/[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 1462 by ph10, Mon Mar 10 16:28:54 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 1524  for (;;) Line 1664  for (;;)
1664    
1665      case OP_CALLOUT:      case OP_CALLOUT:
1666      case OP_CREF:      case OP_CREF:
1667      case OP_NCREF:      case OP_DNCREF:
1668      case OP_RREF:      case OP_RREF:
1669      case OP_NRREF:      case OP_DNRREF:
1670      case OP_DEF:      case OP_DEF:
1671      code += PRIV(OP_lengths)[*code];      code += PRIV(OP_lengths)[*code];
1672      break;      break;
# Line 1663  for (;;) Line 1803  for (;;)
1803      case OP_COMMIT:      case OP_COMMIT:
1804      case OP_CREF:      case OP_CREF:
1805      case OP_DEF:      case OP_DEF:
1806        case OP_DNCREF:
1807        case OP_DNRREF:
1808      case OP_DOLL:      case OP_DOLL:
1809      case OP_DOLLM:      case OP_DOLLM:
1810      case OP_EOD:      case OP_EOD:
1811      case OP_EODN:      case OP_EODN:
1812      case OP_FAIL:      case OP_FAIL:
     case OP_NCREF:  
     case OP_NRREF:  
1813      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1814      case OP_PRUNE:      case OP_PRUNE:
1815      case OP_REVERSE:      case OP_REVERSE:
# Line 1764  for (;;) Line 1904  for (;;)
1904    
1905      switch (*cc)      switch (*cc)
1906        {        {
       case OP_CRPLUS:  
       case OP_CRMINPLUS:  
1907        case OP_CRSTAR:        case OP_CRSTAR:
1908        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1909          case OP_CRPLUS:
1910          case OP_CRMINPLUS:
1911        case OP_CRQUERY:        case OP_CRQUERY:
1912        case OP_CRMINQUERY:        case OP_CRMINQUERY:
1913          case OP_CRPOSSTAR:
1914          case OP_CRPOSPLUS:
1915          case OP_CRPOSQUERY:
1916        return -1;        return -1;
1917    
1918        case OP_CRRANGE:        case OP_CRRANGE:
1919        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1920          case OP_CRPOSRANGE:
1921        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1922        branchlength += (int)GET2(cc,1);        branchlength += (int)GET2(cc,1);
1923        cc += 1 + 2 * IMM2_SIZE;        cc += 1 + 2 * IMM2_SIZE;
# Line 2366  for (code = first_significant_code(code Line 2510  for (code = first_significant_code(code
2510        case OP_CRMINSTAR:        case OP_CRMINSTAR:
2511        case OP_CRQUERY:        case OP_CRQUERY:
2512        case OP_CRMINQUERY:        case OP_CRMINQUERY:
2513          case OP_CRPOSSTAR:
2514          case OP_CRPOSQUERY:
2515        break;        break;
2516    
2517        default:                   /* Non-repeat => class must match */        default:                   /* Non-repeat => class must match */
2518        case OP_CRPLUS:            /* These repeats aren't empty */        case OP_CRPLUS:            /* These repeats aren't empty */
2519        case OP_CRMINPLUS:        case OP_CRMINPLUS:
2520          case OP_CRPOSPLUS:
2521        return FALSE;        return FALSE;
2522    
2523        case OP_CRRANGE:        case OP_CRRANGE:
2524        case OP_CRMINRANGE:        case OP_CRMINRANGE:
2525          case OP_CRPOSRANGE:
2526        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */
2527        break;        break;
2528        }        }
# Line 2650  switch(ptype) Line 2798  switch(ptype)
2798    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||    return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2799            PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;            PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2800    
2801    case PT_SPACE:    /* Perl space */    /* Perl space used to exclude VT, but from Perl 5.18 it is included, which
2802    return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||    means that Perl space and POSIX space are now identical. PCRE was changed
2803            c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)    at release 8.34. */
           == negated;  
2804    
2805      case PT_SPACE:    /* Perl space */
2806    case PT_PXSPACE:  /* POSIX space */    case PT_PXSPACE:  /* POSIX space */
2807    return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||    switch(c)
2808            c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||      {
2809            c == CHAR_FF || c == CHAR_CR)      HSPACE_CASES:
2810            == negated;      VSPACE_CASES:
2811        return negated;
2812    
2813        default:
2814        return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z) == negated;
2815        }
2816      break;  /* Control never reaches here */
2817    
2818    case PT_WORD:    case PT_WORD:
2819    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 2862  get_chr_property_list(const pcre_uchar *
2862    const pcre_uint8 *fcc, pcre_uint32 *list)    const pcre_uint8 *fcc, pcre_uint32 *list)
2863  {  {
2864  pcre_uchar c = *code;  pcre_uchar c = *code;
2865    pcre_uchar base;
2866  const pcre_uchar *end;  const pcre_uchar *end;
 const pcre_uint32 *clist_src;  
 pcre_uint32 *clist_dest;  
2867  pcre_uint32 chr;  pcre_uint32 chr;
2868  pcre_uchar base;  
2869    #ifdef SUPPORT_UCP
2870    pcre_uint32 *clist_dest;
2871    const pcre_uint32 *clist_src;
2872    #else
2873    utf = utf;  /* Suppress "unused parameter" compiler warning */
2874    #endif
2875    
2876  list[0] = c;  list[0] = c;
2877  list[1] = FALSE;  list[1] = FALSE;
# Line 2818  switch(c) Line 2977  switch(c)
2977      return code + 2;      return code + 2;
2978      }      }
2979    
2980    /* Convert only if we have anough space. */    /* Convert only if we have enough space. */
2981    
2982    clist_src = PRIV(ucd_caseless_sets) + code[1];    clist_src = PRIV(ucd_caseless_sets) + code[1];
2983    clist_dest = list + 2;    clist_dest = list + 2;
2984    code += 2;    code += 2;
2985    
2986    do {    do {
      /* Early return if there is not enough space. */  
2987       if (clist_dest >= list + 8)       if (clist_dest >= list + 8)
2988         {         {
2989           /* Early return if there is not enough space. This should never
2990           happen, since all clists are shorter than 5 character now. */
2991         list[2] = code[0];         list[2] = code[0];
2992         list[3] = code[1];         list[3] = code[1];
2993         return code;         return code;
2994         }         }
2995       *clist_dest++ = *clist_src;       *clist_dest++ = *clist_src;
2996       }       }
2997     while(*clist_src++ != NOTACHAR);    while(*clist_src++ != NOTACHAR);
2998    
2999    /* Enough space to store all characters. */    /* All characters are stored. The terminating NOTACHAR
3000      is copied form the clist itself. */
3001    
3002    list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;    list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
3003    return code;    return code;
# Line 2846  switch(c) Line 3007  switch(c)
3007    case OP_CLASS:    case OP_CLASS:
3008  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3009    case OP_XCLASS:    case OP_XCLASS:
   
3010    if (c == OP_XCLASS)    if (c == OP_XCLASS)
3011      end = code + GET(code, 0);      end = code + GET(code, 0) - 1;
3012    else    else
3013  #endif  #endif
3014      end = code + 32 / sizeof(pcre_uchar);      end = code + 32 / sizeof(pcre_uchar);
# Line 2859  switch(c) Line 3019  switch(c)
3019      case OP_CRMINSTAR:      case OP_CRMINSTAR:
3020      case OP_CRQUERY:      case OP_CRQUERY:
3021      case OP_CRMINQUERY:      case OP_CRMINQUERY:
3022        case OP_CRPOSSTAR:
3023        case OP_CRPOSQUERY:
3024      list[1] = TRUE;      list[1] = TRUE;
3025      end++;      end++;
3026      break;      break;
3027    
3028        case OP_CRPLUS:
3029        case OP_CRMINPLUS:
3030        case OP_CRPOSPLUS:
3031        end++;
3032        break;
3033    
3034      case OP_CRRANGE:      case OP_CRRANGE:
3035      case OP_CRMINRANGE:      case OP_CRMINRANGE:
3036        case OP_CRPOSRANGE:
3037      list[1] = (GET2(end, 1) == 0);      list[1] = (GET2(end, 1) == 0);
3038      end += 1 + 2 * IMM2_SIZE;      end += 1 + 2 * IMM2_SIZE;
3039      break;      break;
# Line 2895  Returns:      TRUE if the auto-possessif Line 3064  Returns:      TRUE if the auto-possessif
3064    
3065  static BOOL  static BOOL
3066  compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,  compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
3067    const pcre_uint32* base_list)    const pcre_uint32 *base_list, const pcre_uchar *base_end)
3068  {  {
3069  pcre_uchar c;  pcre_uchar c;
3070  pcre_uint32 list[8];  pcre_uint32 list[8];
3071  const pcre_uint32* chr_ptr;  const pcre_uint32 *chr_ptr;
3072  const pcre_uint32* ochr_ptr;  const pcre_uint32 *ochr_ptr;
3073  const pcre_uint32* list_ptr;  const pcre_uint32 *list_ptr;
3074    const pcre_uchar *next_code;
3075    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3076    const pcre_uchar *xclass_flags;
3077    #endif
3078    const pcre_uint8 *class_bitset;
3079    const pcre_uint8 *set1, *set2, *set_end;
3080  pcre_uint32 chr;  pcre_uint32 chr;
3081    BOOL accepted, invert_bits;
3082    
3083    /* Note: the base_list[1] contains whether the current opcode has greedy
3084    (represented by a non-zero value) quantifier. This is a different from
3085    other character type lists, which stores here that the character iterator
3086    matches to an empty string (also represented by a non-zero value). */
3087    
3088  for(;;)  for(;;)
3089    {    {
3090      /* All operations move the code pointer forward.
