/[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 1364 by ph10, Sat Oct 5 15:45:11 2013 UTC revision 1534 by ph10, Tue Mar 24 10:33:21 2015 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 47  supporting internal functions that are n Line 47  supporting internal functions that are n
47  #endif  #endif
48    
49  #define NLBLOCK cd             /* Block containing newline information */  #define NLBLOCK cd             /* Block containing newline information */
50  #define PSSTART start_pattern  /* Field containing processed string start */  #define PSSTART start_pattern  /* Field containing pattern start */
51  #define PSEND   end_pattern    /* Field containing processed string end */  #define PSEND   end_pattern    /* Field containing pattern end */
52    
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
# 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      "digits missing in \\x{} or \\o{}\0"
553    ;    ;
554    
555  /* 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 807  static const pcre_uint8 posspropstab[3][
807    { ucp_L, ucp_N, ucp_P, ucp_Po }   /* WORD */    { ucp_L, ucp_N, ucp_P, ucp_Po }   /* WORD */
808  };  };
809    
810    /* This table is used when converting repeating opcodes into possessified
811    versions as a result of an explicit possessive quantifier such as ++. A zero
812    value means there is no possessified version - in those cases the item in
813    question must be wrapped in ONCE brackets. The table is truncated at OP_CALLOUT
814    because all relevant opcodes are less than that. */
815    
816    static const pcre_uint8 opcode_possessify[] = {
817      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 0 - 15  */
818      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,   /* 16 - 31 */
819    
820      0,                       /* NOTI */
821      OP_POSSTAR, 0,           /* STAR, MINSTAR */
822      OP_POSPLUS, 0,           /* PLUS, MINPLUS */
823      OP_POSQUERY, 0,          /* QUERY, MINQUERY */
824      OP_POSUPTO, 0,           /* UPTO, MINUPTO */
825      0,                       /* EXACT */
826      0, 0, 0, 0,              /* POS{STAR,PLUS,QUERY,UPTO} */
827    
828      OP_POSSTARI, 0,          /* STARI, MINSTARI */
829      OP_POSPLUSI, 0,          /* PLUSI, MINPLUSI */
830      OP_POSQUERYI, 0,         /* QUERYI, MINQUERYI */
831      OP_POSUPTOI, 0,          /* UPTOI, MINUPTOI */
832      0,                       /* EXACTI */
833      0, 0, 0, 0,              /* POS{STARI,PLUSI,QUERYI,UPTOI} */
834    
835      OP_NOTPOSSTAR, 0,        /* NOTSTAR, NOTMINSTAR */
836      OP_NOTPOSPLUS, 0,        /* NOTPLUS, NOTMINPLUS */
837      OP_NOTPOSQUERY, 0,       /* NOTQUERY, NOTMINQUERY */
838      OP_NOTPOSUPTO, 0,        /* NOTUPTO, NOTMINUPTO */
839      0,                       /* NOTEXACT */
840      0, 0, 0, 0,              /* NOTPOS{STAR,PLUS,QUERY,UPTO} */
841    
842      OP_NOTPOSSTARI, 0,       /* NOTSTARI, NOTMINSTARI */
843      OP_NOTPOSPLUSI, 0,       /* NOTPLUSI, NOTMINPLUSI */
844      OP_NOTPOSQUERYI, 0,      /* NOTQUERYI, NOTMINQUERYI */
845      OP_NOTPOSUPTOI, 0,       /* NOTUPTOI, NOTMINUPTOI */
846      0,                       /* NOTEXACTI */
847      0, 0, 0, 0,              /* NOTPOS{STARI,PLUSI,QUERYI,UPTOI} */
848    
849      OP_TYPEPOSSTAR, 0,       /* TYPESTAR, TYPEMINSTAR */
850      OP_TYPEPOSPLUS, 0,       /* TYPEPLUS, TYPEMINPLUS */
851      OP_TYPEPOSQUERY, 0,      /* TYPEQUERY, TYPEMINQUERY */
852      OP_TYPEPOSUPTO, 0,       /* TYPEUPTO, TYPEMINUPTO */
853      0,                       /* TYPEEXACT */
854      0, 0, 0, 0,              /* TYPEPOS{STAR,PLUS,QUERY,UPTO} */
855    
856      OP_CRPOSSTAR, 0,         /* CRSTAR, CRMINSTAR */
857      OP_CRPOSPLUS, 0,         /* CRPLUS, CRMINPLUS */
858      OP_CRPOSQUERY, 0,        /* CRQUERY, CRMINQUERY */
859      OP_CRPOSRANGE, 0,        /* CRRANGE, CRMINRANGE */
860      0, 0, 0, 0,              /* CRPOS{STAR,PLUS,QUERY,RANGE} */
861    
862      0, 0, 0,                 /* CLASS, NCLASS, XCLASS */
863      0, 0,                    /* REF, REFI */
864      0, 0,                    /* DNREF, DNREFI */
865      0, 0                     /* RECURSE, CALLOUT */
866    };
867    
868    
869    
870  /*************************************************  /*************************************************
# Line 879  return (*p == CHAR_RIGHT_CURLY_BRACKET); Line 971  return (*p == CHAR_RIGHT_CURLY_BRACKET);
971  *************************************************/  *************************************************/
972    
973  /* 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
974  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
975  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.
976  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
977  be returned in chptr.  chptr. On entry, ptr is pointing at the \. On exit, it is on the final
978  On entry,ptr is pointing at the \. On exit, it is on the final character of the  character of the escape sequence.
 escape sequence.  
979    
980  Arguments:  Arguments:
981    ptrptr         points to the pattern position pointer    ptrptr         points to the pattern position pointer
982    chptr          points to the data character    chptr          points to a returned data character
983    errorcodeptr   points to the errorcode variable    errorcodeptr   points to the errorcode variable
984    bracount       number of previous extracting brackets    bracount       number of previous extracting brackets
985    options        the options bits    options        the options bits
# Line 1092  else Line 1183  else
1183      break;      break;
1184    
1185      /* The handling of escape sequences consisting of a string of digits      /* The handling of escape sequences consisting of a string of digits
1186      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
1187      the way Perl works seems to be as follows:      over the years. Nowadays \g{} for backreferences and \o{} for octal are
1188        recommended to avoid the ambiguities in the old syntax.
1189    
1190      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
1191      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
1192      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
1193      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
1194      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
1195      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
1196      character class, \ followed by a digit is always an octal number. */      taken. \8 and \9 are treated as the literal characters 8 and 9.
1197    
1198        Inside a character class, \ followed by a digit is always either a literal
1199        8 or 9 or an octal number. */
1200    
1201      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:
1202      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 1223  else
1223          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
1224          break;          break;
1225          }          }
1226        if (s < 10 || s <= bracount)        if (s < 8 || s <= bracount)  /* Check for back reference */
1227          {          {
1228          escape = -s;          escape = -s;
1229          break;          break;
# Line 1136  else Line 1231  else
1231        ptr = oldptr;      /* Put the pointer back and fall through */        ptr = oldptr;      /* Put the pointer back and fall through */
1232        }        }
1233    
1234      /* 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
1235      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
1236      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
1237        changed so as not to insert the binary zero. */
1238    
1239      if ((c = *ptr) >= CHAR_8)      if ((c = *ptr) >= CHAR_8) break;
1240        {  
1241        ptr--;      /* Fall through with a digit less than 8 */
       c = 0;  
       break;  
       }  
1242    
1243      /* \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
1244      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 1255  else
1255  #endif  #endif
1256      break;      break;
1257    
1258      /* \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
1259      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}. */
1260      If not, { is treated as a data character. */  
1261        case CHAR_o:
1262        if (ptr[1] != CHAR_LEFT_CURLY_BRACKET) *errorcodeptr = ERR81; else
1263        if (ptr[2] == CHAR_RIGHT_CURLY_BRACKET) *errorcodeptr = ERR86; else
1264          {
1265          ptr += 2;
1266          c = 0;
1267          overflow = FALSE;
1268          while (*ptr >= CHAR_0 && *ptr <= CHAR_7)
1269            {
1270            register pcre_uint32 cc = *ptr++;
1271            if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1272    #ifdef COMPILE_PCRE32
1273            if (c >= 0x20000000l) { overflow = TRUE; break; }
1274    #endif
1275            c = (c << 3) + cc - CHAR_0 ;
1276    #if defined COMPILE_PCRE8
1277            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1278    #elif defined COMPILE_PCRE16
1279            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1280    #elif defined COMPILE_PCRE32
1281            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1282    #endif
1283            }
1284          if (overflow)
1285            {
1286            while (*ptr >= CHAR_0 && *ptr <= CHAR_7) ptr++;
1287            *errorcodeptr = ERR34;
1288            }
1289          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1290            {
1291            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1292            }
1293          else *errorcodeptr = ERR80;
1294          }
1295        break;
1296    
1297        /* \x is complicated. In JavaScript, \x must be followed by two hexadecimal
1298        numbers. Otherwise it is a lowercase x letter. */
1299    
1300      case CHAR_x:      case CHAR_x:
1301      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)      if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1302        {        {
       /* In JavaScript, \x must be followed by two hexadecimal numbers.  
