/[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 1363 by ph10, Tue Oct 1 16:54:40 2013 UTC revision 1472 by ph10, Mon Apr 21 16:11:50 2014 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) 1997-2013 University of Cambridge             Copyright (c) 1997-2014 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 260  static const verbitem verbs[] = { Line 260  static const verbitem verbs[] = {
260  static const int verbcount = sizeof(verbs)/sizeof(verbitem);  static const int verbcount = sizeof(verbs)/sizeof(verbitem);
261    
262    
263    /* Substitutes for [[:<:]] and [[:>:]], which mean start and end of word in
264    another regex library. */
265    
266    static const pcre_uchar sub_start_of_word[] = {
267      CHAR_BACKSLASH, CHAR_b, CHAR_LEFT_PARENTHESIS, CHAR_QUESTION_MARK,
268      CHAR_EQUALS_SIGN, CHAR_BACKSLASH, CHAR_w, CHAR_RIGHT_PARENTHESIS, '\0' };
269    
270    static const pcre_uchar sub_end_of_word[] = {
271      CHAR_BACKSLASH, CHAR_b, CHAR_LEFT_PARENTHESIS, CHAR_QUESTION_MARK,
272      CHAR_LESS_THAN_SIGN, CHAR_EQUALS_SIGN, CHAR_BACKSLASH, CHAR_w,
273      CHAR_RIGHT_PARENTHESIS, '\0' };
274    
275    
276  /* 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
277  now all in a single string, to reduce the number of relocations when a shared  now all in a single string, to reduce the number of relocations when a shared
278  library is dynamically loaded. The list of lengths is terminated by a zero  library is dynamically loaded. The list of lengths is terminated by a zero
279  length entry. The first three must be alpha, lower, upper, as this is assumed  length entry. The first three must be alpha, lower, upper, as this is assumed
280  for handling case independence. */  for handling case independence. The indices for graph, print, and punct are
281    needed, so identify them. */
282    
283  static const char posix_names[] =  static const char posix_names[] =
284    STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0    STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
# Line 275  static const char posix_names[] = Line 289  static const char posix_names[] =
289  static const pcre_uint8 posix_name_lengths[] = {  static const pcre_uint8 posix_name_lengths[] = {
290    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 };
291    
292    #define PC_GRAPH  8
293    #define PC_PRINT  9
294    #define PC_PUNCT 10
295    
296    
297  /* Table of class bit maps for each POSIX class. Each class is formed from a  /* Table of class bit maps for each POSIX class. Each class is formed from a
298  base map, with an optional addition or removal of another map. Then, for some  base map, with an optional addition or removal of another map. Then, for some
299  classes, there is some additional tweaking: for [:blank:] the vertical space  classes, there is some additional tweaking: for [:blank:] the vertical space
# Line 302  static const int posix_class_maps[] = { Line 321  static const int posix_class_maps[] = {
321    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
322  };  };
323    
324  /* Table of substitutes for \d etc when PCRE_UCP is set. The POSIX class  /* Table of substitutes for \d etc when PCRE_UCP is set. They are replaced by
325  substitutes must be in the order of the names, defined above, and there are  Unicode property escapes. */
 both positive and negative cases. NULL means no substitute. */  
326    
327  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
328  static const pcre_uchar string_PNd[]  = {  static const pcre_uchar string_PNd[]  = {
# Line 329  static const pcre_uchar string_pXwd[] = Line 347  static const pcre_uchar string_pXwd[] =
347  static const pcre_uchar *substitutes[] = {  static const pcre_uchar *substitutes[] = {
348    string_PNd,           /* \D */    string_PNd,           /* \D */
349    string_pNd,           /* \d */    string_pNd,           /* \d */
350    string_PXsp,          /* \S */       /* NOTE: Xsp is Perl space */    string_PXsp,          /* \S */   /* Xsp is Perl space, but from 8.34, Perl */
351    string_pXsp,          /* \s */    string_pXsp,          /* \s */   /* space and POSIX space are the same. */
352    string_PXwd,          /* \W */    string_PXwd,          /* \W */
353    string_pXwd           /* \w */    string_pXwd           /* \w */
354  };  };
355    
356    /* The POSIX class substitutes must be in the order of the POSIX class names,
357    defined above, and there are both positive and negative cases. NULL means no
358    general substitute of a Unicode property escape (\p or \P). However, for some
359    POSIX classes (e.g. graph, print, punct) a special property code is compiled
360    directly. */
361    
362  static const pcre_uchar string_pL[] =   {  static const pcre_uchar string_pL[] =   {
363    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,    CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
364    CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };    CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
# Line 382  static const pcre_uchar *posix_substitut Line 406  static const pcre_uchar *posix_substitut
406    NULL,                 /* graph */    NULL,                 /* graph */
407    NULL,                 /* print */    NULL,                 /* print */
408    NULL,                 /* punct */    NULL,                 /* punct */
409    string_pXps,          /* space */    /* NOTE: Xps is POSIX space */    string_pXps,          /* space */   /* Xps is POSIX space, but from 8.34 */
410    string_pXwd,          /* word */    string_pXwd,          /* word  */   /* Perl and POSIX space are the same */
411    NULL,                 /* xdigit */    NULL,                 /* xdigit */
412    /* Negated cases */    /* Negated cases */
413    string_PL,            /* ^alpha */    string_PL,            /* ^alpha */
# Line 397  static const pcre_uchar *posix_substitut Line 421  static const pcre_uchar *posix_substitut
421    NULL,                 /* ^graph */    NULL,                 /* ^graph */
422    NULL,                 /* ^print */    NULL,                 /* ^print */
423    NULL,                 /* ^punct */    NULL,                 /* ^punct */
424    string_PXps,          /* ^space */   /* NOTE: Xps is POSIX space */    string_PXps,          /* ^space */  /* Xps is POSIX space, but from 8.34 */
425    string_PXwd,          /* ^word */    string_PXwd,          /* ^word */   /* Perl and POSIX space are the same */
426    NULL                  /* ^xdigit */    NULL                  /* ^xdigit */
427  };  };
428  #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))  #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))
# Line 462  static const char error_texts[] = Line 486  static const char error_texts[] =
486    "POSIX collating elements are not supported\0"    "POSIX collating elements are not supported\0"
487    "this version of PCRE is compiled without UTF support\0"    "this version of PCRE is compiled without UTF support\0"
488    "spare error\0"  /** DEAD **/    "spare error\0"  /** DEAD **/
489    "character value in \\x{...} sequence is too large\0"    "character value in \\x{} or \\o{} is too large\0"
490    /* 35 */    /* 35 */
491    "invalid condition (?(0)\0"    "invalid condition (?(0)\0"
492    "\\C not allowed in lookbehind assertion\0"    "\\C not allowed in lookbehind assertion\0"
# Line 516  static const char error_texts[] = Line 540  static const char error_texts[] =
540    "character value in \\u.... sequence is too large\0"    "character value in \\u.... sequence is too large\0"
541    "invalid UTF-32 string\0"    "invalid UTF-32 string\0"
542    "setting UTF is disabled by the application\0"    "setting UTF is disabled by the application\0"
543      "non-hex character in \\x{} (closing brace missing?)\0"
544      /* 80 */
545      "non-octal character in \\o{} (closing brace missing?)\0"
546      "missing opening brace after \\o\0"
547      "parentheses are too deeply nested\0"
548      "invalid range in character class\0"
549      "group name must start with a non-digit\0"
550      /* 85 */
551      "parentheses are too deeply nested (stack check)\0"
552    ;    ;
553    
554  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
# Line 773  static const pcre_uint8 posspropstab[3][ Line 806  static const pcre_uint8 posspropstab[3][
806    { ucp_L, ucp_N, ucp_P, ucp_Po }   /* WORD */    { ucp_L, ucp_N, ucp_P, ucp_Po }   /* WORD */
807  };  };
808    
809    /* This table is used when converting repeating opcodes into possessified
810    versions as a result of an explicit possessive quantifier such as ++. A zero
811    value means there is no possessified version - in those cases the item in
812    question must be wrapped in ONCE brackets. The table is truncated at OP_CALLOUT
813    because all relevant opcodes are less than that. */
814    
815    static const pcre_uint8 opcode_possessify[] = {
816      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 0 - 15  */
817      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 16 - 31 */
818    
819      0,                       /* NOTI */
820      OP_POSSTAR, 0,           /* STAR, MINSTAR */
821      OP_POSPLUS, 0,           /* PLUS, MINPLUS */
822      OP_POSQUERY, 0,          /* QUERY, MINQUERY */
823      OP_POSUPTO, 0,           /* UPTO, MINUPTO */
824      0,                       /* EXACT */
825      0, 0, 0, 0,              /* POS{STAR,PLUS,QUERY,UPTO} */
826    
827      OP_POSSTARI, 0,          /* STARI, MINSTARI */
828      OP_POSPLUSI, 0,          /* PLUSI, MINPLUSI */
829      OP_POSQUERYI, 0,         /* QUERYI, MINQUERYI */
830      OP_POSUPTOI, 0,          /* UPTOI, MINUPTOI */
831      0,                       /* EXACTI */
832      0, 0, 0, 0,              /* POS{STARI,PLUSI,QUERYI,UPTOI} */
833    
834      OP_NOTPOSSTAR, 0,        /* NOTSTAR, NOTMINSTAR */
835      OP_NOTPOSPLUS, 0,        /* NOTPLUS, NOTMINPLUS */
836      OP_NOTPOSQUERY, 0,       /* NOTQUERY, NOTMINQUERY */
837      OP_NOTPOSUPTO, 0,        /* NOTUPTO, NOTMINUPTO */
838      0,                       /* NOTEXACT */
839      0, 0, 0, 0,              /* NOTPOS{STAR,PLUS,QUERY,UPTO} */
840    
841      OP_NOTPOSSTARI, 0,       /* NOTSTARI, NOTMINSTARI */
842      OP_NOTPOSPLUSI, 0,       /* NOTPLUSI, NOTMINPLUSI */
843      OP_NOTPOSQUERYI, 0,      /* NOTQUERYI, NOTMINQUERYI */
844      OP_NOTPOSUPTOI, 0,       /* NOTUPTOI, NOTMINUPTOI */
845      0,                       /* NOTEXACTI */
846      0, 0, 0, 0,              /* NOTPOS{STARI,PLUSI,QUERYI,UPTOI} */
847    
848      OP_TYPEPOSSTAR, 0,       /* TYPESTAR, TYPEMINSTAR */
849      OP_TYPEPOSPLUS, 0,       /* TYPEPLUS, TYPEMINPLUS */
850      OP_TYPEPOSQUERY, 0,      /* TYPEQUERY, TYPEMINQUERY */
851      OP_TYPEPOSUPTO, 0,       /* TYPEUPTO, TYPEMINUPTO */
852      0,                       /* TYPEEXACT */
853      0, 0, 0, 0,              /* TYPEPOS{STAR,PLUS,QUERY,UPTO} */
854    
855      OP_CRPOSSTAR, 0,         /* CRSTAR, CRMINSTAR */
856      OP_CRPOSPLUS, 0,         /* CRPLUS, CRMINPLUS */
857      OP_CRPOSQUERY, 0,        /* CRQUERY, CRMINQUERY */
858      OP_CRPOSRANGE, 0,        /* CRRANGE, CRMINRANGE */
859      0, 0, 0, 0,              /* CRPOS{STAR,PLUS,QUERY,RANGE} */
860    
861      0, 0, 0,                 /* CLASS, NCLASS, XCLASS */
862      0, 0,                    /* REF, REFI */
863      0, 0,                    /* DNREF, DNREFI */
864      0, 0                     /* RECURSE, CALLOUT */
865    };
866    
867    
868    
869  /*************************************************  /*************************************************
# Line 879  return (*p == CHAR_RIGHT_CURLY_BRACKET); Line 970  return (*p == CHAR_RIGHT_CURLY_BRACKET);
970  *************************************************/  *************************************************/
971    
972  /* This function is called when a \ has been encountered. It either returns a  /* This function is called when a \ has been encountered. It either returns a
973  positive value for a simple escape such as \n, or 0 for a data character  positive value for a simple escape such as \n, or 0 for a data character which
974  which will be placed in chptr. A backreference to group n is returned as  will be placed in chptr. A backreference to group n is returned as negative n.
975  negative n. When UTF-8 is enabled, a positive value greater than 255 may  When UTF-8 is enabled, a positive value greater than 255 may be returned in
976  be returned in chptr.  chptr. On entry, ptr is pointing at the \. On exit, it is on the final
977  On entry,ptr is pointing at the \. On exit, it is on the final character of the  character of the escape sequence.
 escape sequence.  
978    
979  Arguments:  Arguments:
980    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
981    chptr          points to the data character    chptr          points to a returned data character
982    errorcodeptr   points to the errorcode variable    errorcodeptr   points to the errorcode variable
983    bracount       number of previous extracting brackets    bracount       number of previous extracting brackets
984    options        the options bits    options        the options bits
# Line 1092  else Line 1182  else
1182      break;      break;
1183    
1184      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
1185      starting with one that is not zero is not straightforward. By experiment,      starting with one that is not zero is not straightforward. Perl has changed
1186      the way Perl works seems to be as follows:      over the years. Nowadays \g{} for backreferences and \o{} for octal are
1187        recommended to avoid the ambiguities in the old syntax.