3091      Therefore infinite recursions are not possible. */
3092    
3093    c = *code;    c = *code;
3094    
3095    /* Skip over callouts */    /* Skip over callouts */
# Line 2925  for(;;) Line 3109  for(;;)
3109    switch(c)    switch(c)
3110      {      {
3111      case OP_END:      case OP_END:
3112      /* TRUE only in greedy case. The non-greedy case could be replaced by an      case OP_KETRPOS:
3113      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
3114      more memory, which we cannot get at this stage.) */      an OP_EXACT, but it is probably not worth it. (And note that OP_EXACT
3115        uses more memory, which we cannot get at this stage.) */
3116    
3117      return base_list[1] != 0;      return base_list[1] != 0;
3118    
3119      case OP_KET:      case OP_KET:
3120      /* If the bracket is capturing, and referenced by an OP_RECURSE, the      /* If the bracket is capturing, and referenced by an OP_RECURSE, or
3121      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
3122      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.) */  
3123    
3124      if (base_list[1] == 0) return FALSE;      if (base_list[1] == 0) return FALSE;
3125    
3126        switch(*(code - GET(code, 1)))
3127          {
3128          case OP_ASSERT:
3129          case OP_ASSERT_NOT:
3130          case OP_ASSERTBACK:
3131          case OP_ASSERTBACK_NOT:
3132          case OP_ONCE:
3133          case OP_ONCE_NC:
3134          /* Atomic sub-patterns and assertions can always auto-possessify their
3135          last iterator. */
3136          return TRUE;
3137          }
3138    
3139        code += PRIV(OP_lengths)[c];
3140        continue;
3141    
3142        case OP_ONCE:
3143        case OP_ONCE_NC:
3144        case OP_BRA:
3145        case OP_CBRA:
3146        next_code = code + GET(code, 1);
3147        code += PRIV(OP_lengths)[c];
3148    
3149        while (*next_code == OP_ALT)
3150          {
3151          if (!compare_opcodes(code, utf, cd, base_list, base_end)) return FALSE;
3152          code = next_code + 1 + LINK_SIZE;
3153          next_code += GET(next_code, 1);
3154          }
3155        continue;
3156    
3157        case OP_BRAZERO:
3158        case OP_BRAMINZERO:
3159    
3160        next_code = code + 1;
3161        if (*next_code != OP_BRA && *next_code != OP_CBRA
3162            && *next_code != OP_ONCE && *next_code != OP_ONCE_NC) return FALSE;
3163    
3164        do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
3165    
3166        /* The bracket content will be checked by the
3167        OP_BRA/OP_CBRA case above. */
3168        next_code += 1 + LINK_SIZE;
3169        if (!compare_opcodes(next_code, utf, cd, base_list, base_end))
3170          return FALSE;
3171    
3172      code += PRIV(OP_lengths)[c];      code += PRIV(OP_lengths)[c];
3173      continue;      continue;
3174      }      }
# Line 2961  for(;;) Line 3192  for(;;)
3192      list_ptr = base_list;      list_ptr = base_list;
3193      }      }
3194    
3195      /* Character bitsets can also be compared to certain opcodes. */
3196    
3197      else if (base_list[0] == OP_CLASS || list[0] == OP_CLASS
3198    #ifdef COMPILE_PCRE8
3199          /* In 8 bit, non-UTF mode, OP_CLASS and OP_NCLASS are the same. */
3200          || (!utf && (base_list[0] == OP_NCLASS || list[0] == OP_NCLASS))
3201    #endif
3202          )
3203        {
3204    #ifdef COMPILE_PCRE8
3205        if (base_list[0] == OP_CLASS || (!utf && base_list[0] == OP_NCLASS))
3206    #else
3207        if (base_list[0] == OP_CLASS)
3208    #endif
3209          {
3210          set1 = (pcre_uint8 *)(base_end - base_list[2]);
3211          list_ptr = list;
3212          }
3213        else
3214          {
3215          set1 = (pcre_uint8 *)(code - list[2]);
3216          list_ptr = base_list;
3217          }
3218    
3219        invert_bits = FALSE;
3220        switch(list_ptr[0])
3221          {
3222          case OP_CLASS:
3223          case OP_NCLASS:
3224          set2 = (pcre_uint8 *)
3225            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3226          break;
3227    
3228    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3229          case OP_XCLASS:
3230          xclass_flags = (list_ptr == list ? code : base_end) - list_ptr[2] + LINK_SIZE;
3231          if ((*xclass_flags & XCL_HASPROP) != 0) return FALSE;
3232          if ((*xclass_flags & XCL_MAP) == 0)
3233            {
3234            /* No bits are set for characters < 256. */
3235            if (list[1] == 0) return TRUE;
3236            /* Might be an empty repeat. */
3237            continue;
3238            }
3239          set2 = (pcre_uint8 *)(xclass_flags + 1);
3240          break;
3241    #endif
3242    
3243          case OP_NOT_DIGIT:
3244          invert_bits = TRUE;
3245          /* Fall through */
3246          case OP_DIGIT:
3247          set2 = (pcre_uint8 *)(cd->cbits + cbit_digit);
3248          break;
3249    
3250          case OP_NOT_WHITESPACE:
3251          invert_bits = TRUE;
3252          /* Fall through */
3253          case OP_WHITESPACE:
3254          set2 = (pcre_uint8 *)(cd->cbits + cbit_space);
3255          break;
3256    
3257          case OP_NOT_WORDCHAR:
3258          invert_bits = TRUE;
3259          /* Fall through */
3260          case OP_WORDCHAR:
3261          set2 = (pcre_uint8 *)(cd->cbits + cbit_word);
3262          break;
3263    
3264          default:
3265          return FALSE;
3266          }
3267    
3268        /* Because the sets are unaligned, we need
3269        to perform byte comparison here. */
3270        set_end = set1 + 32;
3271        if (invert_bits)
3272          {
3273          do
3274            {
3275            if ((*set1++ & ~(*set2++)) != 0) return FALSE;
3276            }
3277          while (set1 < set_end);
3278          }
3279        else
3280          {
3281          do
3282            {
3283            if ((*set1++ & *set2++) != 0) return FALSE;
3284            }
3285          while (set1 < set_end);
3286          }
3287    
3288        if (list[1] == 0) return TRUE;
3289        /* Might be an empty repeat. */
3290        continue;
3291        }
3292    
3293    /* Some property combinations also acceptable. Unicode property opcodes are    /* Some property combinations also acceptable. Unicode property opcodes are
3294    processed specially; the rest can be handled with a lookup table. */    processed specially; the rest can be handled with a lookup table. */
3295    
# Line 2968  for(;;) Line 3297  for(;;)
3297      {      {
3298      pcre_uint32 leftop, rightop;      pcre_uint32 leftop, rightop;
3299    
     if (list[1] != 0) return FALSE;   /* Must match at least one character */  
3300      leftop = base_list[0];      leftop = base_list[0];
3301      rightop = list[0];      rightop = list[0];
3302    
3303  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3304        accepted = FALSE; /* Always set in non-unicode case. */
3305      if (leftop == OP_PROP || leftop == OP_NOTPROP)      if (leftop == OP_PROP || leftop == OP_NOTPROP)
3306        {        {
3307        if (rightop == OP_EOD) return TRUE;        if (rightop == OP_EOD)
3308        if (rightop == OP_PROP || rightop == OP_NOTPROP)          accepted = TRUE;
3309          else if (rightop == OP_PROP || rightop == OP_NOTPROP)
3310          {          {
3311          int n;          int n;
3312          const pcre_uint8 *p;          const pcre_uint8 *p;
# Line 2997  for(;;) Line 3327  for(;;)
3327          n = propposstab[base_list[2]][list[2]];          n = propposstab[base_list[2]][list[2]];
3328          switch(n)          switch(n)
3329            {            {
3330            case 0: return FALSE;            case 0: break;
3331            case 1: return bothprop;            case 1: accepted = bothprop; break;
3332            case 2: return (base_list[3] == list[3]) != same;            case 2: accepted = (base_list[3] == list[3]) != same; break;
3333            case 3: return !same;            case 3: accepted = !same; break;
3334    
3335            case 4:  /* Left general category, right particular category */            case 4:  /* Left general category, right particular category */
3336            return risprop && catposstab[base_list[3]][list[3]] == same;            accepted = risprop && catposstab[base_list[3]][list[3]] == same;
3337              break;
3338    
3339            case 5:  /* Right general category, left particular category */            case 5:  /* Right general category, left particular category */
3340            return lisprop && catposstab[list[3]][base_list[3]] == same;            accepted = lisprop && catposstab[list[3]][base_list[3]] == same;
3341              break;
3342    
3343            /* This code is logically tricky. Think hard before fiddling with it.            /* This code is logically tricky. Think hard before fiddling with it.
3344            The posspropstab table has four entries per row. Each row relates to            The posspropstab table has four entries per row. Each row relates to
3345            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.
