != 0
4567              || *ptr == '_') ptr++;
4568            arglen = ptr - arg;
4569          }          }
4570        if (*ptr != ')')  
4571          if (*ptr != CHAR_RIGHT_PARENTHESIS)
4572          {          {
4573          *errorcodeptr = ERR60;          *errorcodeptr = ERR60;
4574          goto FAILED;          goto FAILED;
4575          }          }
4576        namelen = ptr - name;  
4577          /* Scan the table of verb names */
4578    
4579        for (i = 0; i < verbcount; i++)        for (i = 0; i < verbcount; i++)
4580          {          {
4581          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
4582              strncmp((char *)name, vn, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
4583            {            {
4584            *code = verbs[i].op;            /* Check for open captures before ACCEPT */
4585            if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;  
4586            break;            if (verbs[i].op == OP_ACCEPT)
4587                {
4588                open_capitem *oc;
4589                cd->had_accept = TRUE;
4590                for (oc = cd->open_caps; oc != NULL; oc = oc->next)
4591                  {
4592                  *code++ = OP_CLOSE;
4593                  PUT2INC(code, 0, oc->number);
4594                  }
4595                }
4596    
4597              /* Handle the cases with/without an argument */
4598    
4599              if (arglen == 0)
4600                {
4601                if (verbs[i].op < 0)   /* Argument is mandatory */
4602                  {
4603                  *errorcodeptr = ERR66;
4604                  goto FAILED;
4605                  }
4606                *code++ = verbs[i].op;
4607                }
4608    
4609              else
4610                {
4611                if (verbs[i].op_arg < 0)   /* Argument is forbidden */
4612                  {
4613                  *errorcodeptr = ERR59;
4614                  goto FAILED;
4615                  }
4616                *code++ = verbs[i].op_arg;
4617                *code++ = arglen;
4618                memcpy(code, arg, arglen);
4619                code += arglen;
4620                *code++ = 0;
4621                }
4622    
4623              break;  /* Found verb, exit loop */
4624            }            }
4625    
4626          vn += verbs[i].len + 1;          vn += verbs[i].len + 1;
4627          }          }
4628        if (i < verbcount) continue;  
4629        *errorcodeptr = ERR60;        if (i < verbcount) continue;    /* Successfully handled a verb */
4630          *errorcodeptr = ERR60;          /* Verb not recognized */
4631        goto FAILED;        goto FAILED;
4632        }        }
4633    
4634      /* Deal with the extended parentheses; all are introduced by '?', and the      /* Deal with the extended parentheses; all are introduced by '?', and the
4635      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. */
4636    
4637      else if (*ptr == '?')      else if (*ptr == CHAR_QUESTION_MARK)
4638        {        {
4639        int i, set, unset, namelen;        int i, set, unset, namelen;
4640        int *optset;        int *optset;
# Line 4262  we set the flag only if there is a liter Line 4643  we set the flag only if there is a liter
4643    
4644        switch (*(++ptr))        switch (*(++ptr))
4645          {          {
4646          case '#':                 /* Comment; skip to ket */          case CHAR_NUMBER_SIGN:                 /* Comment; skip to ket */
4647          ptr++;          ptr++;
4648          while (*ptr != 0 && *ptr != ')') ptr++;          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
4649          if (*ptr == 0)          if (*ptr == 0)
4650            {            {
4651            *errorcodeptr = ERR18;            *errorcodeptr = ERR18;
# Line 4274  we set the flag only if there is a liter Line 4655  we set the flag only if there is a liter
4655    
4656    
4657          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4658          case '|':                 /* Reset capture count for each branch */          case CHAR_VERTICAL_LINE:  /* Reset capture count for each branch */
4659          reset_bracount = TRUE;          reset_bracount = TRUE;
4660          /* Fall through */          /* Fall through */
4661    
4662          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4663          case ':':                 /* Non-capturing bracket */          case CHAR_COLON:          /* Non-capturing bracket */
4664          bravalue = OP_BRA;          bravalue = OP_BRA;
4665          ptr++;          ptr++;
4666          break;          break;
4667    
4668    
4669          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4670          case '(':          case CHAR_LEFT_PARENTHESIS:
4671          bravalue = OP_COND;       /* Conditional group */          bravalue = OP_COND;       /* Conditional group */
4672    
4673          /* 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 4306  we set the flag only if there is a liter Line 4687  we set the flag only if there is a liter
4687          the switch. This will take control down to where bracketed groups,          the switch. This will take control down to where bracketed groups,
4688          including assertions, are processed. */          including assertions, are processed. */
4689    
4690          if (ptr[1] == '?' && (ptr[2] == '=' || ptr[2] == '!' || ptr[2] == '<'))          if (ptr[1] == CHAR_QUESTION_MARK && (ptr[2] == CHAR_EQUALS_SIGN ||
4691                ptr[2] == CHAR_EXCLAMATION_MARK || ptr[2] == CHAR_LESS_THAN_SIGN))
4692            break;            break;
4693    
4694          /* 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 4318  we set the flag only if there is a liter Line 4700  we set the flag only if there is a liter
4700    
4701          /* Check for a test for recursion in a named group. */          /* Check for a test for recursion in a named group. */
4702    
4703          if (ptr[1] == 'R' && ptr[2] == '&')          if (ptr[1] == CHAR_R && ptr[2] == CHAR_AMPERSAND)
4704            {            {
4705            terminator = -1;            terminator = -1;
4706            ptr += 2;            ptr += 2;
# Line 4328  we set the flag only if there is a liter Line 4710  we set the flag only if there is a liter
4710          /* 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
4711          syntax (?(<name>) or (?('name') */          syntax (?(<name>) or (?('name') */
4712    
4713          else if (ptr[1] == '<')          else if (ptr[1] == CHAR_LESS_THAN_SIGN)
4714            {            {
4715            terminator = '>';            terminator = CHAR_GREATER_THAN_SIGN;
4716            ptr++;            ptr++;
4717            }            }
4718          else if (ptr[1] == '\'')          else if (ptr[1] == CHAR_APOSTROPHE)
4719            {            {
4720            terminator = '\'';            terminator = CHAR_APOSTROPHE;
4721            ptr++;            ptr++;
4722            }            }
4723          else          else
4724            {            {
4725            terminator = 0;            terminator = 0;
4726            if (ptr[1] == '-' || ptr[1] == '+') refsign = *(++ptr);            if (ptr[1] == CHAR_MINUS || ptr[1] == CHAR_PLUS) refsign = *(++ptr);
4727            }            }
4728    
4729          /* 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 4361  we set the flag only if there is a liter Line 4743  we set the flag only if there is a liter
4743            {            {
4744            if (recno >= 0)            if (recno >= 0)
4745              recno = ((digitab[*ptr] & ctype_digit) != 0)?              recno = ((digitab[*ptr] & ctype_digit) != 0)?