1188    
1189      Outside a character class, the digits are read as a decimal number. If the      Outside a character class, the digits are read as a decimal number. If the
1190      number is less than 10, or if there are that many previous extracting      number is less than 8 (used to be 10), or if there are that many previous
1191      left brackets, then it is a back reference. Otherwise, up to three octal      extracting left brackets, then it is a back reference. Otherwise, up to
1192      digits are read to form an escaped byte. Thus \123 is likely to be octal      three octal digits are read to form an escaped byte. Thus \123 is likely to
1193      123 (cf \0123, which is octal 012 followed by the literal 3). If the octal      be octal 123 (cf \0123, which is octal 012 followed by the literal 3). If
1194      value is greater than 377, the least significant 8 bits are taken. Inside a      the octal value is greater than 377, the least significant 8 bits are
1195      character class, \ followed by a digit is always an octal number. */      taken. \8 and \9 are treated as the literal characters 8 and 9.
1196    
1197        Inside a character class, \ followed by a digit is always either a literal
1198        8 or 9 or an octal number. */
1199    
1200      case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:      case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:
1201      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
# Line 1128  else Line 1222  else
1222          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
1223          break;          break;
1224          }          }
1225        if (s < 10 || s <= bracount)        if (s < 8 || s <= bracount)  /* Check for back reference */
1226          {          {
1227          escape = -s;          escape = -s;
1228          break;          break;
# Line 1136  else Line 1230  else
1230        ptr = oldptr;      /* Put the pointer back and fall through */        ptr = oldptr;      /* Put the pointer back and fall through */
1231        }        }
1232    
1233      /* Handle an octal number following \. If the first digit is 8 or 9, Perl      /* Handle a digit following \ when the number is not a back reference. If
1234      generates a binary zero byte and treats the digit as a following literal.      the first digit is 8 or 9, Perl used to generate a binary zero byte and
1235      Thus we have to pull back the pointer by one. */      then treat the digit as a following literal. At least by Perl 5.18 this
1236        changed so as not to insert the binary zero. */
1237    
1238      if ((c = *ptr) >= CHAR_8)      if ((c = *ptr) >= CHAR_8) break;
1239        {  
1240        ptr--;      /* Fall through with a digit less than 8 */
       c = 0;  
       break;  
       }  
1241    
1242      /* \0 always starts an octal number, but we may drop through to here with a      /* \0 always starts an octal number, but we may drop through to here with a
1243      larger first octal digit. The original code used just to take the least      larger first octal digit. The original code used just to take the least
# Line 1162  else Line 1254  else
1254  #endif  #endif
1255      break;      break;
1256    
1257      /* \x is complicated. \x{ddd} is a character number which can be greater      /* \o is a relatively new Perl feature, supporting a more general way of
1258      than 0xff in utf or non-8bit mode, but only if the ddd are hex digits.      specifying character codes in octal. The only supported form is \o{ddd}. */
1259      If not, { is treated as a data character. */  
1260        case CHAR_o:
1261        if (ptr[1] != CHAR_LEFT_CURLY_BRACKET) *errorcodeptr = ERR81; else
1262          {
1263          ptr += 2;
1264          c = 0;
1265          overflow = FALSE;
1266          while (*ptr >= CHAR_0 && *ptr <= CHAR_7)
1267            {
1268            register pcre_uint32 cc = *ptr++;
1269            if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1270    #ifdef COMPILE_PCRE32
1271            if (c >= 0x20000000l) { overflow = TRUE; break; }
1272    #endif
1273            c = (c << 3) + cc - CHAR_0 ;
1274    #if defined COMPILE_PCRE8
1275            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1276    #elif defined COMPILE_PCRE16
1277            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1278    #elif defined COMPILE_PCRE32
1279            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1280    #endif
1281            }
1282          if (overflow)
1283            {
1284            while (*ptr >= CHAR_0 && *ptr <= CHAR_7) ptr++;
1285            *errorcodeptr = ERR34;
1286            }
1287          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1288            {
1289            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1290            }
1291          else *errorcodeptr = ERR80;
1292          }
1293        break;
1294    
1295        /* \x is complicated. In JavaScript, \x must be followed by two hexadecimal
1296        numbers. Otherwise it is a lowercase x letter. */
1297    
1298      case CHAR_x:      case CHAR_x:
1299      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1300        {        {
       /* In JavaScript, \x must be followed by two hexadecimal numbers.  
       Otherwise it is a lowercase x letter. */  
1301        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1302          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1303          {          {
# Line 1187  else Line 1314  else
1314  #endif  #endif
1315            }            }
1316          }          }
1317        break;        }    /* End JavaScript handling */
       }  
1318    
1319      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      /* Handle \x in Perl's style. \x{ddd} is a character number which can be
1320        {      greater than 0xff in utf or non-8bit mode, but only if the ddd are hex
1321        const pcre_uchar *pt = ptr + 2;      digits. If not, { used to be treated as a data character. However, Perl
1322        seems to read hex digits up to the first non-such, and ignore the rest, so
1323        that, for example \x{zz} matches a binary zero. This seems crazy, so PCRE
1324        now gives an error. */
1325    
1326        c = 0;      else
1327        overflow = FALSE;        {
1328        while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0)        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1329          {          {
1330          register pcre_uint32 cc = *pt++;          ptr += 2;
1331          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */          c = 0;
1332            overflow = FALSE;
1333            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0)
1334              {
1335              register pcre_uint32 cc = *ptr++;
1336              if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1337    
1338  #ifdef COMPILE_PCRE32  #ifdef COMPILE_PCRE32
1339          if (c >= 0x10000000l) { overflow = TRUE; break; }            if (c >= 0x10000000l) { overflow = TRUE; break; }
1340  #endif  #endif
1341    
1342  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1343          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1344          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1345  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1346          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */            if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
1347          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1348  #endif  #endif
1349    
1350  #if defined COMPILE_PCRE8  #if defined COMPILE_PCRE8
1351          if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1352  #elif defined COMPILE_PCRE16  #elif defined COMPILE_PCRE16
1353          if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1354  #elif defined COMPILE_PCRE32  #elif defined COMPILE_PCRE32
1355          if (utf && c > 0x10ffffU) { overflow = TRUE; break; }            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1356  #endif  #endif
1357          }            }
1358    
1359        if (overflow)          if (overflow)
1360          {            {
1361          while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0) pt++;            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0) ptr++;
1362          *errorcodeptr = ERR34;            *errorcodeptr = ERR34;
1363          }            }
1364    
1365        if (*pt == CHAR_RIGHT_CURLY_BRACKET)          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1366          {            {
1367          if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1368          ptr = pt;            }
         break;  
         }  
1369    
1370        /* If the sequence of hex digits does not end with '}', then we don't          /* If the sequence of hex digits does not end with '}', give an error.
1371        recognize this construct; fall through to the normal \x handling. */          We used just to recognize this construct and fall through to the normal
1372        }          \x handling, but nowadays Perl gives an error, which seems much more
1373            sensible, so we do too. */
1374    
1375      /* Read just a single-byte hex-defined char */          else *errorcodeptr = ERR79;
1376            }   /* End of \x{} processing */
1377    
1378      c = 0;        /* Read a single-byte hex-defined char (up to two hex digits after \x) */
1379      while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)  
1380        {        else
1381        pcre_uint32 cc;                          /* Some compilers don't like */          {
1382        cc = *(++ptr);                           /* ++ in initializers */          c = 0;
1383            while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1384              {
1385              pcre_uint32 cc;                          /* Some compilers don't like */
1386              cc = *(++ptr);                           /* ++ in initializers */
1387  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1388        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1389        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1390  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1391        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */            if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1392        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1393  #endif  #endif
1394        }            }
1395            }     /* End of \xdd handling */
1396          }       /* End of Perl-style \x handling */
1397      break;      break;
1398    
1399      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
# Line 1443  read_repeat_counts(const pcre_uchar *p, Line 1583  read_repeat_counts(const pcre_uchar *p,
1583  int min = 0;  int min = 0;
1584  int max = -1;  int max = -1;
1585    
1586  /* Read the minimum value and do a paranoid check: a negative value indicates  while (IS_DIGIT(*p))
 an integer overflow. */  
   
 while (IS_DIGIT(*p)) min = min * 10 + (int)(*p++ - CHAR_0);  
 if (min < 0 || min > 65535)  
1587    {    {
1588    *errorcodeptr = ERR5;    min = min * 10 + (int)(*p++ - CHAR_0);
1589    return p;    if (min > 65535)
1590    }      {
1591        *errorcodeptr = ERR5;
1592  /* Read the maximum value if there is one, and again do a paranoid on its size.      return p;
1593  Also, max must not be less than min. */      }
1594      }
1595    
1596  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1597    {    {
1598    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1599      {      {
1600      max = 0;      max = 0;
1601      while(IS_DIGIT(*p)) max = max * 10 + (int)(*p++ - CHAR_0);      while(IS_DIGIT(*p))
     if (max < 0 || max > 65535)  
1602        {        {
1603        *errorcodeptr = ERR5;        max = max * 10 + (int)(*p++ - CHAR_0);
1604        return p;        if (max > 65535)
1605        }          {
1606            *errorcodeptr = ERR5;
1607            return p;
1608            }
1609          }
1610      if (max < min)      if (max < min)
1611        {        {
1612        *errorcodeptr = ERR4;        *errorcodeptr = ERR4;
# Line 1475  if (*p == CHAR_RIGHT_CURLY_BRACKET) max Line 1615  if (*p == CHAR_RIGHT_CURLY_BRACKET) max
1615      }      }
1616    }    }
1617    
 /* Fill in the required variables, and pass back the pointer to the terminating  
 '}'. */  
   
1618  *minp = min;  *minp = min;
1619  *maxp = max;  *maxp = max;
1620  return p;  return p;
# Line 1524  for (;;) Line 1661  for (;;)
1661    
1662      case OP_CALLOUT:      case OP_CALLOUT:
1663      case OP_CREF:      case OP_CREF:
1664      case OP_NCREF:      case OP_DNCREF:
1665      case OP_RREF:      case OP_RREF:
1666      case OP_NRREF:      case OP_DNRREF:
1667      case OP_DEF:      case OP_DEF:
1668      code += PRIV(OP_lengths)[*code];      code += PRIV(OP_lengths)[*code];
1669      break;      break;
# Line 1663  for (;;) Line 1800  for (;;)
1800      case OP_COMMIT:      case OP_COMMIT:
1801      case OP_CREF:      case OP_CREF:
1802      case OP_DEF:      case OP_DEF:
1803        case OP_DNCREF:
1804        case OP_DNRREF:
1805      case OP_DOLL:      case OP_DOLL:
1806      case OP_DOLLM:      case OP_DOLLM:
1807      case OP_EOD:      case OP_EOD:
1808      case OP_EODN:      case OP_EODN:
1809      case OP_FAIL:      case OP_FAIL:
     case OP_NCREF:  
     case OP_NRREF:  
1810      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1811      case OP_PRUNE:      case OP_PRUNE:
1812      case OP_REVERSE:      case OP_REVERSE:
# Line 1764  for (;;) Line 1901  for (;;)
1901    
1902      switch (*cc)      switch (*cc)
1903        {        {
       case OP_CRPLUS:  
       case OP_CRMINPLUS:  
1904        case OP_CRSTAR:        case OP_CRSTAR:
1905        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1906          case OP_CRPLUS:
1907          case OP_CRMINPLUS:
1908        case OP_CRQUERY:        case OP_CRQUERY:
1909        case OP_CRMINQUERY:        case OP_CRMINQUERY:
1910          case OP_CRPOSSTAR:
1911          case OP_CRPOSPLUS:
1912          case OP_CRPOSQUERY:
1913        return -1;        return -1;
1914    
1915        case OP_CRRANGE:        case OP_CRRANGE:
1916        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1917          case OP_CRPOSRANGE:
1918        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1919        branchlength += (int)GET2(cc,1);        branchlength += (int)GET2(cc,1);
1920        cc += 1 + 2 * IMM2_SIZE;        cc += 1 + 2 * IMM2_SIZE;
# Line 2366  for (code = first_significant_code(code Line 2507  for (code = first_significant_code(code
2507        case OP_CRMINSTAR:        case OP_CRMINSTAR:
2508        case OP_CRQUERY:        case OP_CRQUERY:
2509        case OP_CRMINQUERY:        case OP_CRMINQUERY:
2510          case OP_CRPOSSTAR:
2511          case OP_CRPOSQUERY:
2512        break;        break;
2513    
2514        default:                   /* Non-repeat => class must match */        default:                   /* Non-repeat => class must match */
2515        case OP_CRPLUS:            /* These repeats aren't empty */        case OP_CRPLUS:            /* These repeats aren't empty */
2516        case OP_CRMINPLUS:        case OP_CRMINPLUS:
2517          case OP_CRPOSPLUS:
2518        return FALSE;        return FALSE;
2519    
2520        case OP_CRRANGE:        case OP_CRRANGE:
2521        case OP_CRMINRANGE:        case OP_CRMINRANGE:
2522          case OP_CRPOSRANGE:
2523        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */
2524        break;        break;
2525        }        }
# Line 2650  switch(ptype) Line 2795  switch(ptype)
2795    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2796            PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;            PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2797    
2798    case PT_SPACE:    /* Perl space */    /* Perl space used to exclude VT, but from Perl 5.18 it is included, which
2799    return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||    means that Perl space and POSIX space are now identical. PCRE was changed
2800            c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)    at release 8.34. */
           == negated;  
2801    
2802      case PT_SPACE:    /* Perl space */
2803    case PT_PXSPACE:  /* POSIX space */    case PT_PXSPACE:  /* POSIX space */
2804    return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||    switch(c)
2805            c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||      {
2806            c == CHAR_FF || c == CHAR_CR)      HSPACE_CASES:
2807            == negated;      VSPACE_CASES:
2808        return negated;
2809    
2810        default:
2811        return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z) == negated;
2812        }
2813      break;  /* Control never reaches here */
2814    
2815    case PT_WORD:    case PT_WORD:
2816    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
# Line 2708  get_chr_property_list(const pcre_uchar * Line 2859  get_chr_property_list(const pcre_uchar *
2859    const pcre_uint8 *fcc, pcre_uint32 *list)    const pcre_uint8 *fcc, pcre_uint32 *list)
2860  {  {
2861  pcre_uchar c = *code;  pcre_uchar c = *code;
2862    pcre_uchar base;
2863  const pcre_uchar *end;  const pcre_uchar *end;
 const pcre_uint32 *clist_src;  
 pcre_uint32 *clist_dest;  
2864  pcre_uint32 chr;  pcre_uint32 chr;
2865  pcre_uchar base;  
2866    #ifdef SUPPORT_UCP
2867    pcre_uint32 *clist_dest;
2868    const pcre_uint32 *clist_src;
2869    #else
2870    utf = utf;  /* Suppress "unused parameter" compiler warning */
2871    #endif
2872    
2873  list[0] = c;  list[0] = c;
2874  list[1] = FALSE;  list[1] = FALSE;
# Line 2818  switch(c) Line 2974  switch(c)
2974      return code + 2;      return code + 2;
2975      }      }
2976    
2977    /* Convert only if we have anough space. */    /* Convert only if we have enough space. */
2978    
2979    clist_src = PRIV(ucd_caseless_sets) + code[1];    clist_src = PRIV(ucd_caseless_sets) + code[1];
2980    clist_dest = list + 2;    clist_dest = list + 2;
2981    code += 2;    code += 2;
2982    
2983    do {    do {
      /* Early return if there is not enough space. */  
2984       if (clist_dest >= list + 8)       if (clist_dest >= list + 8)
2985         {         {
2986           /* Early return if there is not enough space. This should never
2987           happen, since all clists are shorter than 5 character now. */
2988         list[2] = code[0];         list[2] = code[0];
2989         list[3] = code[1];         list[3] = code[1];
2990         return code;         return code;
2991         }         }
2992       *clist_dest++ = *clist_src;       *clist_dest++ = *clist_src;
2993       }       }
2994     while(*clist_src++ != NOTACHAR);    while(*clist_src++ != NOTACHAR);
2995    
2996    /* Enough space to store all characters. */    /* All characters are stored. The terminating NOTACHAR
2997      is copied form the clist itself. */
2998    
2999    list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;    list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
3000    return code;    return code;
# Line 2846  switch(c) Line 3004  switch(c)
3004    case OP_CLASS:    case OP_CLASS:
3005  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3006    case OP_XCLASS:    case OP_XCLASS:
   
3007    if (c == OP_XCLASS)    if (c == OP_XCLASS)
3008      end = code + GET(code, 0);      end = code + GET(code, 0) - 1;
3009    else    else
3010  #endif  #endif
3011      end = code + 32 / sizeof(pcre_uchar);      end = code + 32 / sizeof(pcre_uchar);
# Line 2859  switch(c) Line 3016  switch(c)
3016      case OP_CRMINSTAR:      case OP_CRMINSTAR:
3017      case OP_CRQUERY:      case OP_CRQUERY:
3018      case OP_CRMINQUERY:      case OP_CRMINQUERY:
3019        case OP_CRPOSSTAR:
3020        case OP_CRPOSQUERY:
3021      list[1] = TRUE;      list[1] = TRUE;
3022      end++;      end++;
3023      break;      break;
3024    
3025        case OP_CRPLUS:
3026        case OP_CRMINPLUS:
3027        case OP_CRPOSPLUS:
3028        end++;
3029        break;
3030    
3031      case OP_CRRANGE:      case OP_CRRANGE:
3032      case OP_CRMINRANGE:      case OP_CRMINRANGE:
3033        case OP_CRPOSRANGE:
3034      list[1] = (GET2(end, 1) == 0);      list[1] = (GET2(end, 1) == 0);
3035      end += 1 + 2 * IMM2_SIZE;      end += 1 + 2 * IMM2_SIZE;
3036      break;      break;
# Line 2895  Returns:      TRUE if the auto-possessif Line 3061  Returns:      TRUE if the auto-possessif
3061    
3062  static BOOL  static BOOL
3063  compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,  compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
3064    const pcre_uint32* base_list)    const pcre_uint32 *base_list, const pcre_uchar *base_end)
3065  {  {
3066  pcre_uchar c;  pcre_uchar c;
3067  pcre_uint32 list[8];  pcre_uint32 list[8];
3068  const pcre_uint32* chr_ptr;  const pcre_uint32 *chr_ptr;
3069  const pcre_uint32* ochr_ptr;  const pcre_uint32 *ochr_ptr;
3070  const pcre_uint32* list_ptr;  const pcre_uint32 *list_ptr;
3071    const pcre_uchar *next_code;
3072    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3073    const pcre_uchar *xclass_flags;
3074    #endif
3075    const pcre_uint8 *class_bitset;
3076    const pcre_uint8 *set1, *set2, *set_end;
3077  pcre_uint32 chr;  pcre_uint32 chr;
3078    BOOL accepted, invert_bits;
3079    
3080    /* Note: the base_list[1] contains whether the current opcode has greedy
3081    (represented by a non-zero value) quantifier. This is a different from
3082    other character type lists, which stores here that the character iterator
3083    matches to an empty string (also represented by a non-zero value). */
3084    
3085  for(;;)  for(;;)
3086    {    {
3087      /* All operations move the code pointer forward.
3088      Therefore infinite recursions are not possible. */
3089    
3090    c = *code;    c = *code;
3091    
3092    /* Skip over callouts */    /* Skip over callouts */
# Line 2925  for(;;) Line 3106  for(;;)
3106    switch(c)    switch(c)
3107      {      {
3108      case OP_END:      case OP_END:
3109      /* TRUE only in greedy case. The non-greedy case could be replaced by an      case OP_KETRPOS:
3110      OP_EXACT, but it is probably not worth it. (And note that OP_EXACT uses      /* TRUE only in greedy case. The non-greedy case could be replaced by
3111      more memory, which we cannot get at this stage.) */      an OP_EXACT, but it is probably not worth it. (And note that OP_EXACT
3112        uses more memory, which we cannot get at this stage.) */
3113    
3114      return base_list[1] != 0;      return base_list[1] != 0;
3115    
3116      case OP_KET:      case OP_KET:
3117      /* If the bracket is capturing, and referenced by an OP_RECURSE, the      /* If the bracket is capturing, and referenced by an OP_RECURSE, or
3118      non-greedy case cannot be converted to a possessive form. We do not test      it is an atomic sub-pattern (assert, once, etc.) the non-greedy case
3119      the bracket type at the moment, but we might do it in the future to improve      cannot be converted to a possessive form. */
     this condition. (But note that recursive calls are always atomic.) */  
3120    
3121      if (base_list[1] == 0) return FALSE;      if (base_list[1] == 0) return FALSE;
3122    
3123        switch(*(code - GET(code, 1)))
3124          {
3125          case OP_ASSERT:
3126          case OP_ASSERT_NOT:
3127          case OP_ASSERTBACK:
3128          case OP_ASSERTBACK_NOT:
3129          case OP_ONCE:
3130          case OP_ONCE_NC:
3131          /* Atomic sub-patterns and assertions can always auto-possessify their
3132          last iterator. */
3133          return TRUE;
3134          }
3135    
3136        code += PRIV(OP_lengths)[c];
3137        continue;
3138    
3139        case OP_ONCE:
3140        case OP_ONCE_NC:
3141        case OP_BRA:
3142        case OP_CBRA:
3143        next_code = code + GET(code, 1);
3144        code += PRIV(OP_lengths)[c];
3145    
3146        while (*next_code == OP_ALT)
3147          {
3148          if (!compare_opcodes(code, utf, cd, base_list, base_end)) return FALSE;
3149          code = next_code + 1 + LINK_SIZE;
3150          next_code += GET(next_code, 1);
3151          }
3152        continue;
3153    
3154        case OP_BRAZERO:
3155        case OP_BRAMINZERO:
3156    
3157        next_code = code + 1;
3158        if (*next_code != OP_BRA && *next_code != OP_CBRA
3159            && *next_code != OP_ONCE && *next_code != OP_ONCE_NC) return FALSE;
3160    
3161        do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
3162    
3163        /* The bracket content will be checked by the
3164        OP_BRA/OP_CBRA case above. */
3165        next_code += 1 + LINK_SIZE;
3166        if (!compare_opcodes(next_code, utf, cd, base_list, base_end))
3167          return FALSE;
3168    
3169      code += PRIV(OP_lengths)[c];      code += PRIV(OP_lengths)[c];
3170      continue;      continue;
3171      }      }
# Line 2961  for(;;) Line 3189  for(;;)
3189      list_ptr = base_list;      list_ptr = base_list;
3190      }      }
3191    
3192      /* Character bitsets can also be compared to certain opcodes. */
3193    
3194      else if (base_list[0] == OP_CLASS || list[0] == OP_CLASS
3195    #ifdef COMPILE_PCRE8
3196          /* In 8 bit, non-UTF mode, OP_CLASS and OP_NCLASS are the same. */
3197          || (!utf && (base_list[0] == OP_NCLASS || list[0] == OP_NCLASS))
3198    #endif
3199          )
3200        {
3201    #ifdef COMPILE_PCRE8
3202        if (base_list[0] == OP_CLASS || (!utf && base_list[0] == OP_NCLASS))
3203    #else
3204        if (base_list[0] == OP_CLASS)
3205    #endif
3206          {
3207          set1 = (pcre_uint8 *)(base_end - base_list[2]);
3208          list_ptr = list;
3209          }
3210        else
3211          {
3212          set1 = (pcre_uint8 *)(code - list[2]);
3213          list_ptr = base_list;
3214          }
3215    
3216        invert_bits = FALSE;
3217        switch(list_ptr[0])
3218          {
3219          case OP_CLASS:
3220          case OP_NCLASS:
3221          set2 = (pcre_uint8 *)
3222            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3223          break;
3224    
3225    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3226          case OP_XCLASS:
3227          xclass_flags = (list_ptr == list ? code : base_end) - list_ptr[2] + LINK_SIZE;
3228          if ((*xclass_flags & XCL_HASPROP) != 0) return FALSE;
3229          if ((*xclass_flags & XCL_MAP) == 0)
3230            {
3231            /* No bits are set for characters < 256. */
3232            if (list[1] == 0) return TRUE;
3233            /* Might be an empty repeat. */
3234            continue;
3235            }
3236          set2 = (pcre_uint8 *)(xclass_flags + 1);
3237          break;
3238    #endif
3239    
3240          case OP_NOT_DIGIT:
3241          invert_bits = TRUE;
3242          /* Fall through */
3243          case OP_DIGIT:
3244          set2 = (pcre_uint8 *)(cd->cbits + cbit_digit);
3245          break;
3246    
3247          case OP_NOT_WHITESPACE:
3248          invert_bits = TRUE;
3249          /* Fall through */
3250          case OP_WHITESPACE:
3251          set2 = (pcre_uint8 *)(cd->cbits + cbit_space);
3252          break;
3253    
3254          case OP_NOT_WORDCHAR:
3255          invert_bits = TRUE;
3256          /* Fall through */
3257          case OP_WORDCHAR:
3258          set2 = (pcre_uint8 *)(cd->cbits + cbit_word);
3259          break;
3260    
3261          default:
3262          return FALSE;
3263          }
3264    
3265        /* Because the sets are unaligned, we need
3266        to perform byte comparison here. */
3267        set_end = set1 + 32;
3268        if (invert_bits)
3269          {
3270          do
3271            {
3272            if ((*set1++ & ~(*set2++)) != 0) return FALSE;
3273            }
3274          while (set1 < set_end);
3275          }
3276        else
3277          {
3278          do
3279            {
3280            if ((*set1++ & *set2++) != 0) return FALSE;
3281            }
3282          while (set1 < set_end);
3283          }
3284    
3285        if (list[1] == 0) return TRUE;
3286        /* Might be an empty repeat. */
3287        continue;
3288        }
3289    
3290    /* Some property combinations also acceptable. Unicode property opcodes are    /* Some property combinations also acceptable. Unicode property opcodes are
3291    processed specially; the rest can be handled with a lookup table. */    processed specially; the rest can be handled with a lookup table. */
3292    
# Line 2968  for(;;) Line 3294  for(;;)
3294      {      {
3295      pcre_uint32 leftop, rightop;      pcre_uint32 leftop, rightop;
3296    
     if (list[1] != 0) return FALSE;   /* Must match at least one character */  
3297      leftop = base_list[0];      leftop = base_list[0];
3298      rightop = list[0];      rightop = list[0];
3299    
3300  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3301        accepted = FALSE; /* Always set in non-unicode case. */
3302      if (leftop == OP_PROP || leftop == OP_NOTPROP)      if (leftop == OP_PROP || leftop == OP_NOTPROP)
3303        {        {
3304        if (rightop == OP_EOD) return TRUE;        if (rightop == OP_EOD)
3305        if (rightop == OP_PROP || rightop == OP_NOTPROP)          accepted = TRUE;
3306          else if (rightop == OP_PROP || rightop == OP_NOTPROP)
3307          {          {
3308          int n;          int n;
3309          const pcre_uint8 *p;          const pcre_uint8 *p;
# Line 2997  for(;;) Line 3324  for(;;)
3324          n = propposstab[base_list[2]][list[2]];          n = propposstab[base_list[2]][list[2]];
3325          switch(n)          switch(n)
3326            {            {
3327            case 0: return FALSE;            case 0: break;
3328            case 1: return bothprop;            case 1: accepted = bothprop; break;
3329            case 2: return (base_list[3] == list[3]) != same;            case 2: accepted = (base_list[3] == list[3]) != same; break;
3330            case 3: return !same;            case 3: accepted = !same; break;
3331    
3332            case 4:  /* Left general category, right particular category */            case 4:  /* Left general category, right particular category */
3333            return risprop && catposstab[base_list[3]][list[3]] == same;            accepted = risprop && catposstab[base_list[3]][list[3]] == same;
3334              break;
3335    
3336            case 5:  /* Right general category, left particular category */            case 5:  /* Right general category, left particular category */
3337            return lisprop && catposstab[list[3]][base_list[3]] == same;            accepted = lisprop && catposstab[list[3]][base_list[3]] == same;
3338              break;
3339    
3340            /* This code is logically tricky. Think hard before fiddling with it.            /* This code is logically tricky. Think hard before fiddling with it.