3346            Only WORD actually needs all four entries, but using repeats for the            Only WORD actually needs all four entries, but using repeats for the
3347            others means they can all use the same code below.            others means they can all use the same code below.
3348    
3349            The first two entries in each row are Unicode general categories, and            The first two entries in each row are Unicode general categories, and
3350            apply always, because all the characters they include are part of the            apply always, because all the characters they include are part of the
3351            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 3355  for(;;)
3355            category contains more characters than the specials that are defined            category contains more characters than the specials that are defined
3356            for the property being tested against. Therefore, it cannot be used            for the property being tested against. Therefore, it cannot be used
3357            in a NOTPROP case.            in a NOTPROP case.
3358    
3359            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.
3360            Underscore is covered by ucp_P or ucp_Po. */            Underscore is covered by ucp_P or ucp_Po. */
3361    
# Line 3031  for(;;) Line 3363  for(;;)
3363            case 7:  /* Left space vs right general category */            case 7:  /* Left space vs right general category */
3364            case 8:  /* Left word vs right general category */            case 8:  /* Left word vs right general category */
3365            p = posspropstab[n-6];            p = posspropstab[n-6];
3366            return risprop && lisprop ==            accepted = risprop && lisprop ==
3367              (list[3] != p[0] &&              (list[3] != p[0] &&
3368               list[3] != p[1] &&               list[3] != p[1] &&
3369              (list[3] != p[2] || !lisprop));              (list[3] != p[2] || !lisprop));
3370              break;
3371    
3372            case 9:   /* Right alphanum vs left general category */            case 9:   /* Right alphanum vs left general category */
3373            case 10:  /* Right space vs left general category */            case 10:  /* Right space vs left general category */
3374            case 11:  /* Right word vs left general category */            case 11:  /* Right word vs left general category */
3375            p = posspropstab[n-9];            p = posspropstab[n-9];
3376            return lisprop && risprop ==            accepted = lisprop && risprop ==
3377              (base_list[3] != p[0] &&              (base_list[3] != p[0] &&
3378               base_list[3] != p[1] &&               base_list[3] != p[1] &&
3379              (base_list[3] != p[2] || !risprop));              (base_list[3] != p[2] || !risprop));
3380              break;
3381    
3382            case 12:  /* Left alphanum vs right particular category */            case 12:  /* Left alphanum vs right particular category */
3383            case 13:  /* Left space vs right particular category */            case 13:  /* Left space vs right particular category */
3384            case 14:  /* Left word vs right particular category */            case 14:  /* Left word vs right particular category */
3385            p = posspropstab[n-12];            p = posspropstab[n-12];
3386            return risprop && lisprop ==            accepted = risprop && lisprop ==
3387              (catposstab[p[0]][list[3]] &&              (catposstab[p[0]][list[3]] &&
3388               catposstab[p[1]][list[3]] &&               catposstab[p[1]][list[3]] &&
3389              (list[3] != p[3] || !lisprop));              (list[3] != p[3] || !lisprop));
3390              break;
3391    
3392            case 15:  /* Right alphanum vs left particular category */            case 15:  /* Right alphanum vs left particular category */
3393            case 16:  /* Right space vs left particular category */            case 16:  /* Right space vs left particular category */
3394            case 17:  /* Right word vs left particular category */            case 17:  /* Right word vs left particular category */
3395            p = posspropstab[n-15];            p = posspropstab[n-15];
3396            return lisprop && risprop ==            accepted = lisprop && risprop ==
3397              (catposstab[p[0]][base_list[3]] &&              (catposstab[p[0]][base_list[3]] &&
3398               catposstab[p[1]][base_list[3]] &&               catposstab[p[1]][base_list[3]] &&
3399              (base_list[3] != p[3] || !risprop));              (base_list[3] != p[3] || !risprop));
3400              break;
3401            }            }
3402          }          }
       return FALSE;  
3403        }        }
3404    
3405      else      else
3406  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3407    
3408      return leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&      accepted = leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&
3409             rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&             rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&
3410             autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];             autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];
3411    
3412        if (!accepted)
3413          return FALSE;
3414    
3415        if (list[1] == 0) return TRUE;
3416        /* Might be an empty repeat. */
3417        continue;
3418      }      }
3419    
3420    /* 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 3528  for(;;)
3528        case OP_EOD:    /* Can always possessify before \z */        case OP_EOD:    /* Can always possessify before \z */
3529        break;        break;
3530    
3531    #ifdef SUPPORT_UCP
3532        case OP_PROP:        case OP_PROP:
3533        case OP_NOTPROP:        case OP_NOTPROP:
3534        if (!check_char_prop(chr, list_ptr[2], list_ptr[3],        if (!check_char_prop(chr, list_ptr[2], list_ptr[3],
3535              list_ptr[0] == OP_NOTPROP))              list_ptr[0] == OP_NOTPROP))
3536          return FALSE;          return FALSE;
3537        break;        break;
3538    #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. */  
3539    
3540        case OP_NCLASS:        case OP_NCLASS:
3541        if (chr > 255) return FALSE;        if (chr > 255) return FALSE;
3542        /* Fall through */        /* Fall through */
3543    
3544        case OP_CLASS:        case OP_CLASS:
       if (list_ptr != list) return FALSE;   /* Class is first opcode */  
3545        if (chr > 255) break;        if (chr > 255) break;
3546        if ((((pcre_uint8 *)(code - list_ptr[2] + 1))[chr >> 3] & (1 << (chr & 7))) != 0)        class_bitset = (pcre_uint8 *)
3547          return FALSE;          ((list_ptr == list ? code : base_end) - list_ptr[2]);
3548          if ((class_bitset[chr >> 3] & (1 << (chr & 7))) != 0) return FALSE;
3549        break;        break;
3550    
3551  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3552        case OP_XCLASS:        case OP_XCLASS:
3553        if (list_ptr != list) return FALSE;   /* Class is first opcode */        if (PRIV(xclass)(chr, (list_ptr == list ? code : base_end) -
3554        if (PRIV(xclass)(chr, code - list_ptr[2] + 1 + LINK_SIZE, utf))            list_ptr[2] + LINK_SIZE, utf)) return FALSE;
         return FALSE;  
3555        break;        break;
3556  #endif  #endif
3557    
# Line 3232  for(;;) Line 3568  for(;;)
3568    if (list[1] == 0) return TRUE;    if (list[1] == 0) return TRUE;
3569    }    }
3570    
3571  return FALSE;  /* Control never reaches here. There used to be a fail-save return FALSE; here,
3572    but some compilers complain about an unreachable statement. */
3573    
3574  }  }
3575    
3576    
# Line 3257  auto_possessify(pcre_uchar *code, BOOL u Line 3595  auto_possessify(pcre_uchar *code, BOOL u
3595  {  {
3596  register pcre_uchar c;  register pcre_uchar c;
3597  const pcre_uchar *end;  const pcre_uchar *end;
3598    pcre_uchar *repeat_opcode;
3599  pcre_uint32 list[8];  pcre_uint32 list[8];
3600    
3601  for (;;)  for (;;)
# Line 3270  for (;;) Line 3609  for (;;)
3609        get_chr_property_list(code, utf, cd->fcc, list) : NULL;        get_chr_property_list(code, utf, cd->fcc, list) : NULL;
3610      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;
3611    
3612      if (end != NULL && compare_opcodes(end, utf, cd, list))      if (end != NULL && compare_opcodes(end, utf, cd, list, end))
3613        {        {
3614        switch(c)        switch(c)
3615          {          {
# Line 3303  for (;;) Line 3642  for (;;)
3642          break;          break;
3643    
3644          case OP_MINUPTO:          case OP_MINUPTO:
3645          *code += OP_MINUPTO - OP_UPTO;          *code += OP_POSUPTO - OP_MINUPTO;
3646          break;          break;
3647          }          }
3648        }        }
3649      c = *code;      c = *code;
3650      }      }
3651      else if (c == OP_CLASS || c == OP_NCLASS || c == OP_XCLASS)
3652        {
3653    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3654        if (c == OP_XCLASS)
3655          repeat_opcode = code + GET(code, 1);
3656        else
3657    #endif
3658          repeat_opcode = code + 1 + (32 / sizeof(pcre_uchar));
3659    
3660        c = *repeat_opcode;
3661        if (c >= OP_CRSTAR && c <= OP_CRMINRANGE)
3662          {
3663          /* end must not be NULL. */
3664          end = get_chr_property_list(code, utf, cd->fcc, list);
3665    
3666          list[1] = (c & 1) == 0;
3667    
3668          if (compare_opcodes(end, utf, cd, list, end))
3669            {
3670            switch (c)
3671              {
3672              case OP_CRSTAR:
3673              case OP_CRMINSTAR:
3674              *repeat_opcode = OP_CRPOSSTAR;
3675              break;
3676    
3677              case OP_CRPLUS:
3678              case OP_CRMINPLUS:
3679              *repeat_opcode = OP_CRPOSPLUS;
3680              break;
3681    
3682              case OP_CRQUERY:
3683              case OP_CRMINQUERY:
3684              *repeat_opcode = OP_CRPOSQUERY;
3685              break;
3686    
3687              case OP_CRRANGE:
3688              case OP_CRMINRANGE:
3689              *repeat_opcode = OP_CRPOSRANGE;
3690              break;
3691              }
3692            }
3693          }
3694        c = *code;
3695        }
3696    
3697    switch(c)    switch(c)
3698      {      {
# Line 3335  for (;;) Line 3719  for (;;)
3719        code += 2;        code += 2;
3720      break;      break;
3721    
3722    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3723      case OP_XCLASS:      case OP_XCLASS:
3724      code += GET(code, 1);      code += GET(code, 1);
3725      break;      break;
3726    #endif
3727    
3728      case OP_MARK:      case OP_MARK:
3729      case OP_PRUNE_ARG:      case OP_PRUNE_ARG:
# Line 3446  class, but [abc[:x\]pqr:]] is (so that a Line 3832  class, but [abc[:x\]pqr:]] is (so that a
3832  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
3833  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:]