4746                recno * 10 + *ptr - '0' : -1;                recno * 10 + *ptr - CHAR_0 : -1;
4747            ptr++;            ptr++;
4748            }            }
4749          namelen = ptr - name;          namelen = ptr - name;
4750    
4751          if ((terminator > 0 && *ptr++ != terminator) || *ptr++ != ')')          if ((terminator > 0 && *ptr++ != terminator) ||
4752                *ptr++ != CHAR_RIGHT_PARENTHESIS)
4753            {            {
4754            ptr--;      /* Error offset */            ptr--;      /* Error offset */
4755            *errorcodeptr = ERR26;            *errorcodeptr = ERR26;
# Line 4388  we set the flag only if there is a liter Line 4771  we set the flag only if there is a liter
4771              *errorcodeptr = ERR58;              *errorcodeptr = ERR58;
4772              goto FAILED;              goto FAILED;
4773              }              }
4774            recno = (refsign == '-')?            recno = (refsign == CHAR_MINUS)?
4775              cd->bracount - recno + 1 : recno +cd->bracount;              cd->bracount - recno + 1 : recno +cd->bracount;
4776            if (recno <= 0 || recno > cd->final_bracount)            if (recno <= 0 || recno > cd->final_bracount)
4777              {              {
# Line 4400  we set the flag only if there is a liter Line 4783  we set the flag only if there is a liter
4783            }            }
4784    
4785          /* Otherwise (did not start with "+" or "-"), start by looking for the          /* Otherwise (did not start with "+" or "-"), start by looking for the
4786          name. */          name. If we find a name, add one to the opcode to change OP_CREF or
4787            OP_RREF into OP_NCREF or OP_NRREF. These behave exactly the same,
4788            except they record that the reference was originally to a name. The
4789            information is used to check duplicate names. */
4790    
4791          slot = cd->name_table;          slot = cd->name_table;
4792          for (i = 0; i < cd->names_found; i++)          for (i = 0; i < cd->names_found; i++)
# Line 4415  we set the flag only if there is a liter Line 4801  we set the flag only if there is a liter
4801            {            {
4802            recno = GET2(slot, 0);            recno = GET2(slot, 0);
4803            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4804              code[1+LINK_SIZE]++;
4805            }            }
4806    
4807          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
4808    
4809          else if ((i = find_parens(ptr, cd, name, namelen,          else if ((i = find_parens(cd, name, namelen,
4810                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0)) > 0)
4811            {            {
4812            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
4813              code[1+LINK_SIZE]++;
4814            }            }
4815    
4816          /* If terminator == 0 it means that the name followed directly after          /* If terminator == 0 it means that the name followed directly after
# Line 4440  we set the flag only if there is a liter Line 4828  we set the flag only if there is a liter
4828          /* Check for (?(R) for recursion. Allow digits after R to specify a          /* Check for (?(R) for recursion. Allow digits after R to specify a
4829          specific group number. */          specific group number. */
4830    
4831          else if (*name == 'R')          else if (*name == CHAR_R)
4832            {            {
4833            recno = 0;            recno = 0;
4834            for (i = 1; i < namelen; i++)            for (i = 1; i < namelen; i++)
# Line 4450  we set the flag only if there is a liter Line 4838  we set the flag only if there is a liter
4838                *errorcodeptr = ERR15;                *errorcodeptr = ERR15;
4839                goto FAILED;                goto FAILED;
4840                }                }
4841              recno = recno * 10 + name[i] - '0';              recno = recno * 10 + name[i] - CHAR_0;
4842              }              }
4843            if (recno == 0) recno = RREF_ANY;            if (recno == 0) recno = RREF_ANY;
4844            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */
# Line 4460  we set the flag only if there is a liter Line 4848  we set the flag only if there is a liter
4848          /* Similarly, check for the (?(DEFINE) "condition", which is always          /* Similarly, check for the (?(DEFINE) "condition", which is always
4849          false. */          false. */
4850    
4851          else if (namelen == 6 && strncmp((char *)name, "DEFINE", 6) == 0)          else if (namelen == 6 && strncmp((char *)name, STRING_DEFINE, 6) == 0)
4852            {            {
4853            code[1+LINK_SIZE] = OP_DEF;            code[1+LINK_SIZE] = OP_DEF;
4854            skipbytes = 1;            skipbytes = 1;
# Line 4485  we set the flag only if there is a liter Line 4873  we set the flag only if there is a liter
4873    
4874    
4875          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4876          case '=':                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
4877          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
4878          ptr++;          ptr++;
4879          break;          break;
4880    
4881    
4882          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4883          case '!':                 /* Negative lookahead */          case CHAR_EXCLAMATION_MARK:            /* Negative lookahead */
4884          ptr++;          ptr++;
4885          if (*ptr == ')')          /* Optimize (?!) */          if (*ptr == CHAR_RIGHT_PARENTHESIS)    /* Optimize (?!) */
4886            {            {
4887            *code++ = OP_FAIL;            *code++ = OP_FAIL;
4888            previous = NULL;            previous = NULL;
# Line 4505  we set the flag only if there is a liter Line 4893  we set the flag only if there is a liter
4893    
4894    
4895          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4896          case '<':                 /* Lookbehind or named define */          case CHAR_LESS_THAN_SIGN:              /* Lookbehind or named define */
4897          switch (ptr[1])          switch (ptr[1])
4898            {            {
4899            case '=':               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
4900            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
4901            ptr += 2;            ptr += 2;
4902            break;            break;
4903    
4904            case '!':               /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
4905            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
4906            ptr += 2;            ptr += 2;
4907            break;            break;
# Line 4528  we set the flag only if there is a liter Line 4916  we set the flag only if there is a liter
4916    
4917    
4918          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4919          case '>':                 /* One-time brackets */          case CHAR_GREATER_THAN_SIGN:           /* One-time brackets */
4920          bravalue = OP_ONCE;          bravalue = OP_ONCE;
4921          ptr++;          ptr++;
4922          break;          break;
4923    
4924    
4925          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4926          case 'C':                 /* Callout - may be followed by digits; */          case CHAR_C:                 /* Callout - may be followed by digits; */
4927          previous_callout = code;  /* Save for later completion */          previous_callout = code;  /* Save for later completion */
4928          after_manual_callout = 1; /* Skip one item before completing */          after_manual_callout = 1; /* Skip one item before completing */
4929          *code++ = OP_CALLOUT;          *code++ = OP_CALLOUT;
4930            {            {
4931            int n = 0;            int n = 0;
4932            while ((digitab[*(++ptr)] & ctype_digit) != 0)            while ((digitab[*(++ptr)] & ctype_digit) != 0)
4933              n = n * 10 + *ptr - '0';              n = n * 10 + *ptr - CHAR_0;
4934            if (*ptr != ')')            if (*ptr != CHAR_RIGHT_PARENTHESIS)
4935              {              {
4936              *errorcodeptr = ERR39;              *errorcodeptr = ERR39;
4937              goto FAILED;              goto FAILED;
# Line 4563  we set the flag only if there is a liter Line 4951  we set the flag only if there is a liter
4951    
4952    
4953          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4954          case 'P':                 /* Python-style named subpattern handling */          case CHAR_P:              /* Python-style named subpattern handling */
4955          if (*(++ptr) == '=' || *ptr == '>')  /* Reference or recursion */          if (*(++ptr) == CHAR_EQUALS_SIGN ||
4956                *ptr == CHAR_GREATER_THAN_SIGN)  /* Reference or recursion */
4957            {            {
4958            is_recurse = *ptr == '>';            is_recurse = *ptr == CHAR_GREATER_THAN_SIGN;
4959            terminator = ')';            terminator = CHAR_RIGHT_PARENTHESIS;
4960            goto NAMED_REF_OR_RECURSE;            goto NAMED_REF_OR_RECURSE;
4961            }            }
4962          else if (*ptr != '<')    /* Test for Python-style definition */          else if (*ptr != CHAR_LESS_THAN_SIGN)  /* Test for Python-style defn */
4963            {            {
4964            *errorcodeptr = ERR41;            *errorcodeptr = ERR41;
4965            goto FAILED;            goto FAILED;
# Line 4580  we set the flag only if there is a liter Line 4969  we set the flag only if there is a liter
4969    
4970          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4971          DEFINE_NAME:    /* Come here from (?< handling */          DEFINE_NAME:    /* Come here from (?< handling */
4972          case '\'':          case CHAR_APOSTROPHE:
4973            {            {
4974            terminator = (*ptr == '<')? '>' : '\'';            terminator = (*ptr == CHAR_LESS_THAN_SIGN)?