3341            The posspropstab table has four entries per row. Each row relates to            The posspropstab table has four entries per row. Each row relates to
3342            one of PCRE's special properties such as ALNUM or SPACE or WORD.            one of PCRE's special properties such as ALNUM or SPACE or WORD.
3343            Only WORD actually needs all four entries, but using repeats for the            Only WORD actually needs all four entries, but using repeats for the
3344            others means they can all use the same code below.            others means they can all use the same code below.
3345    
3346            The first two entries in each row are Unicode general categories, and            The first two entries in each row are Unicode general categories, and
3347            apply always, because all the characters they include are part of the            apply always, because all the characters they include are part of the
3348            PCRE character set. The third and fourth entries are a general and a            PCRE character set. The third and fourth entries are a general and a
# Line 3023  for(;;) Line 3352  for(;;)
3352            category contains more characters than the specials that are defined            category contains more characters than the specials that are defined
3353            for the property being tested against. Therefore, it cannot be used            for the property being tested against. Therefore, it cannot be used
3354            in a NOTPROP case.            in a NOTPROP case.
3355    
3356            Example: the row for WORD contains ucp_L, ucp_N, ucp_P, ucp_Po.            Example: the row for WORD contains ucp_L, ucp_N, ucp_P, ucp_Po.
3357            Underscore is covered by ucp_P or ucp_Po. */            Underscore is covered by ucp_P or ucp_Po. */
3358    
# Line 3031  for(;;) Line 3360  for(;;)
3360            case 7:  /* Left space vs right general category */            case 7:  /* Left space vs right general category */
3361            case 8:  /* Left word vs right general category */            case 8:  /* Left word vs right general category */
3362            p = posspropstab[n-6];            p = posspropstab[n-6];
3363            return risprop && lisprop ==            accepted = risprop && lisprop ==
3364              (list[3] != p[0] &&              (list[3] != p[0] &&
3365               list[3] != p[1] &&               list[3] != p[1] &&
3366              (list[3] != p[2] || !lisprop));              (list[3] != p[2] || !lisprop));
3367              break;
3368    
3369            case 9:   /* Right alphanum vs left general category */            case 9:   /* Right alphanum vs left general category */
3370            case 10:  /* Right space vs left general category */            case 10:  /* Right space vs left general category */
3371            case 11:  /* Right word vs left general category */            case 11:  /* Right word vs left general category */
3372            p = posspropstab[n-9];            p = posspropstab[n-9];
3373            return lisprop && risprop ==            accepted = lisprop && risprop ==
3374              (base_list[3] != p[0] &&              (base_list[3] != p[0] &&
3375               base_list[3] != p[1] &&               base_list[3] != p[1] &&
3376              (base_list[3] != p[2] || !risprop));              (base_list[3] != p[2] || !risprop));
3377              break;
3378    
3379            case 12:  /* Left alphanum vs right particular category */            case 12:  /* Left alphanum vs right particular category */
3380            case 13:  /* Left space vs right particular category */            case 13:  /* Left space vs right particular category */
3381            case 14:  /* Left word vs right particular category */            case 14:  /* Left word vs right particular category */
3382            p = posspropstab[n-12];            p = posspropstab[n-12];
3383            return risprop && lisprop ==            accepted = risprop && lisprop ==
3384              (catposstab[p[0]][list[3]] &&              (catposstab[p[0]][list[3]] &&
3385               catposstab[p[1]][list[3]] &&               catposstab[p[1]][list[3]] &&
3386              (list[3] != p[3] || !lisprop));              (list[3] != p[3] || !lisprop));
3387              break;
3388    
3389            case 15:  /* Right alphanum vs left particular category */            case 15:  /* Right alphanum vs left particular category */
3390            case 16:  /* Right space vs left particular category */            case 16:  /* Right space vs left particular category */
3391            case 17:  /* Right word vs left particular category */            case 17:  /* Right word vs left particular category */
3392            p = posspropstab[n-15];            p = posspropstab[n-15];
3393            return lisprop && risprop ==            accepted = lisprop && risprop ==
3394              (catposstab[p[0]][base_list[3]] &&              (catposstab[p[0]][base_list[3]] &&
3395               catposstab[p[1]][base_list[3]] &&               catposstab[p[1]][base_list[3]] &&
3396              (base_list[3] != p[3] || !risprop));              (base_list[3] != p[3] || !risprop));
3397              break;
3398            }            }
3399          }          }
       return FALSE;  
3400        }        }
3401    
3402      else      else
3403  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3404    
3405      return leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&      accepted = leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&
3406             rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&             rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&
3407             autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];             autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];
3408    
3409        if (!accepted)
3410          return FALSE;
3411    
3412        if (list[1] == 0) return TRUE;
3413        /* Might be an empty repeat. */
3414        continue;
3415      }      }
3416    
3417    /* Control reaches here only if one of the items is a small character list.    /* Control reaches here only if one of the items is a small character list.
# Line 3186  for(;;) Line 3525  for(;;)
3525        case OP_EOD:    /* Can always possessify before \z */        case OP_EOD:    /* Can always possessify before \z */
3526        break;        break;
3527    
3528    #ifdef SUPPORT_UCP
3529        case OP_PROP:        case OP_PROP:
3530        case OP_NOTPROP:        case OP_NOTPROP:
3531        if (!check_char_prop(chr, list_ptr[2], list_ptr[3],        if (!check_char_prop(chr, list_ptr[2], list_ptr[3],
3532              list_ptr[0] == OP_NOTPROP))              list_ptr[0] == OP_NOTPROP))
3533          return FALSE;          return FALSE;
3534        break;        break;
3535    #endif
       /* The class comparisons work only when the class is the second item  
       of the pair, because there are at present no possessive forms of the  
       class opcodes. Note also that the "code" variable that is used below  
       points after the second item, and that the pointer for the first item  
       is not available, so even if there were possessive forms of the class  
       opcodes, the correct comparison could not be done. */  
3536    
3537        case OP_NCLASS:        case OP_NCLASS:
3538        if (chr > 255) return FALSE;        if (chr > 255) return FALSE;
3539        /* Fall through */        /* Fall through */
3540    
3541        case OP_CLASS:        case OP_CLASS:
       if (list_ptr != list) return FALSE;   /* Class is first opcode */  
3542        if (chr > 255) break;        if (chr > 255) break;
3543        if ((((pcre_uint8 *)(code - list_ptr[2] + 1))[chr >> 3] & (1 << (chr & 7))) != 0)        class_bitset = (pcre_uint8 *)
3544          return FALSE;          ((list_ptr == list ? code : base_end) - list_ptr[2]);
3545          if ((class_bitset[chr >> 3] & (1 << (chr & 7))) != 0) return FALSE;
3546        break;        break;
3547    
3548  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3549        case OP_XCLASS:        case OP_XCLASS:
3550        if (list_ptr != list) return FALSE;   /* Class is first opcode */        if (PRIV(xclass)(chr, (list_ptr == list ? code : base_end) -
3551        if (PRIV(xclass)(chr, code - list_ptr[2] + 1 + LINK_SIZE, utf))            list_ptr[2] + LINK_SIZE, utf)) return FALSE;
         return FALSE;  
3552        break;        break;
3553  #endif  #endif
3554    
# Line 3232  for(;;) Line 3565  for(;;)
3565    if (list[1] == 0) return TRUE;    if (list[1] == 0) return TRUE;
3566    }    }
3567    
3568  return FALSE;  /* Control never reaches here. There used to be a fail-save return FALSE; here,
3569    but some compilers complain about an unreachable statement. */
3570    
3571  }  }
3572    
3573    
# Line 3257  auto_possessify(pcre_uchar *code, BOOL u Line 3592  auto_possessify(pcre_uchar *code, BOOL u
3592  {  {
3593  register pcre_uchar c;  register pcre_uchar c;
3594  const pcre_uchar *end;  const pcre_uchar *end;
3595    pcre_uchar *repeat_opcode;
3596  pcre_uint32 list[8];  pcre_uint32 list[8];
3597    
3598  for (;;)  for (;;)
# Line 3270  for (;;) Line 3606  for (;;)
3606        get_chr_property_list(code, utf, cd->fcc, list) : NULL;        get_chr_property_list(code, utf, cd->fcc, list) : NULL;
3607      list[1] = c == OP_STAR || c == OP_PLUS || c == OP_QUERY || c == OP_UPTO;      list[1] = c == OP_STAR || c == OP_PLUS || c == OP_QUERY || c == OP_UPTO;
3608    
3609      if (end != NULL && compare_opcodes(end, utf, cd, list))      if (end != NULL && compare_opcodes(end, utf, cd, list, end))
3610        {        {
3611        switch(c)        switch(c)
3612          {          {
# Line 3303  for (;;) Line 3639  for (;;)
3639          break;          break;
3640    
3641          case OP_MINUPTO:          case OP_MINUPTO:
3642          *code += OP_MINUPTO - OP_UPTO;          *code += OP_POSUPTO - OP_MINUPTO;
3643          break;          break;
3644          }          }
3645        }        }
3646      c = *code;      c = *code;
3647      }      }
3648      else if (c == OP_CLASS || c == OP_NCLASS || c == OP_XCLASS)
3649        {
3650    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3651        if (c == OP_XCLASS)
3652          repeat_opcode = code + GET(code, 1);
3653        else
3654    #endif
3655          repeat_opcode = code + 1 + (32 / sizeof(pcre_uchar));
3656    
3657        c = *repeat_opcode;
3658        if (c >= OP_CRSTAR && c <= OP_CRMINRANGE)
3659          {
3660          /* end must not be NULL. */
3661          end = get_chr_property_list(code, utf, cd->fcc, list);
3662    
3663          list[1] = (c & 1) == 0;
3664    
3665          if (compare_opcodes(end, utf, cd, list, end))
3666            {
3667            switch (c)
3668              {
3669              case OP_CRSTAR:
3670              case OP_CRMINSTAR:
3671              *repeat_opcode = OP_CRPOSSTAR;
3672              break;
3673    
3674              case OP_CRPLUS:
3675              case OP_CRMINPLUS:
3676              *repeat_opcode = OP_CRPOSPLUS;
3677              break;
3678    
3679              case OP_CRQUERY:
3680              case OP_CRMINQUERY:
3681              *repeat_opcode = OP_CRPOSQUERY;
3682              break;
3683    
3684              case OP_CRRANGE:
3685              case OP_CRMINRANGE:
3686              *repeat_opcode = OP_CRPOSRANGE;
3687              break;
3688              }
3689            }
3690          }
3691        c = *code;
3692        }
3693    
3694    switch(c)    switch(c)
3695      {      {
# Line 3335  for (;;) Line 3716  for (;;)
3716        code += 2;        code += 2;
3717      break;      break;
3718    
3719    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3720      case OP_XCLASS:      case OP_XCLASS:
3721      code += GET(code, 1);      code += GET(code, 1);
3722      break;      break;
3723    #endif
3724    
3725      case OP_MARK:      case OP_MARK:
3726      case OP_PRUNE_ARG:      case OP_PRUNE_ARG:
# Line 3446  class, but [abc[:x\]pqr:]] is (so that a Line 3829  class, but [abc[:x\]pqr:]] is (so that a
3829  below handles the special case of \], but does not try to do any other escape  below handles the special case of \], but does not try to do any other escape
3830  processing. This makes it different from Perl for cases such as [:l\ower:]  processing. This makes it different from Perl for cases such as [:l\ower:]
3831  where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize  where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
3832  "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,  "l\ower". This is a lesser evil than not diagnosing bad classes when Perl does,