3834  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
3835  "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,
3836  I think.  I think.
3837    
3838  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 4081  for (c = *cptr; c <= d; c++)
4081    
4082  if (c > d) return -1;  /* Reached end of range */  if (c > d) return -1;  /* Reached end of range */
4083    
4084    /* Found a character that has a single other case. Search for the end of the
4085    range, which is either the end of the input range, or a character that has zero
4086    or more than one other cases. */
4087    
4088  *ocptr = othercase;  *ocptr = othercase;
4089  next = othercase + 1;  next = othercase + 1;
4090    
4091  for (++c; c <= d; c++)  for (++c; c <= d; c++)
4092    {    {
4093    if (UCD_OTHERCASE(c) != next) break;    if ((co = UCD_CASESET(c)) != 0 || UCD_OTHERCASE(c) != next) break;
4094    next++;    next++;
4095    }    }
4096    
# Line 3738  add_to_class(pcre_uint8 *classbits, pcre Line 4128  add_to_class(pcre_uint8 *classbits, pcre
4128    compile_data *cd, pcre_uint32 start, pcre_uint32 end)    compile_data *cd, pcre_uint32 start, pcre_uint32 end)
4129  {  {
4130  pcre_uint32 c;  pcre_uint32 c;
4131    pcre_uint32 classbits_end = (end <= 0xff ? end : 0xff);
4132  int n8 = 0;  int n8 = 0;
4133    
4134  /* 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 4172  if ((options & PCRE_CASELESS) != 0)
4172    
4173    /* Not UTF-mode, or no UCP */    /* Not UTF-mode, or no UCP */
4174    
4175    for (c = start; c <= end && c < 256; c++)    for (c = start; c <= classbits_end; c++)
4176      {      {
4177      SETBIT(classbits, cd->fcc[c]);      SETBIT(classbits, cd->fcc[c]);
4178      n8++;      n8++;
# Line 3806  in all cases. */ Line 4197  in all cases. */
4197    
4198  #endif /* COMPILE_PCRE[8|16] */  #endif /* COMPILE_PCRE[8|16] */
4199    
4200  /* 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. */  
4201    
4202  if (end < 0x100)  for (c = start; c <= classbits_end; c++)
4203    {    {
4204    for (c = start; c <= end; c++)    /* Regardless of start, c will always be <= 255. */
4205      {    SETBIT(classbits, c);
4206      n8++;    n8++;
     SETBIT(classbits, c);  
     }  
4207    }    }
4208    
4209  else  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4210    if (start <= 0xff) start = 0xff + 1;
4211    
4212    if (end >= start)
4213    {    {
4214    pcre_uchar *uchardata = *uchardptr;    pcre_uchar *uchardata = *uchardptr;
   
4215  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4216    if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */    if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */
4217      {      {
# Line 3861  else Line 4251  else
4251    
4252    *uchardptr = uchardata;   /* Updata extra data pointer */    *uchardptr = uchardata;   /* Updata extra data pointer */
4253    }    }
4254    #endif /* SUPPORT_UTF || !COMPILE_PCRE8 */
4255    
4256  return n8;    /* Number of 8-bit characters */  return n8;    /* Number of 8-bit characters */
4257  }  }
# Line 4082  for (;; ptr++) Line 4473  for (;; ptr++)
4473    BOOL reset_bracount;    BOOL reset_bracount;
4474    int class_has_8bitchar;    int class_has_8bitchar;
4475    int class_one_char;    int class_one_char;
4476    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4477      BOOL xclass_has_prop;
4478    #endif
4479    int newoptions;    int newoptions;
4480    int recno;    int recno;
4481    int refsign;    int refsign;
# Line 4199  for (;; ptr++) Line 4593  for (;; ptr++)
4593          }          }
4594        goto NORMAL_CHAR;        goto NORMAL_CHAR;
4595        }        }
4596        /* Control does not reach here. */
4597      }      }
4598    
4599    /* 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
4600    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. */  
4601    
4602    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
4603      {      {
4604      if (MAX_255(*ptr) && (cd->ctypes[c] & ctype_space) != 0) continue;      for (;;)
     if (c == CHAR_NUMBER_SIGN)  
4605        {        {
4606          while (MAX_255(c) && (cd->ctypes[c] & ctype_space) != 0) c = *(++ptr);
4607          if (c != CHAR_NUMBER_SIGN) break;
4608        ptr++;        ptr++;
4609        while (*ptr != CHAR_NULL)        while (*ptr != CHAR_NULL)
4610          {          {
4611          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr))         /* For non-fixed-length newline cases, */
4612              {                          /* IS_NEWLINE sets cd->nllen. */
4613              ptr += cd->nllen;
4614              break;
4615              }
4616          ptr++;          ptr++;
4617  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4618          if (utf) FORWARDCHAR(ptr);          if (utf) FORWARDCHAR(ptr);
4619  #endif  #endif
4620          }          }
4621        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;  
4622        }        }
4623      }      }
4624    
4625    /* No auto callout for quantifiers. */    /* See if the next thing is a quantifier. */
4626    
4627      is_quantifier =
4628        c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
4629        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
4630    
4631      /* Fill in length of a previous callout, except when the next thing is a
4632      quantifier or when processing a property substitution string in UCP mode. */
4633    
4634    if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier)    if (!is_quantifier && previous_callout != NULL && nestptr == NULL &&
4635           after_manual_callout-- <= 0)
4636        {
4637        if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */
4638          complete_callout(previous_callout, ptr, cd);
4639        previous_callout = NULL;
4640        }
4641    
4642      /* Create auto callout, except for quantifiers, or while processing property
4643      strings that are substituted for \w etc in UCP mode. */
4644    
4645      if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier && nestptr == NULL)
4646      {      {
4647      previous_callout = code;      previous_callout = code;
4648      code = auto_callout(code, ptr, cd);      code = auto_callout(code, ptr, cd);
4649      }      }
4650    
4651      /* Process the next pattern item. */
4652    
4653    switch(c)    switch(c)
4654      {      {
4655      /* ===================================================================*/      /* ===================================================================*/
4656      case 0:                        /* The branch terminates at string end */      case CHAR_NULL:                /* The branch terminates at string end */
4657      case CHAR_VERTICAL_LINE:       /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
4658      case CHAR_RIGHT_PARENTHESIS:      case CHAR_RIGHT_PARENTHESIS:
4659      *firstcharptr = firstchar;      *firstcharptr = firstchar;
# Line 4328  for (;; ptr++) Line 4731  for (;; ptr++)
4731        }        }
4732      goto NORMAL_CHAR;      goto NORMAL_CHAR;
4733    
4734        /* In another (POSIX) regex library, the ugly syntax [[:<:]] and [[:>:]] is
4735        used for "start of word" and "end of word". As these are otherwise illegal
4736        sequences, we don't break anything by recognizing them. They are replaced
4737        by \b(?=\w) and \b(?<=\w) respectively. Sequences like [a[:<:]] are
4738        erroneous and are handled by the normal code below. */
4739    
4740      case CHAR_LEFT_SQUARE_BRACKET:      case CHAR_LEFT_SQUARE_BRACKET:
4741        if (STRNCMP_UC_C8(ptr+1, STRING_WEIRD_STARTWORD, 6) == 0)
4742          {
4743          nestptr = ptr + 7;
4744          ptr = sub_start_of_word - 1;
4745          continue;
4746          }
4747    
4748        if (STRNCMP_UC_C8(ptr+1, STRING_WEIRD_ENDWORD, 6) == 0)
4749          {
4750          nestptr = ptr + 7;
4751          ptr = sub_end_of_word - 1;
4752          continue;
4753          }
4754    
4755        /* Handle a real character class. */
4756    
4757      previous = code;      previous = code;
4758    
4759      /* 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 4810  for (;; ptr++)
4810    
4811      should_flip_negation = FALSE;      should_flip_negation = FALSE;
4812    
4813        /* Extended class (xclass) will be used when characters > 255
4814        might match. */
4815    
4816    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4817        xclass = FALSE;
4818        class_uchardata = code + LINK_SIZE + 2;   /* For XCLASS items */
4819        class_uchardata_base = class_uchardata;   /* Save the start */
4820    #endif
4821    
4822      /* For optimization purposes, we track some properties of the class:      /* For optimization purposes, we track some properties of the class:
4823      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 <
4824      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
4825      character. */      character; xclass_has_prop will be TRUE if unicode property checks
4826        are present in the class. */
4827    
4828      class_has_8bitchar = 0;      class_has_8bitchar = 0;
4829      class_one_char = 0;      class_one_char = 0;
4830    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4831        xclass_has_prop = FALSE;
4832    #endif
4833    
4834      /* 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
4835      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 4838  for (;; ptr++)
4838    
4839      memset(classbits, 0, 32 * sizeof(pcre_uint8));      memset(classbits, 0, 32 * sizeof(pcre_uint8));
4840    
 #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  
   
4841      /* Process characters until ] is reached. By writing this as a "do" it      /* Process characters until ] is reached. By writing this as a "do" it
4842      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
4843      loop, c contains the first byte of the character. */      loop, c contains the first byte of the character. */
# Line 4493  for (;; ptr++) Line 4925  for (;; ptr++)
4925            posix_class = 0;            posix_class = 0;
4926    
4927          /* 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