4975                CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
4976            name = ++ptr;            name = ++ptr;
4977    
4978            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
# Line 4613  we set the flag only if there is a liter Line 5003  we set the flag only if there is a liter
5003                }                }
5004              }              }
5005    
5006            /* In the real compile, create the entry in the table */            /* In the real compile, create the entry in the table, maintaining
5007              alphabetical order. Duplicate names for different numbers are
5008              permitted only if PCRE_DUPNAMES is set. Duplicate names for the same
5009              number are always OK. (An existing number can be re-used if (?|
5010              appears in the pattern.) In either event, a duplicate name results in
5011              a duplicate entry in the table, even if the number is the same. This
5012              is because the number of names, and hence the table size, is computed
5013              in the pre-compile, and it affects various numbers and pointers which
5014              would all have to be modified, and the compiled code moved down, if
5015              duplicates with the same number were omitted from the table. This
5016              doesn't seem worth the hassle. However, *different* names for the
5017              same number are not permitted. */
5018    
5019            else            else
5020              {              {
5021                BOOL dupname = FALSE;
5022              slot = cd->name_table;              slot = cd->name_table;
5023    
5024              for (i = 0; i < cd->names_found; i++)              for (i = 0; i < cd->names_found; i++)
5025                {                {
5026                int crc = memcmp(name, slot+2, namelen);                int crc = memcmp(name, slot+2, namelen);
# Line 4625  we set the flag only if there is a liter Line 5028  we set the flag only if there is a liter
5028                  {                  {
5029                  if (slot[2+namelen] == 0)                  if (slot[2+namelen] == 0)
5030                    {                    {
5031                    if ((options & PCRE_DUPNAMES) == 0)                    if (GET2(slot, 0) != cd->bracount + 1 &&
5032                          (options & PCRE_DUPNAMES) == 0)
5033                      {                      {
5034                      *errorcodeptr = ERR43;                      *errorcodeptr = ERR43;
5035                      goto FAILED;                      goto FAILED;
5036                      }                      }
5037                      else dupname = TRUE;
5038                    }                    }
5039                  else crc = -1;      /* Current name is substring */                  else crc = -1;      /* Current name is a substring */
5040                  }                  }
5041    
5042                  /* Make space in the table and break the loop for an earlier
5043                  name. For a duplicate or later name, carry on. We do this for
5044                  duplicates so that in the simple case (when ?(| is not used) they
5045                  are in order of their numbers. */
5046    
5047                if (crc < 0)                if (crc < 0)
5048                  {                  {
5049                  memmove(slot + cd->name_entry_size, slot,                  memmove(slot + cd->name_entry_size, slot,
5050                    (cd->names_found - i) * cd->name_entry_size);                    (cd->names_found - i) * cd->name_entry_size);
5051                  break;                  break;
5052                  }                  }
5053    
5054                  /* Continue the loop for a later or duplicate name */
5055    
5056                slot += cd->name_entry_size;                slot += cd->name_entry_size;
5057                }                }
5058    
5059                /* For non-duplicate names, check for a duplicate number before
5060                adding the new name. */
5061    
5062                if (!dupname)
5063                  {
5064                  uschar *cslot = cd->name_table;
5065                  for (i = 0; i < cd->names_found; i++)
5066                    {
5067                    if (cslot != slot)
5068                      {
5069                      if (GET2(cslot, 0) == cd->bracount + 1)
5070                        {
5071                        *errorcodeptr = ERR65;
5072                        goto FAILED;
5073                        }
5074                      }
5075                    else i--;
5076                    cslot += cd->name_entry_size;
5077                    }
5078                  }
5079    
5080              PUT2(slot, 0, cd->bracount + 1);              PUT2(slot, 0, cd->bracount + 1);
5081              memcpy(slot + 2, name, namelen);              memcpy(slot + 2, name, namelen);
5082              slot[2+namelen] = 0;              slot[2+namelen] = 0;
5083              }              }
5084            }            }
5085    
5086          /* In both cases, count the number of names we've encountered. */          /* In both pre-compile and compile, count the number of names we've
5087            encountered. */
5088    
         ptr++;                    /* Move past > or ' */  
5089          cd->names_found++;          cd->names_found++;
5090            ptr++;                    /* Move past > or ' */
5091          goto NUMBERED_GROUP;          goto NUMBERED_GROUP;
5092    
5093    
5094          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5095          case '&':                 /* Perl recursion/subroutine syntax */          case CHAR_AMPERSAND:            /* Perl recursion/subroutine syntax */
5096          terminator = ')';          terminator = CHAR_RIGHT_PARENTHESIS;
5097          is_recurse = TRUE;          is_recurse = TRUE;
5098          /* Fall through */          /* Fall through */
5099    
# Line 4716  we set the flag only if there is a liter Line 5152  we set the flag only if there is a liter
5152              recno = GET2(slot, 0);              recno = GET2(slot, 0);
5153              }              }
5154            else if ((recno =                /* Forward back reference */            else if ((recno =                /* Forward back reference */
5155                      find_parens(ptr, cd, name, namelen,                      find_parens(cd, name, namelen,
5156                        (options & PCRE_EXTENDED) != 0)) <= 0)                        (options & PCRE_EXTENDED) != 0)) <= 0)
5157              {              {
5158              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
# Line 4732  we set the flag only if there is a liter Line 5168  we set the flag only if there is a liter
5168    
5169    
5170          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5171          case 'R':                 /* Recursion */          case CHAR_R:              /* Recursion */
5172          ptr++;                    /* Same as (?0)      */          ptr++;                    /* Same as (?0)      */
5173          /* Fall through */          /* Fall through */
5174    
5175    
5176          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5177          case '-': case '+':          case CHAR_MINUS: case CHAR_PLUS:  /* Recursion or subroutine */
5178          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:
5179          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:
5180            {            {
5181            const uschar *called;            const uschar *called;
5182            terminator = ')';            terminator = CHAR_RIGHT_PARENTHESIS;
5183    
5184            /* Come here from the \g<...> and \g'...' code (Oniguruma            /* Come here from the \g<...> and \g'...' code (Oniguruma
5185            compatibility). However, the syntax has been checked to ensure that            compatibility). However, the syntax has been checked to ensure that
# Line 4753  we set the flag only if there is a liter Line 5189  we set the flag only if there is a liter
5189    
5190            HANDLE_NUMERICAL_RECURSION:            HANDLE_NUMERICAL_RECURSION:
5191    
5192            if ((refsign = *ptr) == '+')            if ((refsign = *ptr) == CHAR_PLUS)
5193              {              {
5194              ptr++;              ptr++;
5195              if ((digitab[*ptr] & ctype_digit) == 0)              if ((digitab[*ptr] & ctype_digit) == 0)
# Line 4762  we set the flag only if there is a liter Line 5198  we set the flag only if there is a liter
5198                goto FAILED;                goto FAILED;
5199                }                }
5200              }              }
5201            else if (refsign == '-')            else if (refsign == CHAR_MINUS)
5202              {              {
5203              if ((digitab[ptr[1]] & ctype_digit) == 0)              if ((digitab[ptr[1]] & ctype_digit) == 0)
5204                goto OTHER_CHAR_AFTER_QUERY;                goto OTHER_CHAR_AFTER_QUERY;
# Line 4771  we set the flag only if there is a liter Line 5207  we set the flag only if there is a liter
5207    
5208            recno = 0;            recno = 0;
5209            while((digitab[*ptr] & ctype_digit) != 0)            while((digitab[*ptr] & ctype_digit) != 0)
5210              recno = recno * 10 + *ptr++ - '0';              recno = recno * 10 + *ptr++ - CHAR_0;
5211    
5212            if (*ptr != terminator)            if (*ptr != terminator)
5213              {              {
# Line 4779  we set the flag only if there is a liter Line 5215  we set the flag only if there is a liter
5215              goto FAILED;              goto FAILED;
5216              }              }
5217    
5218            if (refsign == '-')            if (refsign == CHAR_MINUS)
5219              {              {
5220              if (recno == 0)              if (recno == 0)
5221                {                {
# Line 4793  we set the flag only if there is a liter Line 5229  we set the flag only if there is a liter
5229                goto FAILED;                goto FAILED;
5230                }                }
5231              }              }
5232            else if (refsign == '+')            else if (refsign == CHAR_PLUS)
5233              {              {
5234              if (recno == 0)              if (recno == 0)
5235                {                {
# Line 4820  we set the flag only if there is a liter Line 5256  we set the flag only if there is a liter
5256            if (lengthptr == NULL)            if (lengthptr == NULL)
5257              {              {
5258              *code = OP_END;              *code = OP_END;
5259              if (recno != 0) called = find_bracket(cd->start_code, utf8, recno);              if (recno != 0)
5260                  called = _pcre_find_bracket(cd->start_code, utf8, recno);
5261    
5262              /* Forward reference */              /* Forward reference */
5263    
5264              if (called == NULL)              if (called == NULL)
5265                {                {
5266                if (find_parens(ptr, cd, NULL, recno,                if (find_parens(cd, NULL, recno,
5267                      (options & PCRE_EXTENDED) != 0) < 0)                      (options & PCRE_EXTENDED) != 0) < 0)
5268                  {                  {
5269                  *errorcodeptr = ERR15;                  *errorcodeptr = ERR15;
5270                  goto FAILED;                  goto FAILED;
5271                  }                  }
5272    
5273                  /* Fudge the value of "called" so that when it is inserted as an
5274                  offset below, what it actually inserted is the reference number
5275                  of the group. */
5276    
5277                called = cd->start_code + recno;                called = cd->start_code + recno;
5278                PUTINC(cd->hwm, 0, code + 2 + LINK_SIZE - cd->start_code);                PUTINC(cd->hwm, 0, code + 2 + LINK_SIZE - cd->start_code);
5279                }                }
# Line 4841  we set the flag only if there is a liter Line 5283  we set the flag only if there is a liter
5283              recursion that could loop for ever, and diagnose that case. */              recursion that could loop for ever, and diagnose that case. */
5284    
5285              else if (GET(called, 1) == 0 &&              else if (GET(called, 1) == 0 &&
5286                       could_be_empty(called, code, bcptr, utf8))                       could_be_empty(called, code, bcptr, utf8, cd))
5287                {                {
5288                *errorcodeptr = ERR40;                *errorcodeptr = ERR40;
5289                goto FAILED;                goto FAILED;
# Line 4879  we set the flag only if there is a liter Line 5321  we set the flag only if there is a liter
5321          set = unset = 0;          set = unset = 0;
5322          optset = &set;          optset = &set;
5323    
5324          while (*ptr != ')' && *ptr != ':')          while (*ptr != CHAR_RIGHT_PARENTHESIS && *ptr != CHAR_COLON)
5325            {            {
5326            switch (*ptr++)            switch (*ptr++)
5327              {              {
5328              case '-': optset = &unset; break;              case CHAR_MINUS: optset = &unset; break;
5329    
5330              case 'J':    /* Record that it changed in the external options */              case CHAR_J:    /* Record that it changed in the external options */
5331              *optset |= PCRE_DUPNAMES;              *optset |= PCRE_DUPNAMES;
5332              cd->external_flags |= PCRE_JCHANGED;              cd->external_flags |= PCRE_JCHANGED;
5333              break;              break;
5334    
5335              case 'i': *optset |= PCRE_CASELESS; break;              case CHAR_i: *optset |= PCRE_CASELESS; break;
5336              case 'm': *optset |= PCRE_MULTILINE; break;              case CHAR_m: *optset |= PCRE_MULTILINE; break;
5337              case 's': *optset |= PCRE_DOTALL; break;              case CHAR_s: *optset |= PCRE_DOTALL; break;
5338              case 'x': *optset |= PCRE_EXTENDED; break;              case CHAR_x: *optset |= PCRE_EXTENDED; break;
5339              case 'U': *optset |= PCRE_UNGREEDY; break;              case CHAR_U: *optset |= PCRE_UNGREEDY; break;
5340              case 'X': *optset |= PCRE_EXTRA; break;              case CHAR_X: *optset |= PCRE_EXTRA; break;
5341    
5342              default:  *errorcodeptr = ERR12;              default:  *errorcodeptr = ERR12;
5343                        ptr--;    /* Correct the offset */                        ptr--;    /* Correct the offset */
# Line 4926  we set the flag only if there is a liter Line 5368  we set the flag only if there is a liter
5368          both phases.          both phases.