3833  I think.  I think.
3834    
3835  A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.  A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
# Line 3695  for (c = *cptr; c <= d; c++) Line 4078  for (c = *cptr; c <= d; c++)
4078    
4079  if (c > d) return -1;  /* Reached end of range */  if (c > d) return -1;  /* Reached end of range */
4080    
4081    /* Found a character that has a single other case. Search for the end of the
4082    range, which is either the end of the input range, or a character that has zero
4083    or more than one other cases. */
4084    
4085  *ocptr = othercase;  *ocptr = othercase;
4086  next = othercase + 1;  next = othercase + 1;
4087    
4088  for (++c; c <= d; c++)  for (++c; c <= d; c++)
4089    {    {
4090    if (UCD_OTHERCASE(c) != next) break;    if ((co = UCD_CASESET(c)) != 0 || UCD_OTHERCASE(c) != next) break;
4091    next++;    next++;
4092    }    }
4093    
# Line 3738  add_to_class(pcre_uint8 *classbits, pcre Line 4125  add_to_class(pcre_uint8 *classbits, pcre
4125    compile_data *cd, pcre_uint32 start, pcre_uint32 end)    compile_data *cd, pcre_uint32 start, pcre_uint32 end)
4126  {  {
4127  pcre_uint32 c;  pcre_uint32 c;
4128    pcre_uint32 classbits_end = (end <= 0xff ? end : 0xff);
4129  int n8 = 0;  int n8 = 0;
4130    
4131  /* If caseless matching is required, scan the range and process alternate  /* If caseless matching is required, scan the range and process alternate
# Line 3781  if ((options & PCRE_CASELESS) != 0) Line 4169  if ((options & PCRE_CASELESS) != 0)
4169    
4170    /* Not UTF-mode, or no UCP */    /* Not UTF-mode, or no UCP */
4171    
4172    for (c = start; c <= end && c < 256; c++)    for (c = start; c <= classbits_end; c++)
4173      {      {
4174      SETBIT(classbits, cd->fcc[c]);      SETBIT(classbits, cd->fcc[c]);
4175      n8++;      n8++;
# Line 3806  in all cases. */ Line 4194  in all cases. */
4194    
4195  #endif /* COMPILE_PCRE[8|16] */  #endif /* COMPILE_PCRE[8|16] */
4196    
4197  /* If all characters are less than 256, use the bit map. Otherwise use extra  /* Use the bitmap for characters < 256. Otherwise use extra data.*/
 data. */  
4198    
4199  if (end < 0x100)  for (c = start; c <= classbits_end; c++)
4200    {    {
4201    for (c = start; c <= end; c++)    /* Regardless of start, c will always be <= 255. */
4202      {    SETBIT(classbits, c);
4203      n8++;    n8++;
     SETBIT(classbits, c);  
     }  
4204    }    }
4205    
4206  else  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4207    if (start <= 0xff) start = 0xff + 1;
4208    
4209    if (end >= start)
4210    {    {
4211    pcre_uchar *uchardata = *uchardptr;    pcre_uchar *uchardata = *uchardptr;
   
4212  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4213    if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */    if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */
4214      {      {
# Line 3861  else Line 4248  else
4248    
4249    *uchardptr = uchardata;   /* Updata extra data pointer */    *uchardptr = uchardata;   /* Updata extra data pointer */
4250    }    }
4251    #endif /* SUPPORT_UTF || !COMPILE_PCRE8 */
4252    
4253  return n8;    /* Number of 8-bit characters */  return n8;    /* Number of 8-bit characters */
4254  }  }
# Line 4082  for (;; ptr++) Line 4470  for (;; ptr++)
4470    BOOL reset_bracount;    BOOL reset_bracount;
4471    int class_has_8bitchar;    int class_has_8bitchar;
4472    int class_one_char;    int class_one_char;
4473    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4474      BOOL xclass_has_prop;
4475    #endif
4476    int newoptions;    int newoptions;
4477    int recno;    int recno;
4478    int refsign;    int refsign;
# Line 4199  for (;; ptr++) Line 4590  for (;; ptr++)
4590          }          }
4591        goto NORMAL_CHAR;        goto NORMAL_CHAR;
4592        }        }
4593        /* Control does not reach here. */
4594      }      }
4595    
4596    /* Fill in length of a previous callout, except when the next thing is    /* In extended mode, skip white space and comments. We need a loop in order
4597    a quantifier. */    to check for more white space and more comments after a comment. */
   
   is_quantifier =  
     c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||  
     (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));  
   
   if (!is_quantifier && previous_callout != NULL &&  
        after_manual_callout-- <= 0)  
     {  
     if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */  
       complete_callout(previous_callout, ptr, cd);  
     previous_callout = NULL;  
     }  
   
   /* In extended mode, skip white space and comments. */  
4598    
4599    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
4600      {      {
4601      if (MAX_255(*ptr) && (cd->ctypes[c] & ctype_space) != 0) continue;      for (;;)
     if (c == CHAR_NUMBER_SIGN)  
4602        {        {
4603          while (MAX_255(c) && (cd->ctypes[c] & ctype_space) != 0) c = *(++ptr);
4604          if (c != CHAR_NUMBER_SIGN) break;
4605        ptr++;        ptr++;
4606        while (*ptr != CHAR_NULL)        while (*ptr != CHAR_NULL)
4607          {          {
4608          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr))         /* For non-fixed-length newline cases, */
4609              {                          /* IS_NEWLINE sets cd->nllen. */
4610              ptr += cd->nllen;
4611              break;
4612              }
4613          ptr++;          ptr++;
4614  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4615          if (utf) FORWARDCHAR(ptr);          if (utf) FORWARDCHAR(ptr);
4616  #endif  #endif
4617          }          }
4618        if (*ptr != CHAR_NULL) continue;        c = *ptr;     /* Either NULL or the char after a newline */
   
       /* Else fall through to handle end of string */  
       c = 0;  
4619        }        }
4620      }      }
4621    
4622    /* No auto callout for quantifiers. */    /* See if the next thing is a quantifier. */
4623    
4624    if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier)    is_quantifier =
4625        c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
4626        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
4627    
4628      /* Fill in length of a previous callout, except when the next thing is a
4629      quantifier or when processing a property substitution string in UCP mode. */
4630    
4631      if (!is_quantifier && previous_callout != NULL && nestptr == NULL &&
4632           after_manual_callout-- <= 0)
4633        {
4634        if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */
4635          complete_callout(previous_callout, ptr, cd);
4636        previous_callout = NULL;
4637        }
4638    
4639      /* Create auto callout, except for quantifiers, or while processing property
4640      strings that are substituted for \w etc in UCP mode. */
4641    
4642      if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier && nestptr == NULL)
4643      {      {
4644      previous_callout = code;      previous_callout = code;
4645      code = auto_callout(code, ptr, cd);      code = auto_callout(code, ptr, cd);
4646      }      }
4647    
4648      /* Process the next pattern item. */
4649    
4650    switch(c)    switch(c)
4651      {      {
4652      /* ===================================================================*/      /* ===================================================================*/
4653      case 0:                        /* The branch terminates at string end */      case CHAR_NULL:                /* The branch terminates at string end */
4654      case CHAR_VERTICAL_LINE:       /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
4655      case CHAR_RIGHT_PARENTHESIS:      case CHAR_RIGHT_PARENTHESIS:
4656      *firstcharptr = firstchar;      *firstcharptr = firstchar;
# Line 4328  for (;; ptr++) Line 4728  for (;; ptr++)
4728        }        }
4729      goto NORMAL_CHAR;      goto NORMAL_CHAR;
4730    
4731        /* In another (POSIX) regex library, the ugly syntax [[:<:]] and [[:>:]] is
4732        used for "start of word" and "end of word". As these are otherwise illegal
4733        sequences, we don't break anything by recognizing them. They are replaced
4734        by \b(?=\w) and \b(?<=\w) respectively. Sequences like [a[:<:]] are
4735        erroneous and are handled by the normal code below. */
4736    
4737      case CHAR_LEFT_SQUARE_BRACKET:      case CHAR_LEFT_SQUARE_BRACKET:
4738        if (STRNCMP_UC_C8(ptr+1, STRING_WEIRD_STARTWORD, 6) == 0)
4739          {
4740          nestptr = ptr + 7;
4741          ptr = sub_start_of_word - 1;
4742          continue;
4743          }
4744    
4745        if (STRNCMP_UC_C8(ptr+1, STRING_WEIRD_ENDWORD, 6) == 0)
4746          {
4747          nestptr = ptr + 7;
4748          ptr = sub_end_of_word - 1;
4749          continue;
4750          }
4751    
4752        /* Handle a real character class. */
4753    
4754      previous = code;      previous = code;
4755    
4756      /* 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
# Line 4385  for (;; ptr++) Line 4807  for (;; ptr++)
4807    
4808      should_flip_negation = FALSE;      should_flip_negation = FALSE;
4809    
4810        /* Extended class (xclass) will be used when characters > 255
4811        might match. */
4812    
4813    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4814        xclass = FALSE;
4815        class_uchardata = code + LINK_SIZE + 2;   /* For XCLASS items */
4816        class_uchardata_base = class_uchardata;   /* Save the start */
4817    #endif
4818    
4819      /* For optimization purposes, we track some properties of the class:      /* For optimization purposes, we track some properties of the class:
4820      class_has_8bitchar will be non-zero if the class contains at least one <      class_has_8bitchar will be non-zero if the class contains at least one <
4821      256 character; class_one_char will be 1 if the class contains just one      256 character; class_one_char will be 1 if the class contains just one
4822      character. */      character; xclass_has_prop will be TRUE if unicode property checks
4823        are present in the class. */
4824    
4825      class_has_8bitchar = 0;      class_has_8bitchar = 0;
4826      class_one_char = 0;      class_one_char = 0;
4827    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4828        xclass_has_prop = FALSE;
4829    #endif
4830    
4831      /* Initialize the 32-char bit map to all zeros. We build the map in a      /* Initialize the 32-char bit map to all zeros. We build the map in a
4832      temporary bit of memory, in case the class contains fewer than two      temporary bit of memory, in case the class contains fewer than two
# Line 4400  for (;; ptr++) Line 4835  for (;; ptr++)
4835    
4836      memset(classbits, 0, 32 * sizeof(pcre_uint8));      memset(classbits, 0, 32 * sizeof(pcre_uint8));
4837    
 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8  
     xclass = FALSE;  
     class_uchardata = code + LINK_SIZE + 2;   /* For XCLASS items */  
     class_uchardata_base = class_uchardata;   /* Save the start */  
 #endif  
   
4838      /* Process characters until ] is reached. By writing this as a "do" it      /* Process characters until ] is reached. By writing this as a "do" it
4839      means that an initial ] is taken as a data character. At the start of the      means that an initial ] is taken as a data character. At the start of the
4840      loop, c contains the first byte of the character. */      loop, c contains the first byte of the character. */
# Line 4493  for (;; ptr++) Line 4922  for (;; ptr++)
4922            posix_class = 0;            posix_class = 0;
4923    
4924          /* When PCRE_UCP is set, some of the POSIX classes are converted to          /* When PCRE_UCP is set, some of the POSIX classes are converted to
4925          different escape sequences that use Unicode properties. */          different escape sequences that use Unicode properties \p or \P. Others
4926            that are not available via \p or \P generate XCL_PROP/XCL_NOTPROP
4927            directly. */
4928    
4929  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4930          if ((options & PCRE_UCP) != 0)          if ((options & PCRE_UCP) != 0)
4931            {            {
4932              unsigned int ptype = 0;
4933            int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);            int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
4934    
4935              /* The posix_substitutes table specifies which POSIX classes can be
4936              converted to \p or \P items. */
4937    
4938            if (posix_substitutes[pc] != NULL)            if (posix_substitutes[pc] != NULL)
4939              {              {
4940              nestptr = tempptr + 1;              nestptr = tempptr + 1;
4941              ptr = posix_substitutes[pc] - 1;              ptr = posix_substitutes[pc] - 1;
4942              continue;              continue;
4943              }              }
4944    
4945              /* There are three other classes that generate special property calls
4946              that are recognized only in an XCLASS. */
4947    
4948              else switch(posix_class)
4949                {
4950                case PC_GRAPH:
4951                ptype = PT_PXGRAPH;
4952                /* Fall through */
4953                case PC_PRINT:
4954                if (ptype == 0) ptype = PT_PXPRINT;
4955                /* Fall through */
4956                case PC_PUNCT:
4957                if (ptype == 0) ptype = PT_PXPUNCT;
4958                *class_uchardata++ = local_negate? XCL_NOTPROP : XCL_PROP;
4959                *class_uchardata++ = ptype;
4960                *class_uchardata++ = 0;
4961                xclass_has_prop = TRUE;
4962                ptr = tempptr + 1;
4963                continue;
4964    
4965                /* For all other POSIX classes, no special action is taken in UCP
4966                mode. Fall through to the non_UCP case. */
4967    
4968                default:
4969                break;
4970                }
4971            }            }
4972  #endif  #endif
4973          /* In the non-UCP case, we build the bit map for the POSIX class in a          /* In the non-UCP case, or when UCP makes no difference, we build the
4974          chunk of local store because we may be adding and subtracting from it,          bit map for the POSIX class in a chunk of local store because we may be
4975          and we don't want to subtract bits that may be in the main map already.          adding and subtracting from it, and we don't want to subtract bits that
4976          At the end we or the result into the bit map that is being built. */          may be in the main map already. At the end we or the result into the
4977            bit map that is being built. */
4978    
4979          posix_class *= 3;          posix_class *= 3;
4980    
# Line 4627  for (;; ptr++) Line 5091  for (;; ptr++)
5091              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
5092              continue;              continue;
5093    
5094              /* Perl 5.004 onwards omits VT from \s, but we must preserve it              /* Perl 5.004 onwards omitted VT from \s, but restored it at Perl
5095              if it was previously set by something earlier in the character              5.18. Before PCRE 8.34, we had to preserve the VT bit if it was
5096              class. Luckily, the value of CHAR_VT is 0x0b in both ASCII and              previously set by something earlier in the character class.
