/[pcre]/code/trunk/pcre_compile.c
ViewVC logotype

Diff of /code/trunk/pcre_compile.c

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

revision 323 by ph10, Wed Mar 5 17:23:42 2008 UTC revision 504 by ph10, Mon Mar 8 08:57:04 2010 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2008 University of Cambridge             Copyright (c) 1997-2010 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 53  supporting internal functions that are n Line 53  supporting internal functions that are n
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56  /* When DEBUG is defined, we need the pcre_printint() function, which is also  /* When PCRE_DEBUG is defined, we need the pcre_printint() function, which is
57  used by pcretest. DEBUG is not defined when building a production library. */  also used by pcretest. PCRE_DEBUG is not defined when building a production
58    library. */
59    
60  #ifdef DEBUG  #ifdef PCRE_DEBUG
61  #include "pcre_printint.src"  #include "pcre_printint.src"
62  #endif  #endif
63    
# Line 97  are simple data values; negative values Line 98  are simple data values; negative values
98  on. Zero means further processing is needed (for things like \x), or the escape  on. Zero means further processing is needed (for things like \x), or the escape
99  is invalid. */  is invalid. */
100    
101  #ifndef EBCDIC  /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC
102    
103    /* This is the "normal" table for ASCII systems or for EBCDIC systems running
104    in UTF-8 mode. */
105    
106  static const short int escapes[] = {  static const short int escapes[] = {
107       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,                       0,
108       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,                       0,
109     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */       0,                       0,
110  -ESC_H,      0,      0, -ESC_K,      0,      0,      0,      0,   /* H - O */       0,                       0,
111  -ESC_P, -ESC_Q, -ESC_R, -ESC_S,      0,      0, -ESC_V, -ESC_W,   /* P - W */       0,                       0,
112  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */       CHAR_COLON,              CHAR_SEMICOLON,
113     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
114  -ESC_h,      0,      0, -ESC_k,      0,      0,  ESC_n,      0,   /* h - o */       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
115  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0, -ESC_v, -ESC_w,   /* p - w */       CHAR_COMMERCIAL_AT,      -ESC_A,
116       0,      0, -ESC_z                                            /* x - z */       -ESC_B,                  -ESC_C,
117         -ESC_D,                  -ESC_E,
118         0,                       -ESC_G,
119         -ESC_H,                  0,
120         0,                       -ESC_K,
121         0,                       0,
122         0,                       0,
123         -ESC_P,                  -ESC_Q,
124         -ESC_R,                  -ESC_S,
125         0,                       0,
126         -ESC_V,                  -ESC_W,
127         -ESC_X,                  0,
128         -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,
129         CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,
130         CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,
131         CHAR_GRAVE_ACCENT,       7,
132         -ESC_b,                  0,
133         -ESC_d,                  ESC_e,
134         ESC_f,                   0,
135         -ESC_h,                  0,
136         0,                       -ESC_k,
137         0,                       0,
138         ESC_n,                   0,
139         -ESC_p,                  0,
140         ESC_r,                   -ESC_s,
141         ESC_tee,                 0,
142         -ESC_v,                  -ESC_w,
143         0,                       0,
144         -ESC_z
145  };  };
146    
147  #else           /* This is the "abnormal" table for EBCDIC systems */  #else
148    
149    /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
150    
151  static const short int escapes[] = {  static const short int escapes[] = {
152  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
153  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 142  static const short int escapes[] = { Line 178  static const short int escapes[] = {
178    
179  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
180  searched linearly. Put all the names into a single string, in order to reduce  searched linearly. Put all the names into a single string, in order to reduce
181  the number of relocations when a shared library is dynamically linked. */  the number of relocations when a shared library is dynamically linked. The
182    string is built from string macros so that it works in UTF-8 mode on EBCDIC
183    platforms. */
184    
185  typedef struct verbitem {  typedef struct verbitem {
186    int   len;    int   len;
# Line 150  typedef struct verbitem { Line 188  typedef struct verbitem {
188  } verbitem;  } verbitem;
189    
190  static const char verbnames[] =  static const char verbnames[] =
191    "ACCEPT\0"    STRING_ACCEPT0
192    "COMMIT\0"    STRING_COMMIT0
193    "F\0"    STRING_F0
194    "FAIL\0"    STRING_FAIL0
195    "PRUNE\0"    STRING_PRUNE0
196    "SKIP\0"    STRING_SKIP0
197    "THEN";    STRING_THEN;
198    
199  static verbitem verbs[] = {  static const verbitem verbs[] = {
200    { 6, OP_ACCEPT },    { 6, OP_ACCEPT },
201    { 6, OP_COMMIT },    { 6, OP_COMMIT },
202    { 1, OP_FAIL },    { 1, OP_FAIL },
# Line 168  static verbitem verbs[] = { Line 206  static verbitem verbs[] = {
206    { 4, OP_THEN  }    { 4, OP_THEN  }
207  };  };
208    
209  static int verbcount = sizeof(verbs)/sizeof(verbitem);  static const int verbcount = sizeof(verbs)/sizeof(verbitem);
210    
211    
212  /* Tables of names of POSIX character classes and their lengths. The names are  /* Tables of names of POSIX character classes and their lengths. The names are
# Line 178  length entry. The first three must be al Line 216  length entry. The first three must be al
216  for handling case independence. */  for handling case independence. */
217    
218  static const char posix_names[] =  static const char posix_names[] =
219    "alpha\0"  "lower\0"  "upper\0"  "alnum\0"  "ascii\0"  "blank\0"    STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
220    "cntrl\0"  "digit\0"  "graph\0"  "print\0"  "punct\0"  "space\0"    STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
221    "word\0"   "xdigit";    STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
222      STRING_word0  STRING_xdigit;
223    
224  static const uschar posix_name_lengths[] = {  static const uschar posix_name_lengths[] = {
225    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
# Line 224  the number of relocations needed when a Line 263  the number of relocations needed when a
263  it is now one long string. We cannot use a table of offsets, because the  it is now one long string. We cannot use a table of offsets, because the
264  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
265  simply count through to the one we want - this isn't a performance issue  simply count through to the one we want - this isn't a performance issue
266  because these strings are used only when there is a compilation error. */  because these strings are used only when there is a compilation error.
267    
268    Each substring ends with \0 to insert a null character. This includes the final
269    substring, so that the whole string ends with \0\0, which can be detected when
270    counting through. */
271    
272  static const char error_texts[] =  static const char error_texts[] =
273    "no error\0"    "no error\0"
# Line 295  static const char error_texts[] = Line 338  static const char error_texts[] =
338    /* 55 */    /* 55 */
339    "repeating a DEFINE group is not allowed\0"    "repeating a DEFINE group is not allowed\0"
340    "inconsistent NEWLINE options\0"    "inconsistent NEWLINE options\0"
341    "\\g is not followed by a braced name or an optionally braced non-zero number\0"    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
342    "(?+ or (?- or (?(+ or (?(- must be followed by a non-zero number\0"    "a numbered reference must not be zero\0"
343    "(*VERB) with an argument is not supported\0"    "(*VERB) with an argument is not supported\0"
344    /* 60 */    /* 60 */
345    "(*VERB) not recognized\0"    "(*VERB) not recognized\0"
346    "number is too big\0"    "number is too big\0"
347    "subpattern name expected\0"    "subpattern name expected\0"
348    "digit expected after (?+";    "digit expected after (?+\0"
349      "] is an invalid data character in JavaScript compatibility mode\0"
350      /* 65 */
351      "different names for subpatterns of the same number are not allowed\0";
352    
353  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
354  patterns. Note that the tables in chartables are dependent on the locale, and  patterns. Note that the tables in chartables are dependent on the locale, and
# Line 321  For convenience, we use the same bit def Line 366  For convenience, we use the same bit def
366    
367  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
368    
369  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC
370    
371    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
372    UTF-8 mode. */
373    
374  static const unsigned char digitab[] =  static const unsigned char digitab[] =
375    {    {
376    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
# Line 357  static const unsigned char digitab[] = Line 406  static const unsigned char digitab[] =
406    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
407    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
408    
409  #else           /* This is the "abnormal" case, for EBCDIC systems */  #else
410    
411    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
412    
413  static const unsigned char digitab[] =  static const unsigned char digitab[] =
414    {    {
415    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
# Line 454  static const char * Line 506  static const char *
506  find_error_text(int n)  find_error_text(int n)
507  {  {
508  const char *s = error_texts;  const char *s = error_texts;
509  for (; n > 0; n--) while (*s++ != 0);  for (; n > 0; n--)
510      {
511      while (*s++ != 0) {};
512      if (*s == 0) return "Error text not found (please report)";
513      }
514  return s;  return s;
515  }  }
516    
# Line 502  if (c == 0) *errorcodeptr = ERR1; Line 558  if (c == 0) *errorcodeptr = ERR1;
558  in a table. A non-zero result is something that can be returned immediately.  in a table. A non-zero result is something that can be returned immediately.
559  Otherwise further processing may be required. */  Otherwise further processing may be required. */
560    
561  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
562  else if (c < '0' || c > 'z') {}                           /* Not alphanumeric */  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */
563  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0) c = i;
564    
565  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
566  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */
# Line 523  else Line 579  else
579      /* A number of Perl escapes are not handled by PCRE. We give an explicit      /* A number of Perl escapes are not handled by PCRE. We give an explicit
580      error. */      error. */
581    
582      case 'l':      case CHAR_l:
583      case 'L':      case CHAR_L:
584      case 'N':      case CHAR_N:
585      case 'u':      case CHAR_u:
586      case 'U':      case CHAR_U:
587      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
588      break;      break;
589    
590      /* \g must be followed by a number, either plain or braced. If positive, it      /* \g must be followed by one of a number of specific things:
591      is an absolute backreference. If negative, it is a relative backreference.  
592      This is a Perl 5.10 feature. Perl 5.10 also supports \g{name} as a      (1) A number, either plain or braced. If positive, it is an absolute
593      reference to a named group. This is part of Perl's movement towards a      backreference. If negative, it is a relative backreference. This is a Perl
594      unified syntax for back references. As this is synonymous with \k{name}, we      5.10 feature.
595      fudge it up by pretending it really was \k. */  
596        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
597        is part of Perl's movement towards a unified syntax for back references. As
598        this is synonymous with \k{name}, we fudge it up by pretending it really
599        was \k.
600    
601        (3) For Oniguruma compatibility we also support \g followed by a name or a
602        number either in angle brackets or in single quotes. However, these are
603        (possibly recursive) subroutine calls, _not_ backreferences. Just return
604        the -ESC_g code (cf \k). */
605    
606        case CHAR_g:
607        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
608          {
609          c = -ESC_g;
610          break;
611          }
612    
613        /* Handle the Perl-compatible cases */
614    
615      case 'g':      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
     if (ptr[1] == '{')  
616        {        {
617        const uschar *p;        const uschar *p;
618        for (p = ptr+2; *p != 0 && *p != '}'; p++)        for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
619          if (*p != '-' && (digitab[*p] & ctype_digit) == 0) break;          if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
620        if (*p != 0 && *p != '}')        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
621          {          {
622          c = -ESC_k;          c = -ESC_k;
623          break;          break;
# Line 554  else Line 627  else
627        }        }
628      else braced = FALSE;      else braced = FALSE;
629    
630      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
631        {        {
632        negated = TRUE;        negated = TRUE;
633        ptr++;        ptr++;
# Line 563  else Line 636  else
636    
637      c = 0;      c = 0;
638      while ((digitab[ptr[1]] & ctype_digit) != 0)      while ((digitab[ptr[1]] & ctype_digit) != 0)
639        c = c * 10 + *(++ptr) - '0';        c = c * 10 + *(++ptr) - CHAR_0;
640    
641      if (c < 0)      if (c < 0)   /* Integer overflow */
642        {        {
643        *errorcodeptr = ERR61;        *errorcodeptr = ERR61;
644        break;        break;
645        }        }
646    
647      if (c == 0 || (braced && *(++ptr) != '}'))      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
648        {        {
649        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
650        break;        break;
651        }        }
652    
653        if (c == 0)
654          {
655          *errorcodeptr = ERR58;
656          break;
657          }
658    
659      if (negated)      if (negated)
660        {        {
661        if (c > bracount)        if (c > bracount)
# Line 602  else Line 681  else
681      value is greater than 377, the least significant 8 bits are taken. Inside a      value is greater than 377, the least significant 8 bits are taken. Inside a
682      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
683    
684      case '1': case '2': case '3': case '4': case '5':      case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:
685      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
686    
687      if (!isclass)      if (!isclass)
688        {        {
689        oldptr = ptr;        oldptr = ptr;
690        c -= '0';        c -= CHAR_0;
691        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
692          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - CHAR_0;
693        if (c < 0)        if (c < 0)    /* Integer overflow */
694          {          {
695          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
696          break;          break;
# Line 628  else Line 707  else
707      generates a binary zero byte and treats the digit as a following literal.      generates a binary zero byte and treats the digit as a following literal.