5369    
5370          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
5371          options if this setting actually changes any of them, and reset the          options if this setting actually changes any of them, and reset the
5372          greedy defaults and the case value for firstbyte and reqbyte. */          greedy defaults and the case value for firstbyte and reqbyte. */
5373    
5374          if (*ptr == ')')          if (*ptr == CHAR_RIGHT_PARENTHESIS)
5375            {            {
5376            if (code == cd->start_code + 1 + LINK_SIZE &&            if (code == cd->start_code + 1 + LINK_SIZE &&
5377                 (lengthptr == NULL || *lengthptr == 2 + 2*LINK_SIZE))                 (lengthptr == NULL || *lengthptr == 2 + 2*LINK_SIZE))
5378              {              {
5379              cd->external_options = newoptions;              cd->external_options = newoptions;
5380              }              }
5381           else            else
5382              {              {
5383              if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))              if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))
5384                {                {
# Line 4950  we set the flag only if there is a liter Line 5392  we set the flag only if there is a liter
5392    
5393            /* Change options at this level, and pass them back for use            /* Change options at this level, and pass them back for use
5394            in subsequent branches. When not at the start of the pattern, this            in subsequent branches. When not at the start of the pattern, this
5395            information is also necessary so that a resetting item can be            information is also necessary so that a resetting item can be
5396            compiled at the end of a group (if we are in a group). */            compiled at the end of a group (if we are in a group). */
5397    
5398            *optionsptr = options = newoptions;            *optionsptr = options = newoptions;
# Line 4968  we set the flag only if there is a liter Line 5410  we set the flag only if there is a liter
5410          }     /* End of switch for character following (? */          }     /* End of switch for character following (? */
5411        }       /* End of (? handling */        }       /* End of (? handling */
5412    
5413      /* Opening parenthesis not followed by '?'. If PCRE_NO_AUTO_CAPTURE is set,      /* Opening parenthesis not followed by '*' or '?'. If PCRE_NO_AUTO_CAPTURE
5414      all unadorned brackets become non-capturing and behave like (?:...)      is set, all unadorned brackets become non-capturing and behave like (?:...)
5415      brackets. */      brackets. */
5416    
5417      else if ((options & PCRE_NO_AUTO_CAPTURE) != 0)      else if ((options & PCRE_NO_AUTO_CAPTURE) != 0)
# Line 5069  we set the flag only if there is a liter Line 5511  we set the flag only if there is a liter
5511    
5512      /* Error if hit end of pattern */      /* Error if hit end of pattern */
5513    
5514      if (*ptr != ')')      if (*ptr != CHAR_RIGHT_PARENTHESIS)
5515        {        {
5516        *errorcodeptr = ERR14;        *errorcodeptr = ERR14;
5517        goto FAILED;        goto FAILED;
# Line 5167  we set the flag only if there is a liter Line 5609  we set the flag only if there is a liter
5609      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
5610      have to change if any new ones are ever created. */      have to change if any new ones are ever created. */
5611    
5612      case '\\':      case CHAR_BACKSLASH:
5613      tempptr = ptr;      tempptr = ptr;
5614      c = check_escape(&ptr, errorcodeptr, cd->bracount, options, FALSE);      c = check_escape(&ptr, errorcodeptr, cd->bracount, options, FALSE);
5615      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
# Line 5176  we set the flag only if there is a liter Line 5618  we set the flag only if there is a liter
5618        {        {
5619        if (-c == ESC_Q)            /* Handle start of quoted string */        if (-c == ESC_Q)            /* Handle start of quoted string */
5620          {          {
5621          if (ptr[1] == '\\' && ptr[2] == 'E') ptr += 2; /* avoid empty string */          if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
5622            else inescq = TRUE;            ptr += 2;               /* avoid empty string */
5623                else inescq = TRUE;
5624          continue;          continue;
5625          }          }
5626    
# Line 5205  we set the flag only if there is a liter Line 5648  we set the flag only if there is a liter
5648          {          {
5649          const uschar *p;          const uschar *p;
5650          save_hwm = cd->hwm;   /* Normally this is set when '(' is read */          save_hwm = cd->hwm;   /* Normally this is set when '(' is read */
5651          terminator = (*(++ptr) == '<')? '>' : '\'';          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
5652              CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
5653    
5654          /* These two statements stop the compiler for warning about possibly          /* These two statements stop the compiler for warning about possibly
5655          unset variables caused by the jump to HANDLE_NUMERICAL_RECURSION. In          unset variables caused by the jump to HANDLE_NUMERICAL_RECURSION. In
# Line 5217  we set the flag only if there is a liter Line 5661  we set the flag only if there is a liter
5661    
5662          /* Test for a name */          /* Test for a name */
5663    
5664          if (ptr[1] != '+' && ptr[1] != '-')          if (ptr[1] != CHAR_PLUS && ptr[1] != CHAR_MINUS)
5665            {            {
5666            BOOL isnumber = TRUE;            BOOL isnumber = TRUE;
5667            for (p = ptr + 1; *p != 0 && *p != terminator; p++)            for (p = ptr + 1; *p != 0 && *p != terminator; p++)
# Line 5255  we set the flag only if there is a liter Line 5699  we set the flag only if there is a liter
5699        /* \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).
5700        We also support \k{name} (.NET syntax) */        We also support \k{name} (.NET syntax) */
5701    
5702        if (-c == ESC_k && (ptr[1] == '<' || ptr[1] == '\'' || ptr[1] == '{'))        if (-c == ESC_k && (ptr[1] == CHAR_LESS_THAN_SIGN ||
5703              ptr[1] == CHAR_APOSTROPHE || ptr[1] == CHAR_LEFT_CURLY_BRACKET))
5704          {          {
5705          is_recurse = FALSE;          is_recurse = FALSE;
5706          terminator = (*(++ptr) == '<')? '>' : (*ptr == '\'')? '\'' : '}';          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