5097              EBCDIC, so we lazily just adjust the appropriate bit. */              Luckily, the value of CHAR_VT is 0x0b in both ASCII and EBCDIC, so
5098                we could just adjust the appropriate bit. From PCRE 8.34 we no
5099                longer treat \s and \S specially. */
5100    
5101              case ESC_s:              case ESC_s:
5102              classbits[0] |= cbits[cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
             classbits[1] |= cbits[cbit_space+1] & ~0x08;  
             for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];  
5103              continue;              continue;
5104    
5105              case ESC_S:              case ESC_S:
5106              should_flip_negation = TRUE;              should_flip_negation = TRUE;
5107              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
             classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */  
5108              continue;              continue;
5109    
5110              /* The rest apply in both UCP and non-UCP cases. */              /* The rest apply in both UCP and non-UCP cases. */
# Line 4678  for (;; ptr++) Line 5141  for (;; ptr++)
5141                  XCL_PROP : XCL_NOTPROP;                  XCL_PROP : XCL_NOTPROP;
5142                *class_uchardata++ = ptype;                *class_uchardata++ = ptype;
5143                *class_uchardata++ = pdata;                *class_uchardata++ = pdata;
5144                  xclass_has_prop = TRUE;
5145                class_has_8bitchar--;                /* Undo! */                class_has_8bitchar--;                /* Undo! */
5146                continue;                continue;
5147                }                }
# Line 4763  for (;; ptr++) Line 5227  for (;; ptr++)
5227  #endif  #endif
5228          d = *ptr;  /* Not UTF-8 mode */          d = *ptr;  /* Not UTF-8 mode */
5229    
5230          /* The second part of a range can be a single-character escape, but          /* The second part of a range can be a single-character escape
5231          not any of the other escapes. Perl 5.6 treats a hyphen as a literal          sequence, but not any of the other escapes. Perl treats a hyphen as a
5232          in such circumstances. */          literal in such circumstances. However, in Perl's warning mode, a
5233            warning is given, so PCRE now faults it as it is almost certainly a
5234            mistake on the user's part. */
5235    
5236          if (!inescq && d == CHAR_BACKSLASH)          if (!inescq)
5237            {            {
5238            int descape;            if (d == CHAR_BACKSLASH)
5239            descape = check_escape(&ptr, &d, errorcodeptr, cd->bracount, options, TRUE);              {
5240            if (*errorcodeptr != 0) goto FAILED;              int descape;
5241                descape = check_escape(&ptr, &d, errorcodeptr, cd->bracount, options, TRUE);
5242                if (*errorcodeptr != 0) goto FAILED;
5243    
5244            /* \b is backspace; any other special means the '-' was literal. */              /* 0 means a character was put into d; \b is backspace; any other
5245                special causes an error. */
5246    
5247            if (descape != 0)              if (descape != 0)
             {  
             if (descape == ESC_b) d = CHAR_BS; else  
5248                {                {
5249                ptr = oldptr;                if (descape == ESC_b) d = CHAR_BS; else
5250                goto CLASS_SINGLE_CHARACTER;  /* A few lines below */                  {
5251                    *errorcodeptr = ERR83;
5252                    goto FAILED;
5253                    }
5254                }                }
5255              }              }
5256    
5257              /* A hyphen followed by a POSIX class is treated in the same way. */
5258    
5259              else if (d == CHAR_LEFT_SQUARE_BRACKET &&
5260                       (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
5261                        ptr[1] == CHAR_EQUALS_SIGN) &&
5262                       check_posix_syntax(ptr, &tempptr))
5263                {
5264                *errorcodeptr = ERR83;
5265                goto FAILED;
5266                }
5267            }            }
5268    
5269          /* Check that the two values are in the correct order. Optimize          /* Check that the two values are in the correct order. Optimize
# Line 4955  for (;; ptr++) Line 5436  for (;; ptr++)
5436        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
5437        code += LINK_SIZE;        code += LINK_SIZE;
5438        *code = negate_class? XCL_NOT:0;        *code = negate_class? XCL_NOT:0;
5439          if (xclass_has_prop) *code |= XCL_HASPROP;
5440    
5441        /* If the map is required, move up the extra data to make room for it;        /* If the map is required, move up the extra data to make room for it;
5442        otherwise just move the code pointer to the end of the extra data. */        otherwise just move the code pointer to the end of the extra data. */
# Line 4964  for (;; ptr++) Line 5446  for (;; ptr++)
5446          *code++ |= XCL_MAP;          *code++ |= XCL_MAP;
5447          memmove(code + (32 / sizeof(pcre_uchar)), code,          memmove(code + (32 / sizeof(pcre_uchar)), code,
5448            IN_UCHARS(class_uchardata - code));            IN_UCHARS(class_uchardata - code));
5449            if (negate_class && !xclass_has_prop)
5450              for (c = 0; c < 32; c++) classbits[c] = ~classbits[c];
5451          memcpy(code, classbits, 32);          memcpy(code, classbits, 32);
5452          code = class_uchardata + (32 / sizeof(pcre_uchar));          code = class_uchardata + (32 / sizeof(pcre_uchar));
5453          }          }
# Line 5046  for (;; ptr++) Line 5530  for (;; ptr++)
5530    
5531      tempcode = previous;      tempcode = previous;
5532    
5533        /* Before checking for a possessive quantifier, we must skip over
5534        whitespace and comments in extended mode because Perl allows white space at
5535        this point. */
5536    
5537        if ((options & PCRE_EXTENDED) != 0)
5538          {
5539          const pcre_uchar *p = ptr + 1;
5540          for (;;)
5541            {
5542            while (MAX_255(*p) && (cd->ctypes[*p] & ctype_space) != 0) p++;
5543            if (*p != CHAR_NUMBER_SIGN) break;
5544            p++;
5545            while (*p != CHAR_NULL)
5546              {
5547              if (IS_NEWLINE(p))         /* For non-fixed-length newline cases, */
5548                {                        /* IS_NEWLINE sets cd->nllen. */
5549                p += cd->nllen;
5550                break;
5551                }
5552              p++;
5553    #ifdef SUPPORT_UTF
5554              if (utf) FORWARDCHAR(p);
5555    #endif
5556              }           /* Loop for comment characters */
5557            }             /* Loop for multiple comments */
5558          ptr = p - 1;    /* Character before the next significant one. */
5559          }
5560    
5561      /* If the next character is '+', we have a possessive quantifier. This      /* If the next character is '+', we have a possessive quantifier. This
5562      implies greediness, whatever the setting of the PCRE_UNGREEDY option.      implies greediness, whatever the setting of the PCRE_UNGREEDY option.
5563      If the next character is '?' this is a minimizing repeat, by default,      If the next character is '?' this is a minimizing repeat, by default,
# Line 5338  for (;; ptr++) Line 5850  for (;; ptr++)
5850      opcodes such as BRA and CBRA, as this is the place where they get converted      opcodes such as BRA and CBRA, as this is the place where they get converted
5851      into the more special varieties such as BRAPOS and SBRA. A test for >=      into the more special varieties such as BRAPOS and SBRA. A test for >=
5852      OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,      OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
5853      ASSERTBACK_NOT, ONCE, BRA, CBRA, and COND. Originally, PCRE did not allow      ASSERTBACK_NOT, ONCE, ONCE_NC, BRA, BRAPOS, CBRA, CBRAPOS, and COND.
5854      repetition of assertions, but now it does, for Perl compatibility. */      Originally, PCRE did not allow repetition of assertions, but now it does,
5855        for Perl compatibility. */
5856    
5857      else if (*previous >= OP_ASSERT && *previous <= OP_COND)      else if (*previous >= OP_ASSERT && *previous <= OP_COND)
5858        {        {
# Line 5357  for (;; ptr++) Line 5870  for (;; ptr++)
5870        /* There is no sense in actually repeating assertions. The only potential        /* There is no sense in actually repeating assertions. The only potential
5871        use of repetition is in cases when the assertion is optional. Therefore,        use of repetition is in cases when the assertion is optional. Therefore,
5872        if the minimum is greater than zero, just ignore the repeat. If the        if the minimum is greater than zero, just ignore the repeat. If the
5873        maximum is not not zero or one, set it to 1. */        maximum is not zero or one, set it to 1. */
5874    
5875        if (*previous < OP_ONCE)    /* Assertion */        if (*previous < OP_ONCE)    /* Assertion */
5876          {          {
# Line 5730  for (;; ptr++) Line 6243  for (;; ptr++)
6243        goto FAILED;        goto FAILED;
6244        }        }
6245    
6246      /* If the character following a repeat is '+', or if certain optimization      /* If the character following a repeat is '+', possessive_quantifier is
6247      tests above succeeded, possessive_quantifier is TRUE. For some opcodes,      TRUE. For some opcodes, there are special alternative opcodes for this
6248      there are special alternative opcodes for this case. For anything else, we      case. For anything else, we wrap the entire repeated item inside OP_ONCE
6249      wrap the entire repeated item inside OP_ONCE brackets. Logically, the '+'      brackets. Logically, the '+' notation is just syntactic sugar, taken from
6250      notation is just syntactic sugar, taken from Sun's Java package, but the      Sun's Java package, but the special opcodes can optimize it.
     special opcodes can optimize it.  
6251    
6252      Some (but not all) possessively repeated subpatterns have already been      Some (but not all) possessively repeated subpatterns have already been
6253      completely handled in the code just above. For them, possessive_quantifier      completely handled in the code just above. For them, possessive_quantifier
6254      is always FALSE at this stage.      is always FALSE at this stage. Note that the repeated item starts at
6255        tempcode, not at previous, which might be the first part of a string whose
6256      Note that the repeated item starts at tempcode, not at previous, which      (former) last char we repeated. */
     might be the first part of a string whose (former) last char we repeated.  
   
     Possessifying an 'exact' quantifier has no effect, so we can ignore it. But  
     an 'upto' may follow. We skip over an 'exact' item, and then test the  
     length of what remains before proceeding. */  
6257    
6258      if (possessive_quantifier)      if (possessive_quantifier)
6259        {        {
6260        int len;        int len;
6261    
6262        if (*tempcode == OP_TYPEEXACT)        /* Possessifying an EXACT quantifier has no effect, so we can ignore it.
6263          However, QUERY, STAR, or UPTO may follow (for quantifiers such as {5,6},
6264          {5,}, or {5,10}). We skip over an EXACT item; if the length of what
6265          remains is greater than zero, there's a further opcode that can be
6266          handled. If not, do nothing, leaving the EXACT alone. */
6267    
6268          switch(*tempcode)
6269            {
6270            case OP_TYPEEXACT:
6271          tempcode += PRIV(OP_lengths)[*tempcode] +          tempcode += PRIV(OP_lengths)[*tempcode] +
6272            ((tempcode[1 + IMM2_SIZE] == OP_PROP            ((tempcode[1 + IMM2_SIZE] == OP_PROP
6273            || tempcode[1 + IMM2_SIZE] == OP_NOTPROP)? 2 : 0);            || tempcode[1 + IMM2_SIZE] == OP_NOTPROP)? 2 : 0);
6274            break;
6275    
6276        else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)          /* CHAR opcodes are used for exacts whose count is 1. */
6277          {  
6278            case OP_CHAR:
6279            case OP_CHARI:
6280            case OP_NOT:
6281            case OP_NOTI:
6282            case OP_EXACT:
6283            case OP_EXACTI:
6284            case OP_NOTEXACT:
6285            case OP_NOTEXACTI:
6286          tempcode += PRIV(OP_lengths)[*tempcode];          tempcode += PRIV(OP_lengths)[*tempcode];
6287  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
6288          if (utf && HAS_EXTRALEN(tempcode[-1]))          if (utf && HAS_EXTRALEN(tempcode[-1]))
6289            tempcode += GET_EXTRALEN(tempcode[-1]);            tempcode += GET_EXTRALEN(tempcode[-1]);
6290  #endif  #endif
6291            break;
6292    
6293            /* For the class opcodes, the repeat operator appears at the end;
6294            adjust tempcode to point to it. */
6295    
6296            case OP_CLASS:
6297            case OP_NCLASS:
6298            tempcode += 1 + 32/sizeof(pcre_uchar);
6299            break;
6300    
6301    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
6302            case OP_XCLASS:
6303            tempcode += GET(tempcode, 1);
6304            break;
6305    #endif
6306          }          }
6307    
6308          /* If tempcode is equal to code (which points to the end of the repeated
6309          item), it means we have skipped an EXACT item but there is no following
6310          QUERY, STAR, or UPTO; the value of len will be 0, and we do nothing. In
6311          all other cases, tempcode will be pointing to the repeat opcode, and will
6312          be less than code, so the value of len will be greater than 0. */
6313    
6314        len = (int)(code - tempcode);        len = (int)(code - tempcode);
6315          if (len > 0)
6316            {
6317            unsigned int repcode = *tempcode;
6318    
6319            /* There is a table for possessifying opcodes, all of which are less
6320            than OP_CALLOUT. A zero entry means there is no possessified version.