708      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
709    
710      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
711        {        {
712        ptr--;        ptr--;
713        c = 0;        c = 0;
# Line 641  else Line 720  else
720      to do). Nowadays we allow for larger numbers in UTF-8 mode, but no more      to do). Nowadays we allow for larger numbers in UTF-8 mode, but no more
721      than 3 octal digits. */      than 3 octal digits. */
722    
723      case '0':      case CHAR_0:
724      c -= '0';      c -= CHAR_0;
725      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
726          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
727      if (!utf8 && c > 255) *errorcodeptr = ERR51;      if (!utf8 && c > 255) *errorcodeptr = ERR51;
728      break;      break;
729    
# Line 652  else Line 731  else
731      than 0xff in utf8 mode, but only if the ddd are hex digits. If not, { is      than 0xff in utf8 mode, but only if the ddd are hex digits. If not, { is
732      treated as a data character. */      treated as a data character. */
733    
734      case 'x':      case CHAR_x:
735      if (ptr[1] == '{')      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
736        {        {
737        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
738        int count = 0;        int count = 0;
# Line 662  else Line 741  else
741        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
742          {          {
743          register int cc = *pt++;          register int cc = *pt++;
744          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
745          count++;          count++;
746    
747  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
748          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
749          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
750  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
751          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
752          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
753  #endif  #endif
754          }          }
755    
756        if (*pt == '}')        if (*pt == CHAR_RIGHT_CURLY_BRACKET)
757          {          {
758          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
759          ptr = pt;          ptr = pt;
# Line 690  else Line 769  else
769      c = 0;      c = 0;
770      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
771        {        {
772        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
773        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
774  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
775        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
776        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
777  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
778        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
779        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
780  #endif  #endif
781        }        }
782      break;      break;
# Line 706  else Line 785  else
785      This coding is ASCII-specific, but then the whole concept of \cx is      This coding is ASCII-specific, but then the whole concept of \cx is
786      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
787    
788      case 'c':      case CHAR_c:
789      c = *(++ptr);      c = *(++ptr);
790      if (c == 0)      if (c == 0)
791        {        {
# Line 714  else Line 793  else
793        break;        break;
794        }        }
795    
796  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
797      if (c >= 'a' && c <= 'z') c -= 32;      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
798      c ^= 0x40;      c ^= 0x40;
799  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
800      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
801      c ^= 0xC0;      c ^= 0xC0;
802  #endif  #endif
803      break;      break;
# Line 780  if (c == 0) goto ERROR_RETURN; Line 859  if (c == 0) goto ERROR_RETURN;
859  /* \P or \p can be followed by a name in {}, optionally preceded by ^ for  /* \P or \p can be followed by a name in {}, optionally preceded by ^ for
860  negation. */  negation. */
861    
862  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
863    {    {
864    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
865      {      {
866      *negptr = TRUE;      *negptr = TRUE;
867      ptr++;      ptr++;
# Line 791  if (c == '{') Line 870  if (c == '{')
870      {      {
871      c = *(++ptr);      c = *(++ptr);
872      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
873      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
874      name[i] = c;      name[i] = c;
875      }      }
876    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
877    name[i] = 0;    name[i] = 0;
878    }    }
879    
# Line 859  is_counted_repeat(const uschar *p) Line 938  is_counted_repeat(const uschar *p)
938  {  {
939  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
940  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
941  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
942    
943  if (*p++ != ',') return FALSE;  if (*p++ != CHAR_COMMA) return FALSE;
944  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
945    
946  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
947  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
948    
949  return (*p == '}');  return (*p == CHAR_RIGHT_CURLY_BRACKET);
950  }  }
951    
952    
# Line 900  int max = -1; Line 979  int max = -1;
979  /* Read the minimum value and do a paranoid check: a negative value indicates  /* Read the minimum value and do a paranoid check: a negative value indicates
980  an integer overflow. */  an integer overflow. */
981    
982  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;
983  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
984    {    {
985    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 910  if (min < 0 || min > 65535) Line 989  if (min < 0 || min > 65535)
989  /* Read the maximum value if there is one, and again do a paranoid on its size.  /* Read the maximum value if there is one, and again do a paranoid on its size.
990  Also, max must not be less than min. */  Also, max must not be less than min. */
991    
992  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
993    {    {
994    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
995      {      {
996      max = 0;      max = 0;
997      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;
998      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
999        {        {
1000        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 940  return p; Line 1019  return p;
1019    
1020    
1021  /*************************************************  /*************************************************
1022  *       Find forward referenced subpattern       *  *  Subroutine for finding forward reference      *
1023  *************************************************/  *************************************************/
1024    
1025  /* This function scans along a pattern's text looking for capturing  /* This recursive function is called only from find_parens() below. The
1026    top-level call starts at the beginning of the pattern. All other calls must
1027    start at a parenthesis. It scans along a pattern's text looking for capturing
1028  subpatterns, and counting them. If it finds a named pattern that matches the  subpatterns, and counting them. If it finds a named pattern that matches the
1029  name it is given, it returns its number. Alternatively, if the name is NULL, it  name it is given, it returns its number. Alternatively, if the name is NULL, it
1030  returns when it reaches a given numbered subpattern. This is used for forward  returns when it reaches a given numbered subpattern. We know that if (?P< is
1031  references to subpatterns. We know that if (?P< is encountered, the name will  encountered, the name will be terminated by '>' because that is checked in the
1032  be terminated by '>' because that is checked in the first pass.  first pass. Recursion is used to keep track of subpatterns that reset the
1033    capturing group numbers - the (?| feature.
1034    
1035  Arguments:  Arguments:
1036    ptr          current position in the pattern    ptrptr       address of the current character pointer (updated)
1037    count        current count of capturing parens so far encountered    cd           compile background data
1038    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1039    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1040    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1041      count        pointer to the current capturing subpattern number (updated)
1042    
1043  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1044  */  */
1045    
1046  static int  static int
1047  find_parens(const uschar *ptr, int count, const uschar *name, int lorn,  find_parens_sub(uschar **ptrptr, compile_data *cd, const uschar *name, int lorn,
1048    BOOL xmode)    BOOL xmode, int *count)
1049  {  {
1050  const uschar *thisname;  uschar *ptr = *ptrptr;
1051    int start_count = *count;
1052    int hwm_count = start_count;
1053    BOOL dup_parens = FALSE;
1054    
1055  for (; *ptr != 0; ptr++)  /* If the first character is a parenthesis, check on the type of group we are
1056    dealing with. The very first call may not start with a parenthesis. */
1057    
1058    if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1059    {    {
1060    int term;    if (ptr[1] == CHAR_QUESTION_MARK &&
1061          ptr[2] == CHAR_VERTICAL_LINE)
1062        {
1063        ptr += 3;
1064        dup_parens = TRUE;
1065        }
1066    
1067      /* Handle a normal, unnamed capturing parenthesis */
1068    
1069      else if (ptr[1] != CHAR_QUESTION_MARK && ptr[1] != CHAR_ASTERISK)
1070        {
1071        *count += 1;
1072        if (name == NULL && *count == lorn) return *count;
1073        ptr++;
1074        }
1075    
1076      /* Handle a condition. If it is an assertion, just carry on so that it
1077      is processed as normal. If not, skip to the closing parenthesis of the
1078      condition (there can't be any nested parens. */
1079    
1080      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1081        {
1082        ptr += 2;
1083        if (ptr[1] != CHAR_QUESTION_MARK)
1084          {
1085          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1086          if (*ptr != 0) ptr++;
1087          }
1088        }
1089    
1090      /* We have either (? or (* and not a condition */
1091    
1092      else
1093        {
1094        ptr += 2;
1095        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1096    
1097        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1098    
1099        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1100            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1101          {
1102          int term;
1103          const uschar *thisname;
1104          *count += 1;
1105          if (name == NULL && *count == lorn) return *count;
1106          term = *ptr++;
1107          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1108          thisname = ptr;
1109          while (*ptr != term) ptr++;
1110          if (name != NULL && lorn == ptr - thisname &&
1111              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1112            return *count;
1113          term++;
1114          }
1115        }
1116      }
1117    
1118    /* Past any initial parenthesis handling, scan for parentheses or vertical
1119    bars. */
1120    
1121    for (; *ptr != 0; ptr++)
1122      {
1123    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1124    
1125    if (*ptr == '\\')    if (*ptr == CHAR_BACKSLASH)
1126      {      {
1127      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1128      if (*ptr == 'Q') for (;;)      if (*ptr == CHAR_Q) for (;;)
1129        {        {
1130        while (*(++ptr) != 0 && *ptr != '\\');        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1131        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1132        if (*(++ptr) == 'E') break;        if (*(++ptr) == CHAR_E) break;
1133        }        }
1134      continue;      continue;
1135      }      }
1136    
1137    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1138      are handled for real. If the first character is '^', skip it. Also, if the
1139      first few characters (either before or after ^) are \Q\E or \E we skip them
1140      too. This makes for compatibility with Perl. Note the use of STR macros to
1141      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1142    
1143    if (*ptr == '[')    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1144      {      {
1145      while (*(++ptr) != ']')      BOOL negate_class = FALSE;
1146        for (;;)
1147          {
1148          if (ptr[1] == CHAR_BACKSLASH)
1149            {
1150            if (ptr[2] == CHAR_E)
1151              ptr+= 2;
1152            else if (strncmp((const char *)ptr+2,
1153                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1154              ptr += 4;
1155            else
1156              break;
1157            }
1158          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1159            {
1160            negate_class = TRUE;
1161            ptr++;
1162            }
1163          else break;
1164          }
1165    
1166        /* If the next character is ']', it is a data character that must be
1167        skipped, except in JavaScript compatibility mode. */
1168    
1169        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1170            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1171          ptr++;
1172    
1173        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1174        {        {
1175        if (*ptr == 0) return -1;        if (*ptr == 0) return -1;
1176        if (*ptr == '\\')        if (*ptr == CHAR_BACKSLASH)
1177          {          {
1178          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) goto FAIL_EXIT;
1179          if (*ptr == 'Q') for (;;)          if (*ptr == CHAR_Q) for (;;)
1180            {            {
1181            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1182            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1183            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1184            }            }
1185          continue;          continue;
1186          }          }
# Line 1008  for (; *ptr != 0; ptr++) Line 1190  for (; *ptr != 0; ptr++)
1190    
1191    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1192    
1193    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1194      {      {
1195      while (*(++ptr) != 0 && *ptr != '\n');      while (*(++ptr) != 0 && *ptr != CHAR_NL) {};
1196      if (*ptr == 0) return -1;      if (*ptr == 0) goto FAIL_EXIT;
1197      continue;      continue;
1198      }      }
1199    
1200    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1201    
1202    if (*ptr != '(') continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != '?' && ptr[1] != '*')  
1203      {      {
1204      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, count);
1205      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1206      continue;      if (*ptr == 0) goto FAIL_EXIT;
1207      }      }
1208    
1209    ptr += 2;    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1210    if (*ptr == 'P') ptr++;                      /* Allow optional P */      {
1211        if (dup_parens && *count < hwm_count) *count = hwm_count;
1212        *ptrptr = ptr;
1213        return -1;
1214        }
1215    
1216    /* We have to disambiguate (?<! and (?<= from (?<name> */    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1217        {
1218        if (*count > hwm_count) hwm_count = *count;
1219        *count = start_count;
1220        }
1221      }
1222    
1223    if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  FAIL_EXIT:
1224         *ptr != '\'')  *ptrptr = ptr;
1225      continue;  return -1;
1226    }
1227    
1228    
1229    
1230    
1231    /*************************************************
1232    *       Find forward referenced subpattern       *
1233    *************************************************/
1234    
1235    /* This function scans along a pattern's text looking for capturing
1236    subpatterns, and counting them. If it finds a named pattern that matches the
1237    name it is given, it returns its number. Alternatively, if the name is NULL, it
1238    returns when it reaches a given numbered subpattern. This is used for forward
1239    references to subpatterns. We used to be able to start this scan from the
1240    current compiling point, using the current count value from cd->bracount, and
1241    do it all in a single loop, but the addition of the possibility of duplicate
1242    subpattern numbers means that we have to scan from the very start, in order to
1243    take account of such duplicates, and to use a recursive function to keep track
1244    of the different types of group.
1245    
1246    Arguments:
1247      cd           compile background data
1248      name         name to seek, or NULL if seeking a numbered subpattern
1249      lorn         name length, or subpattern number if name is NULL
1250      xmode        TRUE if we are in /x mode
1251    
1252    Returns:       the number of the found subpattern, or -1 if not found
1253    */
1254    
1255    count++;  static int
1256    find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode)
1257    {
1258    uschar *ptr = (uschar *)cd->start_pattern;
1259    int count = 0;
1260    int rc;
1261    
1262    /* If the pattern does not start with an opening parenthesis, the first call
1263    to find_parens_sub() will scan right to the end (if necessary). However, if it
1264    does start with a parenthesis, find_parens_sub() will return when it hits the
1265    matching closing parens. That is why we have to have a loop. */
1266    
1267    if (name == NULL && count == lorn) return count;  for (;;)
1268    term = *ptr++;    {
1269    if (term == '<') term = '>';    rc = find_parens_sub(&ptr, cd, name, lorn, xmode, &count);
1270    thisname = ptr;    if (rc > 0 || *ptr++ == 0) break;
   while (*ptr != term) ptr++;  
   if (name != NULL && lorn == ptr - thisname &&  
       strncmp((const char *)name, (const char *)thisname, lorn) == 0)  
     return count;  
1271    }    }
1272    
1273  return -1;  return rc;
1274  }  }
1275    
1276    
1277    
1278    
1279  /*************************************************  /*************************************************
1280  *      Find first significant op code            *  *      Find first significant op code            *
1281  *************************************************/  *************************************************/
# Line 1100  for (;;) Line 1325  for (;;)
1325    
1326      case OP_CALLOUT:      case OP_CALLOUT:
1327      case OP_CREF:      case OP_CREF:
1328        case OP_NCREF:
1329      case OP_RREF:      case OP_RREF:
1330        case OP_NRREF:
1331      case OP_DEF:      case OP_DEF:
1332      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1333      break;      break;
# Line 1116  for (;;) Line 1343  for (;;)
1343    
1344    
1345  /*************************************************  /*************************************************
1346  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1347  *************************************************/  *************************************************/
1348    
1349  /* Scan a pattern and compute the fixed length of subject that will match it,  /* Scan a branch and compute the fixed length of subject that will match it,
1350  if the length is fixed. This is needed for dealing with backward assertions.  if the length is fixed. This is needed for dealing with backward assertions.
1351  In UTF8 mode, the result is in characters rather than bytes.  In UTF8 mode, the result is in characters rather than bytes. The branch is
1352    temporarily terminated with OP_END when this function is called.
1353    
1354    This function is called when a backward assertion is encountered, so that if it
1355    fails, the error message can point to the correct place in the pattern.
1356    However, we cannot do this when the assertion contains subroutine calls,
1357    because they can be forward references. We solve this by remembering this case
1358    and doing the check at the end; a flag specifies which mode we are running in.
1359    
1360  Arguments:  Arguments:
1361    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1362    options  the compiling options    options  the compiling options
1363      atend    TRUE if called when the pattern is complete
1364      cd       the "compile data" structure
1365    
1366  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1367                 or -1 if there is no fixed length,
1368               or -2 if \C was encountered               or -2 if \C was encountered
1369                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1370  */  */
1371    
1372  static int  static int
1373  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)
1374  {  {
1375  int length = -1;  int length = -1;
1376    
# Line 1145  branch, check the length against that of Line 1383  branch, check the length against that of
1383  for (;;)  for (;;)
1384    {    {
1385    int d;    int d;
1386      uschar *ce, *cs;
1387    register int op = *cc;    register int op = *cc;
1388    switch (op)    switch (op)
1389      {      {
# Line 1152  for (;;) Line 1391  for (;;)
1391      case OP_BRA:      case OP_BRA:
1392      case OP_ONCE:      case OP_ONCE:
1393      case OP_COND:      case OP_COND:
1394      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);
1395      if (d < 0) return d;      if (d < 0) return d;
1396      branchlength += d;      branchlength += d;
1397      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1175  for (;;) Line 1414  for (;;)
1414      branchlength = 0;      branchlength = 0;
1415      break;      break;
1416    
1417        /* A true recursion implies not fixed length, but a subroutine call may
1418        be OK. If the subroutine is a forward reference, we can't deal with
1419        it until the end of the pattern, so return -3. */
1420    
1421        case OP_RECURSE:
1422        if (!atend) return -3;
1423        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1424        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1425        if (cc > cs && cc < ce) return -1;                /* Recursion */
1426        d = find_fixedlength(cs + 2, options, atend, cd);
1427        if (d < 0) return d;
1428        branchlength += d;
1429        cc += 1 + LINK_SIZE;
1430        break;
1431    
1432      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1433    
1434      case OP_ASSERT:      case OP_ASSERT:
# Line 1188  for (;;) Line 1442  for (;;)
1442    
1443      case OP_REVERSE:      case OP_REVERSE:
1444      case OP_CREF:      case OP_CREF:
1445        case OP_NCREF:
1446      case OP_RREF:      case OP_RREF:
1447        case OP_NRREF:
1448      case OP_DEF:      case OP_DEF:
1449      case OP_OPT:      case OP_OPT:
1450      case OP_CALLOUT:      case OP_CALLOUT:
1451      case OP_SOD:      case OP_SOD:
1452      case OP_SOM:      case OP_SOM:
1453        case OP_SET_SOM:
1454      case OP_EOD:      case OP_EOD:
1455      case OP_EODN:      case OP_EODN:
1456      case OP_CIRC:      case OP_CIRC:
# Line 1211  for (;;) Line 1468  for (;;)
1468      branchlength++;      branchlength++;
1469      cc += 2;      cc += 2;
1470  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1471      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1472        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1473  #endif  #endif
1474      break;      break;
1475    
# Line 1225  for (;;) Line 1480  for (;;)
1480      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1481      cc += 4;      cc += 4;
1482  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1483      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1484        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1485  #endif  #endif
1486      break;      break;
1487    
# Line 1252  for (;;) Line 1505  for (;;)
1505      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1506      case OP_WORDCHAR:      case OP_WORDCHAR:
1507      case OP_ANY:      case OP_ANY:
1508        case OP_ALLANY:
1509      branchlength++;      branchlength++;
1510      cc++;      cc++;
1511      break;      break;
# Line 1306  for (;;) Line 1560  for (;;)
1560    
1561    
1562  /*************************************************  /*************************************************
1563  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1564  *************************************************/  *************************************************/
1565    
1566  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1567  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1568    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1569    so that it can be called from pcre_study() when finding the minimum matching
1570    length.