       Otherwise it is a lowercase x letter. */  
1303        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0        if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1304          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)          && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1305          {          {
# Line 1187  else Line 1316  else
1316  #endif  #endif
1317            }            }
1318          }          }
1319        break;        }    /* End JavaScript handling */
       }  
1320    
1321      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)      /* Handle \x in Perl's style. \x{ddd} is a character number which can be
1322        {      greater than 0xff in utf or non-8bit mode, but only if the ddd are hex
1323        const pcre_uchar *pt = ptr + 2;      digits. If not, { used to be treated as a data character. However, Perl
1324        seems to read hex digits up to the first non-such, and ignore the rest, so
1325        that, for example \x{zz} matches a binary zero. This seems crazy, so PCRE
1326        now gives an error. */
1327    
1328        c = 0;      else
1329        overflow = FALSE;        {
1330        while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0)        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1331          {          {
1332          register pcre_uint32 cc = *pt++;          ptr += 2;
1333          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */          if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1334              {
1335              *errorcodeptr = ERR86;
1336              break;
1337              }
1338            c = 0;
1339            overflow = FALSE;
1340            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0)
1341              {
1342              register pcre_uint32 cc = *ptr++;
1343              if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
1344    
1345  #ifdef COMPILE_PCRE32  #ifdef COMPILE_PCRE32
1346          if (c >= 0x10000000l) { overflow = TRUE; break; }            if (c >= 0x10000000l) { overflow = TRUE; break; }
1347  #endif  #endif
1348    
1349  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1350          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
1351          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1352  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1353          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 */
1354          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1355  #endif  #endif
1356    
1357  #if defined COMPILE_PCRE8  #if defined COMPILE_PCRE8
1358          if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }            if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1359  #elif defined COMPILE_PCRE16  #elif defined COMPILE_PCRE16
1360          if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }            if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1361  #elif defined COMPILE_PCRE32  #elif defined COMPILE_PCRE32
1362          if (utf && c > 0x10ffffU) { overflow = TRUE; break; }            if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1363  #endif  #endif
1364          }            }
1365    
1366        if (overflow)          if (overflow)
1367          {            {
1368          while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0) pt++;            while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0) ptr++;
1369          *errorcodeptr = ERR34;            *errorcodeptr = ERR34;
1370          }            }
1371    
1372        if (*pt == CHAR_RIGHT_CURLY_BRACKET)          else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1373          {            {
1374          if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;            if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1375          ptr = pt;            }
         break;  
         }  
1376    
1377        /* 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.
1378        recognize this construct; fall through to the normal \x handling. */          We used just to recognize this construct and fall through to the normal
1379        }          \x handling, but nowadays Perl gives an error, which seems much more
1380            sensible, so we do too. */
1381    
1382      /* Read just a single-byte hex-defined char */          else *errorcodeptr = ERR79;
1383            }   /* End of \x{} processing */
1384    
1385      c = 0;        /* Read a single-byte hex-defined char (up to two hex digits after \x) */
1386      while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)  
1387        {        else
1388        pcre_uint32 cc;                          /* Some compilers don't like */          {
1389        cc = *(++ptr);                           /* ++ in initializers */          c = 0;
1390            while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1391              {
1392              pcre_uint32 cc;                          /* Some compilers don't like */
1393              cc = *(++ptr);                           /* ++ in initializers */
1394  #ifndef EBCDIC  /* ASCII/UTF-8 coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
1395        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */            if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
1396        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));            c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1397  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
1398        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */            if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
1399        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));            c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1400  #endif  #endif
1401        }            }
1402            }     /* End of \xdd handling */
1403          }       /* End of Perl-style \x handling */
1404      break;      break;
1405    
1406      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
# Line 1443  read_repeat_counts(const pcre_uchar *p, Line 1590  read_repeat_counts(const pcre_uchar *p,
1590  int min = 0;  int min = 0;
1591  int max = -1;  int max = -1;
1592    
1593  /* Read the minimum value and do a paranoid check: a negative value indicates  while (IS_DIGIT(*p))
 an integer overflow. */  
   
 while (IS_DIGIT(*p)) min = min * 10 + (int)(*p++ - CHAR_0);  
 if (min < 0 || min > 65535)  
1594    {    {
1595    *errorcodeptr = ERR5;    min = min * 10 + (int)(*p++ - CHAR_0);
1596    return p;    if (min > 65535)
1597        {
1598        *errorcodeptr = ERR5;
1599        return p;
1600        }
1601    }    }
1602    
 /* Read the maximum value if there is one, and again do a paranoid on its size.  
 Also, max must not be less than min. */  
   
1603  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1604    {    {
1605    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1606      {      {
1607      max = 0;      max = 0;
1608      while(IS_DIGIT(*p)) max = max * 10 + (int)(*p++ - CHAR_0);      while(IS_DIGIT(*p))
     if (max < 0 || max > 65535)  
1609        {        {
1610        *errorcodeptr = ERR5;        max = max * 10 + (int)(*p++ - CHAR_0);
1611        return p;        if (max > 65535)
1612            {
1613            *errorcodeptr = ERR5;
1614            return p;
1615            }
1616        }        }
1617      if (max < min)      if (max < min)
1618        {        {
# Line 1475  if (*p == CHAR_RIGHT_CURLY_BRACKET) max Line 1622  if (*p == CHAR_RIGHT_CURLY_BRACKET) max
1622      }      }
1623    }    }
1624    
 /* Fill in the required variables, and pass back the pointer to the terminating  
 '}'. */  
   
1625  *minp = min;  *minp = min;
1626  *maxp = max;  *maxp = max;
1627  return p;  return p;
# Line 1524  for (;;) Line 1668  for (;;)
1668    
1669      case OP_CALLOUT:      case OP_CALLOUT:
1670      case OP_CREF:      case OP_CREF:
1671      case OP_NCREF:      case OP_DNCREF:
1672      case OP_RREF:      case OP_RREF:
1673      case OP_NRREF:      case OP_DNRREF:
1674      case OP_DEF:      case OP_DEF:
1675      code += PRIV(OP_lengths)[*code];      code += PRIV(OP_lengths)[*code];
1676      break;      break;
# Line 1663  for (;;) Line 1807  for (;;)
1807      case OP_COMMIT:      case OP_COMMIT:
1808      case OP_CREF:      case OP_CREF:
1809      case OP_DEF:      case OP_DEF:
1810        case OP_DNCREF:
1811        case OP_DNRREF:
1812      case OP_DOLL:      case OP_DOLL:
1813      case OP_DOLLM:      case OP_DOLLM:
1814      case OP_EOD:      case OP_EOD:
1815      case OP_EODN:      case OP_EODN:
1816      case OP_FAIL:      case OP_FAIL:
     case OP_NCREF:  
     case OP_NRREF:  
1817      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1818      case OP_PRUNE:      case OP_PRUNE:
1819      case OP_REVERSE:      case OP_REVERSE:
# Line 1764  for (;;) Line 1908  for (;;)
1908    
1909      switch (*cc)      switch (*cc)
1910        {        {
       case OP_CRPLUS:  
       case OP_CRMINPLUS:  
1911        case OP_CRSTAR:        case OP_CRSTAR:
1912        case OP_CRMINSTAR:        case OP_CRMINSTAR:
1913          case OP_CRPLUS:
1914          case OP_CRMINPLUS:
1915        case OP_CRQUERY:        case OP_CRQUERY:
1916        case OP_CRMINQUERY:        case OP_CRMINQUERY:
1917          case OP_CRPOSSTAR:
1918          case OP_CRPOSPLUS:
1919          case OP_CRPOSQUERY:
1920        return -1;        return -1;
1921    
1922        case OP_CRRANGE:        case OP_CRRANGE:
1923        case OP_CRMINRANGE:        case OP_CRMINRANGE:
1924          case OP_CRPOSRANGE:
1925        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;        if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1926        branchlength += (int)GET2(cc,1);        branchlength += (int)GET2(cc,1);
1927        cc += 1 + 2 * IMM2_SIZE;        cc += 1 + 2 * IMM2_SIZE;
# Line 1981  for (;;) Line 2129  for (;;)
2129        {        {
2130        case OP_CHAR:        case OP_CHAR:
2131        case OP_CHARI:        case OP_CHARI:
2132          case OP_NOT:
2133          case OP_NOTI:
2134        case OP_EXACT:        case OP_EXACT:
2135        case OP_EXACTI:        case OP_EXACTI:
2136          case OP_NOTEXACT:
2137          case OP_NOTEXACTI:
2138        case OP_UPTO:        case OP_UPTO:
2139        case OP_UPTOI:        case OP_UPTOI:
2140          case OP_NOTUPTO:
2141          case OP_NOTUPTOI:
2142        case OP_MINUPTO:        case OP_MINUPTO:
2143        case OP_MINUPTOI:        case OP_MINUPTOI:
2144          case OP_NOTMINUPTO:
2145          case OP_NOTMINUPTOI:
2146        case OP_POSUPTO:        case OP_POSUPTO:
2147        case OP_POSUPTOI:        case OP_POSUPTOI:
2148          case OP_NOTPOSUPTO:
2149          case OP_NOTPOSUPTOI:
2150        case OP_STAR:        case OP_STAR:
2151        case OP_STARI:        case OP_STARI:
2152          case OP_NOTSTAR:
2153          case OP_NOTSTARI:
2154        case OP_MINSTAR:        case OP_MINSTAR:
2155        case OP_MINSTARI:        case OP_MINSTARI:
2156          case OP_NOTMINSTAR:
2157          case OP_NOTMINSTARI:
2158        case OP_POSSTAR:        case OP_POSSTAR:
2159        case OP_POSSTARI:        case OP_POSSTARI:
2160          case OP_NOTPOSSTAR:
2161          case OP_NOTPOSSTARI:
2162        case OP_PLUS:        case OP_PLUS:
2163        case OP_PLUSI:        case OP_PLUSI:
2164          case OP_NOTPLUS:
2165          case OP_NOTPLUSI:
2166        case OP_MINPLUS:        case OP_MINPLUS:
2167        case OP_MINPLUSI:        case OP_MINPLUSI:
2168          case OP_NOTMINPLUS:
2169          case OP_NOTMINPLUSI:
2170        case OP_POSPLUS:        case OP_POSPLUS:
2171        case OP_POSPLUSI:        case OP_POSPLUSI:
2172          case OP_NOTPOSPLUS:
2173          case OP_NOTPOSPLUSI:
2174        case OP_QUERY:        case OP_QUERY:
2175        case OP_QUERYI:        case OP_QUERYI:
2176          case OP_NOTQUERY:
2177          case OP_NOTQUERYI:
2178        case OP_MINQUERY:        case OP_MINQUERY:
2179        case OP_MINQUERYI:        case OP_MINQUERYI:
2180          case OP_NOTMINQUERY:
2181          case OP_NOTMINQUERYI:
2182        case OP_POSQUERY:        case OP_POSQUERY:
2183        case OP_POSQUERYI:        case OP_POSQUERYI:
2184          case OP_NOTPOSQUERY:
2185          case OP_NOTPOSQUERYI:
2186        if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);        if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2187        break;        break;
2188        }        }
# Line 2226  for (code = first_significant_code(code Line 2402  for (code = first_significant_code(code
2402    if (c == OP_RECURSE)    if (c == OP_RECURSE)
2403      {      {
2404      const pcre_uchar *scode = cd->start_code + GET(code, 1);      const pcre_uchar *scode = cd->start_code + GET(code, 1);
2405        const pcre_uchar *endgroup = scode;
2406      BOOL empty_branch;      BOOL empty_branch;
2407    
2408      /* Test for forward reference or uncompleted reference. This is disabled      /* Test for forward reference or uncompleted reference. This is disabled
# Line 2240  for (code = first_significant_code(code Line 2417  for (code = first_significant_code(code
2417        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2418        }        }
2419    
2420      /* If we are scanning a completed pattern, there are no forward references      /* If the reference is to a completed group, we need to detect whether this
2421      and all groups are complete. We need to detect whether this is a recursive      is a recursive call, as otherwise there will be an infinite loop. If it is
2422      call, as otherwise there will be an infinite loop. If it is a recursion,      a recursion, just skip over it. Simple recursions are easily detected. For
2423      just skip over it. Simple recursions are easily detected. For mutual      mutual recursions we keep a chain on the stack. */
     recursions we keep a chain on the stack. */  
2424    
2425        do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);
2426        if (code >= scode && code <= endgroup) continue;  /* Simple recursion */
2427      else      else
2428        {        {
2429        recurse_check *r = recurses;        recurse_check *r = recurses;
       const pcre_uchar *endgroup = scode;  
   
       do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);  
       if (code >= scode && code <= endgroup) continue;  /* Simple recursion */  
   
2430        for (r = recurses; r != NULL; r = r->prev)        for (r = recurses; r != NULL; r = r->prev)
2431          if (r->group == scode) break;          if (r->group == scode) break;
2432        if (r != NULL) continue;   /* Mutual recursion */        if (r != NULL) continue;   /* Mutual recursion */
# Line 2366  for (code = first_significant_code(code Line 2539  for (code = first_significant_code(code
2539        case OP_CRMINSTAR:        case OP_CRMINSTAR:
2540        case OP_CRQUERY:        case OP_CRQUERY:
2541        case OP_CRMINQUERY:        case OP_CRMINQUERY:
2542          case OP_CRPOSSTAR:
2543          case OP_CRPOSQUERY:
2544        break;        break;
2545    
2546        default:                   /* Non-repeat => class must match */        default:                   /* Non-repeat => class must match */
2547        case OP_CRPLUS:            /* These repeats aren't empty */        case OP_CRPLUS:            /* These repeats aren't empty */
2548        case OP_CRMINPLUS:        case OP_CRMINPLUS:
2549          case OP_CRPOSPLUS:
2550        return FALSE;        return FALSE;
2551    
2552        case OP_CRRANGE:        case OP_CRRANGE:
2553        case OP_CRMINRANGE:        case OP_CRMINRANGE:
2554          case OP_CRPOSRANGE:
2555        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */        if (GET2(ccode, 1) > 0) return FALSE;  /* Minimum > 0 */
2556        break;        break;
2557        }        }
# Line 2653  switch(ptype) Line 2830  switch(ptype)
2830    /* Perl space used to exclude VT, but from Perl 5.18 it is included, which    /* Perl space used to exclude VT, but from Perl 5.18 it is included, which
2831    means that Perl space and POSIX space are now identical. PCRE was changed    means that Perl space and POSIX space are now identical. PCRE was changed
2832    at release 8.34. */    at release 8.34. */
2833    
2834    case PT_SPACE:    /* Perl space */    case PT_SPACE:    /* Perl space */
2835    case PT_PXSPACE:  /* POSIX space */    case PT_PXSPACE:  /* POSIX space */
2836    return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||    switch(c)
2837            c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||      {
2838            c == CHAR_FF || c == CHAR_CR)      HSPACE_CASES:
2839            == negated;      VSPACE_CASES:
2840        return negated;
2841    
2842        default:
2843        return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z) == negated;
2844        }
2845      break;  /* Control never reaches here */
2846    
2847    case PT_WORD:    case PT_WORD:
2848    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 2891  get_chr_property_list(const pcre_uchar *
2891    const pcre_uint8 *fcc, pcre_uint32 *list)    const pcre_uint8 *fcc, pcre_uint32 *list)
2892  {  {
2893  pcre_uchar c = *code;  pcre_uchar c = *code;
2894    pcre_uchar base;
2895  const pcre_uchar *end;  const pcre_uchar *end;
 const pcre_uint32 *clist_src;  
 pcre_uint32 *clist_dest;  
2896  pcre_uint32 chr;  pcre_uint32 chr;
2897  pcre_uchar base;  
2898    #ifdef SUPPORT_UCP
2899    pcre_uint32 *clist_dest;
2900    const pcre_uint32 *clist_src;
2901    #else
2902    utf = utf;  /* Suppress "unused parameter" compiler warning */
2903    #endif
2904    
2905  list[0] = c;  list[0] = c;
2906  list[1] = FALSE;  list[1] = FALSE;
# Line 2818  switch(c) Line 3006  switch(c)
3006      return code + 2;      return code + 2;
3007      }      }
3008    
3009    /* Convert only if we have anough space. */    /* Convert only if we have enough space. */
3010    
3011    clist_src = PRIV(ucd_caseless_sets) + code[1];    clist_src = PRIV(ucd_caseless_sets) + code[1];
3012    clist_dest = list + 2;    clist_dest = list + 2;
3013    code += 2;    code += 2;
3014    
3015    do {    do {
      /* Early return if there is not enough space. */  
3016       if (clist_dest >= list + 8)       if (clist_dest >= list + 8)
3017         {         {
3018           /* Early return if there is not enough space. This should never
3019           happen, since all clists are shorter than 5 character now. */
3020         list[2] = code[0];         list[2] = code[0];
3021         list[3] = code[1];         list[3] = code[1];
3022         return code;         return code;
3023         }         }
3024       *clist_dest++ = *clist_src;       *clist_dest++ = *clist_src;
3025       }       }
3026     while(*clist_src++ != NOTACHAR);    while(*clist_src++ != NOTACHAR);
3027    
3028    /* Enough space to store all characters. */    /* All characters are stored. The terminating NOTACHAR
3029      is copied form the clist itself. */
3030    
3031    list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;    list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
3032    return code;    return code;
# Line 2846  switch(c) Line 3036  switch(c)
3036    case OP_CLASS:    case OP_CLASS:
3037  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3038    case OP_XCLASS:    case OP_XCLASS:
   
3039    if (c == OP_XCLASS)    if (c == OP_XCLASS)
3040      end = code + GET(code, 0);      end = code + GET(code, 0) - 1;
3041    else    else
3042  #endif  #endif
3043      end = code + 32 / sizeof(pcre_uchar);      end = code + 32 / sizeof(pcre_uchar);
# Line 2859  switch(c) Line 3048  switch(c)
3048      case OP_CRMINSTAR:      case OP_CRMINSTAR:
3049      case OP_CRQUERY:      case OP_CRQUERY:
3050      case OP_CRMINQUERY:      case OP_CRMINQUERY:
3051        case OP_CRPOSSTAR:
3052        case OP_CRPOSQUERY:
3053      list[1] = TRUE;      list[1] = TRUE;
3054      end++;      end++;
3055      break;      break;
3056    
3057        case OP_CRPLUS:
3058        case OP_CRMINPLUS:
3059        case OP_CRPOSPLUS:
3060        end++;
3061        break;
3062    
3063      case OP_CRRANGE:      case OP_CRRANGE:
3064      case OP_CRMINRANGE:      case OP_CRMINRANGE:
3065        case OP_CRPOSRANGE:
3066      list[1] = (GET2(end, 1) == 0);      list[1] = (GET2(end, 1) == 0);
3067      end += 1 + 2 * IMM2_SIZE;      end += 1 + 2 * IMM2_SIZE;
3068      break;      break;
3069      }      }
3070    list[2] = end - code;    list[2] = (pcre_uint32)(end - code);
3071    return end;    return end;
3072    }    }
3073  return NULL;    /* Opcode not accepted */  return NULL;    /* Opcode not accepted */
# Line 2895  Returns:      TRUE if the auto-possessif Line 3093  Returns:      TRUE if the auto-possessif
3093    
3094  static BOOL  static BOOL
3095  compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,  compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
3096    const pcre_uint32* base_list)    const pcre_uint32 *base_list, const pcre_uchar *base_end)
3097  {  {
3098  pcre_uchar c;  pcre_uchar c;
3099  pcre_uint32 list[8];  pcre_uint32 list[8];
3100  const pcre_uint32* chr_ptr;  const pcre_uint32 *chr_ptr;
3101  const pcre_uint32* ochr_ptr;  const pcre_uint32 *ochr_ptr;
3102  const pcre_uint32* list_ptr;  const pcre_uint32 *list_ptr;
3103    const pcre_uchar *next_code;
3104    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3105    const pcre_uchar *xclass_flags;
3106    #endif
3107    const pcre_uint8 *class_bitset;
3108    const pcre_uint8 *set1, *set2, *set_end;
3109  pcre_uint32 chr;  pcre_uint32 chr;
3110    BOOL accepted, invert_bits;
3111    BOOL entered_a_group = FALSE;
3112    
3113    /* Note: the base_list[1] contains whether the current opcode has greedy
3114    (represented by a non-zero value) quantifier. This is a different from
3115    other character type lists, which stores here that the character iterator
3116    matches to an empty string (also represented by a non-zero value). */
3117    
3118  for(;;)  for(;;)
3119    {    {
3120      /* All operations move the code pointer forward.