6321            */
6322    
6323            if (repcode < OP_CALLOUT && opcode_possessify[repcode] > 0)
6324              *tempcode = opcode_possessify[repcode];
6325    
6326            /* For opcode without a special possessified version, wrap the item in
6327            ONCE brackets. Because we are moving code along, we must ensure that any
6328            pending recursive references are updated. */
6329    
6330            else
6331              {
6332              *code = OP_END;
6333              adjust_recurse(tempcode, 1 + LINK_SIZE, utf, cd, save_hwm);
6334              memmove(tempcode + 1 + LINK_SIZE, tempcode, IN_UCHARS(len));
6335              code += 1 + LINK_SIZE;
6336              len += 1 + LINK_SIZE;
6337              tempcode[0] = OP_ONCE;
6338              *code++ = OP_KET;
6339              PUTINC(code, 0, len);
6340              PUT(tempcode, 1, len);
6341              }
6342            }
6343    
6344    #ifdef NEVER
6345        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
6346          {          {
6347          case OP_STAR:  *tempcode = OP_POSSTAR; break;          case OP_STAR:  *tempcode = OP_POSSTAR; break;
# Line 5794  for (;; ptr++) Line 6369  for (;; ptr++)
6369          case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;          case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
6370          case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;          case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;
6371    
6372            case OP_CRSTAR:   *tempcode = OP_CRPOSSTAR; break;
6373            case OP_CRPLUS:   *tempcode = OP_CRPOSPLUS; break;
6374            case OP_CRQUERY:  *tempcode = OP_CRPOSQUERY; break;
6375            case OP_CRRANGE:  *tempcode = OP_CRPOSRANGE; break;
6376    
6377          /* Because we are moving code along, we must ensure that any          /* Because we are moving code along, we must ensure that any
6378          pending recursive references are updated. */          pending recursive references are updated. */
6379    
# Line 5809  for (;; ptr++) Line 6389  for (;; ptr++)
6389          PUT(tempcode, 1, len);          PUT(tempcode, 1, len);
6390          break;          break;
6391          }          }
6392    #endif
6393        }        }
6394    
6395      /* In all case we no longer have a previous item. We also set the      /* In all case we no longer have a previous item. We also set the
# Line 5997  for (;; ptr++) Line 6578  for (;; ptr++)
6578          tempptr = ptr;          tempptr = ptr;
6579    
6580          /* A condition can be an assertion, a number (referring to a numbered          /* A condition can be an assertion, a number (referring to a numbered
6581          group), a name (referring to a named group), or 'R', referring to          group's having been set), a name (referring to a named group), or 'R',
6582          recursion. R<digits> and R&name are also permitted for recursion tests.          referring to recursion. R<digits> and R&name are also permitted for
6583            recursion tests.
6584          There are several syntaxes for testing a named group: (?(name)) is used  
6585          by Python; Perl 5.10 onwards uses (?(<name>) or (?('name')).          There are ways of testing a named group: (?(name)) is used by Python;
6586            Perl 5.10 onwards uses (?(<name>) or (?('name')).
6587          There are two unfortunate ambiguities, caused by history. (a) 'R' can  
6588          be the recursive thing or the name 'R' (and similarly for 'R' followed          There is one unfortunate ambiguity, caused by history. 'R' can be the
6589          by digits), and (b) a number could be a name that consists of digits.          recursive thing or the name 'R' (and similarly for 'R' followed by
6590          In both cases, we look for a name first; if not found, we try the other          digits). We look for a name first; if not found, we try the other case.
         cases.  
6591    
6592          For compatibility with auto-callouts, we allow a callout to be          For compatibility with auto-callouts, we allow a callout to be
6593          specified before a condition that is an assertion. First, check for the          specified before a condition that is an assertion. First, check for the
# Line 6031  for (;; ptr++) Line 6611  for (;; ptr++)
6611                 tempptr[2] == CHAR_LESS_THAN_SIGN))                 tempptr[2] == CHAR_LESS_THAN_SIGN))
6612            break;            break;
6613    
6614          /* Most other conditions use OP_CREF (a couple change to OP_RREF          /* Other conditions use OP_CREF/OP_DNCREF/OP_RREF/OP_DNRREF, and all
6615          below), and all need to skip 1+IMM2_SIZE bytes at the start of the group. */          need to skip at least 1+IMM2_SIZE bytes at the start of the group. */
6616    
6617          code[1+LINK_SIZE] = OP_CREF;          code[1+LINK_SIZE] = OP_CREF;
6618          skipbytes = 1+IMM2_SIZE;          skipbytes = 1+IMM2_SIZE;
6619          refsign = -1;          refsign = -1;     /* => not a number */
6620            namelen = -1;     /* => not a name; must set to avoid warning */
6621            name = NULL;      /* Always set to avoid warning */
6622            recno = 0;        /* Always set to avoid warning */
6623    
6624          /* Check for a test for recursion in a named group. */          /* Check for a test for recursion in a named group. */
6625    
6626          if (ptr[1] == CHAR_R && ptr[2] == CHAR_AMPERSAND)          ptr++;
6627            if (*ptr == CHAR_R && ptr[1] == CHAR_AMPERSAND)
6628            {            {
6629            terminator = -1;            terminator = -1;
6630            ptr += 2;            ptr += 2;
# Line 6048  for (;; ptr++) Line 6632  for (;; ptr++)
6632            }            }
6633    
6634          /* 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
6635          syntax (?(<name>) or (?('name') */          syntax (?(<name>) or (?('name'), and also allow for the original PCRE
6636            syntax of (?(name) or for (?(+n), (?(-n), and just (?(n). */
6637    
6638          else if (ptr[1] == CHAR_LESS_THAN_SIGN)          else if (*ptr == CHAR_LESS_THAN_SIGN)
6639            {            {
6640            terminator = CHAR_GREATER_THAN_SIGN;            terminator = CHAR_GREATER_THAN_SIGN;
6641            ptr++;            ptr++;
6642            }            }
6643          else if (ptr[1] == CHAR_APOSTROPHE)          else if (*ptr == CHAR_APOSTROPHE)
6644            {            {
6645            terminator = CHAR_APOSTROPHE;            terminator = CHAR_APOSTROPHE;
6646            ptr++;            ptr++;
# Line 6063  for (;; ptr++) Line 6648  for (;; ptr++)
6648          else          else
6649            {            {
6650            terminator = CHAR_NULL;            terminator = CHAR_NULL;
6651            if (ptr[1] == CHAR_MINUS || ptr[1] == CHAR_PLUS) refsign = *(++ptr);            if (*ptr == CHAR_MINUS || *ptr == CHAR_PLUS) refsign = *ptr++;
6652                else if (IS_DIGIT(*ptr)) refsign = 0;
6653            }            }
6654    
6655          /* We now expect to read a name; any thing else is an error */          /* Handle a number */
6656    
6657          if (!MAX_255(ptr[1]) || (cd->ctypes[ptr[1]] & ctype_word) == 0)          if (refsign >= 0)
6658            {            {
6659            ptr += 1;  /* To get the right offset */            while (IS_DIGIT(*ptr))
6660            *errorcodeptr = ERR28;              {
6661            goto FAILED;              recno = recno * 10 + (int)(*ptr - CHAR_0);
6662                ptr++;
6663                }
6664            }            }
6665    
6666          /* Read the name, but also get it as a number if it's all digits */          /* Otherwise we expect to read a name; anything else is an error. When
6667            a name is one of a number of duplicates, a different opcode is used and
6668            it needs more memory. Unfortunately we cannot tell whether a name is a
6669            duplicate in the first pass, so we have to allow for more memory. */
6670    
6671          recno = 0;          else
         name = ++ptr;  
         while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_word) != 0)  
6672            {            {
6673            if (recno >= 0)            if (IS_DIGIT(*ptr))
6674              recno = (IS_DIGIT(*ptr))? recno * 10 + (int)(*ptr - CHAR_0) : -1;              {
6675            ptr++;              *errorcodeptr = ERR84;
6676                goto FAILED;
6677                }
6678              if (!MAX_255(*ptr) || (cd->ctypes[*ptr] & ctype_word) == 0)
6679                {
6680                *errorcodeptr = ERR28;   /* Assertion expected */
6681                goto FAILED;
6682                }
6683              name = ptr++;
6684              while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_word) != 0)
6685                {
6686                ptr++;
6687                }
6688              namelen = (int)(ptr - name);
6689              if (lengthptr != NULL) *lengthptr += IMM2_SIZE;
6690            }            }
6691          namelen = (int)(ptr - name);  
6692            /* Check the terminator */
6693    
6694          if ((terminator > 0 && *ptr++ != (pcre_uchar)terminator) ||          if ((terminator > 0 && *ptr++ != (pcre_uchar)terminator) ||
6695              *ptr++ != CHAR_RIGHT_PARENTHESIS)              *ptr++ != CHAR_RIGHT_PARENTHESIS)
6696            {            {
6697            ptr--;      /* Error offset */            ptr--;                  /* Error offset */
6698            *errorcodeptr = ERR26;            *errorcodeptr = ERR26;  /* Malformed number or name */
6699            goto FAILED;            goto FAILED;
6700            }            }
6701    
# Line 6100  for (;; ptr++) Line 6704  for (;; ptr++)
6704          if (lengthptr != NULL) break;          if (lengthptr != NULL) break;
6705    
6706          /* In the real compile we do the work of looking for the actual          /* In the real compile we do the work of looking for the actual
6707          reference. If the string started with "+" or "-" we require the rest to          reference. If refsign is not negative, it means we have a number in
6708          be digits, in which case recno will be set. */          recno. */
6709    
6710          if (refsign > 0)          if (refsign >= 0)
6711            {            {
6712            if (recno <= 0)            if (recno <= 0)
6713              {              {
6714              *errorcodeptr = ERR58;              *errorcodeptr = ERR35;
6715              goto FAILED;              goto FAILED;
6716              }              }
6717            recno = (refsign == CHAR_MINUS)?            if (refsign != 0) recno = (refsign == CHAR_MINUS)?
6718              cd->bracount - recno + 1 : recno +cd->bracount;              cd->bracount - recno + 1 : recno + cd->bracount;
6719            if (recno <= 0 || recno > cd->final_bracount)            if (recno <= 0 || recno > cd->final_bracount)
6720              {              {
6721              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
# Line 6121  for (;; ptr++) Line 6725  for (;; ptr++)
6725            break;            break;
6726            }            }
6727    
6728          /* Otherwise (did not start with "+" or "-"), start by looking for the          /* Otherwise look for the name. */
         name. If we find a name, add one to the opcode to change OP_CREF or  
         OP_RREF into OP_NCREF or OP_NRREF. These behave exactly the same,  
         except they record that the reference was originally to a name. The  
         information is used to check duplicate names. */  
6729    
6730          slot = cd->name_table;          slot = cd->name_table;
6731          for (i = 0; i < cd->names_found; i++)          for (i = 0; i < cd->names_found; i++)
# Line 6134  for (;; ptr++) Line 6734  for (;; ptr++)
6734            slot += cd->name_entry_size;            slot += cd->name_entry_size;
6735            }            }
6736    
6737          /* Found the named subpattern */          /* Found the named subpattern. If the name is duplicated, add one to
6738            the opcode to change CREF/RREF into DNCREF/DNRREF and insert
6739            appropriate data values. Otherwise, just insert the unique subpattern
6740            number. */
6741    
6742          if (i < cd->names_found)          if (i < cd->names_found)
6743            {            {
6744            recno = GET2(slot, 0);            int offset = i++;
6745            PUT2(code, 2+LINK_SIZE, recno);            int count = 1;
6746            code[1+LINK_SIZE]++;            recno = GET2(slot, 0);   /* Number from first found */
6747              for (; i < cd->names_found; i++)
6748                {
6749                slot += cd->name_entry_size;
6750                if (STRNCMP_UC_UC(name, slot+IMM2_SIZE, namelen) != 0) break;
6751                count++;
6752                }
6753              if (count > 1)
6754                {
6755                PUT2(code, 2+LINK_SIZE, offset);
6756                PUT2(code, 2+LINK_SIZE+IMM2_SIZE, count);
6757                skipbytes += IMM2_SIZE;
6758                code[1+LINK_SIZE]++;
6759                }
6760              else  /* Not a duplicated name */
6761                {
6762                PUT2(code, 2+LINK_SIZE, recno);
6763                }
6764            }            }
6765    
6766          /* If terminator == CHAR_NULL it means that the name followed directly          /* If terminator == CHAR_NULL it means that the name followed directly
6767          after the opening parenthesis [e.g. (?(abc)...] and in this case there          after the opening parenthesis [e.g. (?(abc)...] and in this case there
6768          are some further alternatives to try. For the cases where terminator !=          are some further alternatives to try. For the cases where terminator !=
6769          0 [things like (?(<name>... or (?('name')... or (?(R&name)... ] we have          CHAR_NULL [things like (?(<name>... or (?('name')... or (?(R&name)... ]
6770          now checked all the possibilities, so give an error. */          we have now checked all the possibilities, so give an error. */
6771    
6772          else if (terminator != CHAR_NULL)          else if (terminator != CHAR_NULL)
6773            {            {
# Line 6184  for (;; ptr++) Line 6804  for (;; ptr++)
6804            skipbytes = 1;            skipbytes = 1;
6805            }            }
6806    
6807          /* Check for the "name" actually being a subpattern number. We are          /* Reference to an unidentified subpattern. */
         in the second pass here, so final_bracount is set. */  
   
         else if (recno > 0 && recno <= cd->final_bracount)  
           {  
           PUT2(code, 2+LINK_SIZE, recno);  
           }  
   
         /* Either an unidentified subpattern, or a reference to (?(0) */  
6808    
6809          else          else
6810            {            {
6811            *errorcodeptr = (recno == 0)? ERR35: ERR15;            *errorcodeptr = ERR15;
6812            goto FAILED;            goto FAILED;
6813            }            }
6814          break;          break;
# Line 6209  for (;; ptr++) Line 6821  for (;; ptr++)
6821          ptr++;          ptr++;
6822          break;          break;
6823    
6824            /* Optimize (?!) to (*FAIL) unless it is quantified - which is a weird
6825            thing to do, but Perl allows all assertions to be quantified, and when
6826            they contain capturing parentheses there may be a potential use for
6827            this feature. Not that that applies to a quantified (?!) but we allow
6828            it for uniformity. */
6829    
6830          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
6831          case CHAR_EXCLAMATION_MARK:            /* Negative lookahead */          case CHAR_EXCLAMATION_MARK:            /* Negative lookahead */
6832          ptr++;          ptr++;
6833          if (*ptr == CHAR_RIGHT_PARENTHESIS)    /* Optimize (?!) */          if (*ptr == CHAR_RIGHT_PARENTHESIS && ptr[1] != CHAR_ASTERISK &&
6834                 ptr[1] != CHAR_PLUS && ptr[1] != CHAR_QUESTION_MARK &&
6835                (ptr[1] != CHAR_LEFT_CURLY_BRACKET || !is_counted_repeat(ptr+2)))
6836            {            {
6837            *code++ = OP_FAIL;            *code++ = OP_FAIL;
6838            previous = NULL;            previous = NULL;
# Line 6309  for (;; ptr++) Line 6928  for (;; ptr++)
6928          terminator = (*ptr == CHAR_LESS_THAN_SIGN)?          terminator = (*ptr == CHAR_LESS_THAN_SIGN)?