1571    
1572  Arguments:  Arguments:
1573    code        points to start of expression    code        points to start of expression
1574    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1575    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1576    
1577  Returns:      pointer to the opcode for the bracket, or NULL if not found  Returns:      pointer to the opcode for the bracket, or NULL if not found
1578  */  */
1579    
1580  static const uschar *  const uschar *
1581  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1582  {  {
1583  for (;;)  for (;;)
1584    {    {
# Line 1334  for (;;) Line 1591  for (;;)
1591    
1592    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1593    
1594      /* Handle recursion */
1595    
1596      else if (c == OP_REVERSE)
1597        {
1598        if (number < 0) return (uschar *)code;
1599        code += _pcre_OP_lengths[c];
1600        }
1601    
1602    /* Handle capturing bracket */    /* Handle capturing bracket */
1603    
1604    else if (c == OP_CBRA)    else if (c == OP_CBRA)
# Line 1400  for (;;) Line 1665  for (;;)
1665        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1666        break;        break;
1667        }        }
1668    #else
1669        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1670  #endif  #endif
1671      }      }
1672    }    }
# Line 1493  for (;;) Line 1760  for (;;)
1760        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1761        break;        break;
1762        }        }
1763    #else
1764        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1765  #endif  #endif
1766      }      }
1767    }    }
# Line 1516  Arguments: Line 1785  Arguments:
1785    code        points to start of search    code        points to start of search
1786    endcode     points to where to stop    endcode     points to where to stop
1787    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
1788      cd          contains pointers to tables etc.
1789    
1790  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
1791  */  */
1792    
1793  static BOOL  static BOOL
1794  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
1795      compile_data *cd)
1796  {  {
1797  register int c;  register int c;
1798  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);
# Line 1531  for (code = first_significant_code(code Line 1802  for (code = first_significant_code(code
1802    const uschar *ccode;    const uschar *ccode;
1803    
1804    c = *code;    c = *code;
1805    
1806    /* Skip over forward assertions; the other assertions are skipped by    /* Skip over forward assertions; the other assertions are skipped by
1807    first_significant_code() with a TRUE final argument. */    first_significant_code() with a TRUE final argument. */
1808    
# Line 1544  for (code = first_significant_code(code Line 1815  for (code = first_significant_code(code
1815    
1816    /* Groups with zero repeats can of course be empty; skip them. */    /* Groups with zero repeats can of course be empty; skip them. */
1817    
1818    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)
1819      {      {
1820      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1821      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
1822      c = *code;      c = *code;
1823      continue;      continue;
1824      }      }
1825    
1826      /* For a recursion/subroutine call, if its end has been reached, which
1827      implies a subroutine call, we can scan it. */
1828    
1829      if (c == OP_RECURSE)
1830        {
1831        BOOL empty_branch = FALSE;
1832        const uschar *scode = cd->start_code + GET(code, 1);
1833        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
1834        do
1835          {
1836          if (could_be_empty_branch(scode, endcode, utf8, cd))
1837            {
1838            empty_branch = TRUE;
1839            break;
1840            }
1841          scode += GET(scode, 1);
1842          }
1843        while (*scode == OP_ALT);
1844        if (!empty_branch) return FALSE;  /* All branches are non-empty */
1845        continue;
1846        }
1847    
1848    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
1849    
# Line 1559  for (code = first_significant_code(code Line 1852  for (code = first_significant_code(code
1852      BOOL empty_branch;      BOOL empty_branch;
1853      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
1854    
1855      /* Scan a closed bracket */      /* If a conditional group has only one branch, there is a second, implied,
1856        empty branch, so just skip over the conditional, because it could be empty.
1857        Otherwise, scan the individual branches of the group. */
1858    
1859      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;  
1860        code += GET(code, 1);        code += GET(code, 1);
1861        else
1862          {
1863          empty_branch = FALSE;
1864          do
1865            {
1866            if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
1867              empty_branch = TRUE;
1868            code += GET(code, 1);
1869            }
1870          while (*code == OP_ALT);
1871          if (!empty_branch) return FALSE;   /* All branches are non-empty */
1872        }        }
1873      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
1874      c = *code;      c = *code;
1875      continue;      continue;
1876      }      }
# Line 1630  for (code = first_significant_code(code Line 1931  for (code = first_significant_code(code
1931      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1932      case OP_WORDCHAR:      case OP_WORDCHAR:
1933      case OP_ANY:      case OP_ANY:
1934        case OP_ALLANY:
1935      case OP_ANYBYTE:      case OP_ANYBYTE:
1936      case OP_CHAR:      case OP_CHAR:
1937      case OP_CHARNC:      case OP_CHARNC:
# Line 1686  for (code = first_significant_code(code Line 1988  for (code = first_significant_code(code
1988      case OP_QUERY:      case OP_QUERY:
1989      case OP_MINQUERY:      case OP_MINQUERY:
1990      case OP_POSQUERY:      case OP_POSQUERY:
1991        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
1992        break;
1993    
1994      case OP_UPTO:      case OP_UPTO:
1995      case OP_MINUPTO:      case OP_MINUPTO:
1996      case OP_POSUPTO:      case OP_POSUPTO:
1997      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
1998      break;      break;
1999  #endif  #endif
2000    
2001        /* None of the remaining opcodes are required to match a character. */
2002    
2003        default:
2004        break;
2005      }      }
2006    }    }
2007    
# Line 1714  Arguments: Line 2024  Arguments:
2024    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2025    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2026    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2027      cd          pointers to tables etc
2028    
2029  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2030  */  */
2031    
2032  static BOOL  static BOOL
2033  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2034    BOOL utf8)    BOOL utf8, compile_data *cd)
2035  {  {
2036  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2037    {    {
2038    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2039        return FALSE;
2040    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2041    }    }
2042  return TRUE;  return TRUE;
# Line 1770  int terminator; /* Don't combin Line 2082  int terminator; /* Don't combin
2082  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2083  for (++ptr; *ptr != 0; ptr++)  for (++ptr; *ptr != 0; ptr++)
2084    {    {
2085    if (*ptr == '\\' && ptr[1] == ']') ptr++; else    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else
2086      {      {
2087      if (*ptr == ']') return FALSE;      if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2088      if (*ptr == terminator && ptr[1] == ']')      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2089        {        {
2090        *endptr = ptr;        *endptr = ptr;
2091        return TRUE;        return TRUE;
# Line 1824  return -1; Line 2136  return -1;
2136  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2137  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2138  earlier groups that are outside the current group). However, when a group is  earlier groups that are outside the current group). However, when a group is
2139  optional (i.e. the minimum quantifier is zero), OP_BRAZERO is inserted before  optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is
2140  it, after it has been compiled. This means that any OP_RECURSE items within it  inserted before it, after it has been compiled. This means that any OP_RECURSE
2141  that refer to the group itself or any contained groups have to have their  items within it that refer to the group itself or any contained groups have to
2142  offsets adjusted. That one of the jobs of this function. Before it is called,  have their offsets adjusted. That one of the jobs of this function. Before it
2143  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2144    OP_END.
2145    
2146  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2147  recursions and subroutine calls. It must also check the list of such references  recursions and subroutine calls. It must also check the list of such references
# Line 1963  get_othercase_range(unsigned int *cptr, Line 2276  get_othercase_range(unsigned int *cptr,
2276  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2277    
2278  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2279    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2280    
2281  if (c > d) return FALSE;  if (c > d) return FALSE;
2282    
# Line 1972  next = othercase + 1; Line 2285  next = othercase + 1;
2285    
2286  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2287    {    {
2288    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2289    next++;    next++;
2290    }    }
2291    
# Line 2018  if ((options & PCRE_EXTENDED) != 0) Line 2331  if ((options & PCRE_EXTENDED) != 0)
2331    for (;;)    for (;;)
2332      {      {
2333      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2334      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2335        {        {
2336        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2337          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
# Line 2030  if ((options & PCRE_EXTENDED) != 0) Line 2343  if ((options & PCRE_EXTENDED) != 0)
2343  /* 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
2344  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2345    
2346  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2347    {    {
2348    int temperrorcode = 0;    int temperrorcode = 0;
2349    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
# Line 2055  if ((options & PCRE_EXTENDED) != 0) Line 2368  if ((options & PCRE_EXTENDED) != 0)
2368    for (;;)    for (;;)
2369      {      {
2370      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2371      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2372        {        {
2373        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2374          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
# Line 2066  if ((options & PCRE_EXTENDED) != 0) Line 2379  if ((options & PCRE_EXTENDED) != 0)
2379    
2380  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
2381    
2382  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2383    return FALSE;    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2384        return FALSE;
2385    
2386  /* 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
2387  positive single character match, "item" either contains the character or, if  positive single character match, "item" either contains the character or, if
# Line 2082  if (next >= 0) switch(op_code) Line 2396  if (next >= 0) switch(op_code)
2396    case OP_CHAR:    case OP_CHAR:
2397  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2398    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2399    #else
2400      (void)(utf8_char);  /* Keep compiler happy by referencing function argument */
2401  #endif  #endif
2402    return item != next;    return item != next;
2403    
# Line 2100  if (next >= 0) switch(op_code) Line 2416  if (next >= 0) switch(op_code)
2416      unsigned int othercase;      unsigned int othercase;
2417      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2418  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2419      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
2420  #else  #else
2421      othercase = NOTACHAR;      othercase = NOTACHAR;
2422  #endif  #endif
# Line 2121  if (next >= 0) switch(op_code) Line 2437  if (next >= 0) switch(op_code)
2437      unsigned int othercase;      unsigned int othercase;
2438      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2439  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2440      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
2441  #else  #else
2442      othercase = NOTACHAR;      othercase = NOTACHAR;
2443  #endif  #endif
# Line 2382  BOOL utf8 = FALSE; Line 2698  BOOL utf8 = FALSE;
2698  uschar *utf8_char = NULL;  uschar *utf8_char = NULL;
2699  #endif  #endif
2700    
2701  #ifdef DEBUG  #ifdef PCRE_DEBUG
2702  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2703  #endif  #endif
2704    
# Line 2441  for (;; ptr++) Line 2757  for (;; ptr++)
2757    
2758    if (lengthptr != NULL)    if (lengthptr != NULL)
2759      {      {
2760  #ifdef DEBUG  #ifdef PCRE_DEBUG
2761      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
2762  #endif  #endif
2763      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */
# Line 2503  for (;; ptr++) Line 2819  for (;; ptr++)
2819    
2820    if (inescq && c != 0)    if (inescq && c != 0)
2821      {      {
2822      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
2823        {        {
2824        inescq = FALSE;        inescq = FALSE;
2825        ptr++;        ptr++;
# Line 2529  for (;; ptr++) Line 2845  for (;; ptr++)
2845    /* 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
2846    a quantifier. */    a quantifier. */
2847    
2848    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
2849      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
2850        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
2851    
2852    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
2853         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2545  for (;; ptr++) Line 2862  for (;; ptr++)
2862    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
2863      {      {
2864      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
2865      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
2866        {        {
2867        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2868          {          {
# Line 2570  for (;; ptr++) Line 2887  for (;; ptr++)
2887      {      {
2888      /* ===================================================================*/      /* ===================================================================*/
2889      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
2890      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
2891      case ')':      case CHAR_RIGHT_PARENTHESIS:
2892      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
2893      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
2894      *codeptr = code;      *codeptr = code;
# Line 2593  for (;; ptr++) Line 2910  for (;; ptr++)
2910      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
2911      the setting of any following char as a first character. */      the setting of any following char as a first character. */
2912    
2913      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
2914      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
2915        {        {
2916        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
# Line 2602  for (;; ptr++) Line 2919  for (;; ptr++)
2919      *code++ = OP_CIRC;      *code++ = OP_CIRC;
2920      break;      break;
2921    
2922      case '$':      case CHAR_DOLLAR_SIGN:
2923      previous = NULL;      previous = NULL;
2924      *code++ = OP_DOLL;      *code++ = OP_DOLL;
2925      break;      break;
# Line 2610  for (;; ptr++) Line 2927  for (;; ptr++)
2927      /* 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
2928      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
2929    
2930      case '.':      case CHAR_DOT:
2931      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
2932      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
2933      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
2934      previous = code;      previous = code;
2935      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
2936      break;      break;
2937    
2938    
# Line 2630  for (;; ptr++) Line 2947  for (;; ptr++)
2947      opcode is compiled. It may optionally have a bit map for characters < 256,      opcode is compiled. It may optionally have a bit map for characters < 256,
2948      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
2949      whether the bitmap is present, and whether this is a negated class or not.      whether the bitmap is present, and whether this is a negated class or not.