3121      Therefore infinite recursions are not possible. */
3122    
3123    c = *code;    c = *code;
3124    
3125    /* Skip over callouts */    /* Skip over callouts */
# Line 2925  for(;;) Line 3139  for(;;)
3139    switch(c)    switch(c)
3140      {      {
3141      case OP_END:      case OP_END:
3142      /* TRUE only in greedy case. The non-greedy case could be replaced by an      case OP_KETRPOS:
3143      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
3144      more memory, which we cannot get at this stage.) */      an OP_EXACT, but it is probably not worth it. (And note that OP_EXACT
3145        uses more memory, which we cannot get at this stage.) */
3146    
3147      return base_list[1] != 0;      return base_list[1] != 0;
3148    
3149      case OP_KET:      case OP_KET:
3150      /* If the bracket is capturing, and referenced by an OP_RECURSE, the      /* If the bracket is capturing, and referenced by an OP_RECURSE, or
3151      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
3152      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.) */  
3153    
3154      if (base_list[1] == 0) return FALSE;      if (base_list[1] == 0) return FALSE;
3155    
3156        switch(*(code - GET(code, 1)))
3157          {
3158          case OP_ASSERT:
3159          case OP_ASSERT_NOT:
3160          case OP_ASSERTBACK:
3161          case OP_ASSERTBACK_NOT:
3162          case OP_ONCE:
3163          case OP_ONCE_NC:
3164          /* Atomic sub-patterns and assertions can always auto-possessify their
3165          last iterator. However, if the group was entered as a result of checking
3166          a previous iterator, this is not possible. */
3167    
3168          return !entered_a_group;
3169          }
3170    
3171        code += PRIV(OP_lengths)[c];
3172        continue;
3173    
3174        case OP_ONCE:
3175        case OP_ONCE_NC:
3176        case OP_BRA:
3177        case OP_CBRA:
3178        next_code = code + GET(code, 1);
3179      code += PRIV(OP_lengths)[c];      code += PRIV(OP_lengths)[c];
3180    
3181        while (*next_code == OP_ALT)
3182          {
3183          if (!compare_opcodes(code, utf, cd, base_list, base_end)) return FALSE;
3184          code = next_code + 1 + LINK_SIZE;
3185          next_code += GET(next_code, 1);
3186          }
3187    
3188        entered_a_group = TRUE;
3189      continue;      continue;
3190    
3191        case OP_BRAZERO:
3192        case OP_BRAMINZERO:
3193    
3194        next_code = code + 1;
3195        if (*next_code != OP_BRA && *next_code != OP_CBRA
3196            && *next_code != OP_ONCE && *next_code != OP_ONCE_NC) return FALSE;
3197    
3198        do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
3199    
3200        /* The bracket content will be checked by the
3201        OP_BRA/OP_CBRA case above. */
3202        next_code += 1 + LINK_SIZE;
3203        if (!compare_opcodes(next_code, utf, cd, base_list, base_end))
3204          return FALSE;
3205    
3206        code += PRIV(OP_lengths)[c];
3207        continue;
3208    
3209        default:
3210        break;
3211      }      }
3212    
3213    /* Check for a supported opcode, and load its properties. */    /* Check for a supported opcode, and load its properties. */
# Line 2961  for(;;) Line 3229  for(;;)
3229      list_ptr = base_list;      list_ptr = base_list;
3230      }      }
3231    
3232      /* Character bitsets can also be compared to certain opcodes. */
3233    
3234      else if (base_list[0] == OP_CLASS || list[0] == OP_CLASS
3235    #ifdef COMPILE_PCRE8
3236          /* In 8 bit, non-UTF mode, OP_CLASS and OP_NCLASS are the same. */
3237          || (!utf && (base_list[0] == OP_NCLASS || list[0] == OP_NCLASS))
3238    #endif
3239          )
3240        {
3241    #ifdef COMPILE_PCRE8
3242        if (base_list[0] == OP_CLASS || (!utf && base_list[0] == OP_NCLASS))
3243    #else
3244        if (base_list[0] == OP_CLASS)
3245    #endif
3246          {
3247          set1 = (pcre_uint8 *)(base_end - base_list[2]);
3248          list_ptr = list;
3249          }
3250        else
3251          {
3252          set1 = (pcre_uint8 *)(code - list[2]);
3253          list_ptr = base_list;
3254          }
3255    
3256        invert_bits = FALSE;
3257        switch(list_ptr[0])
3258          {
3259          case OP_CLASS:
3260          case OP_NCLASS:
3261          set2 = (pcre_uint8 *)
3262            ((list_ptr == list ? code : base_end) - list_ptr[2]);
3263          break;
3264    
3265    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3266          case OP_XCLASS:
3267          xclass_flags = (list_ptr == list ? code : base_end) - list_ptr[2] + LINK_SIZE;
3268          if ((*xclass_flags & XCL_HASPROP) != 0) return FALSE;
3269          if ((*xclass_flags & XCL_MAP) == 0)
3270            {
3271            /* No bits are set for characters < 256. */
3272            if (list[1] == 0) return TRUE;
3273            /* Might be an empty repeat. */
3274            continue;
3275            }
3276          set2 = (pcre_uint8 *)(xclass_flags + 1);
3277          break;
3278    #endif
3279    
3280          case OP_NOT_DIGIT:
3281          invert_bits = TRUE;
3282          /* Fall through */
3283          case OP_DIGIT:
3284          set2 = (pcre_uint8 *)(cd->cbits + cbit_digit);
3285          break;
3286    
3287          case OP_NOT_WHITESPACE:
3288          invert_bits = TRUE;
3289          /* Fall through */
3290          case OP_WHITESPACE:
3291          set2 = (pcre_uint8 *)(cd->cbits + cbit_space);
3292          break;
3293    
3294          case OP_NOT_WORDCHAR:
3295          invert_bits = TRUE;
3296          /* Fall through */
3297          case OP_WORDCHAR:
3298          set2 = (pcre_uint8 *)(cd->cbits + cbit_word);
3299          break;
3300    
3301          default:
3302          return FALSE;
3303          }
3304    
3305        /* Because the sets are unaligned, we need
3306        to perform byte comparison here. */
3307        set_end = set1 + 32;
3308        if (invert_bits)
3309          {
3310          do
3311            {
3312            if ((*set1++ & ~(*set2++)) != 0) return FALSE;
3313            }
3314          while (set1 < set_end);
3315          }
3316        else
3317          {
3318          do
3319            {
3320            if ((*set1++ & *set2++) != 0) return FALSE;
3321            }
3322          while (set1 < set_end);
3323          }
3324    
3325        if (list[1] == 0) return TRUE;
3326        /* Might be an empty repeat. */
3327        continue;
3328        }
3329    
3330    /* Some property combinations also acceptable. Unicode property opcodes are    /* Some property combinations also acceptable. Unicode property opcodes are
3331    processed specially; the rest can be handled with a lookup table. */    processed specially; the rest can be handled with a lookup table. */
3332    
# Line 2968  for(;;) Line 3334  for(;;)
3334      {      {
3335      pcre_uint32 leftop, rightop;      pcre_uint32 leftop, rightop;
3336    
     if (list[1] != 0) return FALSE;   /* Must match at least one character */  
3337      leftop = base_list[0];      leftop = base_list[0];
3338      rightop = list[0];      rightop = list[0];
3339    
3340  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3341        accepted = FALSE; /* Always set in non-unicode case. */
3342      if (leftop == OP_PROP || leftop == OP_NOTPROP)      if (leftop == OP_PROP || leftop == OP_NOTPROP)
3343        {        {
3344        if (rightop == OP_EOD) return TRUE;        if (rightop == OP_EOD)
3345        if (rightop == OP_PROP || rightop == OP_NOTPROP)          accepted = TRUE;
3346          else if (rightop == OP_PROP || rightop == OP_NOTPROP)
3347          {          {
3348          int n;          int n;
3349          const pcre_uint8 *p;          const pcre_uint8 *p;
# Line 2997  for(;;) Line 3364  for(;;)
3364          n = propposstab[base_list[2]][list[2]];          n = propposstab[base_list[2]][list[2]];
3365          switch(n)          switch(n)
3366            {            {
3367            case 0: return FALSE;            case 0: break;
3368            case 1: return bothprop;            case 1: accepted = bothprop; break;
3369            case 2: return (base_list[3] == list[3]) != same;            case 2: accepted = (base_list[3] == list[3]) != same; break;
3370            case 3: return !same;            case 3: accepted = !same; break;
3371    
3372            case 4:  /* Left general category, right particular category */            case 4:  /* Left general category, right particular category */
3373            return risprop && catposstab[base_list[3]][list[3]] == same;            accepted = risprop && catposstab[base_list[3]][list[3]] == same;
3374              break;
3375    
3376            case 5:  /* Right general category, left particular category */            case 5:  /* Right general category, left particular category */
3377            return lisprop && catposstab[list[3]][base_list[3]] == same;            accepted = lisprop && catposstab[list[3]][base_list[3]] == same;
3378              break;
3379    
3380            /* This code is logically tricky. Think hard before fiddling with it.            /* This code is logically tricky. Think hard before fiddling with it.
3381            The posspropstab table has four entries per row. Each row relates to            The posspropstab table has four entries per row. Each row relates to
3382            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.
3383            Only WORD actually needs all four entries, but using repeats for the            Only WORD actually needs all four entries, but using repeats for the
3384            others means they can all use the same code below.            others means they can all use the same code below.
3385    
3386            The first two entries in each row are Unicode general categories, and            The first two entries in each row are Unicode general categories, and
3387            apply always, because all the characters they include are part of the            apply always, because all the characters they include are part of the
3388            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 3392  for(;;)
3392            category contains more characters than the specials that are defined            category contains more characters than the specials that are defined
3393            for the property being tested against. Therefore, it cannot be used            for the property being tested against. Therefore, it cannot be used
3394            in a NOTPROP case.            in a NOTPROP case.