6929            CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;            CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
6930          name = ++ptr;          name = ++ptr;
6931            if (IS_DIGIT(*ptr))
6932              {
6933              *errorcodeptr = ERR84;   /* Group name must start with non-digit */
6934              goto FAILED;
6935              }
6936          while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_word) != 0) ptr++;          while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
6937          namelen = (int)(ptr - name);          namelen = (int)(ptr - name);
6938    
# Line 6423  for (;; ptr++) Line 7046  for (;; ptr++)
7046    
7047          NAMED_REF_OR_RECURSE:          NAMED_REF_OR_RECURSE:
7048          name = ++ptr;          name = ++ptr;
7049            if (IS_DIGIT(*ptr))
7050              {
7051              *errorcodeptr = ERR84;   /* Group name must start with non-digit */
7052              goto FAILED;
7053              }
7054          while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_word) != 0) ptr++;          while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
7055          namelen = (int)(ptr - name);          namelen = (int)(ptr - name);
7056    
# Line 6815  for (;; ptr++) Line 7443  for (;; ptr++)
7443        skipbytes = IMM2_SIZE;        skipbytes = IMM2_SIZE;
7444        }        }
7445    
7446      /* Process nested bracketed regex. Assertions used not to be repeatable,      /* Process nested bracketed regex. First check for parentheses nested too
7447      but this was changed for Perl compatibility, so all kinds can now be      deeply. */
7448      repeated. We copy code into a non-register variable (tempcode) in order to  
7449      be able to pass its address because some compilers complain otherwise. */      if ((cd->parens_depth += 1) > PARENS_NEST_LIMIT)
7450          {
7451          *errorcodeptr = ERR82;
7452          goto FAILED;
7453          }
7454    
7455        /* Assertions used not to be repeatable, but this was changed for Perl
7456        compatibility, so all kinds can now be repeated. We copy code into a
7457        non-register variable (tempcode) in order to be able to pass its address
7458        because some compilers complain otherwise. */
7459    
7460      previous = code;                      /* For handling repetition */      previous = code;                      /* For handling repetition */
7461      *code = bravalue;      *code = bravalue;
# Line 6849  for (;; ptr++) Line 7486  for (;; ptr++)
7486           ))           ))
7487        goto FAILED;        goto FAILED;
7488    
7489        cd->parens_depth -= 1;
7490    
7491      /* If this was an atomic group and there are no capturing groups within it,      /* If this was an atomic group and there are no capturing groups within it,
7492      generate OP_ONCE_NC instead of OP_ONCE. */      generate OP_ONCE_NC instead of OP_ONCE. */
7493    
# Line 7063  for (;; ptr++) Line 7702  for (;; ptr++)
7702        if (escape == ESC_g)        if (escape == ESC_g)
7703          {          {
7704          const pcre_uchar *p;          const pcre_uchar *p;
7705            pcre_uint32 cf;
7706    
7707          save_hwm = cd->hwm;   /* Normally this is set when '(' is read */          save_hwm = cd->hwm;   /* Normally this is set when '(' is read */
7708          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
7709            CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;            CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
7710    
7711          /* These two statements stop the compiler for warning about possibly          /* These two statements stop the compiler for warning about possibly
7712          unset variables caused by the jump to HANDLE_NUMERICAL_RECURSION. In          unset variables caused by the jump to HANDLE_NUMERICAL_RECURSION. In
7713          fact, because we actually check for a number below, the paths that          fact, because we do the check for a number below, the paths that
7714          would actually be in error are never taken. */          would actually be in error are never taken. */
7715    
7716          skipbytes = 0;          skipbytes = 0;
7717          reset_bracount = FALSE;          reset_bracount = FALSE;
7718    
7719          /* Test for a name */          /* If it's not a signed or unsigned number, treat it as a name. */
7720    
7721          if (ptr[1] != CHAR_PLUS && ptr[1] != CHAR_MINUS)          cf = ptr[1];
7722            if (cf != CHAR_PLUS && cf != CHAR_MINUS && !IS_DIGIT(cf))
7723            {            {
           BOOL is_a_number = TRUE;  
           for (p = ptr + 1; *p != CHAR_NULL && *p != (pcre_uchar)terminator; p++)  
             {  
             if (!MAX_255(*p)) { is_a_number = FALSE; break; }  
             if ((cd->ctypes[*p] & ctype_digit) == 0) is_a_number = FALSE;  
             if ((cd->ctypes[*p] & ctype_word) == 0) break;  
             }  
           if (*p != (pcre_uchar)terminator)  
             {  
             *errorcodeptr = ERR57;  
             break;  
             }  
           if (is_a_number)  
             {  
             ptr++;  
             goto HANDLE_NUMERICAL_RECURSION;  
             }  
7724            is_recurse = TRUE;            is_recurse = TRUE;
7725            goto NAMED_REF_OR_RECURSE;            goto NAMED_REF_OR_RECURSE;
7726            }            }
7727    
7728          /* Test a signed number in angle brackets or quotes. */          /* Signed or unsigned number (cf = ptr[1]) is known to be plus or minus
7729            or a digit. */
7730    
7731          p = ptr + 2;          p = ptr + 2;
7732          while (IS_DIGIT(*p)) p++;          while (IS_DIGIT(*p)) p++;
# Line 7240  for (;; ptr++) Line 7866  for (;; ptr++)
7866    
7867      /* ===================================================================*/      /* ===================================================================*/
7868      /* Handle a literal character. It is guaranteed not to be whitespace or #      /* Handle a literal character. It is guaranteed not to be whitespace or #
7869      when the extended flag is set. If we are in UTF-8 mode, it may be a      when the extended flag is set. If we are in a UTF mode, it may be a
7870      multi-byte literal character. */      multi-unit literal character. */
7871    
7872      default:      default:
7873      NORMAL_CHAR:      NORMAL_CHAR:
# Line 7406  unsigned int orig_bracount; Line 8032  unsigned int orig_bracount;
8032  unsigned int max_bracount;  unsigned int max_bracount;
8033  branch_chain bc;  branch_chain bc;
8034    
8035    /* If set, call the external function that checks for stack availability. */
8036    
8037    if (PUBL(stack_guard) != NULL && PUBL(stack_guard)())
8038      {
8039      *errorcodeptr= ERR85;
8040      return FALSE;
8041      }
8042    
8043    /* Miscellaneous initialization */
8044    
8045  bc.outer = bcptr;  bc.outer = bcptr;
8046  bc.current_branch = code;  bc.current_branch = code;
8047    
# Line 7830  do { Line 8466  do {
8466       switch (*scode)       switch (*scode)
8467         {         {
8468         case OP_CREF:         case OP_CREF:
8469         case OP_NCREF:         case OP_DNCREF:
8470         case OP_RREF:         case OP_RREF:
8471         case OP_NRREF:         case OP_DNRREF:
8472         case OP_DEF:         case OP_DEF:
8473         return FALSE;         return FALSE;
8474    
# Line 8229  PCRE_UTF8 == PCRE_UTF16 == PCRE_UTF32. * Line 8865  PCRE_UTF8 == PCRE_UTF16 == PCRE_UTF32. *
8865      { skipatstart += 6; options |= PCRE_UTF8; continue; }      { skipatstart += 6; options |= PCRE_UTF8; continue; }
8866    else if (STRNCMP_UC_C8(ptr+skipatstart+2, STRING_UCP_RIGHTPAR, 4) == 0)    else if (STRNCMP_UC_C8(ptr+skipatstart+2, STRING_UCP_RIGHTPAR, 4) == 0)
8867      { skipatstart += 6; options |= PCRE_UCP; continue; }      { skipatstart += 6; options |= PCRE_UCP; continue; }
8868      else if (STRNCMP_UC_C8(ptr+skipatstart+2, STRING_NO_AUTO_POSSESS_RIGHTPAR, 16) == 0)
8869        { skipatstart += 18; options |= PCRE_NO_AUTO_POSSESS; continue; }
8870    else if (STRNCMP_UC_C8(ptr+skipatstart+2, STRING_NO_START_OPT_RIGHTPAR, 13) == 0)    else if (STRNCMP_UC_C8(ptr+skipatstart+2, STRING_NO_START_OPT_RIGHTPAR, 13) == 0)
8871      { skipatstart += 15; options |= PCRE_NO_START_OPTIMIZE; continue; }      { skipatstart += 15; options |= PCRE_NO_START_OPTIMIZE; continue; }
8872    
# Line 8423  cd->named_group_list_size = NAMED_GROUP_ Line 9061  cd->named_group_list_size = NAMED_GROUP_
9061  cd->start_pattern = (const pcre_uchar *)pattern;  cd->start_pattern = (const pcre_uchar *)pattern;
9062  cd->end_pattern = (const pcre_uchar *)(pattern + STRLEN_UC((const pcre_uchar *)pattern));  cd->end_pattern = (const pcre_uchar *)(pattern + STRLEN_UC((const pcre_uchar *)pattern));
9063  cd->req_varyopt = 0;  cd->req_varyopt = 0;
9064    cd->parens_depth = 0;
9065  cd->assert_depth = 0;  cd->assert_depth = 0;
9066  cd->max_lookbehind = 0;  cd->max_lookbehind = 0;
9067  cd->external_options = options;  cd->external_options = options;
# Line 8437  outside can help speed up starting point Line 9076  outside can help speed up starting point
9076  ptr += skipatstart;  ptr += skipatstart;
9077  code = cworkspace;  code = cworkspace;
9078  *code = OP_BRA;  *code = OP_BRA;
9079    
9080  (void)compile_regex(cd->external_options, &code, &ptr, &errorcode, FALSE,  (void)compile_regex(cd->external_options, &code, &ptr, &errorcode, FALSE,
9081    FALSE, 0, 0, &firstchar, &firstcharflags, &reqchar, &reqcharflags, NULL,    FALSE, 0, 0, &firstchar, &firstcharflags, &reqchar, &reqcharflags, NULL,
9082    cd, &length);    cd, &length);
# Line 8508  field; this time it's used for rememberi Line 9148  field; this time it's used for rememberi
9148  */  */
9149    
9150  cd->final_bracount = cd->bracount;  /* Save for checking forward references */  cd->final_bracount = cd->bracount;  /* Save for checking forward references */
9151    cd->parens_depth = 0;
9152  cd->assert_depth = 0;  cd->assert_depth = 0;
9153  cd->bracount = 0;  cd->bracount = 0;
9154  cd->max_lookbehind = 0;  cd->max_lookbehind = 0;
# Line 8613  if (errorcode == 0 && re->top_backref > Line 9254  if (errorcode == 0 && re->top_backref >
9254  /* Unless disabled, check whether single character iterators can be  /* Unless disabled, check whether single character iterators can be
9255  auto-possessified. The function overwrites the appropriate opcode values. */  auto-possessified. The function overwrites the appropriate opcode values. */
9256    
9257  if ((options & PCRE_NO_AUTO_POSSESSIFY) == 0)  if ((options & PCRE_NO_AUTO_POSSESS) == 0)
9258    auto_possessify((pcre_uchar *)codestart, utf, cd);    auto_possessify((pcre_uchar *)codestart, utf, cd);
9259    
9260  /* If there were any lookbehind assertions that contained OP_RECURSE  /* If there were any lookbehind assertions that contained OP_RECURSE
# Line 8837  return (pcre32 *)re; Line 9478  return (pcre32 *)re;
9478  }  }
9479    
9480  /* End of pcre_compile.c */  /* End of pcre_compile.c */
9481    

Legend:
Removed from v.1363  
changed lines
  Added in v.1472

  ViewVC Help
Powered by ViewVC 1.1.5