     */  
2950    
2951      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
2952        default (Perl) mode, it is treated as a data character. */
2953    
2954        case CHAR_RIGHT_SQUARE_BRACKET:
2955        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2956          {
2957          *errorcodeptr = ERR64;
2958          goto FAILED;
2959          }
2960        goto NORMAL_CHAR;
2961    
2962        case CHAR_LEFT_SQUARE_BRACKET:
2963      previous = code;      previous = code;
2964    
2965      /* 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
2966      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. */
2967    
2968      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2969             ptr[1] == CHAR_EQUALS_SIGN) &&
2970          check_posix_syntax(ptr, &tempptr))          check_posix_syntax(ptr, &tempptr))
2971        {        {
2972        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
2973        goto FAILED;        goto FAILED;
2974        }        }
2975    
# Line 2653  for (;; ptr++) Line 2981  for (;; ptr++)
2981      for (;;)      for (;;)
2982        {        {
2983        c = *(++ptr);        c = *(++ptr);
2984        if (c == '\\')        if (c == CHAR_BACKSLASH)
2985          {          {
2986          if (ptr[1] == 'E') ptr++;          if (ptr[1] == CHAR_E)
2987            else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;            ptr++;
2988              else break;          else if (strncmp((const char *)ptr+1,
2989                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
2990              ptr += 3;
2991            else
2992              break;
2993          }          }
2994        else if (!negate_class && c == '^')        else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
2995          negate_class = TRUE;          negate_class = TRUE;
2996        else break;        else break;
2997        }        }
2998    
2999        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
3000        an initial ']' is taken as a data character -- the code below handles
3001        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3002        [^] must match any character, so generate OP_ALLANY. */
3003    
3004        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3005            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3006          {
3007          *code++ = negate_class? OP_ALLANY : OP_FAIL;
3008          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3009          zerofirstbyte = firstbyte;
3010          break;
3011          }
3012    
3013      /* If a class contains a negative special such as \S, we need to flip the      /* If a class contains a negative special such as \S, we need to flip the
3014      negation flag at the end, so that support for characters > 255 works      negation flag at the end, so that support for characters > 255 works
3015      correctly (they are all included in the class). */      correctly (they are all included in the class). */
# Line 2721  for (;; ptr++) Line 3067  for (;; ptr++)
3067    
3068        if (inescq)        if (inescq)
3069          {          {
3070          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3071            {            {
3072            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
3073            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2736  for (;; ptr++) Line 3082  for (;; ptr++)
3082        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3083        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3084    
3085        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3086            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3087            check_posix_syntax(ptr, &tempptr))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3088          {          {
3089          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3090          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3091          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3092          uschar pbits[32];          uschar pbits[32];
3093    
3094          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3095            {            {
3096            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3097            goto FAILED;            goto FAILED;
3098            }            }
3099    
3100          ptr += 2;          ptr += 2;
3101          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3102            {            {
3103            local_negate = TRUE;            local_negate = TRUE;
3104            should_flip_negation = TRUE;  /* Note negative special */            should_flip_negation = TRUE;  /* Note negative special */
# Line 2825  for (;; ptr++) Line 3171  for (;; ptr++)
3171        to 'or' into the one we are building. We assume they have more than one        to 'or' into the one we are building. We assume they have more than one
3172        character in them, so set class_charcount bigger than one. */        character in them, so set class_charcount bigger than one. */
3173    
3174        if (c == '\\')        if (c == CHAR_BACKSLASH)
3175          {          {
3176          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3177          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3178    
3179          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 */
3180          else if (-c == ESC_X) c = 'X';   /* \X is literal X in a class */          else if (-c == ESC_X) c = CHAR_X;   /* \X is literal X in a class */
3181          else if (-c == ESC_R) c = 'R';   /* \R is literal R in a class */          else if (-c == ESC_R) c = CHAR_R;   /* \R is literal R in a class */
3182          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3183            {            {
3184            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3185              {              {
3186              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3187              }              }
# Line 3061  for (;; ptr++) Line 3407  for (;; ptr++)
3407        entirely. The code for handling \Q and \E is messy. */        entirely. The code for handling \Q and \E is messy. */
3408    
3409        CHECK_RANGE:        CHECK_RANGE:
3410        while (ptr[1] == '\\' && ptr[2] == 'E')        while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3411          {          {
3412          inescq = FALSE;          inescq = FALSE;
3413          ptr += 2;          ptr += 2;
# Line 3071  for (;; ptr++) Line 3417  for (;; ptr++)
3417    
3418        /* Remember \r or \n */        /* Remember \r or \n */
3419    
3420        if (c == '\r' || c == '\n') cd->external_flags |= PCRE_HASCRORLF;        if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3421    
3422        /* Check for range */        /* Check for range */
3423    
3424        if (!inescq && ptr[1] == '-')        if (!inescq && ptr[1] == CHAR_MINUS)
3425          {          {
3426          int d;          int d;
3427          ptr += 2;          ptr += 2;
3428          while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3429    
3430          /* 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
3431          mode. */          mode. */
3432    
3433          while (*ptr == '\\' && ptr[1] == 'Q')          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3434            {            {
3435            ptr += 2;            ptr += 2;
3436            if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }            if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3437                { ptr += 2; continue; }
3438            inescq = TRUE;            inescq = TRUE;
3439            break;            break;
3440            }            }
3441    
3442          if (*ptr == 0 || (!inescq && *ptr == ']'))          if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3443            {            {
3444            ptr = oldptr;            ptr = oldptr;
3445            goto LONE_SINGLE_CHARACTER;            goto LONE_SINGLE_CHARACTER;
# Line 3111  for (;; ptr++) Line 3458  for (;; ptr++)
3458          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
3459          in such circumstances. */          in such circumstances. */
3460    
3461          if (!inescq && d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3462            {            {
3463            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3464            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
# Line 3121  for (;; ptr++) Line 3468  for (;; ptr++)
3468    
3469            if (d < 0)            if (d < 0)
3470              {              {
3471              if (d == -ESC_b) d = '\b';              if (d == -ESC_b) d = CHAR_BS;
3472              else if (d == -ESC_X) d = 'X';              else if (d == -ESC_X) d = CHAR_X;
3473              else if (d == -ESC_R) d = 'R'; else              else if (d == -ESC_R) d = CHAR_R; else
3474                {                {
3475                ptr = oldptr;                ptr = oldptr;
3476                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 3144  for (;; ptr++) Line 3491  for (;; ptr++)
3491    
3492          /* Remember \r or \n */          /* Remember \r or \n */
3493    
3494          if (d == '\r' || d == '\n') cd->external_flags |= PCRE_HASCRORLF;          if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3495    
3496          /* 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
3497          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 3264  for (;; ptr++) Line 3611  for (;; ptr++)
3611          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3612            {            {
3613            unsigned int othercase;            unsigned int othercase;
3614            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)            if ((othercase = UCD_OTHERCASE(c)) != c)
3615              {              {
3616              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3617              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 3291  for (;; ptr++) Line 3638  for (;; ptr++)
3638    
3639      /* Loop until ']' reached. This "while" is the end of the "do" above. */      /* Loop until ']' reached. This "while" is the end of the "do" above. */
3640    
3641      while ((c = *(++ptr)) != 0 && (c != ']' || inescq));      while ((c = *(++ptr)) != 0 && (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
3642    
3643      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)                          /* Missing terminating ']' */
3644        {        {
# Line 3436  we set the flag only if there is a liter Line 3783  we set the flag only if there is a liter
3783      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
3784      has been tested above. */      has been tested above. */
3785    
3786      case '{':      case CHAR_LEFT_CURLY_BRACKET:
3787      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
3788      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
3789      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
3790      goto REPEAT;      goto REPEAT;
3791    
3792      case '*':      case CHAR_ASTERISK:
3793      repeat_min = 0;      repeat_min = 0;
3794      repeat_max = -1;      repeat_max = -1;
3795      goto REPEAT;      goto REPEAT;
3796    
3797      case '+':      case CHAR_PLUS:
3798      repeat_min = 1;      repeat_min = 1;
3799      repeat_max = -1;      repeat_max = -1;
3800      goto REPEAT;      goto REPEAT;
3801    
3802      case '?':      case CHAR_QUESTION_MARK:
3803      repeat_min = 0;      repeat_min = 0;
3804      repeat_max = 1;      repeat_max = 1;
3805    
# Line 3487  we set the flag only if there is a liter Line 3834  we set the flag only if there is a liter
3834      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
3835      repeat type to the non-default. */      repeat type to the non-default. */
3836    
3837      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
3838        {        {
3839        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
3840        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
3841        ptr++;        ptr++;
3842        }        }
3843      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
3844        {        {
3845        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
3846        ptr++;        ptr++;
# Line 3608  we set the flag only if there is a liter Line 3955  we set the flag only if there is a liter
3955    
3956        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
3957    
3958          /*--------------------------------------------------------------------*/
3959          /* This code is obsolete from release 8.00; the restriction was finally
3960          removed: */
3961    
3962        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
3963        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
3964    
3965        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
3966          /*--------------------------------------------------------------------*/
3967    
3968        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
3969    
# Line 3758  we set the flag only if there is a liter Line 4110  we set the flag only if there is a liter
4110          goto END_REPEAT;          goto END_REPEAT;
4111          }          }
4112    
4113          /*--------------------------------------------------------------------*/
4114          /* This code is obsolete from release 8.00; the restriction was finally
4115          removed: */
4116    
4117        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4118        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4119    
4120        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4121          /*--------------------------------------------------------------------*/
4122    
4123        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4124          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 3819  we set the flag only if there is a liter Line 4176  we set the flag only if there is a liter
4176    
4177        if (repeat_min == 0)        if (repeat_min == 0)
4178          {          {
4179          /* If the maximum is also zero, we just omit the group from the output          /* If the maximum is also zero, we used to just omit the group from the
4180          altogether. */          output altogether, like this:
4181    
4182          if (repeat_max == 0)          ** if (repeat_max == 0)
4183            {          **   {
4184            code = previous;          **   code = previous;
4185            goto END_REPEAT;          **   goto END_REPEAT;
4186            }          **   }
4187    
4188            However, that fails when a group is referenced as a subroutine from
4189            elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it
4190            so that it is skipped on execution. As we don't have a list of which
4191            groups are referenced, we cannot do this selectively.
4192    
4193            If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4194            and do no more at this point. However, we do need to adjust any
4195            OP_RECURSE calls inside the group that refer to the group itself or any
4196            internal or forward referenced group, because the offset is from the
4197            start of the whole regex. Temporarily terminate the pattern while doing
4198            this. */
4199    
4200          /* If the maximum is 1 or unlimited, we just have to stick in the          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
         BRAZERO and do no more at this point. However, we do need to adjust  
         any OP_RECURSE calls inside the group that refer to the group itself or  
         any internal or forward referenced group, because the offset is from  
         the start of the whole regex. Temporarily terminate the pattern while  
         doing this. */  
   
         if (repeat_max <= 1)  
4201            {            {
4202            *code = OP_END;            *code = OP_END;
4203            adjust_recurse(previous, 1, utf8, cd, save_hwm);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
4204            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
4205            code++;            code++;
4206              if (repeat_max == 0)
4207                {
4208                *previous++ = OP_SKIPZERO;
4209                goto END_REPEAT;
4210                }
4211            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4212            }            }
4213    
# Line 3886  we set the flag only if there is a liter Line 4253  we set the flag only if there is a liter
4253            {            {
4254            /* 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
4255            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
4256            potential integer overflow. */            potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
4257              integer type when available, otherwise double. */
4258    
4259            if (lengthptr != NULL)            if (lengthptr != NULL)
4260              {              {
4261              int delta = (repeat_min - 1)*length_prevgroup;              int delta = (repeat_min - 1)*length_prevgroup;
4262              if ((double)(repeat_min - 1)*(double)length_prevgroup >              if ((INT64_OR_DOUBLE)(repeat_min - 1)*
4263                                                              (double)INT_MAX ||                    (INT64_OR_DOUBLE)length_prevgroup >
4264                        (INT64_OR_DOUBLE)INT_MAX ||
4265                  OFLOW_MAX - *lengthptr < delta)                  OFLOW_MAX - *lengthptr < delta)
4266                {                {
4267                *errorcodeptr = ERR20;                *errorcodeptr = ERR20;
# Line 3938  we set the flag only if there is a liter Line 4307  we set the flag only if there is a liter
4307          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
4308          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
4309          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
4310          paranoid checks to avoid integer overflow. */          paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
4311            a 64-bit integer type when available, otherwise double. */
4312    
4313          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4314            {            {
4315            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4316                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4317            if ((double)repeat_max *            if ((INT64_OR_DOUBLE)repeat_max *
4318                  (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)                  (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4319                    > (double)INT_MAX ||                    > (INT64_OR_DOUBLE)INT_MAX ||
4320                OFLOW_MAX - *lengthptr < delta)                OFLOW_MAX - *lengthptr < delta)
4321              {              {
4322              *errorcodeptr = ERR20;              *errorcodeptr = ERR20;
# Line 4023  we set the flag only if there is a liter Line 4393  we set the flag only if there is a liter
4393            uschar *scode = bracode;            uschar *scode = bracode;
4394            do            do
4395              {              {
4396              if (could_be_empty_branch(scode, ketcode, utf8))              if (could_be_empty_branch(scode, ketcode, utf8, cd))
4397                {                {
4398                *bracode += OP_SBRA - OP_BRA;                *bracode += OP_SBRA - OP_BRA;
4399                break;                break;
# Line 4035  we set the flag only if there is a liter Line 4405  we set the flag only if there is a liter
4405          }          }
4406        }        }
4407    
4408        /* If previous is OP_FAIL, it was generated by an empty class [] in
4409        JavaScript mode. The other ways in which OP_FAIL can be generated, that is
4410        by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
4411        error above. We can just ignore the repeat in JS case. */
4412    
4413        else if (*previous == OP_FAIL) goto END_REPEAT;