3395    
3396            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.
3397            Underscore is covered by ucp_P or ucp_Po. */            Underscore is covered by ucp_P or ucp_Po. */
3398    
# Line 3031  for(;;) Line 3400  for(;;)
3400            case 7:  /* Left space vs right general category */            case 7:  /* Left space vs right general category */
3401            case 8:  /* Left word vs right general category */            case 8:  /* Left word vs right general category */
3402            p = posspropstab[n-6];            p = posspropstab[n-6];
3403            return risprop && lisprop ==            accepted = risprop && lisprop ==
3404              (list[3] != p[0] &&              (list[3] != p[0] &&
3405               list[3] != p[1] &&               list[3] != p[1] &&
3406              (list[3] != p[2] || !lisprop));              (list[3] != p[2] || !lisprop));
3407              break;
3408    
3409            case 9:   /* Right alphanum vs left general category */            case 9:   /* Right alphanum vs left general category */
3410            case 10:  /* Right space vs left general category */            case 10:  /* Right space vs left general category */
3411            case 11:  /* Right word vs left general category */            case 11:  /* Right word vs left general category */
3412            p = posspropstab[n-9];            p = posspropstab[n-9];
3413            return lisprop && risprop ==            accepted = lisprop && risprop ==
3414              (base_list[3] != p[0] &&              (base_list[3] != p[0] &&
3415               base_list[3] != p[1] &&               base_list[3] != p[1] &&
3416              (base_list[3] != p[2] || !risprop));              (base_list[3] != p[2] || !risprop));
3417              break;
3418    
3419            case 12:  /* Left alphanum vs right particular category */            case 12:  /* Left alphanum vs right particular category */
3420            case 13:  /* Left space vs right particular category */            case 13:  /* Left space vs right particular category */
3421            case 14:  /* Left word vs right particular category */            case 14:  /* Left word vs right particular category */
3422            p = posspropstab[n-12];            p = posspropstab[n-12];
3423            return risprop && lisprop ==            accepted = risprop && lisprop ==
3424              (catposstab[p[0]][list[3]] &&              (catposstab[p[0]][list[3]] &&
3425               catposstab[p[1]][list[3]] &&               catposstab[p[1]][list[3]] &&
3426              (list[3] != p[3] || !lisprop));              (list[3] != p[3] || !lisprop));
3427              break;
3428    
3429            case 15:  /* Right alphanum vs left particular category */            case 15:  /* Right alphanum vs left particular category */
3430            case 16:  /* Right space vs left particular category */            case 16:  /* Right space vs left particular category */
3431            case 17:  /* Right word vs left particular category */            case 17:  /* Right word vs left particular category */
3432            p = posspropstab[n-15];            p = posspropstab[n-15];
3433            return lisprop && risprop ==            accepted = lisprop && risprop ==
3434              (catposstab[p[0]][base_list[3]] &&              (catposstab[p[0]][base_list[3]] &&
3435               catposstab[p[1]][base_list[3]] &&               catposstab[p[1]][base_list[3]] &&
3436              (base_list[3] != p[3] || !risprop));              (base_list[3] != p[3] || !risprop));
3437              break;
3438            }            }
3439          }          }
       return FALSE;  
3440        }        }
3441    
3442      else      else
3443  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
3444    
3445      return leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&      accepted = leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&
3446             rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&             rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&
3447             autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];             autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];
3448    
3449        if (!accepted) return FALSE;
3450    
3451        if (list[1] == 0) return TRUE;
3452        /* Might be an empty repeat. */
3453        continue;
3454      }      }
3455    
3456    /* 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 3564  for(;;)
3564        case OP_EOD:    /* Can always possessify before \z */        case OP_EOD:    /* Can always possessify before \z */
3565        break;        break;
3566    
3567    #ifdef SUPPORT_UCP
3568        case OP_PROP:        case OP_PROP:
3569        case OP_NOTPROP:        case OP_NOTPROP:
3570        if (!check_char_prop(chr, list_ptr[2], list_ptr[3],        if (!check_char_prop(chr, list_ptr[2], list_ptr[3],
3571              list_ptr[0] == OP_NOTPROP))              list_ptr[0] == OP_NOTPROP))
3572          return FALSE;          return FALSE;
3573        break;        break;
3574    #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. */  
3575    
3576        case OP_NCLASS:        case OP_NCLASS:
3577        if (chr > 255) return FALSE;        if (chr > 255) return FALSE;
3578        /* Fall through */        /* Fall through */
3579    
3580        case OP_CLASS:        case OP_CLASS:
       if (list_ptr != list) return FALSE;   /* Class is first opcode */  
3581        if (chr > 255) break;        if (chr > 255) break;
3582        if ((((pcre_uint8 *)(code - list_ptr[2] + 1))[chr >> 3] & (1 << (chr & 7))) != 0)        class_bitset = (pcre_uint8 *)
3583          return FALSE;          ((list_ptr == list ? code : base_end) - list_ptr[2]);
3584          if ((class_bitset[chr >> 3] & (1 << (chr & 7))) != 0) return FALSE;
3585        break;        break;
3586    
3587  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3588        case OP_XCLASS:        case OP_XCLASS:
3589        if (list_ptr != list) return FALSE;   /* Class is first opcode */        if (PRIV(xclass)(chr, (list_ptr == list ? code : base_end) -
3590        if (PRIV(xclass)(chr, code - list_ptr[2] + 1 + LINK_SIZE, utf))            list_ptr[2] + LINK_SIZE, utf)) return FALSE;
         return FALSE;  
3591        break;        break;
3592  #endif  #endif
3593    
# Line 3232  for(;;) Line 3604  for(;;)
3604    if (list[1] == 0) return TRUE;    if (list[1] == 0) return TRUE;
3605    }    }
3606    
3607  return FALSE;  /* Control never reaches here. There used to be a fail-save return FALSE; here,
3608    but some compilers complain about an unreachable statement. */
3609    
3610  }  }
3611    
3612    
# Line 3257  auto_possessify(pcre_uchar *code, BOOL u Line 3631  auto_possessify(pcre_uchar *code, BOOL u
3631  {  {
3632  register pcre_uchar c;  register pcre_uchar c;
3633  const pcre_uchar *end;  const pcre_uchar *end;
3634    pcre_uchar *repeat_opcode;
3635  pcre_uint32 list[8];  pcre_uint32 list[8];
3636    
3637  for (;;)  for (;;)
3638    {    {
3639    c = *code;    c = *code;
3640    
3641      /* When a pattern with bad UTF-8 encoding is compiled with NO_UTF_CHECK,
3642      it may compile without complaining, but may get into a loop here if the code
3643      pointer points to a bad value. This is, of course a documentated possibility,
3644      when NO_UTF_CHECK is set, so it isn't a bug, but we can detect this case and
3645      just give up on this optimization. */
3646    
3647      if (c >= OP_TABLE_LENGTH) return;
3648    
3649    if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)    if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
3650      {      {
3651      c -= get_repeat_base(c) - OP_STAR;      c -= get_repeat_base(c) - OP_STAR;
# Line 3270  for (;;) Line 3653  for (;;)
3653        get_chr_property_list(code, utf, cd->fcc, list) : NULL;        get_chr_property_list(code, utf, cd->fcc, list) : NULL;
3654      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;
3655    
3656      if (end != NULL && compare_opcodes(end, utf, cd, list))      if (end != NULL && compare_opcodes(end, utf, cd, list, end))
3657        {        {
3658        switch(c)        switch(c)
3659          {          {
# Line 3303  for (;;) Line 3686  for (;;)
3686          break;          break;
3687    
3688          case OP_MINUPTO:          case OP_MINUPTO:
3689          *code += OP_MINUPTO - OP_UPTO;          *code += OP_POSUPTO - OP_MINUPTO;
3690          break;          break;
3691          }          }
3692        }        }
3693      c = *code;      c = *code;
3694      }      }
3695      else if (c == OP_CLASS || c == OP_NCLASS || c == OP_XCLASS)
3696        {
3697    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3698        if (c == OP_XCLASS)
3699          repeat_opcode = code + GET(code, 1);
3700        else
3701    #endif
3702          repeat_opcode = code + 1 + (32 / sizeof(pcre_uchar));
3703    
3704        c = *repeat_opcode;
3705        if (c >= OP_CRSTAR && c <= OP_CRMINRANGE)
3706          {
3707          /* end must not be NULL. */
3708          end = get_chr_property_list(code, utf, cd->fcc, list);
3709    
3710          list[1] = (c & 1) == 0;
3711    
3712          if (compare_opcodes(end, utf, cd, list, end))
3713            {
3714            switch (c)
3715              {
3716              case OP_CRSTAR:
3717              case OP_CRMINSTAR:
3718              *repeat_opcode = OP_CRPOSSTAR;
3719              break;
3720    
3721              case OP_CRPLUS:
3722              case OP_CRMINPLUS:
3723              *repeat_opcode = OP_CRPOSPLUS;
3724              break;
3725    
3726              case OP_CRQUERY:
3727              case OP_CRMINQUERY:
3728              *repeat_opcode = OP_CRPOSQUERY;
3729              break;
3730    
3731              case OP_CRRANGE:
3732              case OP_CRMINRANGE:
3733              *repeat_opcode = OP_CRPOSRANGE;
3734              break;
3735              }
3736            }
3737          }
3738        c = *code;
3739        }
3740    
3741    switch(c)    switch(c)
3742      {      {
# Line 3335  for (;;) Line 3763  for (;;)
3763        code += 2;        code += 2;
3764      break;      break;
3765    
3766    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3767      case OP_XCLASS:      case OP_XCLASS:
3768      code += GET(code, 1);      code += GET(code, 1);
3769      break;      break;
3770    #endif
3771    
3772      case OP_MARK:      case OP_MARK:
3773      case OP_PRUNE_ARG:      case OP_PRUNE_ARG:
# Line 3446  class, but [abc[:x\]pqr:]] is (so that a Line 3876  class, but [abc[:x\]pqr:]] is (so that a
3876  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
3877  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:]
3878  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
3879  "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,
3880  I think.  I think.
3881    
3882  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 3553  Arguments: Line 3983  Arguments:
3983    adjust     the amount by which the group is to be moved    adjust     the amount by which the group is to be moved
3984    utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode    utf        TRUE in UTF-8 / UTF-16 / UTF-32 mode
3985    cd         contains pointers to tables etc.    cd         contains pointers to tables etc.