4928          different escape sequences that use Unicode properties. */          different escape sequences that use Unicode properties \p or \P. Others
4929            that are not available via \p or \P generate XCL_PROP/XCL_NOTPROP
4930            directly. */
4931    
4932  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4933          if ((options & PCRE_UCP) != 0)          if ((options & PCRE_UCP) != 0)
4934            {            {
4935              unsigned int ptype = 0;
4936            int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);            int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
4937    
4938              /* The posix_substitutes table specifies which POSIX classes can be
4939              converted to \p or \P items. */
4940    
4941            if (posix_substitutes[pc] != NULL)            if (posix_substitutes[pc] != NULL)
4942              {              {
4943              nestptr = tempptr + 1;              nestptr = tempptr + 1;
4944              ptr = posix_substitutes[pc] - 1;              ptr = posix_substitutes[pc] - 1;
4945              continue;              continue;
4946              }              }
4947    
4948              /* There are three other classes that generate special property calls
4949              that are recognized only in an XCLASS. */
4950    
4951              else switch(posix_class)
4952                {
4953                case PC_GRAPH:
4954                ptype = PT_PXGRAPH;
4955                /* Fall through */
4956                case PC_PRINT:
4957                if (ptype == 0) ptype = PT_PXPRINT;
4958                /* Fall through */
4959                case PC_PUNCT:
4960                if (ptype == 0) ptype = PT_PXPUNCT;
4961                *class_uchardata++ = local_negate? XCL_NOTPROP : XCL_PROP;
4962                *class_uchardata++ = ptype;
4963                *class_uchardata++ = 0;
4964                xclass_has_prop = TRUE;
4965                ptr = tempptr + 1;
4966                continue;
4967    
4968                /* For all other POSIX classes, no special action is taken in UCP
4969                mode. Fall through to the non_UCP case. */
4970    
4971                default:
4972                break;
4973                }
4974            }            }
4975  #endif  #endif
4976          /* 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
4977          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
4978          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
4979          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
4980            bit map that is being built. */
4981    
4982          posix_class *= 3;          posix_class *= 3;
4983    
# Line 4627  for (;; ptr++) Line 5094  for (;; ptr++)
5094              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
5095              continue;              continue;
5096    
5097              /* 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
5098              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
5099              class. Luckily, the value of CHAR_VT is 0x0b in both ASCII and              previously set by something earlier in the character class.
5100              EBCDIC, so we lazily just adjust the appropriate bit. */              Luckily, the value of CHAR_VT is 0x0b in both ASCII and EBCDIC, so
5101                we could just adjust the appropriate bit. From PCRE 8.34 we no
5102                longer treat \s and \S specially. */
5103    
5104              case ESC_s:              case ESC_s:
5105              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];  
5106              continue;              continue;
5107    
5108              case ESC_S:              case ESC_S:
5109              should_flip_negation = TRUE;              should_flip_negation = TRUE;
5110              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 */  
5111              continue;              continue;
5112    
5113              /* 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 5144  for (;; ptr++)
5144                  XCL_PROP : XCL_NOTPROP;                  XCL_PROP : XCL_NOTPROP;
5145                *class_uchardata++ = ptype;                *class_uchardata++ = ptype;
5146                *class_uchardata++ = pdata;                *class_uchardata++ = pdata;
5147                  xclass_has_prop = TRUE;
5148                class_has_8bitchar--;                /* Undo! */                class_has_8bitchar--;                /* Undo! */
5149                continue;                continue;
5150                }                }
# Line 4763  for (;; ptr++) Line 5230  for (;; ptr++)
5230  #endif  #endif
5231          d = *ptr;  /* Not UTF-8 mode */          d = *ptr;  /* Not UTF-8 mode */
5232    
5233          /* 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
5234          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
5235          in such circumstances. */          literal in such circumstances. However, in Perl's warning mode, a
5236            warning is given, so PCRE now faults it as it is almost certainly a
5237            mistake on the user's part. */
5238    
5239          if (!inescq && d == CHAR_BACKSLASH)          if (!inescq)
5240            {            {
5241            int descape;            if (d == CHAR_BACKSLASH)
5242            descape = check_escape(&ptr, &d, errorcodeptr, cd->bracount, options, TRUE);              {
5243            if (*errorcodeptr != 0) goto FAILED;              int descape;
5244                descape = check_escape(&ptr, &d, errorcodeptr, cd->bracount, options, TRUE);
5245                if (*errorcodeptr != 0) goto FAILED;
5246    
5247            /* \b is backspace; any other special means the '-' was literal. */              /* 0 means a character was put into d; \b is backspace; any other
5248                special causes an error. */
5249    
5250            if (descape != 0)              if (descape != 0)
             {  
             if (descape == ESC_b) d = CHAR_BS; else  
5251                {                {
5252                ptr = oldptr;                if (descape == ESC_b) d = CHAR_BS; else
5253                goto CLASS_SINGLE_CHARACTER;  /* A few lines below */                  {
5254                    *errorcodeptr = ERR83;
5255                    goto FAILED;
5256                    }
5257                }                }
5258              }              }
5259    
5260              /* A hyphen followed by a POSIX class is treated in the same way. */
5261    
5262              else if (d == CHAR_LEFT_SQUARE_BRACKET &&
5263                       (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
5264                        ptr[1] == CHAR_EQUALS_SIGN) &&
5265                       check_posix_syntax(ptr, &tempptr))
5266                {
5267                *errorcodeptr = ERR83;
5268                goto FAILED;
5269                }
5270            }            }
5271    
5272          /* 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 5439  for (;; ptr++)
5439        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
5440        code += LINK_SIZE;        code += LINK_SIZE;
5441        *code = negate_class? XCL_NOT:0;        *code = negate_class? XCL_NOT:0;
5442          if (xclass_has_prop) *code |= XCL_HASPROP;
5443    
5444        /* 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;
5445        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 5449  for (;; ptr++)
5449          *code++ |= XCL_MAP;          *code++ |= XCL_MAP;
5450          memmove(code + (32 / sizeof(pcre_uchar)), code,          memmove(code + (32 / sizeof(pcre_uchar)), code,
5451            IN_UCHARS(class_uchardata - code));            IN_UCHARS(class_uchardata - code));
5452            if (negate_class && !xclass_has_prop)
5453              for (c = 0; c < 32; c++) classbits[c] = ~classbits[c];
5454          memcpy(code, classbits, 32);          memcpy(code, classbits, 32);
5455          code = class_uchardata + (32 / sizeof(pcre_uchar));          code = class_uchardata + (32 / sizeof(pcre_uchar));
5456          }          }
# Line 5046  for (;; ptr++) Line 5533  for (;; ptr++)
5533    
5534      tempcode = previous;      tempcode = previous;
5535    
5536        /* Before checking for a possessive quantifier, we must skip over
5537        whitespace and comments in extended mode because Perl allows white space at
5538        this point. */
5539    
5540        if ((options & PCRE_EXTENDED) != 0)
5541          {
5542          const pcre_uchar *p = ptr + 1;
5543          for (;;)
5544            {
5545            while (MAX_255(*p) && (cd->ctypes[*p] & ctype_space) != 0) p++;
5546            if (*p != CHAR_NUMBER_SIGN) break;
5547            p++;
5548            while (*p != CHAR_NULL)
5549              {
5550              if (IS_NEWLINE(p))         /* For non-fixed-length newline cases, */
5551                {                        /* IS_NEWLINE sets cd->nllen. */
5552                p += cd->nllen;
5553                break;
5554                }
5555              p++;
5556    #ifdef SUPPORT_UTF
5557              if (utf) FORWARDCHAR(p);
5558    #endif
5559              }           /* Loop for comment characters */
5560            }             /* Loop for multiple comments */
5561          ptr = p - 1;    /* Character before the next significant one. */
5562          }
5563    
5564      /* If the next character is '+', we have a possessive quantifier. This      /* If the next character is '+', we have a possessive quantifier. This
5565      implies greediness, whatever the setting of the PCRE_UNGREEDY option.      implies greediness, whatever the setting of the PCRE_UNGREEDY option.
5566      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 5853  for (;; ptr++)
5853      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
5854      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 >=
5855      OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,      OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
5856      ASSERTBACK_NOT, ONCE, BRA, CBRA, and COND. Originally, PCRE did not allow      ASSERTBACK_NOT, ONCE, ONCE_NC, BRA, BRAPOS, CBRA, CBRAPOS, and COND.