4414    
4415      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
4416    
4417      else      else
# Line 4059  we set the flag only if there is a liter Line 4436  we set the flag only if there is a liter
4436      if (possessive_quantifier)      if (possessive_quantifier)
4437        {        {
4438        int len;        int len;
4439        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||  
4440            *tempcode == OP_NOTEXACT)        if (*tempcode == OP_TYPEEXACT)
4441          tempcode += _pcre_OP_lengths[*tempcode] +          tempcode += _pcre_OP_lengths[*tempcode] +
4442            ((*tempcode == OP_TYPEEXACT &&            ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
4443               (tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP))? 2:0);  
4444          else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
4445            {
4446            tempcode += _pcre_OP_lengths[*tempcode];
4447    #ifdef SUPPORT_UTF8
4448            if (utf8 && tempcode[-1] >= 0xc0)
4449              tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];
4450    #endif
4451            }
4452    
4453        len = code - tempcode;        len = code - tempcode;
4454        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4455          {          {
# Line 4082  we set the flag only if there is a liter Line 4468  we set the flag only if there is a liter
4468          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4469          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4470    
4471            /* Because we are moving code along, we must ensure that any
4472            pending recursive references are updated. */
4473    
4474          default:          default:
4475            *code = OP_END;
4476            adjust_recurse(tempcode, 1 + LINK_SIZE, utf8, cd, save_hwm);
4477          memmove(tempcode + 1+LINK_SIZE, tempcode, len);          memmove(tempcode + 1+LINK_SIZE, tempcode, len);
4478          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
4479          len += 1 + LINK_SIZE;          len += 1 + LINK_SIZE;
# Line 4109  we set the flag only if there is a liter Line 4500  we set the flag only if there is a liter
4500      lookbehind or option setting or condition or all the other extended      lookbehind or option setting or condition or all the other extended
4501      parenthesis forms.  */      parenthesis forms.  */
4502    
4503      case '(':      case CHAR_LEFT_PARENTHESIS:
4504      newoptions = options;      newoptions = options;
4505      skipbytes = 0;      skipbytes = 0;
4506      bravalue = OP_CBRA;      bravalue = OP_CBRA;
# Line 4118  we set the flag only if there is a liter Line 4509  we set the flag only if there is a liter
4509    
4510      /* First deal with various "verbs" that can be introduced by '*'. */      /* First deal with various "verbs" that can be introduced by '*'. */
4511    
4512      if (*(++ptr) == '*' && (cd->ctypes[ptr[1]] & ctype_letter) != 0)      if (*(++ptr) == CHAR_ASTERISK && (cd->ctypes[ptr[1]] & ctype_letter) != 0)
4513        {        {
4514        int i, namelen;        int i, namelen;
4515        const char *vn = verbnames;        const char *vn = verbnames;
4516        const uschar *name = ++ptr;        const uschar *name = ++ptr;
4517        previous = NULL;        previous = NULL;
4518        while ((cd->ctypes[*++ptr] & ctype_letter) != 0);        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
4519        if (*ptr == ':')        if (*ptr == CHAR_COLON)
4520          {          {
4521          *errorcodeptr = ERR59;   /* Not supported */          *errorcodeptr = ERR59;   /* Not supported */
4522          goto FAILED;          goto FAILED;
4523          }          }
4524        if (*ptr != ')')        if (*ptr != CHAR_RIGHT_PARENTHESIS)
4525          {          {
4526          *errorcodeptr = ERR60;          *errorcodeptr = ERR60;
4527          goto FAILED;          goto FAILED;
# Line 4141  we set the flag only if there is a liter Line 4532  we set the flag only if there is a liter
4532          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
4533              strncmp((char *)name, vn, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
4534            {            {
4535            *code = verbs[i].op;            /* Check for open captures before ACCEPT */
4536            if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;  
4537              if (verbs[i].op == OP_ACCEPT)
4538                {
4539                open_capitem *oc;
4540                cd->had_accept = TRUE;
4541                for (oc = cd->open_caps; oc != NULL; oc = oc->next)
4542                  {
4543                  *code++ = OP_CLOSE;
4544                  PUT2INC(code, 0, oc->number);
4545                  }
4546                }
4547              *code++ = verbs[i].op;
4548            break;            break;
4549            }            }
4550          vn += verbs[i].len + 1;          vn += verbs[i].len + 1;
# Line 4155  we set the flag only if there is a liter Line 4557  we set the flag only if there is a liter
4557      /* Deal with the extended parentheses; all are introduced by '?', and the      /* Deal with the extended parentheses; all are introduced by '?', and the
4558      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. */
4559    
4560      else if (*ptr == '?')      else if (*ptr == CHAR_QUESTION_MARK)
4561        {        {
4562        int i, set, unset, namelen;        int i, set, unset, namelen;
4563        int *optset;        int *optset;
# Line 4164  we set the flag only if there is a liter Line 4566  we set the flag only if there is a liter
4566    
4567        switch (*(++ptr))        switch (*(++ptr))
4568          {          {
4569          case '#':                 /* Comment; skip to ket */          case CHAR_NUMBER_SIGN:                 /* Comment; skip to ket */
4570          ptr++;          ptr++;
4571          while (*ptr != 0 && *ptr != ')') ptr++;          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
4572          if (*ptr == 0)          if (*ptr == 0)
4573            {            {
4574            *errorcodeptr = ERR18;            *errorcodeptr = ERR18;
# Line 4176  we set the flag only if there is a liter Line 4578  we set the flag only if there is a liter
4578    
4579    
4580          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4581          case '|':                 /* Reset capture count for each branch */          case CHAR_VERTICAL_LINE:  /* Reset capture count for each branch */
4582          reset_bracount = TRUE;          reset_bracount = TRUE;
4583          /* Fall through */          /* Fall through */
4584    
4585          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4586          case ':':                 /* Non-capturing bracket */          case CHAR_COLON:          /* Non-capturing bracket */
4587          bravalue = OP_BRA;          bravalue = OP_BRA;
4588          ptr++;          ptr++;
4589          break;          break;
4590    
4591    
4592          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4593          case '(':          case CHAR_LEFT_PARENTHESIS:
4594          bravalue = OP_COND;       /* Conditional group */          bravalue = OP_COND;       /* Conditional group */
4595    
4596          /* 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 4208  we set the flag only if there is a liter Line 4610  we set the flag only if there is a liter
4610          the switch. This will take control down to where bracketed groups,          the switch. This will take control down to where bracketed groups,
4611          including assertions, are processed. */          including assertions, are processed. */
4612    
4613          if (ptr[1] == '?' && (ptr[2] == '=' || ptr[2] == '!' || ptr[2] == '<'))          if (ptr[1] == CHAR_QUESTION_MARK && (ptr[2] == CHAR_EQUALS_SIGN ||
4614                ptr[2] == CHAR_EXCLAMATION_MARK || ptr[2] == CHAR_LESS_THAN_SIGN))
4615            break;            break;
4616    
4617          /* 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 4220  we set the flag only if there is a liter Line 4623  we set the flag only if there is a liter
4623    
4624          /* Check for a test for recursion in a named group. */          /* Check for a test for recursion in a named group. */
4625    
4626          if (ptr[1] == 'R' && ptr[2] == '&')          if (ptr[1] == CHAR_R && ptr[2] == CHAR_AMPERSAND)
4627            {            {
4628            terminator = -1;            terminator = -1;
4629            ptr += 2;            ptr += 2;
# Line 4230  we set the flag only if there is a liter Line 4633  we set the flag only if there is a liter
4633          /* 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
4634          syntax (?(<name>) or (?('name') */          syntax (?(<name>) or (?('name') */
4635    
4636          else if (ptr[1] == '<')          else if (ptr[1] == CHAR_LESS_THAN_SIGN)
4637            {            {
4638            terminator = '>';            terminator = CHAR_GREATER_THAN_SIGN;
4639            ptr++;            ptr++;
4640            }            }
4641          else if (ptr[1] == '\'')          else if (ptr[1] == CHAR_APOSTROPHE)
4642            {            {
4643            terminator = '\'';            terminator = CHAR_APOSTROPHE;
4644            ptr++;            ptr++;
4645            }            }
4646          else          else
4647            {            {
4648            terminator = 0;            terminator = 0;
4649            if (ptr[1] == '-' || ptr[1] == '+') refsign = *(++ptr);            if (ptr[1] == CHAR_MINUS || ptr[1] == CHAR_PLUS) refsign = *(++ptr);
4650            }            }
4651    
4652          /* 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 4263  we set the flag only if there is a liter Line 4666  we set the flag only if there is a liter
4666            {            {
4667            if (recno >= 0)            if (recno >= 0)
4668              recno = ((digitab[*ptr] & ctype_digit) != 0)?              recno = ((digitab[*ptr] & ctype_digit) != 0)?
4669                recno * 10 + *ptr - '0' : -1;                recno * 10 + *ptr - CHAR_0 : -1;
4670            ptr++;            ptr++;
4671            }            }
4672          namelen = ptr - name;          namelen = ptr - name;
4673    
4674          if ((terminator > 0 && *ptr++ != terminator) || *ptr++ != ')')          if ((terminator > 0 && *ptr++ != terminator) ||
4675                *ptr++ != CHAR_RIGHT_PARENTHESIS)
4676            {            {
4677            ptr--;      /* Error offset */            ptr--;      /* Error offset */
4678            *errorcodeptr = ERR26;            *errorcodeptr = ERR26;
# Line 4290  we set the flag only if there is a liter Line 4694  we set the flag only if there is a liter
4694              *errorcodeptr = ERR58;              *errorcodeptr = ERR58;
4695              goto FAILED;              goto FAILED;
4696              }              }
4697            recno = (refsign == '-')?            recno = (refsign == CHAR_MINUS)?
4698              cd->bracount - recno + 1 : recno +cd->bracount;              cd->bracount - recno + 1 : recno +cd->bracount;
4699            if (recno <= 0 || recno > cd->final_bracount)            if (recno <= 0 || recno > cd->final_bracount)
4700              {              {
# Line 4302  we set the flag only if there is a liter Line 4706  we set the flag only if there is a liter
4706            }            }
4707    
4708          /* Otherwise (did not start with "+" or "-"), start by looking for the          /* Otherwise (did not start with "+" or "-"), start by looking for the
4709          name. */          name. If we find a name, add one to the opcode to change OP_CREF or
4710            OP_RREF into OP_NCREF or OP_NRREF. These behave exactly the same,
4711            except they record that the reference was originally to a name. The
4712            information is used to check duplicate names. */
4713    
4714          slot = cd->name_table;          slot = cd->name_table;
4715          for (i = 0; i < cd->names_found; i++)          for (i = 0; i < cd->names_found; i++)
# Line 4317  we set the flag only if there is a liter Line 4724  we set the flag only if there is a liter
4724            {            {
4725            recno = GET2(slot, 0);            recno = GET2(slot, 0);
4726            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4727              code[1+LINK_SIZE]++;
4728            }            }
4729    
4730          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
4731    
4732          else if ((i = find_parens(ptr, cd->bracount, name, namelen,          else if ((i = find_parens(cd, name, namelen,
4733                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0)) > 0)
4734            {            {
4735            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
4736              code[1+LINK_SIZE]++;
4737            }            }
4738    
4739          /* If terminator == 0 it means that the name followed directly after          /* If terminator == 0 it means that the name followed directly after
# Line 4342  we set the flag only if there is a liter Line 4751  we set the flag only if there is a liter
4751          /* Check for (?(R) for recursion. Allow digits after R to specify a          /* Check for (?(R) for recursion. Allow digits after R to specify a
4752          specific group number. */          specific group number. */
4753    
4754          else if (*name == 'R')          else if (*name == CHAR_R)
4755            {            {
4756            recno = 0;            recno = 0;
4757            for (i = 1; i < namelen; i++)            for (i = 1; i < namelen; i++)
# Line 4352  we set the flag only if there is a liter Line 4761  we set the flag only if there is a liter
4761                *errorcodeptr = ERR15;                *errorcodeptr = ERR15;
4762                goto FAILED;                goto FAILED;
4763                }                }
4764              recno = recno * 10 + name[i] - '0';              recno = recno * 10 + name[i] - CHAR_0;
4765              }              }
4766            if (recno == 0) recno = RREF_ANY;            if (recno == 0) recno = RREF_ANY;
4767            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */
# Line 4362  we set the flag only if there is a liter Line 4771  we set the flag only if there is a liter
4771          /* Similarly, check for the (?(DEFINE) "condition", which is always          /* Similarly, check for the (?(DEFINE) "condition", which is always
4772          false. */          false. */
4773    
4774          else if (namelen == 6 && strncmp((char *)name, "DEFINE", 6) == 0)          else if (namelen == 6 && strncmp((char *)name, STRING_DEFINE, 6) == 0)
4775            {            {
4776            code[1+LINK_SIZE] = OP_DEF;            code[1+LINK_SIZE] = OP_DEF;
4777            skipbytes = 1;            skipbytes = 1;
# Line 4387  we set the flag only if there is a liter Line 4796  we set the flag only if there is a liter
4796    
4797    
4798          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4799          case '=':                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
4800          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
4801          ptr++;          ptr++;
4802          break;          break;
4803    
4804    
4805          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4806          case '!':                 /* Negative lookahead */          case CHAR_EXCLAMATION_MARK:            /* Negative lookahead */
4807          ptr++;          ptr++;
4808          if (*ptr == ')')          /* Optimize (?!) */          if (*ptr == CHAR_RIGHT_PARENTHESIS)    /* Optimize (?!) */
4809            {            {
4810            *code++ = OP_FAIL;            *code++ = OP_FAIL;
4811            previous = NULL;            previous = NULL;
# Line 4407  we set the flag only if there is a liter Line 4816  we set the flag only if there is a liter
4816    
4817    
4818          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4819          case '<':                 /* Lookbehind or named define */          case CHAR_LESS_THAN_SIGN:              /* Lookbehind or named define */
4820          switch (ptr[1])          switch (ptr[1])
4821            {            {
4822            case '=':               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
4823            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
4824            ptr += 2;            ptr += 2;
4825            break;            break;
4826    
4827            case '!':               /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
4828            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
4829            ptr += 2;            ptr += 2;
4830            break;            break;
# Line 4430  we set the flag only if there is a liter Line 4839  we set the flag only if there is a liter
4839    
4840    
4841          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4842          case '>':                 /* One-time brackets */          case CHAR_GREATER_THAN_SIGN:           /* One-time brackets */
4843          bravalue = OP_ONCE;          bravalue = OP_ONCE;
4844          ptr++;          ptr++;
4845          break;          break;
4846    
4847    
4848          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4849          case 'C':                 /* Callout - may be followed by digits; */          case CHAR_C:                 /* Callout - may be followed by digits; */
4850          previous_callout = code;  /* Save for later completion */          previous_callout = code;  /* Save for later completion */
4851          after_manual_callout = 1; /* Skip one item before completing */          after_manual_callout = 1; /* Skip one item before completing */
4852          *code++ = OP_CALLOUT;          *code++ = OP_CALLOUT;
4853            {            {
4854            int n = 0;            int n = 0;
4855            while ((digitab[*(++ptr)] & ctype_digit) != 0)            while ((digitab[*(++ptr)] & ctype_digit) != 0)
4856              n = n * 10 + *ptr - '0';              n = n * 10 + *ptr - CHAR_0;
4857            if (*ptr != ')')            if (*ptr != CHAR_RIGHT_PARENTHESIS)
4858              {              {
4859              *errorcodeptr = ERR39;              *errorcodeptr = ERR39;
4860              goto FAILED;              goto FAILED;
# Line 4465  we set the flag only if there is a liter Line 4874  we set the flag only if there is a liter
4874    
4875    
4876          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4877          case 'P':                 /* Python-style named subpattern handling */          case CHAR_P:              /* Python-style named subpattern handling */
4878          if (*(++ptr) == '=' || *ptr == '>')  /* Reference or recursion */          if (*(++ptr) == CHAR_EQUALS_SIGN ||
4879                *ptr == CHAR_GREATER_THAN_SIGN)  /* Reference or recursion */
4880            {            {
4881            is_recurse = *ptr == '>';            is_recurse = *ptr == CHAR_GREATER_THAN_SIGN;
4882            terminator = ')';            terminator = CHAR_RIGHT_PARENTHESIS;
4883            goto NAMED_REF_OR_RECURSE;            goto NAMED_REF_OR_RECURSE;
4884            }            }
4885          else if (*ptr != '<')    /* Test for Python-style definition */          else if (*ptr != CHAR_LESS_THAN_SIGN)  /* Test for Python-style defn */
4886            {            {
4887            *errorcodeptr = ERR41;            *errorcodeptr = ERR41;
4888            goto FAILED;            goto FAILED;
# Line 4482  we set the flag only if there is a liter Line 4892  we set the flag only if there is a liter
4892    
4893          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4894          DEFINE_NAME:    /* Come here from (?< handling */          DEFINE_NAME:    /* Come here from (?< handling */
4895          case '\'':          case CHAR_APOSTROPHE:
4896            {            {
4897            terminator = (*ptr == '<')? '>' : '\'';            terminator = (*ptr == CHAR_LESS_THAN_SIGN)?