5707              CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?
5708              CHAR_APOSTROPHE : CHAR_RIGHT_CURLY_BRACKET;
5709          goto NAMED_REF_OR_RECURSE;          goto NAMED_REF_OR_RECURSE;
5710          }          }
5711    
# Line 5268  we set the flag only if there is a liter Line 5715  we set the flag only if there is a liter
5715    
5716        if (-c >= ESC_REF)        if (-c >= ESC_REF)
5717          {          {
5718            open_capitem *oc;
5719          recno = -c - ESC_REF;          recno = -c - ESC_REF;
5720    
5721          HANDLE_REFERENCE:    /* Come here from named backref handling */          HANDLE_REFERENCE:    /* Come here from named backref handling */
# Line 5277  we set the flag only if there is a liter Line 5725  we set the flag only if there is a liter
5725          PUT2INC(code, 0, recno);          PUT2INC(code, 0, recno);
5726          cd->backref_map |= (recno < 32)? (1 << recno) : 1;          cd->backref_map |= (recno < 32)? (1 << recno) : 1;
5727          if (recno > cd->top_backref) cd->top_backref = recno;          if (recno > cd->top_backref) cd->top_backref = recno;
5728    
5729            /* Check to see if this back reference is recursive, that it, it
5730            is inside the group that it references. A flag is set so that the
5731            group can be made atomic. */
5732    
5733            for (oc = cd->open_caps; oc != NULL; oc = oc->next)
5734              {
5735              if (oc->number == recno)
5736                {
5737                oc->flag = TRUE;
5738                break;
5739                }
5740              }
5741          }          }
5742    
5743        /* So are Unicode property matches, if supported. */        /* So are Unicode property matches, if supported. */
# Line 5361  we set the flag only if there is a liter Line 5822  we set the flag only if there is a liter
5822    
5823      /* Remember if \r or \n were seen */      /* Remember if \r or \n were seen */
5824    
5825      if (mcbuffer[0] == '\r' || mcbuffer[0] == '\n')      if (mcbuffer[0] == CHAR_CR || mcbuffer[0] == CHAR_NL)
5826        cd->external_flags |= PCRE_HASCRORLF;        cd->external_flags |= PCRE_HASCRORLF;
5827    
5828      /* Set the first and required bytes appropriately. If no previous first      /* Set the first and required bytes appropriately. If no previous first
# Line 5459  uschar *code = *codeptr; Line 5920  uschar *code = *codeptr;
5920  uschar *last_branch = code;  uschar *last_branch = code;
5921  uschar *start_bracket = code;  uschar *start_bracket = code;
5922  uschar *reverse_count = NULL;  uschar *reverse_count = NULL;
5923    open_capitem capitem;
5924    int capnumber = 0;
5925  int firstbyte, reqbyte;  int firstbyte, reqbyte;
5926  int branchfirstbyte, branchreqbyte;  int branchfirstbyte, branchreqbyte;
5927  int length;  int length;
5928  int orig_bracount;  int orig_bracount;
5929  int max_bracount;  int max_bracount;
5930    int old_external_options = cd->external_options;
5931  branch_chain bc;  branch_chain bc;
5932    
5933  bc.outer = bcptr;  bc.outer = bcptr;
5934  bc.current = code;  bc.current_branch = code;
5935    
5936  firstbyte = reqbyte = REQ_UNSET;  firstbyte = reqbyte = REQ_UNSET;
5937    
# Line 5485  the code that abstracts option settings Line 5949  the code that abstracts option settings
5949  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
5950  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. */
5951    
5952    /* If this is a capturing subpattern, add to the chain of open capturing items
5953    so that we can detect them if (*ACCEPT) is encountered. This is also used to
5954    detect groups that contain recursive back references to themselves. */
5955    
5956    if (*code == OP_CBRA)
5957      {
5958      capnumber = GET2(code, 1 + LINK_SIZE);
5959      capitem.number = capnumber;
5960      capitem.next = cd->open_caps;
5961      capitem.flag = FALSE;
5962      cd->open_caps = &capitem;
5963      }
5964    
5965  /* Offset is set zero to mark that this bracket is still open */  /* Offset is set zero to mark that this bracket is still open */
5966    
5967  PUT(code, 1, 0);  PUT(code, 1, 0);
# Line 5529  for (;;) Line 6006  for (;;)
6006      return FALSE;      return FALSE;
6007      }      }
6008    
6009      /* If the external options have changed during this branch, it means that we
6010      are at the top level, and a leading option setting has been encountered. We
6011      need to re-set the original option values to take account of this so that,
6012      during the pre-compile phase, we know to allow for a re-set at the start of
6013      subsequent branches. */
6014    
6015      if (old_external_options != cd->external_options)
6016        oldims = cd->external_options & PCRE_IMS;
6017    
6018    /* Keep the highest bracket count in case (?| was used and some branch    /* Keep the highest bracket count in case (?| was used and some branch
6019    has fewer than the rest. */    has fewer than the rest. */
6020    
# Line 5579  for (;;) Line 6065  for (;;)
6065    
6066      /* If lookbehind, check that this branch matches a fixed-length string, and      /* If lookbehind, check that this branch matches a fixed-length string, and
6067      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
6068      branch with OP_END. */      branch with OP_END. If the branch contains OP_RECURSE, the result is -3
6069        because there may be forward references that we can't check here. Set a
6070        flag to cause another lookbehind check at the end. Why not do it all at the
6071        end? Because common, erroneous checks are picked up here and the offset of
6072        the problem can be shown. */
6073    
6074      if (lookbehind)      if (lookbehind)
6075        {        {
6076        int fixed_length;        int fixed_length;
6077        *code = OP_END;        *code = OP_END;
6078        fixed_length = find_fixedlength(last_branch, options);        fixed_length = find_fixedlength(last_branch, options, FALSE, cd);
6079        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
6080        if (fixed_length < 0)        if (fixed_length == -3)
6081            {
6082            cd->check_lookbehind = TRUE;
6083            }
6084          else if (fixed_length < 0)
6085          {          {
6086          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;
6087          *ptrptr = ptr;          *ptrptr = ptr;
6088          return FALSE;          return FALSE;
6089          }          }
6090        PUT(reverse_count, 0, fixed_length);        else { PUT(reverse_count, 0, fixed_length); }
6091        }        }
6092      }      }
6093    
# Line 5606  for (;;) Line 6100  for (;;)
6100    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.
6101    Return leaving the pointer at the terminating char. */    Return leaving the pointer at the terminating char. */
6102    
6103    if (*ptr != '|')    if (*ptr != CHAR_VERTICAL_LINE)
6104      {      {
6105      if (lengthptr == NULL)      if (lengthptr == NULL)
6106        {        {
# Line 5627  for (;;) Line 6121  for (;;)
6121      PUT(code, 1, code - start_bracket);      PUT(code, 1, code - start_bracket);
6122      code += 1 + LINK_SIZE;      code += 1 + LINK_SIZE;
6123    
6124      /* Resetting option if needed */      /* If it was a capturing subpattern, check to see if it contained any
6125        recursive back references. If so, we must wrap it in atomic brackets.