3986    save_hwm   the hwm forward reference pointer at the start of the group    save_hwm_offset   the hwm forward reference offset at the start of the group
3987    
3988  Returns:     nothing  Returns:     nothing
3989  */  */
3990    
3991  static void  static void
3992  adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,  adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
3993    pcre_uchar *save_hwm)    size_t save_hwm_offset)
3994  {  {
3995  pcre_uchar *ptr = group;  pcre_uchar *ptr = group;
3996    
# Line 3572  while ((ptr = (pcre_uchar *)find_recurse Line 4002  while ((ptr = (pcre_uchar *)find_recurse
4002    /* See if this recursion is on the forward reference list. If so, adjust the    /* See if this recursion is on the forward reference list. If so, adjust the
4003    reference. */    reference. */
4004    
4005    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)    for (hc = (pcre_uchar *)cd->start_workspace + save_hwm_offset; hc < cd->hwm;
4006           hc += LINK_SIZE)
4007      {      {
4008      offset = (int)GET(hc, 0);      offset = (int)GET(hc, 0);
4009      if (cd->start_code + offset == ptr + 1)      if (cd->start_code + offset == ptr + 1)
# Line 3695  for (c = *cptr; c <= d; c++) Line 4126  for (c = *cptr; c <= d; c++)
4126    
4127  if (c > d) return -1;  /* Reached end of range */  if (c > d) return -1;  /* Reached end of range */
4128    
4129    /* Found a character that has a single other case. Search for the end of the
4130    range, which is either the end of the input range, or a character that has zero
4131    or more than one other cases. */
4132    
4133  *ocptr = othercase;  *ocptr = othercase;
4134  next = othercase + 1;  next = othercase + 1;
4135    
4136  for (++c; c <= d; c++)  for (++c; c <= d; c++)
4137    {    {
4138    if (UCD_OTHERCASE(c) != next) break;    if ((co = UCD_CASESET(c)) != 0 || UCD_OTHERCASE(c) != next) break;
4139    next++;    next++;
4140    }    }
4141    
# Line 3738  add_to_class(pcre_uint8 *classbits, pcre Line 4173  add_to_class(pcre_uint8 *classbits, pcre
4173    compile_data *cd, pcre_uint32 start, pcre_uint32 end)    compile_data *cd, pcre_uint32 start, pcre_uint32 end)
4174  {  {
4175  pcre_uint32 c;  pcre_uint32 c;
4176    pcre_uint32 classbits_end = (end <= 0xff ? end : 0xff);
4177  int n8 = 0;  int n8 = 0;
4178    
4179  /* If caseless matching is required, scan the range and process alternate  /* If caseless matching is required, scan the range and process alternate
# Line 3772  if ((options & PCRE_CASELESS) != 0) Line 4208  if ((options & PCRE_CASELESS) != 0)
4208        range. Otherwise, use a recursive call to add the additional range. */        range. Otherwise, use a recursive call to add the additional range. */
4209    
4210        else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */        else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */
4211        else if (od > end && oc <= end + 1) end = od;       /* Extend upwards */        else if (od > end && oc <= end + 1)
4212            {
4213            end = od;       /* Extend upwards */
4214            if (end > classbits_end) classbits_end = (end <= 0xff ? end : 0xff);
4215            }
4216        else n8 += add_to_class(classbits, uchardptr, options, cd, oc, od);        else n8 += add_to_class(classbits, uchardptr, options, cd, oc, od);
4217        }        }
4218      }      }
# Line 3781  if ((options & PCRE_CASELESS) != 0) Line 4221  if ((options & PCRE_CASELESS) != 0)
4221    
4222    /* Not UTF-mode, or no UCP */    /* Not UTF-mode, or no UCP */
4223    
4224    for (c = start; c <= end && c < 256; c++)    for (c = start; c <= classbits_end; c++)
4225      {      {
4226      SETBIT(classbits, cd->fcc[c]);      SETBIT(classbits, cd->fcc[c]);
4227      n8++;      n8++;
# Line 3806  in all cases. */ Line 4246  in all cases. */
4246    
4247  #endif /* COMPILE_PCRE[8|16] */  #endif /* COMPILE_PCRE[8|16] */
4248    
4249  /* 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. */  
4250    
4251  if (end < 0x100)  for (c = start; c <= classbits_end; c++)
4252    {    {
4253    for (c = start; c <= end; c++)    /* Regardless of start, c will always be <= 255. */
4254      {    SETBIT(classbits, c);
4255      n8++;    n8++;
     SETBIT(classbits, c);  
     }  
4256    }    }
4257    
4258  else  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4259    if (start <= 0xff) start = 0xff + 1;
4260    
4261    if (end >= start)
4262    {    {
4263    pcre_uchar *uchardata = *uchardptr;    pcre_uchar *uchardata = *uchardptr;
   
4264  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4265    if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */    if ((options & PCRE_UTF8) != 0)  /* All UTFs use the same flag bit */
4266      {      {
# Line 3861  else Line 4300  else
4300    
4301    *uchardptr = uchardata;   /* Updata extra data pointer */    *uchardptr = uchardata;   /* Updata extra data pointer */
4302    }    }
4303    #endif /* SUPPORT_UTF || !COMPILE_PCRE8 */
4304    
4305  return n8;    /* Number of 8-bit characters */  return n8;    /* Number of 8-bit characters */
4306  }  }
# Line 4012  const pcre_uchar *tempptr; Line 4452  const pcre_uchar *tempptr;
4452  const pcre_uchar *nestptr = NULL;  const pcre_uchar *nestptr = NULL;
4453  pcre_uchar *previous = NULL;  pcre_uchar *previous = NULL;
4454  pcre_uchar *previous_callout = NULL;  pcre_uchar *previous_callout = NULL;
4455  pcre_uchar *save_hwm = NULL;  size_t save_hwm_offset = 0;
4456  pcre_uint8 classbits[32];  pcre_uint8 classbits[32];
4457    
4458  /* We can fish out the UTF-8 setting once and for all into a BOOL, but we  /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
# Line 4082  for (;; ptr++) Line 4522  for (;; ptr++)
4522    BOOL reset_bracount;    BOOL reset_bracount;
4523    int class_has_8bitchar;    int class_has_8bitchar;
4524    int class_one_char;    int class_one_char;
4525    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4526      BOOL xclass_has_prop;
4527    #endif
4528    int newoptions;    int newoptions;
4529    int recno;    int recno;
4530    int refsign;    int refsign;
# Line 4199  for (;; ptr++) Line 4642  for (;; ptr++)
4642          }          }
4643        goto NORMAL_CHAR;        goto NORMAL_CHAR;
4644        }        }
4645        /* Control does not reach here. */
4646      }      }
4647    
4648    /* 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
4649    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. */  
4650    
4651    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
4652      {      {
4653      if (MAX_255(*ptr) && (cd->ctypes[c] & ctype_space) != 0) continue;      for (;;)
     if (c == CHAR_NUMBER_SIGN)  
4654        {        {
4655          while (MAX_255(c) && (cd->ctypes[c] & ctype_space) != 0) c = *(++ptr);
4656          if (c != CHAR_NUMBER_SIGN) break;
4657        ptr++;        ptr++;
4658        while (*ptr != CHAR_NULL)        while (*ptr != CHAR_NULL)
4659          {          {
4660          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr))         /* For non-fixed-length newline cases, */
4661              {                          /* IS_NEWLINE sets cd->nllen. */
4662              ptr += cd->nllen;
4663              break;
4664              }
4665          ptr++;          ptr++;
4666  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
4667          if (utf) FORWARDCHAR(ptr);          if (utf) FORWARDCHAR(ptr);
4668  #endif  #endif
4669          }          }
4670        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;  
4671        }        }
4672      }      }
4673    
4674    /* No auto callout for quantifiers. */    /* See if the next thing is a quantifier. */
4675    
4676      is_quantifier =
4677        c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
4678        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
4679    
4680      /* Fill in length of a previous callout, except when the next thing is a
4681      quantifier or when processing a property substitution string in UCP mode. */
4682    
4683      if (!is_quantifier && previous_callout != NULL && nestptr == NULL &&
4684           after_manual_callout-- <= 0)
4685        {
4686        if (lengthptr == NULL)      /* Don't attempt in pre-compile phase */
4687          complete_callout(previous_callout, ptr, cd);
4688        previous_callout = NULL;
4689        }
4690    
4691      /* Create auto callout, except for quantifiers, or while processing property
4692      strings that are substituted for \w etc in UCP mode. */
4693    
4694    if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier)    if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier && nestptr == NULL)
4695      {      {
4696      previous_callout = code;      previous_callout = code;
4697      code = auto_callout(code, ptr, cd);      code = auto_callout(code, ptr, cd);
4698      }      }
4699    
4700      /* Process the next pattern item. */
4701    
4702    switch(c)    switch(c)
4703      {      {
4704      /* ===================================================================*/      /* ===================================================================*/
4705      case 0:                        /* The branch terminates at string end */      case CHAR_NULL:                /* The branch terminates at string end */
4706      case CHAR_VERTICAL_LINE:       /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
4707      case CHAR_RIGHT_PARENTHESIS:      case CHAR_RIGHT_PARENTHESIS:
4708      *firstcharptr = firstchar;      *firstcharptr = firstchar;
# Line 4280  for (;; ptr++) Line 4732  for (;; ptr++)
4732      previous = NULL;      previous = NULL;
4733      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
4734        {        {
4735        if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;        if (firstcharflags == REQ_UNSET)
4736            zerofirstcharflags = firstcharflags = REQ_NONE;
4737        *code++ = OP_CIRCM;        *code++ = OP_CIRCM;
4738        }        }
4739      else *code++ = OP_CIRC;      else *code++ = OP_CIRC;
# Line 4328  for (;; ptr++) Line 4781  for (;; ptr++)
4781        }        }
4782      goto NORMAL_CHAR;      goto NORMAL_CHAR;
4783    
4784        /* In another (POSIX) regex library, the ugly syntax [[:<:]] and [[:>:]] is
4785        used for "start of word" and "end of word". As these are otherwise illegal
4786        sequences, we don't break anything by recognizing them. They are replaced
4787        by \b(?=\w) and \b(?<=\w) respectively. Sequences like [a[:<:]] are
4788        erroneous and are handled by the normal code below. */
4789    
4790      case CHAR_LEFT_SQUARE_BRACKET:      case CHAR_LEFT_SQUARE_BRACKET:
4791        if (STRNCMP_UC_C8(ptr+1, STRING_WEIRD_STARTWORD, 6) == 0)
4792          {
4793          nestptr = ptr + 7;
4794          ptr = sub_start_of_word - 1;
4795          continue;
4796          }
4797    
4798        if (STRNCMP_UC_C8(ptr+1, STRING_WEIRD_ENDWORD, 6) == 0)
4799          {
4800          nestptr = ptr + 7;
4801          ptr = sub_end_of_word - 1;
4802          continue;
4803          }
4804    
4805        /* Handle a real character class. */
4806    
4807      previous = code;      previous = code;
4808    
4809      /* 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 4860  for (;; ptr++)
4860    
4861      should_flip_negation = FALSE;      should_flip_negation = FALSE;
4862    
4863        /* Extended class (xclass) will be used when characters > 255
4864        might match. */
4865    
4866    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4867        xclass = FALSE;
4868        class_uchardata = code + LINK_SIZE + 2;   /* For XCLASS items */
4869        class_uchardata_base = class_uchardata;   /* Save the start */
4870    #endif
4871    
4872      /* For optimization purposes, we track some properties of the class:      /* For optimization purposes, we track some properties of the class:
4873      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 <
4874      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
4875      character. */      character; xclass_has_prop will be TRUE if unicode property checks
4876        are present in the class. */
4877    
4878      class_has_8bitchar = 0;      class_has_8bitchar = 0;
4879      class_one_char = 0;      class_one_char = 0;
4880    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4881        xclass_has_prop = FALSE;
4882    #endif
4883    
4884      /* 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
4885      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 4888  for (;; ptr++)
4888    
4889      memset(classbits, 0, 32 * sizeof(pcre_uint8));      memset(classbits, 0, 32 * sizeof(pcre_uint8));
4890    
 #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  
   
4891      /* Process characters until ] is reached. By writing this as a "do" it      /* Process characters until ] is reached. By writing this as a "do" it
4892      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
4893      loop, c contains the first byte of the character. */      loop, c contains the first byte of the character. */
# Line 4431  for (;; ptr++) Line 4913  for (;; ptr++)
4913        if (lengthptr != NULL && class_uchardata > class_uchardata_base)        if (lengthptr != NULL && class_uchardata > class_uchardata_base)
4914          {          {
4915          xclass = TRUE;          xclass = TRUE;
4916          *lengthptr += class_uchardata - class_uchardata_base;          *lengthptr += (int)(class_uchardata - class_uchardata_base);
4917          class_uchardata = class_uchardata_base;          class_uchardata = class_uchardata_base;
4918          }          }
4919  #endif  #endif
# Line 4493  for (;; ptr++) Line 4975  for (;; ptr++)
4975            posix_class = 0;            posix_class = 0;
4976    
4977          /* 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
4978          different escape sequences that use Unicode properties. */          different escape sequences that use Unicode properties \p or \P. Others
4979            that are not available via \p or \P generate XCL_PROP/XCL_NOTPROP
4980            directly. */
4981    
4982  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4983          if ((options & PCRE_UCP) != 0)          if ((options & PCRE_UCP) != 0)
4984            {            {
4985              unsigned int ptype = 0;
4986            int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);            int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
4987    
4988              /* The posix_substitutes table specifies which POSIX classes can be
4989              converted to \p or \P items. */
4990    
4991            if (posix_substitutes[pc] != NULL)            if (posix_substitutes[pc] != NULL)
4992              {              {
4993              nestptr = tempptr + 1;              nestptr = tempptr + 1;
4994              ptr = posix_substitutes[pc] - 1;              ptr = posix_substitutes[pc] - 1;
4995              continue;              continue;
4996              }              }
4997    
4998              /* There are three other classes that generate special property calls
4999              that are recognized only in an XCLASS. */
5000    
5001              else switch(posix_class)
5002                {
5003                case PC_GRAPH:
5004                ptype = PT_PXGRAPH;
5005                /* Fall through */
5006                case PC_PRINT:
5007                if (ptype == 0) ptype = PT_PXPRINT;
5008                /* Fall through */
5009                case PC_PUNCT:
5010                if (ptype == 0) ptype = PT_PXPUNCT;
5011                *class_uchardata++ = local_negate? XCL_NOTPROP : XCL_PROP;
5012                *class_uchardata++ = ptype;
5013                *class_uchardata++ = 0;
5014                xclass_has_prop = TRUE;
5015                ptr = tempptr + 1;
5016                continue;
5017    
5018                /* For all other POSIX classes, no special action is taken in UCP
5019                mode. Fall through to the non_UCP case. */
5020    
5021                default:
5022                break;
5023                }
5024            }            }
5025  #endif  #endif
5026          /* 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
5027          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
5028          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
5029          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
5030            bit map that is being built. */
5031    
5032          posix_class *= 3;          posix_class *= 3;
5033    
# Line 4631  for (;; ptr++) Line 5148  for (;; ptr++)
5148              5.18. Before PCRE 8.34, we had to preserve the VT bit if it was              5.18. Before PCRE 8.34, we had to preserve the VT bit if it was
5149              previously set by something earlier in the character class.              previously set by something earlier in the character class.
5150              Luckily, the value of CHAR_VT is 0x0b in both ASCII and EBCDIC, so              Luckily, the value of CHAR_VT is 0x0b in both ASCII and EBCDIC, so
5151              we could just adjust the appropriate bit. From PCRE 8.34 we no              we could just adjust the appropriate bit. From PCRE 8.34 we no
5152              longer treat \s and \S specially. */              longer treat \s and \S specially. */
5153    
5154              case ESC_s:              case ESC_s:
# Line 4677  for (;; ptr++) Line 5194  for (;; ptr++)
5194                  XCL_PROP : XCL_NOTPROP;                  XCL_PROP : XCL_NOTPROP;
5195                *class_uchardata++ = ptype;                *class_uchardata++ = ptype;
5196                *class_uchardata++ = pdata;                *class_uchardata++ = pdata;
5197                  xclass_has_prop = TRUE;
5198                class_has_8bitchar--;                /* Undo! */                class_has_8bitchar--;                /* Undo! */
5199                continue;                continue;
5200                }                }
# Line 4762  for (;; ptr++) Line 5280  for (;; ptr++)
5280  #endif  #endif
5281          d = *ptr;  /* Not UTF-8 mode */          d = *ptr;  /* Not UTF-8 mode */
5282    
5283          /* 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
5284          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
5285          in such circumstances. */          literal in such circumstances. However, in Perl's warning mode, a
5286            warning is given, so PCRE now faults it as it is almost certainly a
5287            mistake on the user's part. */
5288    
5289          if (!inescq && d == CHAR_BACKSLASH)          if (!inescq)
5290            {            {
5291            int descape;            if (d == CHAR_BACKSLASH)
5292            descape = check_escape(&ptr, &d, errorcodeptr, cd->bracount, options, TRUE);              {
5293            if (*errorcodeptr != 0) goto FAILED;              int descape;
5294                descape = check_escape(&ptr, &d, errorcodeptr, cd->bracount, options, TRUE);
5295                if (*errorcodeptr != 0) goto FAILED;
5296    
5297            /* \b is backspace; any other special means the '-' was literal. */              /* 0 means a character was put into d; \b is backspace; any other
5298                special causes an error. */
5299    
5300            if (descape != 0)              if (descape != 0)
             {  
             if (descape == ESC_b) d = CHAR_BS; else  
5301                {                {
5302                ptr = oldptr;                if (descape == ESC_b) d = CHAR_BS; else
5303                goto CLASS_SINGLE_CHARACTER;  /* A few lines below */                  {
5304                    *errorcodeptr = ERR83;
5305                    goto FAILED;
5306                    }
5307                }                }
5308              }              }
5309    
5310              /* A hyphen followed by a POSIX class is treated in the same way. */
5311    
5312              else if (d == CHAR_LEFT_SQUARE_BRACKET &&
5313                       (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
5314                        ptr[1] == CHAR_EQUALS_SIGN) &&
5315                       check_posix_syntax(ptr, &tempptr))
5316                {
5317                *errorcodeptr = ERR83;
5318                goto FAILED;
5319                }
5320            }            }
5321    
5322          /* Check that the two values are in the correct order. Optimize          /* Check that the two values are in the correct order. Optimize
# Line 4828  for (;; ptr++) Line 5363  for (;; ptr++)
5363        whatever repeat count may follow. In the case of reqchar, save the        whatever repeat count may follow. In the case of reqchar, save the
5364        previous value for reinstating. */        previous value for reinstating. */
5365    
5366        if (class_one_char == 1 && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)        if (!inescq && class_one_char == 1 && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
5367          {          {
5368          ptr++;          ptr++;
5369          zeroreqchar = reqchar;          zeroreqchar = reqchar;
# Line 4954  for (;; ptr++) Line 5489  for (;; ptr++)
5489        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
5490        code += LINK_SIZE;        code += LINK_SIZE;
5491        *code = negate_class? XCL_NOT:0;        *code = negate_class? XCL_NOT:0;
5492          if (xclass_has_prop) *code |= XCL_HASPROP;
5493    
5494        /* 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;
5495        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 4963  for (;; ptr++) Line 5499  for (;; ptr++)
5499          *code++ |= XCL_MAP;          *code++ |= XCL_MAP;
5500          memmove(code + (32 / sizeof(pcre_uchar)), code,          memmove(code + (32 / sizeof(pcre_uchar)), code,
5501            IN_UCHARS(class_uchardata - code));            IN_UCHARS(class_uchardata - code));
5502            if (negate_class && !xclass_has_prop)
5503              for (c = 0; c < 32; c++) classbits[c] = ~classbits[c];
5504          memcpy(code, classbits, 32);          memcpy(code, classbits, 32);
5505          code = class_uchardata + (32 / sizeof(pcre_uchar));          code = class_uchardata + (32 / sizeof(pcre_uchar));
5506          }          }
# Line 5045  for (;; ptr++) Line 5583  for (;; ptr++)
5583    
5584      tempcode = previous;      tempcode = previous;
5585    
5586        /* Before checking for a possessive quantifier, we must skip over
5587        whitespace and comments in extended mode because Perl allows white space at
5588        this point. */
5589    
5590        if ((options & PCRE_EXTENDED) != 0)
5591          {
5592          const pcre_uchar *p = ptr + 1;
5593          for (;;)
5594            {
5595            while (MAX_255(*p) && (cd->ctypes[*p] & ctype_space) != 0) p++;
5596            if (*p != CHAR_NUMBER_SIGN) break;
5597            p++;
5598            while (*p != CHAR_NULL)
5599              {
5600              if (IS_NEWLINE(p))         /* For non-fixed-length newline cases, */
5601                {                        /* IS_NEWLINE sets cd->nllen. */
5602                p += cd->nllen;
5603                break;
5604                }
5605              p++;
5606    #ifdef SUPPORT_UTF
5607              if (utf) FORWARDCHAR(p);
5608    #endif
5609              }           /* Loop for comment characters */
5610            }             /* Loop for multiple comments */
5611          ptr = p - 1;    /* Character before the next significant one. */
5612          }
5613    
5614      /* If the next character is '+', we have a possessive quantifier. This      /* If the next character is '+', we have a possessive quantifier. This
5615      implies greediness, whatever the setting of the PCRE_UNGREEDY option.      implies greediness, whatever the setting of the PCRE_UNGREEDY option.