4898                CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
4899            name = ++ptr;            name = ++ptr;
4900    
4901            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
# Line 4515  we set the flag only if there is a liter Line 4926  we set the flag only if there is a liter
4926                }                }
4927              }              }
4928    
4929            /* In the real compile, create the entry in the table */            /* In the real compile, create the entry in the table, maintaining
4930              alphabetical order. Duplicate names for different numbers are
4931              permitted only if PCRE_DUPNAMES is set. Duplicate names for the same
4932              number are always OK. (An existing number can be re-used if (?|
4933              appears in the pattern.) In either event, a duplicate name results in
4934              a duplicate entry in the table, even if the number is the same. This
4935              is because the number of names, and hence the table size, is computed
4936              in the pre-compile, and it affects various numbers and pointers which
4937              would all have to be modified, and the compiled code moved down, if
4938              duplicates with the same number were omitted from the table. This
4939              doesn't seem worth the hassle. However, *different* names for the
4940              same number are not permitted. */
4941    
4942            else            else
4943              {              {
4944                BOOL dupname = FALSE;
4945              slot = cd->name_table;              slot = cd->name_table;
4946    
4947              for (i = 0; i < cd->names_found; i++)              for (i = 0; i < cd->names_found; i++)
4948                {                {
4949                int crc = memcmp(name, slot+2, namelen);                int crc = memcmp(name, slot+2, namelen);
# Line 4527  we set the flag only if there is a liter Line 4951  we set the flag only if there is a liter
4951                  {                  {
4952                  if (slot[2+namelen] == 0)                  if (slot[2+namelen] == 0)
4953                    {                    {
4954                    if ((options & PCRE_DUPNAMES) == 0)                    if (GET2(slot, 0) != cd->bracount + 1 &&
4955                          (options & PCRE_DUPNAMES) == 0)
4956                      {                      {
4957                      *errorcodeptr = ERR43;                      *errorcodeptr = ERR43;
4958                      goto FAILED;                      goto FAILED;
4959                      }                      }
4960                      else dupname = TRUE;
4961                    }                    }
4962                  else crc = -1;      /* Current name is substring */                  else crc = -1;      /* Current name is a substring */
4963                  }                  }
4964    
4965                  /* Make space in the table and break the loop for an earlier
4966                  name. For a duplicate or later name, carry on. We do this for
4967                  duplicates so that in the simple case (when ?(| is not used) they
4968                  are in order of their numbers. */
4969    
4970                if (crc < 0)                if (crc < 0)
4971                  {                  {
4972                  memmove(slot + cd->name_entry_size, slot,                  memmove(slot + cd->name_entry_size, slot,
4973                    (cd->names_found - i) * cd->name_entry_size);                    (cd->names_found - i) * cd->name_entry_size);
4974                  break;                  break;
4975                  }                  }
4976    
4977                  /* Continue the loop for a later or duplicate name */
4978    
4979                slot += cd->name_entry_size;                slot += cd->name_entry_size;
4980                }                }
4981    
4982                /* For non-duplicate names, check for a duplicate number before
4983                adding the new name. */
4984    
4985                if (!dupname)
4986                  {
4987                  uschar *cslot = cd->name_table;
4988                  for (i = 0; i < cd->names_found; i++)
4989                    {
4990                    if (cslot != slot)
4991                      {
4992                      if (GET2(cslot, 0) == cd->bracount + 1)
4993                        {
4994                        *errorcodeptr = ERR65;
4995                        goto FAILED;
4996                        }
4997                      }
4998                    else i--;
4999                    cslot += cd->name_entry_size;
5000                    }
5001                  }
5002    
5003              PUT2(slot, 0, cd->bracount + 1);              PUT2(slot, 0, cd->bracount + 1);
5004              memcpy(slot + 2, name, namelen);              memcpy(slot + 2, name, namelen);
5005              slot[2+namelen] = 0;              slot[2+namelen] = 0;
5006              }              }
5007            }            }
5008    
5009          /* In both cases, count the number of names we've encountered. */          /* In both pre-compile and compile, count the number of names we've
5010            encountered. */
5011    
         ptr++;                    /* Move past > or ' */  
5012          cd->names_found++;          cd->names_found++;
5013            ptr++;                    /* Move past > or ' */
5014          goto NUMBERED_GROUP;          goto NUMBERED_GROUP;
5015    
5016    
5017          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5018          case '&':                 /* Perl recursion/subroutine syntax */          case CHAR_AMPERSAND:            /* Perl recursion/subroutine syntax */
5019          terminator = ')';          terminator = CHAR_RIGHT_PARENTHESIS;
5020          is_recurse = TRUE;          is_recurse = TRUE;
5021          /* Fall through */          /* Fall through */
5022    
# Line 4567  we set the flag only if there is a liter Line 5024  we set the flag only if there is a liter
5024          references (?P=name) and recursion (?P>name), as well as falling          references (?P=name) and recursion (?P>name), as well as falling
5025          through from the Perl recursion syntax (?&name). We also come here from          through from the Perl recursion syntax (?&name). We also come here from
5026          the Perl \k<name> or \k'name' back reference syntax and the \k{name}          the Perl \k<name> or \k'name' back reference syntax and the \k{name}
5027          .NET syntax. */          .NET syntax, and the Oniguruma \g<...> and \g'...' subroutine syntax. */
5028    
5029          NAMED_REF_OR_RECURSE:          NAMED_REF_OR_RECURSE:
5030          name = ++ptr;          name = ++ptr;
# Line 4618  we set the flag only if there is a liter Line 5075  we set the flag only if there is a liter
5075              recno = GET2(slot, 0);              recno = GET2(slot, 0);
5076              }              }
5077            else if ((recno =                /* Forward back reference */            else if ((recno =                /* Forward back reference */
5078                      find_parens(ptr, cd->bracount, name, namelen,                      find_parens(cd, name, namelen,
5079                        (options & PCRE_EXTENDED) != 0)) <= 0)                        (options & PCRE_EXTENDED) != 0)) <= 0)
5080              {              {
5081              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
# Line 4634  we set the flag only if there is a liter Line 5091  we set the flag only if there is a liter
5091    
5092    
5093          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5094          case 'R':                 /* Recursion */          case CHAR_R:              /* Recursion */
5095          ptr++;                    /* Same as (?0)      */          ptr++;                    /* Same as (?0)      */
5096          /* Fall through */          /* Fall through */
5097    
5098    
5099          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5100          case '-': case '+':          case CHAR_MINUS: case CHAR_PLUS:  /* Recursion or subroutine */
5101          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:
5102          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:
5103            {            {
5104            const uschar *called;            const uschar *called;
5105              terminator = CHAR_RIGHT_PARENTHESIS;
5106    
5107              /* Come here from the \g<...> and \g'...' code (Oniguruma
5108              compatibility). However, the syntax has been checked to ensure that
5109              the ... are a (signed) number, so that neither ERR63 nor ERR29 will
5110              be called on this path, nor with the jump to OTHER_CHAR_AFTER_QUERY
5111              ever be taken. */
5112    
5113            if ((refsign = *ptr) == '+')            HANDLE_NUMERICAL_RECURSION:
5114    
5115              if ((refsign = *ptr) == CHAR_PLUS)
5116              {              {
5117              ptr++;              ptr++;
5118              if ((digitab[*ptr] & ctype_digit) == 0)              if ((digitab[*ptr] & ctype_digit) == 0)
# Line 4655  we set the flag only if there is a liter Line 5121  we set the flag only if there is a liter
5121                goto FAILED;                goto FAILED;
5122                }                }
5123              }              }
5124            else if (refsign == '-')            else if (refsign == CHAR_MINUS)
5125              {              {
5126              if ((digitab[ptr[1]] & ctype_digit) == 0)              if ((digitab[ptr[1]] & ctype_digit) == 0)
5127                goto OTHER_CHAR_AFTER_QUERY;                goto OTHER_CHAR_AFTER_QUERY;
# Line 4664  we set the flag only if there is a liter Line 5130  we set the flag only if there is a liter
5130    
5131            recno = 0;            recno = 0;
5132            while((digitab[*ptr] & ctype_digit) != 0)            while((digitab[*ptr] & ctype_digit) != 0)
5133              recno = recno * 10 + *ptr++ - '0';              recno = recno * 10 + *ptr++ - CHAR_0;
5134    
5135            if (*ptr != ')')            if (*ptr != terminator)
5136              {              {
5137              *errorcodeptr = ERR29;              *errorcodeptr = ERR29;
5138              goto FAILED;              goto FAILED;
5139              }              }
5140    
5141            if (refsign == '-')            if (refsign == CHAR_MINUS)
5142              {              {
5143              if (recno == 0)              if (recno == 0)
5144                {                {
# Line 4686  we set the flag only if there is a liter Line 5152  we set the flag only if there is a liter
5152                goto FAILED;                goto FAILED;
5153                }                }
5154              }              }
5155            else if (refsign == '+')            else if (refsign == CHAR_PLUS)
5156              {              {
5157              if (recno == 0)              if (recno == 0)
5158                {                {
# Line 4713  we set the flag only if there is a liter Line 5179  we set the flag only if there is a liter
5179            if (lengthptr == NULL)            if (lengthptr == NULL)
5180              {              {
5181              *code = OP_END;              *code = OP_END;
5182              if (recno != 0) called = find_bracket(cd->start_code, utf8, recno);              if (recno != 0)
5183                  called = _pcre_find_bracket(cd->start_code, utf8, recno);
5184    
5185              /* Forward reference */              /* Forward reference */
5186    
5187              if (called == NULL)              if (called == NULL)
5188                {                {
5189                if (find_parens(ptr, cd->bracount, NULL, recno,                if (find_parens(cd, NULL, recno,
5190                     (options & PCRE_EXTENDED) != 0) < 0)                      (options & PCRE_EXTENDED) != 0) < 0)
5191                  {                  {
5192                  *errorcodeptr = ERR15;                  *errorcodeptr = ERR15;
5193                  goto FAILED;                  goto FAILED;
5194                  }                  }
5195    
5196                  /* Fudge the value of "called" so that when it is inserted as an
5197                  offset below, what it actually inserted is the reference number
5198                  of the group. */
5199    
5200                called = cd->start_code + recno;                called = cd->start_code + recno;
5201                PUTINC(cd->hwm, 0, code + 2 + LINK_SIZE - cd->start_code);                PUTINC(cd->hwm, 0, code + 2 + LINK_SIZE - cd->start_code);
5202                }                }
# Line 4734  we set the flag only if there is a liter Line 5206  we set the flag only if there is a liter
5206              recursion that could loop for ever, and diagnose that case. */              recursion that could loop for ever, and diagnose that case. */
5207    
5208              else if (GET(called, 1) == 0 &&              else if (GET(called, 1) == 0 &&
5209                       could_be_empty(called, code, bcptr, utf8))                       could_be_empty(called, code, bcptr, utf8, cd))
5210                {                {
5211                *errorcodeptr = ERR40;                *errorcodeptr = ERR40;
5212                goto FAILED;                goto FAILED;
# Line 4772  we set the flag only if there is a liter Line 5244  we set the flag only if there is a liter
5244          set = unset = 0;          set = unset = 0;
5245          optset = &set;          optset = &set;
5246    
5247          while (*ptr != ')' && *ptr != ':')          while (*ptr != CHAR_RIGHT_PARENTHESIS && *ptr != CHAR_COLON)
5248            {            {
5249            switch (*ptr++)            switch (*ptr++)
5250              {              {
5251              case '-': optset = &unset; break;              case CHAR_MINUS: optset = &unset; break;
5252    
5253              case 'J':    /* Record that it changed in the external options */              case CHAR_J:    /* Record that it changed in the external options */
5254              *optset |= PCRE_DUPNAMES;              *optset |= PCRE_DUPNAMES;
5255              cd->external_flags |= PCRE_JCHANGED;              cd->external_flags |= PCRE_JCHANGED;
5256              break;              break;
5257    
5258              case 'i': *optset |= PCRE_CASELESS; break;              case CHAR_i: *optset |= PCRE_CASELESS; break;
5259              case 'm': *optset |= PCRE_MULTILINE; break;              case CHAR_m: *optset |= PCRE_MULTILINE; break;
5260              case 's': *optset |= PCRE_DOTALL; break;              case CHAR_s: *optset |= PCRE_DOTALL; break;
5261              case 'x': *optset |= PCRE_EXTENDED; break;              case CHAR_x: *optset |= PCRE_EXTENDED; break;
5262              case 'U': *optset |= PCRE_UNGREEDY; break;              case CHAR_U: *optset |= PCRE_UNGREEDY; break;
5263              case 'X': *optset |= PCRE_EXTRA; break;              case CHAR_X: *optset |= PCRE_EXTRA; break;
5264    
5265              default:  *errorcodeptr = ERR12;              default:  *errorcodeptr = ERR12;
5266                        ptr--;    /* Correct the offset */                        ptr--;    /* Correct the offset */
# Line 4819  we set the flag only if there is a liter Line 5291  we set the flag only if there is a liter
5291          both phases.          both phases.