6126        In any event, remove the block from the chain. */
6127    
6128      if ((options & PCRE_IMS) != oldims && *ptr == ')')      if (capnumber > 0)
6129          {
6130          if (cd->open_caps->flag)
6131            {
6132            memmove(start_bracket + 1 + LINK_SIZE, start_bracket,
6133              code - start_bracket);
6134            *start_bracket = OP_ONCE;
6135            code += 1 + LINK_SIZE;
6136            PUT(start_bracket, 1, code - start_bracket);
6137            *code = OP_KET;
6138            PUT(code, 1, code - start_bracket);
6139            code += 1 + LINK_SIZE;
6140            length += 2 + 2*LINK_SIZE;
6141            }
6142          cd->open_caps = cd->open_caps->next;
6143          }
6144    
6145        /* Reset options if needed. */
6146    
6147        if ((options & PCRE_IMS) != oldims && *ptr == CHAR_RIGHT_PARENTHESIS)
6148        {        {
6149        *code++ = OP_OPT;        *code++ = OP_OPT;
6150        *code++ = oldims;        *code++ = oldims;
# Line 5676  for (;;) Line 6191  for (;;)
6191      {      {
6192      *code = OP_ALT;      *code = OP_ALT;
6193      PUT(code, 1, code - last_branch);      PUT(code, 1, code - last_branch);
6194      bc.current = last_branch = code;      bc.current_branch = last_branch = code;
6195      code += 1 + LINK_SIZE;      code += 1 + LINK_SIZE;
6196      }      }
6197    
# Line 5811  do { Line 6326  do {
6326       NULL, 0, FALSE);       NULL, 0, FALSE);
6327     register int op = *scode;     register int op = *scode;
6328    
6329       /* If we are at the start of a conditional assertion group, *both* the
6330       conditional assertion *and* what follows the condition must satisfy the test
6331       for start of line. Other kinds of condition fail. Note that there may be an
6332       auto-callout at the start of a condition. */
6333    
6334       if (op == OP_COND)
6335         {
6336         scode += 1 + LINK_SIZE;
6337         if (*scode == OP_CALLOUT) scode += _pcre_OP_lengths[OP_CALLOUT];
6338         switch (*scode)
6339           {
6340           case OP_CREF:
6341           case OP_NCREF:
6342           case OP_RREF:
6343           case OP_NRREF:
6344           case OP_DEF:
6345           return FALSE;
6346    
6347           default:     /* Assertion */
6348           if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
6349           do scode += GET(scode, 1); while (*scode == OP_ALT);
6350           scode += 1 + LINK_SIZE;
6351           break;
6352           }
6353         scode = first_significant_code(scode, NULL, 0, FALSE);
6354         op = *scode;
6355         }
6356    
6357     /* Non-capturing brackets */     /* Non-capturing brackets */
6358    
6359     if (op == OP_BRA)     if (op == OP_BRA)
# Line 5829  do { Line 6372  do {
6372    
6373     /* Other brackets */     /* Other brackets */
6374    
6375     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_COND)     else if (op == OP_ASSERT || op == OP_ONCE)
6376       { if (!is_startline(scode, bracket_map, backref_map)) return FALSE; }       {
6377         if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
6378         }
6379    
6380     /* .* 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
6381     may be referenced. */     may be referenced. */
# Line 5964  int length = 1; /* For final END opcode Line 6509  int length = 1; /* For final END opcode
6509  int firstbyte, reqbyte, newline;  int firstbyte, reqbyte, newline;
6510  int errorcode = 0;  int errorcode = 0;
6511  int skipatstart = 0;  int skipatstart = 0;
6512  #ifdef SUPPORT_UTF8  BOOL utf8 = (options & PCRE_UTF8) != 0;
 BOOL utf8;  
 #endif  
6513  size_t size;  size_t size;
6514  uschar *code;  uschar *code;
6515  const uschar *codestart;  const uschar *codestart;
# Line 6009  if (erroroffset == NULL) Line 6552  if (erroroffset == NULL)
6552    
6553  *erroroffset = 0;  *erroroffset = 0;
6554    
 /* 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;  
   }  
   
6555  /* Set up pointers to the individual character tables */  /* Set up pointers to the individual character tables */
6556    
6557  if (tables == NULL) tables = _pcre_default_tables;  if (tables == NULL) tables = _pcre_default_tables;
# Line 6041  cd->fcc = tables + fcc_offset; Line 6560  cd->fcc = tables + fcc_offset;
6560  cd->cbits = tables + cbits_offset;  cd->cbits = tables + cbits_offset;
6561  cd->ctypes = tables + ctypes_offset;  cd->ctypes = tables + ctypes_offset;
6562    
6563    /* Check that all undefined public option bits are zero */
6564    
6565    if ((options & ~PUBLIC_COMPILE_OPTIONS) != 0)
6566      {
6567      errorcode = ERR17;
6568      goto PCRE_EARLY_ERROR_RETURN;
6569      }
6570    
6571  /* 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
6572  the offset for later. */  the offset for later. */
6573    
6574  while (ptr[skipatstart] == '(' && ptr[skipatstart+1] == '*')  while (ptr[skipatstart] == CHAR_LEFT_PARENTHESIS &&
6575           ptr[skipatstart+1] == CHAR_ASTERISK)
6576    {    {
6577    int newnl = 0;    int newnl = 0;
6578    int newbsr = 0;    int newbsr = 0;
6579    
6580    if (strncmp((char *)(ptr+skipatstart+2), "CR)", 3) == 0)    if (strncmp((char *)(ptr+skipatstart+2), STRING_UTF8_RIGHTPAR, 5) == 0)
6581        { skipatstart += 7; options |= PCRE_UTF8; continue; }
6582    
6583      if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)
6584      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }
6585    else if (strncmp((char *)(ptr+skipatstart+2), "LF)", 3)  == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_LF_RIGHTPAR, 3)  == 0)
6586      { skipatstart += 5; newnl = PCRE_NEWLINE_LF; }      { skipatstart += 5; newnl = PCRE_NEWLINE_LF; }
6587    else if (strncmp((char *)(ptr+skipatstart+2), "CRLF)", 5)  == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_CRLF_RIGHTPAR, 5)  == 0)
6588      { skipatstart += 7; newnl = PCRE_NEWLINE_CR + PCRE_NEWLINE_LF; }      { skipatstart += 7; newnl = PCRE_NEWLINE_CR + PCRE_NEWLINE_LF; }
6589    else if (strncmp((char *)(ptr+skipatstart+2), "ANY)", 4) == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_ANY_RIGHTPAR, 4) == 0)
6590      { skipatstart += 6; newnl = PCRE_NEWLINE_ANY; }      { skipatstart += 6; newnl = PCRE_NEWLINE_ANY; }
6591    else if (strncmp((char *)(ptr+skipatstart+2), "ANYCRLF)", 8)  == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_ANYCRLF_RIGHTPAR, 8) == 0)
6592      { skipatstart += 10; newnl = PCRE_NEWLINE_ANYCRLF; }      { skipatstart += 10; newnl = PCRE_NEWLINE_ANYCRLF; }
6593    
6594    else if (strncmp((char *)(ptr+skipatstart+2), "BSR_ANYCRLF)", 12) == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_BSR_ANYCRLF_RIGHTPAR, 12) == 0)
6595      { skipatstart += 14; newbsr = PCRE_BSR_ANYCRLF; }      { skipatstart += 14; newbsr = PCRE_BSR_ANYCRLF; }
6596    else if (strncmp((char *)(ptr+skipatstart+2), "BSR_UNICODE)", 12) == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_BSR_UNICODE_RIGHTPAR, 12) == 0)
6597      { skipatstart += 14; newbsr = PCRE_BSR_UNICODE; }      { skipatstart += 14; newbsr = PCRE_BSR_UNICODE; }
6598    
6599    if (newnl != 0)    if (newnl != 0)