5857      repetition of assertions, but now it does, for Perl compatibility. */      Originally, PCRE did not allow repetition of assertions, but now it does,
5858        for Perl compatibility. */
5859    
5860      else if (*previous >= OP_ASSERT && *previous <= OP_COND)      else if (*previous >= OP_ASSERT && *previous <= OP_COND)
5861        {        {
# Line 5357  for (;; ptr++) Line 5873  for (;; ptr++)
5873        /* There is no sense in actually repeating assertions. The only potential        /* There is no sense in actually repeating assertions. The only potential
5874        use of repetition is in cases when the assertion is optional. Therefore,        use of repetition is in cases when the assertion is optional. Therefore,
5875        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
5876        maximum is not not zero or one, set it to 1. */        maximum is not zero or one, set it to 1. */
5877    
5878        if (*previous < OP_ONCE)    /* Assertion */        if (*previous < OP_ONCE)    /* Assertion */
5879          {          {
# Line 5730  for (;; ptr++) Line 6246  for (;; ptr++)
6246        goto FAILED;        goto FAILED;
6247        }        }
6248    
6249      /* If the character following a repeat is '+', or if certain optimization      /* If the character following a repeat is '+', possessive_quantifier is
6250      tests above succeeded, possessive_quantifier is TRUE. For some opcodes,      TRUE. For some opcodes, there are special alternative opcodes for this
6251      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
6252      wrap the entire repeated item inside OP_ONCE brackets. Logically, the '+'      brackets. Logically, the '+' notation is just syntactic sugar, taken from
6253      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.  
6254    
6255      Some (but not all) possessively repeated subpatterns have already been      Some (but not all) possessively repeated subpatterns have already been
6256      completely handled in the code just above. For them, possessive_quantifier      completely handled in the code just above. For them, possessive_quantifier
6257      is always FALSE at this stage.      is always FALSE at this stage. Note that the repeated item starts at
6258        tempcode, not at previous, which might be the first part of a string whose
6259      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. */  
6260    
6261      if (possessive_quantifier)      if (possessive_quantifier)
6262        {        {
6263        int len;        int len;
6264    
6265        if (*tempcode == OP_TYPEEXACT)        /* Possessifying an EXACT quantifier has no effect, so we can ignore it.
6266          However, QUERY, STAR, or UPTO may follow (for quantifiers such as {5,6},
6267          {5,}, or {5,10}). We skip over an EXACT item; if the length of what
6268          remains is greater than zero, there's a further opcode that can be
6269          handled. If not, do nothing, leaving the EXACT alone. */
6270    
6271          switch(*tempcode)
6272            {
6273            case OP_TYPEEXACT:
6274          tempcode += PRIV(OP_lengths)[*tempcode] +          tempcode += PRIV(OP_lengths)[*tempcode] +
6275            ((tempcode[1 + IMM2_SIZE] == OP_PROP            ((tempcode[1 + IMM2_SIZE] == OP_PROP
6276            || tempcode[1 + IMM2_SIZE] == OP_NOTPROP)? 2 : 0);            || tempcode[1 + IMM2_SIZE] == OP_NOTPROP)? 2 : 0);
6277            break;
6278    
6279        else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)          /* CHAR opcodes are used for exacts whose count is 1. */
6280          {  
6281            case OP_CHAR:
6282            case OP_CHARI:
6283            case OP_NOT:
6284            case OP_NOTI:
6285            case OP_EXACT:
6286            case OP_EXACTI:
6287            case OP_NOTEXACT:
6288            case OP_NOTEXACTI:
6289          tempcode += PRIV(OP_lengths)[*tempcode];          tempcode += PRIV(OP_lengths)[*tempcode];
6290  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
6291          if (utf && HAS_EXTRALEN(tempcode[-1]))          if (utf && HAS_EXTRALEN(tempcode[-1]))
6292            tempcode += GET_EXTRALEN(tempcode[-1]);            tempcode += GET_EXTRALEN(tempcode[-1]);
6293  #endif  #endif
6294            break;
6295    
6296            /* For the class opcodes, the repeat operator appears at the end;
6297            adjust tempcode to point to it. */
6298    
6299            case OP_CLASS:
6300            case OP_NCLASS:
6301            tempcode += 1 + 32/sizeof(pcre_uchar);
6302            break;
6303    
6304    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
6305            case OP_XCLASS:
6306            tempcode += GET(tempcode, 1);
6307            break;
6308    #endif
6309          }          }
6310    
6311          /* If tempcode is equal to code (which points to the end of the repeated
6312          item), it means we have skipped an EXACT item but there is no following
6313          QUERY, STAR, or UPTO; the value of len will be 0, and we do nothing. In
6314          all other cases, tempcode will be pointing to the repeat opcode, and will
6315          be less than code, so the value of len will be greater than 0. */
6316    
6317        len = (int)(code - tempcode);        len = (int)(code - tempcode);
6318          if (len > 0)
6319            {
6320            unsigned int repcode = *tempcode;
6321    
6322            /* There is a table for possessifying opcodes, all of which are less
6323            than OP_CALLOUT. A zero entry means there is no possessified version.
6324            */
6325    
6326            if (repcode < OP_CALLOUT && opcode_possessify[repcode] > 0)
6327              *tempcode = opcode_possessify[repcode];
6328    
6329            /* For opcode without a special possessified version, wrap the item in
6330            ONCE brackets. Because we are moving code along, we must ensure that any
6331            pending recursive references are updated. */
6332    
6333            else
6334              {
6335              *code = OP_END;
6336              adjust_recurse(tempcode, 1 + LINK_SIZE, utf, cd, save_hwm);
6337              memmove(tempcode + 1 + LINK_SIZE, tempcode, IN_UCHARS(len));
6338              code += 1 + LINK_SIZE;
6339              len += 1 + LINK_SIZE;
6340              tempcode[0] = OP_ONCE;
6341              *code++ = OP_KET;
6342              PUTINC(code, 0, len);
6343              PUT(tempcode, 1, len);
6344              }
6345            }
6346    
6347    #ifdef NEVER
6348        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
6349          {          {
6350          case OP_STAR:  *tempcode = OP_POSSTAR; break;          case OP_STAR:  *tempcode = OP_POSSTAR; break;
# Line 5794  for (;; ptr++) Line 6372  for (;; ptr++)
6372          case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;          case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
6373          case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;          case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;
6374    
6375            case OP_CRSTAR:   *tempcode = OP_CRPOSSTAR; break;
6376            case OP_CRPLUS:   *tempcode = OP_CRPOSPLUS; break;
6377            case OP_CRQUERY:  *tempcode = OP_CRPOSQUERY; break;
6378            case OP_CRRANGE:  *tempcode = OP_CRPOSRANGE; break;
6379    
6380          /* Because we are moving code along, we must ensure that any          /* Because we are moving code along, we must ensure that any
6381          pending recursive references are updated. */          pending recursive references are updated. */
6382    
# Line 5809  for (;; ptr++) Line 6392  for (;; ptr++)
6392          PUT(tempcode, 1, len);          PUT(tempcode, 1, len);
6393          break;          break;
6394          }          }
6395    #endif
6396        }        }
6397    
6398      /* 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 6581  for (;; ptr++)
6581          tempptr = ptr;          tempptr = ptr;
6582    
6583          /* A condition can be an assertion, a number (referring to a numbered          /* A condition can be an assertion, a number (referring to a numbered
6584          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',
6585          recursion. R<digits> and R&name are also permitted for recursion tests.          referring to recursion. R<digits> and R&name are also permitted for
6586            recursion tests.
6587          There are several syntaxes for testing a named group: (?(name)) is used  
6588          by Python; Perl 5.10 onwards uses (?(<name>) or (?('name')).          There are ways of testing a named group: (?(name)) is used by Python;
6589            Perl 5.10 onwards uses (?(<name>) or (?('name')).
6590          There are two unfortunate ambiguities, caused by history. (a) 'R' can  
6591          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
6592          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
6593          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.  