5292    
5293          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
5294          options if this setting actually changes any of them. We also pass the          options if this setting actually changes any of them, and reset the
5295          new setting back so that it can be put at the start of any following          greedy defaults and the case value for firstbyte and reqbyte. */
         branches, and when this group ends (if we are in a group), a resetting  
         item can be compiled. */  
5296    
5297          if (*ptr == ')')          if (*ptr == CHAR_RIGHT_PARENTHESIS)
5298            {            {
5299            if (code == cd->start_code + 1 + LINK_SIZE &&            if (code == cd->start_code + 1 + LINK_SIZE &&
5300                 (lengthptr == NULL || *lengthptr == 2 + 2*LINK_SIZE))                 (lengthptr == NULL || *lengthptr == 2 + 2*LINK_SIZE))
5301              {              {
5302              cd->external_options = newoptions;              cd->external_options = newoptions;
             options = newoptions;  
5303              }              }
5304           else            else
5305              {              {
5306              if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))              if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))
5307                {                {
5308                *code++ = OP_OPT;                *code++ = OP_OPT;
5309                *code++ = newoptions & PCRE_IMS;                *code++ = newoptions & PCRE_IMS;
5310                }                }
   
             /* Change options at this level, and pass them back for use  
             in subsequent branches. Reset the greedy defaults and the case  
             value for firstbyte and reqbyte. */  
   
             *optionsptr = options = newoptions;  
5311              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);
5312              greedy_non_default = greedy_default ^ 1;              greedy_non_default = greedy_default ^ 1;
5313              req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
5314              }              }
5315    
5316              /* Change options at this level, and pass them back for use
5317              in subsequent branches. When not at the start of the pattern, this
5318              information is also necessary so that a resetting item can be
5319              compiled at the end of a group (if we are in a group). */
5320    
5321              *optionsptr = options = newoptions;
5322            previous = NULL;       /* This item can't be repeated */            previous = NULL;       /* This item can't be repeated */
5323            continue;              /* It is complete */            continue;              /* It is complete */
5324            }            }
# Line 4965  we set the flag only if there is a liter Line 5434  we set the flag only if there is a liter
5434    
5435      /* Error if hit end of pattern */      /* Error if hit end of pattern */
5436    
5437      if (*ptr != ')')      if (*ptr != CHAR_RIGHT_PARENTHESIS)
5438        {        {
5439        *errorcodeptr = ERR14;        *errorcodeptr = ERR14;
5440        goto FAILED;        goto FAILED;
# Line 5063  we set the flag only if there is a liter Line 5532  we set the flag only if there is a liter
5532      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
5533      have to change if any new ones are ever created. */      have to change if any new ones are ever created. */
5534    
5535      case '\\':      case CHAR_BACKSLASH:
5536      tempptr = ptr;      tempptr = ptr;
5537      c = check_escape(&ptr, errorcodeptr, cd->bracount, options, FALSE);      c = check_escape(&ptr, errorcodeptr, cd->bracount, options, FALSE);
5538      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
# Line 5072  we set the flag only if there is a liter Line 5541  we set the flag only if there is a liter
5541        {        {
5542        if (-c == ESC_Q)            /* Handle start of quoted string */        if (-c == ESC_Q)            /* Handle start of quoted string */
5543          {          {
5544          if (ptr[1] == '\\' && ptr[2] == 'E') ptr += 2; /* avoid empty string */          if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
5545            else inescq = TRUE;            ptr += 2;               /* avoid empty string */
5546                else inescq = TRUE;
5547          continue;          continue;
5548          }          }
5549    
# Line 5090  we set the flag only if there is a liter Line 5560  we set the flag only if there is a liter
5560        zerofirstbyte = firstbyte;        zerofirstbyte = firstbyte;
5561        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
5562    
5563          /* \g<name> or \g'name' is a subroutine call by name and \g<n> or \g'n'
5564          is a subroutine call by number (Oniguruma syntax). In fact, the value
5565          -ESC_g is returned only for these cases. So we don't need to check for <
5566          or ' if the value is -ESC_g. For the Perl syntax \g{n} the value is
5567          -ESC_REF+n, and for the Perl syntax \g{name} the result is -ESC_k (as
5568          that is a synonym for a named back reference). */
5569    
5570          if (-c == ESC_g)
5571            {
5572            const uschar *p;
5573            save_hwm = cd->hwm;   /* Normally this is set when '(' is read */
5574            terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
5575              CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
5576    
5577            /* These two statements stop the compiler for warning about possibly
5578            unset variables caused by the jump to HANDLE_NUMERICAL_RECURSION. In
5579            fact, because we actually check for a number below, the paths that
5580            would actually be in error are never taken. */
5581    
5582            skipbytes = 0;
5583            reset_bracount = FALSE;
5584    
5585            /* Test for a name */
5586    
5587            if (ptr[1] != CHAR_PLUS && ptr[1] != CHAR_MINUS)
5588              {
5589              BOOL isnumber = TRUE;
5590              for (p = ptr + 1; *p != 0 && *p != terminator; p++)
5591                {
5592                if ((cd->ctypes[*p] & ctype_digit) == 0) isnumber = FALSE;
5593                if ((cd->ctypes[*p] & ctype_word) == 0) break;
5594                }
5595              if (*p != terminator)
5596                {
5597                *errorcodeptr = ERR57;
5598                break;
5599                }
5600              if (isnumber)
5601                {
5602                ptr++;
5603                goto HANDLE_NUMERICAL_RECURSION;
5604                }
5605              is_recurse = TRUE;
5606              goto NAMED_REF_OR_RECURSE;
5607              }
5608    
5609            /* Test a signed number in angle brackets or quotes. */
5610    
5611            p = ptr + 2;
5612            while ((digitab[*p] & ctype_digit) != 0) p++;
5613            if (*p != terminator)
5614              {
5615              *errorcodeptr = ERR57;
5616              break;
5617              }
5618            ptr++;
5619            goto HANDLE_NUMERICAL_RECURSION;
5620            }
5621    
5622        /* \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).
5623        We also support \k{name} (.NET syntax) */        We also support \k{name} (.NET syntax) */
5624    
5625        if (-c == ESC_k && (ptr[1] == '<' || ptr[1] == '\'' || ptr[1] == '{'))        if (-c == ESC_k && (ptr[1] == CHAR_LESS_THAN_SIGN ||
5626              ptr[1] == CHAR_APOSTROPHE || ptr[1] == CHAR_LEFT_CURLY_BRACKET))
5627          {          {
5628          is_recurse = FALSE;          is_recurse = FALSE;
5629          terminator = (*(++ptr) == '<')? '>' : (*ptr == '\'')? '\'' : '}';          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
5630              CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?
5631              CHAR_APOSTROPHE : CHAR_RIGHT_CURLY_BRACKET;
5632          goto NAMED_REF_OR_RECURSE;          goto NAMED_REF_OR_RECURSE;
5633          }          }
5634    
# Line 5106  we set the flag only if there is a liter Line 5638  we set the flag only if there is a liter
5638    
5639        if (-c >= ESC_REF)        if (-c >= ESC_REF)
5640          {          {
5641            open_capitem *oc;
5642          recno = -c - ESC_REF;          recno = -c - ESC_REF;
5643    
5644          HANDLE_REFERENCE:    /* Come here from named backref handling */          HANDLE_REFERENCE:    /* Come here from named backref handling */
# Line 5115  we set the flag only if there is a liter Line 5648  we set the flag only if there is a liter
5648          PUT2INC(code, 0, recno);          PUT2INC(code, 0, recno);
5649          cd->backref_map |= (recno < 32)? (1 << recno) : 1;          cd->backref_map |= (recno < 32)? (1 << recno) : 1;
5650          if (recno > cd->top_backref) cd->top_backref = recno;          if (recno > cd->top_backref) cd->top_backref = recno;
5651    
5652            /* Check to see if this back reference is recursive, that it, it
5653            is inside the group that it references. A flag is set so that the
5654            group can be made atomic. */
5655    
5656            for (oc = cd->open_caps; oc != NULL; oc = oc->next)
5657              {
5658              if (oc->number == recno)
5659                {
5660                oc->flag = TRUE;
5661                break;
5662                }
5663              }
5664          }          }
5665    
5666        /* So are Unicode property matches, if supported. */        /* So are Unicode property matches, if supported. */
# Line 5199  we set the flag only if there is a liter Line 5745  we set the flag only if there is a liter
5745    
5746      /* Remember if \r or \n were seen */      /* Remember if \r or \n were seen */
5747    
5748      if (mcbuffer[0] == '\r' || mcbuffer[0] == '\n')      if (mcbuffer[0] == CHAR_CR || mcbuffer[0] == CHAR_NL)
5749        cd->external_flags |= PCRE_HASCRORLF;        cd->external_flags |= PCRE_HASCRORLF;
5750    
5751      /* Set the first and required bytes appropriately. If no previous first      /* Set the first and required bytes appropriately. If no previous first
# Line 5297  uschar *code = *codeptr; Line 5843  uschar *code = *codeptr;
5843  uschar *last_branch = code;  uschar *last_branch = code;
5844  uschar *start_bracket = code;  uschar *start_bracket = code;
5845  uschar *reverse_count = NULL;  uschar *reverse_count = NULL;
5846    open_capitem capitem;
5847    int capnumber = 0;
5848  int firstbyte, reqbyte;  int firstbyte, reqbyte;
5849  int branchfirstbyte, branchreqbyte;  int branchfirstbyte, branchreqbyte;
5850  int length;  int length;
5851  int orig_bracount;  int orig_bracount;
5852  int max_bracount;  int max_bracount;
5853    int old_external_options = cd->external_options;
5854  branch_chain bc;  branch_chain bc;
5855    
5856  bc.outer = bcptr;  bc.outer = bcptr;
5857  bc.current = code;  bc.current_branch = code;
5858    
5859  firstbyte = reqbyte = REQ_UNSET;  firstbyte = reqbyte = REQ_UNSET;
5860    
# Line 5323  the code that abstracts option settings Line 5872  the code that abstracts option settings
5872  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
5873  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. */
5874    
5875    /* If this is a capturing subpattern, add to the chain of open capturing items
5876    so that we can detect them if (*ACCEPT) is encountered. This is also used to
5877    detect groups that contain recursive back references to themselves. */
5878    
5879    if (*code == OP_CBRA)
5880      {
5881      capnumber = GET2(code, 1 + LINK_SIZE);
5882      capitem.number = capnumber;
5883      capitem.next = cd->open_caps;
5884      capitem.flag = FALSE;
5885      cd->open_caps = &capitem;
5886      }
5887    
5888  /* Offset is set zero to mark that this bracket is still open */  /* Offset is set zero to mark that this bracket is still open */
5889    
5890  PUT(code, 1, 0);  PUT(code, 1, 0);
# Line 5367  for (;;) Line 5929  for (;;)
5929      return FALSE;      return FALSE;
5930      }      }
5931    
5932      /* If the external options have changed during this branch, it means that we
5933      are at the top level, and a leading option setting has been encountered. We
5934      need to re-set the original option values to take account of this so that,
5935      during the pre-compile phase, we know to allow for a re-set at the start of
5936      subsequent branches. */
5937    
5938      if (old_external_options != cd->external_options)
5939        oldims = cd->external_options & PCRE_IMS;
5940    
5941    /* Keep the highest bracket count in case (?| was used and some branch    /* Keep the highest bracket count in case (?| was used and some branch
5942    has fewer than the rest. */    has fewer than the rest. */
5943    
# Line 5417  for (;;) Line 5988  for (;;)
5988    
5989      /* If lookbehind, check that this branch matches a fixed-length string, and      /* If lookbehind, check that this branch matches a fixed-length string, and
5990      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
5991      branch with OP_END. */      branch with OP_END. If the branch contains OP_RECURSE, the result is -3
5992        because there may be forward references that we can't check here. Set a
5993        flag to cause another lookbehind check at the end. Why not do it all at the
5994        end? Because common, erroneous checks are picked up here and the offset of
5995        the problem can be shown. */
5996    
5997      if (lookbehind)      if (lookbehind)
5998        {        {
5999        int fixed_length;        int fixed_length;
6000        *code = OP_END;        *code = OP_END;
6001        fixed_length = find_fixedlength(last_branch, options);        fixed_length = find_fixedlength(last_branch, options, FALSE, cd);
6002        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
6003        if (fixed_length < 0)        if (fixed_length == -3)
6004            {
6005            cd->check_lookbehind = TRUE;
6006            }
6007          else if (fixed_length < 0)
6008          {          {
6009          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;
6010          *ptrptr = ptr;          *ptrptr = ptr;
6011          return FALSE;          return FALSE;
6012          }          }
6013        PUT(reverse_count, 0, fixed_length);        else { PUT(reverse_count, 0, fixed_length); }
6014        }        }
6015      }      }
6016    
# Line 5444  for (;;) Line 6023  for (;;)
6023    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.
6024    Return leaving the pointer at the terminating char. */    Return leaving the pointer at the terminating char. */
6025    
6026    if (*ptr != '|')    if (*ptr != CHAR_VERTICAL_LINE)
6027      {      {
6028      if (lengthptr == NULL)      if (lengthptr == NULL)
6029        {        {
# Line 5465  for (;;) Line 6044  for (;;)
6044      PUT(code, 1, code - start_bracket);      PUT(code, 1, code - start_bracket);
6045      code += 1 + LINK_SIZE;      code += 1 + LINK_SIZE;
6046    
6047      /* Resetting option if needed */      /* If it was a capturing subpattern, check to see if it contained any
6048        recursive back references. If so, we must wrap it in atomic brackets.