# Line 6072  while (ptr[skipatstart] == '(' && ptr[sk Line 6603  while (ptr[skipatstart] == '(' && ptr[sk
6603    else break;    else break;
6604    }    }
6605    
6606    /* Can't support UTF8 unless PCRE has been compiled to include the code. */
6607    
6608    #ifdef SUPPORT_UTF8
6609    if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&
6610         (*erroroffset = _pcre_valid_utf8((USPTR)pattern, -1)) >= 0)
6611      {
6612      errorcode = ERR44;
6613      goto PCRE_EARLY_ERROR_RETURN2;
6614      }
6615    #else
6616    if (utf8)
6617      {
6618      errorcode = ERR32;
6619      goto PCRE_EARLY_ERROR_RETURN;
6620      }
6621    #endif
6622    
6623  /* Check validity of \R options. */  /* Check validity of \R options. */
6624    
6625  switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))  switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))
# Line 6090  current code allows for fixed one- or tw Line 6638  current code allows for fixed one- or tw
6638  switch (options & PCRE_NEWLINE_BITS)  switch (options & PCRE_NEWLINE_BITS)
6639    {    {
6640    case 0: newline = NEWLINE; break;   /* Build-time default */    case 0: newline = NEWLINE; break;   /* Build-time default */
6641    case PCRE_NEWLINE_CR: newline = '\r'; break;    case PCRE_NEWLINE_CR: newline = CHAR_CR; break;
6642    case PCRE_NEWLINE_LF: newline = '\n'; break;    case PCRE_NEWLINE_LF: newline = CHAR_NL; break;
6643    case PCRE_NEWLINE_CR+    case PCRE_NEWLINE_CR+
6644         PCRE_NEWLINE_LF: newline = ('\r' << 8) | '\n'; break;         PCRE_NEWLINE_LF: newline = (CHAR_CR << 8) | CHAR_NL; break;
6645    case PCRE_NEWLINE_ANY: newline = -1; break;    case PCRE_NEWLINE_ANY: newline = -1; break;
6646    case PCRE_NEWLINE_ANYCRLF: newline = -2; break;    case PCRE_NEWLINE_ANYCRLF: newline = -2; break;
6647    default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;    default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;
# Line 6154  cd->end_pattern = (const uschar *)(patte Line 6702  cd->end_pattern = (const uschar *)(patte
6702  cd->req_varyopt = 0;  cd->req_varyopt = 0;
6703  cd->external_options = options;  cd->external_options = options;
6704  cd->external_flags = 0;  cd->external_flags = 0;
6705    cd->open_caps = NULL;
6706    
6707  /* 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
6708  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 6228  cd->start_code = codestart; Line 6777  cd->start_code = codestart;
6777  cd->hwm = cworkspace;  cd->hwm = cworkspace;
6778  cd->req_varyopt = 0;  cd->req_varyopt = 0;
6779  cd->had_accept = FALSE;  cd->had_accept = FALSE;
6780    cd->check_lookbehind = FALSE;
6781    cd->open_caps = NULL;
6782    
6783  /* Set up a starting, non-extracting bracket, then compile the expression. On  /* Set up a starting, non-extracting bracket, then compile the expression. On
6784  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 6253  if debugging, leave the test till after Line 6804  if debugging, leave the test till after
6804    
6805  *code++ = OP_END;  *code++ = OP_END;
6806    
6807  #ifndef DEBUG  #ifndef PCRE_DEBUG
6808  if (code - codestart > length) errorcode = ERR23;  if (code - codestart > length) errorcode = ERR23;
6809  #endif  #endif
6810    
# Line 6266  while (errorcode == 0 && cd->hwm > cwork Line 6817  while (errorcode == 0 && cd->hwm > cwork
6817    cd->hwm -= LINK_SIZE;    cd->hwm -= LINK_SIZE;
6818    offset = GET(cd->hwm, 0);    offset = GET(cd->hwm, 0);
6819    recno = GET(codestart, offset);    recno = GET(codestart, offset);
6820    groupptr = find_bracket(codestart, (re->options & PCRE_UTF8) != 0, recno);    groupptr = _pcre_find_bracket(codestart, utf8, recno);
6821    if (groupptr == NULL) errorcode = ERR53;    if (groupptr == NULL) errorcode = ERR53;
6822      else PUT(((uschar *)codestart), offset, groupptr - codestart);      else PUT(((uschar *)codestart), offset, groupptr - codestart);
6823    }    }
# Line 6276  subpattern. */ Line 6827  subpattern. */
6827    
6828  if (errorcode == 0 && re->top_backref > re->top_bracket) errorcode = ERR15;  if (errorcode == 0 && re->top_backref > re->top_bracket) errorcode = ERR15;
6829    
6830    /* If there were any lookbehind assertions that contained OP_RECURSE
6831    (recursions or subroutine calls), a flag is set for them to be checked here,
6832    because they may contain forward references. Actual recursions can't be fixed
6833    length, but subroutine calls can. It is done like this so that those without
6834    OP_RECURSE that are not fixed length get a diagnosic with a useful offset. The
6835    exceptional ones forgo this. We scan the pattern to check that they are fixed
6836    length, and set their lengths. */
6837    
6838    if (cd->check_lookbehind)
6839      {
6840      uschar *cc = (uschar *)codestart;
6841    
6842      /* Loop, searching for OP_REVERSE items, and process those that do not have
6843      their length set. (Actually, it will also re-process any that have a length
6844      of zero, but that is a pathological case, and it does no harm.) When we find
6845      one, we temporarily terminate the branch it is in while we scan it. */
6846    
6847      for (cc = (uschar *)_pcre_find_bracket(codestart, utf8, -1);
6848           cc != NULL;
6849           cc = (uschar *)_pcre_find_bracket(cc, utf8, -1))
6850        {
6851        if (GET(cc, 1) == 0)
6852          {
6853          int fixed_length;
6854          uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);
6855          int end_op = *be;
6856          *be = OP_END;
6857          fixed_length = find_fixedlength(cc, re->options, TRUE, cd);
6858          *be = end_op;
6859          DPRINTF(("fixed length = %d\n", fixed_length));
6860          if (fixed_length < 0)
6861            {
6862            errorcode = (fixed_length == -2)? ERR36 : ERR25;
6863            break;
6864            }
6865          PUT(cc, 1, fixed_length);
6866          }
6867        cc += 1 + LINK_SIZE;
6868        }
6869      }
6870    
6871  /* Failed to compile, or error while post-processing */  /* Failed to compile, or error while post-processing */
6872    
6873  if (errorcode != 0)  if (errorcode != 0)
# Line 6336  if (reqbyte >= 0 && Line 6928  if (reqbyte >= 0 &&
6928  /* 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
6929  case when building a production library. */  case when building a production library. */
6930    
6931  #ifdef DEBUG  #ifdef PCRE_DEBUG
   
6932  printf("Length = %d top_bracket = %d top_backref = %d\n",  printf("Length = %d top_bracket = %d top_backref = %d\n",
6933    length, re->top_bracket, re->top_backref);    length, re->top_bracket, re->top_backref);
6934    
# Line 6374  if (code - codestart > length) Line 6965  if (code - codestart > length)
6965    if (errorcodeptr != NULL) *errorcodeptr = ERR23;    if (errorcodeptr != NULL) *errorcodeptr = ERR23;
6966    return NULL;    return NULL;
6967    }    }
6968  #endif   /* DEBUG */  #endif   /* PCRE_DEBUG */
6969    
6970  return (pcre *)re;  return (pcre *)re;
6971  }  }

Legend:
Removed from v.369  
changed lines
  Added in v.514

webmaster@exim.org
ViewVC Help
Powered by ViewVC 1.1.12