6594    
6595          For compatibility with auto-callouts, we allow a callout to be          For compatibility with auto-callouts, we allow a callout to be
6596          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 6614  for (;; ptr++)
6614                 tempptr[2] == CHAR_LESS_THAN_SIGN))                 tempptr[2] == CHAR_LESS_THAN_SIGN))
6615            break;            break;
6616    
6617          /* 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
6618          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. */
6619    
6620          code[1+LINK_SIZE] = OP_CREF;          code[1+LINK_SIZE] = OP_CREF;
6621          skipbytes = 1+IMM2_SIZE;          skipbytes = 1+IMM2_SIZE;
6622          refsign = -1;          refsign = -1;     /* => not a number */
6623            namelen = -1;     /* => not a name; must set to avoid warning */
6624            name = NULL;      /* Always set to avoid warning */
6625            recno = 0;        /* Always set to avoid warning */
6626    
6627          /* Check for a test for recursion in a named group. */          /* Check for a test for recursion in a named group. */
6628    
6629          if (ptr[1] == CHAR_R && ptr[2] == CHAR_AMPERSAND)          ptr++;
6630            if (*ptr == CHAR_R && ptr[1] == CHAR_AMPERSAND)
6631            {            {
6632            terminator = -1;            terminator = -1;
6633            ptr += 2;            ptr += 2;
# Line 6048  for (;; ptr++) Line 6635  for (;; ptr++)
6635            }            }
6636    
6637          /* 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
6638          syntax (?(<name>) or (?('name') */          syntax (?(<name>) or (?('name'), and also allow for the original PCRE
6639            syntax of (?(name) or for (?(+n), (?(-n), and just (?(n). */
6640    
6641          else if (ptr[1] == CHAR_LESS_THAN_SIGN)          else if (*ptr == CHAR_LESS_THAN_SIGN)
6642            {            {
6643            terminator = CHAR_GREATER_THAN_SIGN;            terminator = CHAR_GREATER_THAN_SIGN;
6644            ptr++;            ptr++;
6645            }            }
6646          else if (ptr[1] == CHAR_APOSTROPHE)          else if (*ptr == CHAR_APOSTROPHE)
6647            {            {
6648            terminator = CHAR_APOSTROPHE;            terminator = CHAR_APOSTROPHE;
6649            ptr++;            ptr++;
# Line 6063  for (;; ptr++) Line 6651  for (;; ptr++)
6651          else          else
6652            {            {
6653            terminator = CHAR_NULL;            terminator = CHAR_NULL;
6654            if (ptr[1] == CHAR_MINUS || ptr[1] == CHAR_PLUS) refsign = *(++ptr);            if (*ptr == CHAR_MINUS || *ptr == CHAR_PLUS) refsign = *ptr++;
6655                else if (IS_DIGIT(*ptr)) refsign = 0;
6656            }            }
6657    
6658          /* We now expect to read a name; any thing else is an error */          /* Handle a number */
6659    
6660          if (!MAX_255(ptr[1]) || (cd->ctypes[ptr[1]] & ctype_word) == 0)          if (refsign >= 0)
6661            {            {
6662            ptr += 1;  /* To get the right offset */            while (IS_DIGIT(*ptr))
6663            *errorcodeptr = ERR28;              {
6664            goto FAILED;              recno = recno * 10 + (int)(*ptr - CHAR_0);
6665                ptr++;
6666                }
6667            }            }
6668    
6669          /* 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
6670            a name is one of a number of duplicates, a different opcode is used and
6671            it needs more memory. Unfortunately we cannot tell whether a name is a
6672            duplicate in the first pass, so we have to allow for more memory. */
6673    
6674          recno = 0;          else
         name = ++ptr;  
         while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_word) != 0)  
6675            {            {
6676            if (recno >= 0)            if (IS_DIGIT(*ptr))
6677              recno = (IS_DIGIT(*ptr))? recno * 10 + (int)(*ptr - CHAR_0) : -1;              {
6678            ptr++;              *errorcodeptr = ERR84;
6679                goto FAILED;
6680                }
6681              if (!MAX_255(*ptr) || (cd->ctypes[*ptr] & ctype_word) == 0)
6682                {
6683                *errorcodeptr = ERR28;   /* Assertion expected */
6684                goto FAILED;
6685                }
6686              name = ptr++;
6687              while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_word) != 0)
6688                {
6689                ptr++;
6690                }
6691              namelen = (int)(ptr - name);
6692              if (lengthptr != NULL) *lengthptr += IMM2_SIZE;
6693            }            }
6694          namelen = (int)(ptr - name);  
6695            /* Check the terminator */
6696    
6697          if ((terminator > 0 && *ptr++ != (pcre_uchar)terminator) ||          if ((terminator > 0 && *ptr++ != (pcre_uchar)terminator) ||
6698              *ptr++ != CHAR_RIGHT_PARENTHESIS)              *ptr++ != CHAR_RIGHT_PARENTHESIS)
6699            {            {
6700            ptr--;      /* Error offset */            ptr--;                  /* Error offset */
6701            *errorcodeptr = ERR26;            *errorcodeptr = ERR26;  /* Malformed number or name */
6702            goto FAILED;            goto FAILED;
6703            }            }
6704    
# Line 6100  for (;; ptr++) Line 6707  for (;; ptr++)
6707          if (lengthptr != NULL) break;          if (lengthptr != NULL) break;
6708    
6709          /* 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
6710          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
6711          be digits, in which case recno will be set. */          recno. */
6712    
6713          if (refsign > 0)          if (refsign >= 0)
6714            {            {
6715            if (recno <= 0)            if (recno <= 0)
6716              {              {
6717              *errorcodeptr = ERR58;              *errorcodeptr = ERR35;
6718              goto FAILED;              goto FAILED;
6719              }              }
6720            recno = (refsign == CHAR_MINUS)?            if (refsign != 0) recno = (refsign == CHAR_MINUS)?
6721              cd->bracount - recno + 1 : recno +cd->bracount;              cd->bracount - recno + 1 : recno + cd->bracount;
6722            if (recno <= 0 || recno > cd->final_bracount)            if (recno <= 0 || recno > cd->final_bracount)
6723              {              {
6724              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
# Line 6121  for (;; ptr++) Line 6728  for (;; ptr++)
6728            break;            break;
6729            }            }
6730    
6731          /* 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. */  
6732    
6733          slot = cd->name_table;          slot = cd->name_table;
6734          for (i = 0; i < cd->names_found; i++)          for (i = 0; i < cd->names_found; i++)
# Line 6134  for (;; ptr++) Line 6737  for (;; ptr++)
6737            slot += cd->name_entry_size;            slot += cd->name_entry_size;
6738            }            }
6739    
6740          /* Found the named subpattern */          /* Found the named subpattern. If the name is duplicated, add one to
6741            the opcode to change CREF/RREF into DNCREF/DNRREF and insert
6742            appropriate data values. Otherwise, just insert the unique subpattern
6743            number. */
6744    
6745          if (i < cd->names_found)          if (i < cd->names_found)
6746            {            {
6747            recno = GET2(slot, 0);            int offset = i++;
6748            PUT2(code, 2+LINK_SIZE, recno);            int count = 1;
6749            code[1+LINK_SIZE]++;            recno = GET2(slot, 0);   /* Number from first found */
6750              for (; i < cd->names_found; i++)
6751                {
6752                slot += cd->name_entry_size;
6753                if (STRNCMP_UC_UC(name, slot+IMM2_SIZE, namelen) != 0) break;
6754                count++;
6755                }
6756              if (count > 1)
6757                {
6758                PUT2(code, 2+LINK_SIZE, offset);
6759                PUT2(code, 2+LINK_SIZE+IMM2_SIZE, count);
6760                skipbytes += IMM2_SIZE;
6761                code[1+LINK_SIZE]++;
6762                }
6763              else  /* Not a duplicated name */
6764                {
6765                PUT2(code, 2+LINK_SIZE, recno);
6766                }
6767            }            }
6768    
6769          /* If terminator == CHAR_NULL it means that the name followed directly          /* If terminator == CHAR_NULL it means that the name followed directly
6770          after the opening parenthesis [e.g. (?(abc)...] and in this case there          after the opening parenthesis [e.g. (?(abc)...] and in this case there
6771          are some further alternatives to try. For the cases where terminator !=          are some further alternatives to try. For the cases where terminator !=
6772          0 [things like (?(<name>... or (?('name')... or (?(R&name)... ] we have          CHAR_NULL [things like (?(<name>... or (?('name')... or (?(R&name)... ]
6773          now checked all the possibilities, so give an error. */          we have now checked all the possibilities, so give an error. */
6774    
6775          else if (terminator != CHAR_NULL)          else if (terminator != CHAR_NULL)
6776            {            {
# Line 6184  for (;; ptr++) Line 6807  for (;; ptr++)
6807            skipbytes = 1;            skipbytes = 1;
6808            }            }
6809    
6810          /* 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) */  
6811    
6812          else          else
6813            {            {
6814            *errorcodeptr = (recno == 0)? ERR35: ERR15;            *errorcodeptr = ERR15;
6815            goto FAILED;            goto FAILED;
6816            }            }
6817          break;          break;
# Line 6209  for (;; ptr++) Line 6824  for (;; ptr++)
6824          ptr++;          ptr++;
6825          break;          break;
6826    
6827            /* Optimize (?!) to (*FAIL) unless it is quantified - which is a weird
6828            thing to do, but Perl allows all assertions to be quantified, and when
6829            they contain capturing parentheses there may be a potential use for
6830            this feature. Not that that applies to a quantified (?!) but we allow
6831            it for uniformity. */
6832    
6833          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
6834          case CHAR_EXCLAMATION_MARK:            /* Negative lookahead */          case CHAR_EXCLAMATION_MARK:            /* Negative lookahead */
6835          ptr++;          ptr++;
6836          if (*ptr == CHAR_RIGHT_PARENTHESIS)    /* Optimize (?!) */          if (*ptr == CHAR_RIGHT_PARENTHESIS && ptr[1] != CHAR_ASTERISK &&
6837                 ptr[1] != CHAR_PLUS && ptr[1] != CHAR_QUESTION_MARK &&
6838                (ptr[1] != CHAR_LEFT_CURLY_BRACKET || !is_counted_repeat(ptr+2)))
6839            {            {
6840            *code++ = OP_FAIL;            *code++ = OP_FAIL;
6841            previous = NULL;            previous = NULL;
# Line 6309  for (;; ptr++) Line 6931  for (;; ptr++)
6931          terminator = (*ptr == CHAR_LESS_THAN_SIGN)?          terminator = (*ptr == CHAR_LESS_THAN_SIGN)?