6049        In any event, remove the block from the chain. */
6050    
6051      if ((options & PCRE_IMS) != oldims && *ptr == ')')      if (capnumber > 0)
6052          {
6053          if (cd->open_caps->flag)
6054            {
6055            memmove(start_bracket + 1 + LINK_SIZE, start_bracket,
6056              code - start_bracket);
6057            *start_bracket = OP_ONCE;
6058            code += 1 + LINK_SIZE;
6059            PUT(start_bracket, 1, code - start_bracket);
6060            *code = OP_KET;
6061            PUT(code, 1, code - start_bracket);
6062            code += 1 + LINK_SIZE;
6063            length += 2 + 2*LINK_SIZE;
6064            }
6065          cd->open_caps = cd->open_caps->next;
6066          }
6067    
6068        /* Reset options if needed. */
6069    
6070        if ((options & PCRE_IMS) != oldims && *ptr == CHAR_RIGHT_PARENTHESIS)
6071        {        {
6072        *code++ = OP_OPT;        *code++ = OP_OPT;
6073        *code++ = oldims;        *code++ = oldims;
# Line 5514  for (;;) Line 6114  for (;;)
6114      {      {
6115      *code = OP_ALT;      *code = OP_ALT;
6116      PUT(code, 1, code - last_branch);      PUT(code, 1, code - last_branch);
6117      bc.current = last_branch = code;      bc.current_branch = last_branch = code;
6118      code += 1 + LINK_SIZE;      code += 1 + LINK_SIZE;
6119      }      }
6120    
# Line 5596  do { Line 6196  do {
6196       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;
6197       }       }
6198    
6199     /* .* is not anchored unless DOTALL is set and it isn't in brackets that     /* .* is not anchored unless DOTALL is set (which generates OP_ALLANY) and
6200     are or may be referenced. */     it isn't in brackets that are or may be referenced. */
6201    
6202     else if ((op == OP_TYPESTAR || op == OP_TYPEMINSTAR ||     else if ((op == OP_TYPESTAR || op == OP_TYPEMINSTAR ||
6203               op == OP_TYPEPOSSTAR) &&               op == OP_TYPEPOSSTAR))
             (*options & PCRE_DOTALL) != 0)  
6204       {       {
6205       if (scode[1] != OP_ANY || (bracket_map & backref_map) != 0) return FALSE;       if (scode[1] != OP_ALLANY || (bracket_map & backref_map) != 0)
6206           return FALSE;
6207       }       }
6208    
6209     /* Check for explicit anchoring */     /* Check for explicit anchoring */
# Line 5649  do { Line 6249  do {
6249       NULL, 0, FALSE);       NULL, 0, FALSE);
6250     register int op = *scode;     register int op = *scode;
6251    
6252       /* If we are at the start of a conditional assertion group, *both* the
6253       conditional assertion *and* what follows the condition must satisfy the test
6254       for start of line. Other kinds of condition fail. Note that there may be an
6255       auto-callout at the start of a condition. */
6256    
6257       if (op == OP_COND)
6258         {
6259         scode += 1 + LINK_SIZE;
6260         if (*scode == OP_CALLOUT) scode += _pcre_OP_lengths[OP_CALLOUT];
6261         switch (*scode)
6262           {
6263           case OP_CREF:
6264           case OP_NCREF:
6265           case OP_RREF:
6266           case OP_NRREF:
6267           case OP_DEF:
6268           return FALSE;
6269    
6270           default:     /* Assertion */
6271           if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
6272           do scode += GET(scode, 1); while (*scode == OP_ALT);
6273           scode += 1 + LINK_SIZE;
6274           break;
6275           }
6276         scode = first_significant_code(scode, NULL, 0, FALSE);
6277         op = *scode;
6278         }
6279    
6280     /* Non-capturing brackets */     /* Non-capturing brackets */
6281    
6282     if (op == OP_BRA)     if (op == OP_BRA)
# Line 5667  do { Line 6295  do {
6295    
6296     /* Other brackets */     /* Other brackets */
6297    
6298     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_COND)     else if (op == OP_ASSERT || op == OP_ONCE)
6299       { if (!is_startline(scode, bracket_map, backref_map)) return FALSE; }       {
6300         if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
6301         }
6302    
6303     /* .* 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
6304     may be referenced. */     may be referenced. */
# Line 5785  Returns: pointer to compiled data Line 6415  Returns: pointer to compiled data
6415                  with errorptr and erroroffset set                  with errorptr and erroroffset set
6416  */  */
6417    
6418  PCRE_EXP_DEFN pcre *  PCRE_EXP_DEFN pcre * PCRE_CALL_CONVENTION
6419  pcre_compile(const char *pattern, int options, const char **errorptr,  pcre_compile(const char *pattern, int options, const char **errorptr,
6420    int *erroroffset, const unsigned char *tables)    int *erroroffset, const unsigned char *tables)
6421  {  {
# Line 5793  return pcre_compile2(pattern, options, N Line 6423  return pcre_compile2(pattern, options, N
6423  }  }
6424    
6425    
6426  PCRE_EXP_DEFN pcre *  PCRE_EXP_DEFN pcre * PCRE_CALL_CONVENTION
6427  pcre_compile2(const char *pattern, int options, int *errorcodeptr,  pcre_compile2(const char *pattern, int options, int *errorcodeptr,
6428    const char **errorptr, int *erroroffset, const unsigned char *tables)    const char **errorptr, int *erroroffset, const unsigned char *tables)
6429  {  {
# Line 5802  int length = 1; /* For final END opcode Line 6432  int length = 1; /* For final END opcode
6432  int firstbyte, reqbyte, newline;  int firstbyte, reqbyte, newline;
6433  int errorcode = 0;  int errorcode = 0;
6434  int skipatstart = 0;  int skipatstart = 0;
6435  #ifdef SUPPORT_UTF8  BOOL utf8 = (options & PCRE_UTF8) != 0;
 BOOL utf8;  
 #endif  
6436  size_t size;  size_t size;
6437  uschar *code;  uschar *code;
6438  const uschar *codestart;  const uschar *codestart;
# Line 5847  if (erroroffset == NULL) Line 6475  if (erroroffset == NULL)
6475    
6476  *erroroffset = 0;  *erroroffset = 0;
6477    
 /* 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;  
   }  
   
6478  /* Set up pointers to the individual character tables */  /* Set up pointers to the individual character tables */
6479    
6480  if (tables == NULL) tables = _pcre_default_tables;  if (tables == NULL) tables = _pcre_default_tables;
# Line 5879  cd->fcc = tables + fcc_offset; Line 6483  cd->fcc = tables + fcc_offset;
6483  cd->cbits = tables + cbits_offset;  cd->cbits = tables + cbits_offset;
6484  cd->ctypes = tables + ctypes_offset;  cd->ctypes = tables + ctypes_offset;
6485    
6486    /* Check that all undefined public option bits are zero */
6487    
6488    if ((options & ~PUBLIC_COMPILE_OPTIONS) != 0)
6489      {
6490      errorcode = ERR17;
6491      goto PCRE_EARLY_ERROR_RETURN;
6492      }
6493    
6494  /* 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
6495  the offset for later. */  the offset for later. */
6496    
6497  while (ptr[skipatstart] == '(' && ptr[skipatstart+1] == '*')  while (ptr[skipatstart] == CHAR_LEFT_PARENTHESIS &&
6498           ptr[skipatstart+1] == CHAR_ASTERISK)
6499    {    {
6500    int newnl = 0;    int newnl = 0;
6501    int newbsr = 0;    int newbsr = 0;
6502    
6503    if (strncmp((char *)(ptr+skipatstart+2), "CR)", 3) == 0)    if (strncmp((char *)(ptr+skipatstart+2), STRING_UTF8_RIGHTPAR, 5) == 0)
6504        { skipatstart += 7; options |= PCRE_UTF8; continue; }
6505    
6506      if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)
6507      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }
6508    else if (strncmp((char *)(ptr+skipatstart+2), "LF)", 3)  == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_LF_RIGHTPAR, 3)  == 0)
6509      { skipatstart += 5; newnl = PCRE_NEWLINE_LF; }      { skipatstart += 5; newnl = PCRE_NEWLINE_LF; }
6510    else if (strncmp((char *)(ptr+skipatstart+2), "CRLF)", 5)  == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_CRLF_RIGHTPAR, 5)  == 0)
6511      { skipatstart += 7; newnl = PCRE_NEWLINE_CR + PCRE_NEWLINE_LF; }      { skipatstart += 7; newnl = PCRE_NEWLINE_CR + PCRE_NEWLINE_LF; }
6512    else if (strncmp((char *)(ptr+skipatstart+2), "ANY)", 4) == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_ANY_RIGHTPAR, 4) == 0)
6513      { skipatstart += 6; newnl = PCRE_NEWLINE_ANY; }      { skipatstart += 6; newnl = PCRE_NEWLINE_ANY; }
6514    else if (strncmp((char *)(ptr+skipatstart+2), "ANYCRLF)", 8)  == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_ANYCRLF_RIGHTPAR, 8) == 0)
6515      { skipatstart += 10; newnl = PCRE_NEWLINE_ANYCRLF; }      { skipatstart += 10; newnl = PCRE_NEWLINE_ANYCRLF; }
6516    
6517    else if (strncmp((char *)(ptr+skipatstart+2), "BSR_ANYCRLF)", 12) == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_BSR_ANYCRLF_RIGHTPAR, 12) == 0)
6518      { skipatstart += 14; newbsr = PCRE_BSR_ANYCRLF; }      { skipatstart += 14; newbsr = PCRE_BSR_ANYCRLF; }
6519    else if (strncmp((char *)(ptr+skipatstart+2), "BSR_UNICODE)", 12) == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_BSR_UNICODE_RIGHTPAR, 12) == 0)
6520      { skipatstart += 14; newbsr = PCRE_BSR_UNICODE; }      { skipatstart += 14; newbsr = PCRE_BSR_UNICODE; }
6521    
6522    if (newnl != 0)    if (newnl != 0)
# Line 5910  while (ptr[skipatstart] == '(' && ptr[sk Line 6526  while (ptr[skipatstart] == '(' && ptr[sk
6526    else break;    else break;
6527    }    }
6528    
6529    /* Can't support UTF8 unless PCRE has been compiled to include the code. */
6530    
6531    #ifdef SUPPORT_UTF8
6532    if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&
6533         (*erroroffset = _pcre_valid_utf8((USPTR)pattern, -1)) >= 0)
6534      {
6535      errorcode = ERR44;
6536      goto PCRE_EARLY_ERROR_RETURN2;
6537      }
6538    #else
6539    if (utf8)
6540      {
6541      errorcode = ERR32;
6542      goto PCRE_EARLY_ERROR_RETURN;
6543      }
6544    #endif
6545    
6546  /* Check validity of \R options. */  /* Check validity of \R options. */
6547    
6548  switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))  switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))
# Line 5928  current code allows for fixed one- or tw Line 6561  current code allows for fixed one- or tw
6561  switch (options & PCRE_NEWLINE_BITS)  switch (options & PCRE_NEWLINE_BITS)
6562    {    {
6563    case 0: newline = NEWLINE; break;   /* Build-time default */    case 0: newline = NEWLINE; break;   /* Build-time default */
6564    case PCRE_NEWLINE_CR: newline = '\r'; break;    case PCRE_NEWLINE_CR: newline = CHAR_CR; break;
6565    case PCRE_NEWLINE_LF: newline = '\n'; break;    case PCRE_NEWLINE_LF: newline = CHAR_NL; break;
6566    case PCRE_NEWLINE_CR+    case PCRE_NEWLINE_CR+
6567         PCRE_NEWLINE_LF: newline = ('\r' << 8) | '\n'; break;         PCRE_NEWLINE_LF: newline = (CHAR_CR << 8) | CHAR_NL; break;
6568    case PCRE_NEWLINE_ANY: newline = -1; break;    case PCRE_NEWLINE_ANY: newline = -1; break;
6569    case PCRE_NEWLINE_ANYCRLF: newline = -2; break;    case PCRE_NEWLINE_ANYCRLF: newline = -2; break;
6570    default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;    default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;
# Line 5992  cd->end_pattern = (const uschar *)(patte Line 6625  cd->end_pattern = (const uschar *)(patte
6625  cd->req_varyopt = 0;  cd->req_varyopt = 0;
6626  cd->external_options = options;  cd->external_options = options;
6627  cd->external_flags = 0;  cd->external_flags = 0;
6628    cd->open_caps = NULL;
6629    
6630  /* 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
6631  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 6066  cd->start_code = codestart; Line 6700  cd->start_code = codestart;
6700  cd->hwm = cworkspace;  cd->hwm = cworkspace;
6701  cd->req_varyopt = 0;  cd->req_varyopt = 0;
6702  cd->had_accept = FALSE;  cd->had_accept = FALSE;
6703    cd->check_lookbehind = FALSE;
6704    cd->open_caps = NULL;
6705    
6706  /* Set up a starting, non-extracting bracket, then compile the expression. On  /* Set up a starting, non-extracting bracket, then compile the expression. On
6707  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 6091  if debugging, leave the test till after Line 6727  if debugging, leave the test till after
6727    
6728  *code++ = OP_END;  *code++ = OP_END;
6729    
6730  #ifndef DEBUG  #ifndef PCRE_DEBUG
6731  if (code - codestart > length) errorcode = ERR23;  if (code - codestart > length) errorcode = ERR23;
6732  #endif  #endif
6733    
# Line 6104  while (errorcode == 0 && cd->hwm > cwork Line 6740  while (errorcode == 0 && cd->hwm > cwork
6740    cd->hwm -= LINK_SIZE;    cd->hwm -= LINK_SIZE;
6741    offset = GET(cd->hwm, 0);    offset = GET(cd->hwm, 0);
6742    recno = GET(codestart, offset);    recno = GET(codestart, offset);
6743    groupptr = find_bracket(codestart, (re->options & PCRE_UTF8) != 0, recno);    groupptr = _pcre_find_bracket(codestart, utf8, recno);
6744    if (groupptr == NULL) errorcode = ERR53;    if (groupptr == NULL) errorcode = ERR53;
6745      else PUT(((uschar *)codestart), offset, groupptr - codestart);      else PUT(((uschar *)codestart), offset, groupptr - codestart);
6746    }    }
# Line 6114  subpattern. */ Line 6750  subpattern. */
6750    
6751  if (errorcode == 0 && re->top_backref > re->top_bracket) errorcode = ERR15;  if (errorcode == 0 && re->top_backref > re->top_bracket) errorcode = ERR15;
6752    
6753    /* If there were any lookbehind assertions that contained OP_RECURSE
6754    (recursions or subroutine calls), a flag is set for them to be checked here,
6755    because they may contain forward references. Actual recursions can't be fixed
6756    length, but subroutine calls can. It is done like this so that those without
6757    OP_RECURSE that are not fixed length get a diagnosic with a useful offset. The
6758    exceptional ones forgo this. We scan the pattern to check that they are fixed
6759    length, and set their lengths. */
6760    
6761    if (cd->check_lookbehind)
6762      {
6763      uschar *cc = (uschar *)codestart;
6764    
6765      /* Loop, searching for OP_REVERSE items, and process those that do not have
6766      their length set. (Actually, it will also re-process any that have a length
6767      of zero, but that is a pathological case, and it does no harm.) When we find
6768      one, we temporarily terminate the branch it is in while we scan it. */
6769    
6770      for (cc = (uschar *)_pcre_find_bracket(codestart, utf8, -1);
6771           cc != NULL;
6772           cc = (uschar *)_pcre_find_bracket(cc, utf8, -1))
6773        {
6774        if (GET(cc, 1) == 0)
6775          {
6776          int fixed_length;
6777          uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);
6778          int end_op = *be;
6779          *be = OP_END;
6780          fixed_length = find_fixedlength(cc, re->options, TRUE, cd);
6781          *be = end_op;
6782          DPRINTF(("fixed length = %d\n", fixed_length));
6783          if (fixed_length < 0)
6784            {
6785            errorcode = (fixed_length == -2)? ERR36 : ERR25;
6786            break;
6787            }
6788          PUT(cc, 1, fixed_length);
6789          }
6790        cc += 1 + LINK_SIZE;
6791        }
6792      }
6793    
6794  /* Failed to compile, or error while post-processing */  /* Failed to compile, or error while post-processing */
6795    
6796  if (errorcode != 0)  if (errorcode != 0)
# Line 6174  if (reqbyte >= 0 && Line 6851  if (reqbyte >= 0 &&
6851  /* 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
6852  case when building a production library. */  case when building a production library. */
6853    
6854  #ifdef DEBUG  #ifdef PCRE_DEBUG
   
6855  printf("Length = %d top_bracket = %d top_backref = %d\n",  printf("Length = %d top_bracket = %d top_backref = %d\n",
6856    length, re->top_bracket, re->top_backref);    length, re->top_bracket, re->top_backref);
6857    
# Line 6212  if (code - codestart > length) Line 6888  if (code - codestart > length)
6888    if (errorcodeptr != NULL) *errorcodeptr = ERR23;    if (errorcodeptr != NULL) *errorcodeptr = ERR23;
6889    return NULL;    return NULL;
6890    }    }
6891  #endif   /* DEBUG */  #endif   /* PCRE_DEBUG */
6892    
6893  return (pcre *)re;  return (pcre *)re;
6894  }  }

Legend:
Removed from v.323  
changed lines
  Added in v.504

webmaster@exim.org
ViewVC Help
Powered by ViewVC 1.1.12