6932            CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;            CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
6933          name = ++ptr;          name = ++ptr;
6934            if (IS_DIGIT(*ptr))
6935              {
6936              *errorcodeptr = ERR84;   /* Group name must start with non-digit */
6937              goto FAILED;
6938              }
6939          while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_word) != 0) ptr++;          while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
6940          namelen = (int)(ptr - name);          namelen = (int)(ptr - name);
6941    
# Line 6423  for (;; ptr++) Line 7049  for (;; ptr++)
7049    
7050          NAMED_REF_OR_RECURSE:          NAMED_REF_OR_RECURSE:
7051          name = ++ptr;          name = ++ptr;
7052            if (IS_DIGIT(*ptr))
7053              {
7054              *errorcodeptr = ERR84;   /* Group name must start with non-digit */
7055              goto FAILED;
7056              }
7057          while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_word) != 0) ptr++;          while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
7058          namelen = (int)(ptr - name);          namelen = (int)(ptr - name);
7059    
# Line 6815  for (;; ptr++) Line 7446  for (;; ptr++)
7446        skipbytes = IMM2_SIZE;        skipbytes = IMM2_SIZE;
7447        }        }
7448    
7449      /* Process nested bracketed regex. Assertions used not to be repeatable,      /* Process nested bracketed regex. First check for parentheses nested too
7450      but this was changed for Perl compatibility, so all kinds can now be      deeply. */
7451      repeated. We copy code into a non-register variable (tempcode) in order to  
7452      be able to pass its address because some compilers complain otherwise. */      if ((cd->parens_depth += 1) > PARENS_NEST_LIMIT)
7453          {
7454          *errorcodeptr = ERR82;
7455          goto FAILED;
7456          }
7457    
7458        /* Assertions used not to be repeatable, but this was changed for Perl
7459        compatibility, so all kinds can now be repeated. We copy code into a
7460        non-register variable (tempcode) in order to be able to pass its address
7461        because some compilers complain otherwise. */
7462    
7463      previous = code;                      /* For handling repetition */      previous = code;                      /* For handling repetition */
7464      *code = bravalue;      *code = bravalue;
# Line 6849  for (;; ptr++) Line 7489  for (;; ptr++)
7489           ))           ))
7490        goto FAILED;        goto FAILED;
7491    
7492        cd->parens_depth -= 1;
7493    
7494      /* 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,
7495      generate OP_ONCE_NC instead of OP_ONCE. */      generate OP_ONCE_NC instead of OP_ONCE. */
7496    
# Line 7063  for (;; ptr++) Line 7705  for (;; ptr++)
7705        if (escape == ESC_g)        if (escape == ESC_g)
7706          {          {
7707          const pcre_uchar *p;          const pcre_uchar *p;
7708            pcre_uint32 cf;
7709    
7710          save_hwm = cd->hwm;   /* Normally this is set when '(' is read */          save_hwm = cd->hwm;   /* Normally this is set when '(' is read */
7711          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
7712            CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;            CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
7713    
7714          /* These two statements stop the compiler for warning about possibly          /* These two statements stop the compiler for warning about possibly
7715          unset variables caused by the jump to HANDLE_NUMERICAL_RECURSION. In          unset variables caused by the jump to HANDLE_NUMERICAL_RECURSION. In
7716          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
7717          would actually be in error are never taken. */          would actually be in error are never taken. */
7718    
7719          skipbytes = 0;          skipbytes = 0;
7720          reset_bracount = FALSE;          reset_bracount = FALSE;
7721    
7722          /* Test for a name */          /* If it's not a signed or unsigned number, treat it as a name. */
7723    
7724          if (ptr[1] != CHAR_PLUS && ptr[1] != CHAR_MINUS)          cf = ptr[1];
7725            if (cf != CHAR_PLUS && cf != CHAR_MINUS && !IS_DIGIT(cf))
7726            {            {
           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;  
             }  
7727            is_recurse = TRUE;            is_recurse = TRUE;
7728            goto NAMED_REF_OR_RECURSE;            goto NAMED_REF_OR_RECURSE;
7729            }            }
7730    
7731          /* Test a signed number in angle brackets or quotes. */          /* Signed or unsigned number (cf = ptr[1]) is known to be plus or minus
7732            or a digit. */
7733    
7734          p = ptr + 2;          p = ptr + 2;
7735          while (IS_DIGIT(*p)) p++;          while (IS_DIGIT(*p)) p++;
# Line 7240  for (;; ptr++) Line 7869  for (;; ptr++)
7869    
7870      /* ===================================================================*/      /* ===================================================================*/
7871      /* Handle a literal character. It is guaranteed not to be whitespace or #      /* Handle a literal character. It is guaranteed not to be whitespace or #
7872      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
7873      multi-byte literal character. */      multi-unit literal character. */
7874    
7875      default:      default:
7876      NORMAL_CHAR:      NORMAL_CHAR:
# Line 7406  unsigned int orig_bracount; Line 8035  unsigned int orig_bracount;
8035  unsigned int max_bracount;  unsigned int max_bracount;
8036  branch_chain bc;  branch_chain bc;
8037    
8038    /* If set, call the external function that checks for stack availability. */
8039    
8040    if (PUBL(stack_guard) != NULL && PUBL(stack_guard)())
8041      {
8042      *errorcodeptr= ERR85;
8043      return FALSE;
8044      }
8045    
8046    /* Miscellaneous initialization */
8047    
8048  bc.outer = bcptr;  bc.outer = bcptr;
8049  bc.current_branch = code;  bc.current_branch = code;
8050    
# Line 7830  do { Line 8469  do {
8469       switch (*scode)       switch (*scode)
8470         {         {
8471         case OP_CREF:         case OP_CREF:
8472         case OP_NCREF:         case OP_DNCREF:
8473         case OP_RREF:         case OP_RREF:
8474         case OP_NRREF:         case OP_DNRREF:
8475         case OP_DEF:         case OP_DEF:
8476         return FALSE;         return FALSE;
8477    
# Line 8229  PCRE_UTF8 == PCRE_UTF16 == PCRE_UTF32. * Line 8868  PCRE_UTF8 == PCRE_UTF16 == PCRE_UTF32. *
8868      { skipatstart += 6; options |= PCRE_UTF8; continue; }      { skipatstart += 6; options |= PCRE_UTF8; continue; }
8869    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)
8870      { skipatstart += 6; options |= PCRE_UCP; continue; }      { skipatstart += 6; options |= PCRE_UCP; continue; }
8871      else if (STRNCMP_UC_C8(ptr+skipatstart+2, STRING_NO_AUTO_POSSESS_RIGHTPAR, 16) == 0)
8872        { skipatstart += 18; options |= PCRE_NO_AUTO_POSSESS; continue; }
8873    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)
8874      { skipatstart += 15; options |= PCRE_NO_START_OPTIMIZE; continue; }      { skipatstart += 15; options |= PCRE_NO_START_OPTIMIZE; continue; }
8875    
# Line 8423  cd->named_group_list_size = NAMED_GROUP_ Line 9064  cd->named_group_list_size = NAMED_GROUP_
9064  cd->start_pattern = (const pcre_uchar *)pattern;  cd->start_pattern = (const pcre_uchar *)pattern;
9065  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));
9066  cd->req_varyopt = 0;  cd->req_varyopt = 0;
9067    cd->parens_depth = 0;
9068  cd->assert_depth = 0;  cd->assert_depth = 0;
9069  cd->max_lookbehind = 0;  cd->max_lookbehind = 0;
9070  cd->external_options = options;  cd->external_options = options;
# Line 8437  outside can help speed up starting point Line 9079  outside can help speed up starting point
9079  ptr += skipatstart;  ptr += skipatstart;
9080  code = cworkspace;  code = cworkspace;
9081  *code = OP_BRA;  *code = OP_BRA;
9082    
9083  (void)compile_regex(cd->external_options, &code, &ptr, &errorcode, FALSE,  (void)compile_regex(cd->external_options, &code, &ptr, &errorcode, FALSE,
9084    FALSE, 0, 0, &firstchar, &firstcharflags, &reqchar, &reqcharflags, NULL,    FALSE, 0, 0, &firstchar, &firstcharflags, &reqchar, &reqcharflags, NULL,
9085    cd, &length);    cd, &length);
# Line 8508  field; this time it's used for rememberi Line 9151  field; this time it's used for rememberi
9151  */  */
9152    
9153  cd->final_bracount = cd->bracount;  /* Save for checking forward references */  cd->final_bracount = cd->bracount;  /* Save for checking forward references */
9154    cd->parens_depth = 0;
9155  cd->assert_depth = 0;  cd->assert_depth = 0;
9156  cd->bracount = 0;  cd->bracount = 0;
9157  cd->max_lookbehind = 0;  cd->max_lookbehind = 0;
# Line 8613  if (errorcode == 0 && re->top_backref > Line 9257  if (errorcode == 0 && re->top_backref >
9257  /* Unless disabled, check whether single character iterators can be  /* Unless disabled, check whether single character iterators can be
9258  auto-possessified. The function overwrites the appropriate opcode values. */  auto-possessified. The function overwrites the appropriate opcode values. */
9259    
9260  if ((options & PCRE_NO_AUTO_POSSESSIFY) == 0)  if ((options & PCRE_NO_AUTO_POSSESS) == 0)
9261    auto_possessify((pcre_uchar *)codestart, utf, cd);    auto_possessify((pcre_uchar *)codestart, utf, cd);
9262    
9263  /* 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 9481  return (pcre32 *)re;
9481  }  }
9482    
9483  /* End of pcre_compile.c */  /* End of pcre_compile.c */
9484    

Legend:
Removed from v.1363  
changed lines
  Added in v.1462

  ViewVC Help
Powered by ViewVC 1.1.5