5616      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 5337  for (;; ptr++) Line 5903  for (;; ptr++)
5903      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
5904      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 >=
5905      OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,      OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
5906      ASSERTBACK_NOT, ONCE, BRA, CBRA, and COND. Originally, PCRE did not allow      ASSERTBACK_NOT, ONCE, ONCE_NC, BRA, BRAPOS, CBRA, CBRAPOS, and COND.
5907      repetition of assertions, but now it does, for Perl compatibility. */      Originally, PCRE did not allow repetition of assertions, but now it does,
5908        for Perl compatibility. */
5909    
5910      else if (*previous >= OP_ASSERT && *previous <= OP_COND)      else if (*previous >= OP_ASSERT && *previous <= OP_COND)
5911        {        {
# Line 5356  for (;; ptr++) Line 5923  for (;; ptr++)
5923        /* There is no sense in actually repeating assertions. The only potential        /* There is no sense in actually repeating assertions. The only potential
5924        use of repetition is in cases when the assertion is optional. Therefore,        use of repetition is in cases when the assertion is optional. Therefore,
5925        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
5926        maximum is not not zero or one, set it to 1. */        maximum is not zero or one, set it to 1. */
5927    
5928        if (*previous < OP_ONCE)    /* Assertion */        if (*previous < OP_ONCE)    /* Assertion */
5929          {          {
# Line 5398  for (;; ptr++) Line 5965  for (;; ptr++)
5965          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
5966            {            {
5967            *code = OP_END;            *code = OP_END;
5968            adjust_recurse(previous, 1, utf, cd, save_hwm);            adjust_recurse(previous, 1, utf, cd, save_hwm_offset);
5969            memmove(previous + 1, previous, IN_UCHARS(len));            memmove(previous + 1, previous, IN_UCHARS(len));
5970            code++;            code++;
5971            if (repeat_max == 0)            if (repeat_max == 0)
# Line 5422  for (;; ptr++) Line 5989  for (;; ptr++)
5989            {            {
5990            int offset;            int offset;
5991            *code = OP_END;            *code = OP_END;
5992            adjust_recurse(previous, 2 + LINK_SIZE, utf, cd, save_hwm);            adjust_recurse(previous, 2 + LINK_SIZE, utf, cd, save_hwm_offset);
5993            memmove(previous + 2 + LINK_SIZE, previous, IN_UCHARS(len));            memmove(previous + 2 + LINK_SIZE, previous, IN_UCHARS(len));
5994            code += 2 + LINK_SIZE;            code += 2 + LINK_SIZE;
5995            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
# Line 5485  for (;; ptr++) Line 6052  for (;; ptr++)
6052              for (i = 1; i < repeat_min; i++)              for (i = 1; i < repeat_min; i++)
6053                {                {
6054                pcre_uchar *hc;                pcre_uchar *hc;
6055                pcre_uchar *this_hwm = cd->hwm;                size_t this_hwm_offset = cd->hwm - cd->start_workspace;
6056                memcpy(code, previous, IN_UCHARS(len));                memcpy(code, previous, IN_UCHARS(len));
6057    
6058                while (cd->hwm > cd->start_workspace + cd->workspace_size -                while (cd->hwm > cd->start_workspace + cd->workspace_size -
6059                       WORK_SIZE_SAFETY_MARGIN - (this_hwm - save_hwm))                       WORK_SIZE_SAFETY_MARGIN -
6060                         (this_hwm_offset - save_hwm_offset))
6061                  {                  {
                 int save_offset = save_hwm - cd->start_workspace;  
                 int this_offset = this_hwm - cd->start_workspace;  
6062                  *errorcodeptr = expand_workspace(cd);                  *errorcodeptr = expand_workspace(cd);
6063                  if (*errorcodeptr != 0) goto FAILED;                  if (*errorcodeptr != 0) goto FAILED;
                 save_hwm = (pcre_uchar *)cd->start_workspace + save_offset;  
                 this_hwm = (pcre_uchar *)cd->start_workspace + this_offset;  
6064                  }                  }
6065    
6066                for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)                for (hc = (pcre_uchar *)cd->start_workspace + save_hwm_offset;
6067                       hc < (pcre_uchar *)cd->start_workspace + this_hwm_offset;
6068                       hc += LINK_SIZE)
6069                  {                  {
6070                  PUT(cd->hwm, 0, GET(hc, 0) + len);                  PUT(cd->hwm, 0, GET(hc, 0) + len);
6071                  cd->hwm += LINK_SIZE;                  cd->hwm += LINK_SIZE;
6072                  }                  }
6073                save_hwm = this_hwm;                save_hwm_offset = this_hwm_offset;
6074                code += len;                code += len;
6075                }                }
6076              }              }
# Line 5549  for (;; ptr++) Line 6115  for (;; ptr++)
6115          else for (i = repeat_max - 1; i >= 0; i--)          else for (i = repeat_max - 1; i >= 0; i--)
6116            {            {
6117            pcre_uchar *hc;            pcre_uchar *hc;
6118            pcre_uchar *this_hwm = cd->hwm;            size_t this_hwm_offset = cd->hwm - cd->start_workspace;
6119    
6120            *code++ = OP_BRAZERO + repeat_type;            *code++ = OP_BRAZERO + repeat_type;
6121    
# Line 5571  for (;; ptr++) Line 6137  for (;; ptr++)
6137            copying them. */            copying them. */
6138    
6139            while (cd->hwm > cd->start_workspace + cd->workspace_size -            while (cd->hwm > cd->start_workspace + cd->workspace_size -
6140                   WORK_SIZE_SAFETY_MARGIN - (this_hwm - save_hwm))                   WORK_SIZE_SAFETY_MARGIN -
6141                     (this_hwm_offset - save_hwm_offset))
6142              {              {
             int save_offset = save_hwm - cd->start_workspace;  
             int this_offset = this_hwm - cd->start_workspace;  
6143              *errorcodeptr = expand_workspace(cd);              *errorcodeptr = expand_workspace(cd);
6144              if (*errorcodeptr != 0) goto FAILED;              if (*errorcodeptr != 0) goto FAILED;
             save_hwm = (pcre_uchar *)cd->start_workspace + save_offset;  
             this_hwm = (pcre_uchar *)cd->start_workspace + this_offset;  
6145              }              }
6146    
6147            for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)            for (hc = (pcre_uchar *)cd->start_workspace + save_hwm_offset;
6148                   hc < (pcre_uchar *)cd->start_workspace + this_hwm_offset;
6149                   hc += LINK_SIZE)
6150              {              {
6151              PUT(cd->hwm, 0, GET(hc, 0) + len + ((i != 0)? 2+LINK_SIZE : 1));              PUT(cd->hwm, 0, GET(hc, 0) + len + ((i != 0)? 2+LINK_SIZE : 1));
6152              cd->hwm += LINK_SIZE;              cd->hwm += LINK_SIZE;
6153              }              }
6154            save_hwm = this_hwm;            save_hwm_offset = this_hwm_offset;
6155            code += len;            code += len;
6156            }            }
6157    
# Line 5682  for (;; ptr++) Line 6247  for (;; ptr++)
6247                {                {
6248                int nlen = (int)(code - bracode);                int nlen = (int)(code - bracode);
6249                *code = OP_END;                *code = OP_END;
6250                adjust_recurse(bracode, 1 + LINK_SIZE, utf, cd, save_hwm);                adjust_recurse(bracode, 1 + LINK_SIZE, utf, cd, save_hwm_offset);
6251                memmove(bracode + 1 + LINK_SIZE, bracode, IN_UCHARS(nlen));                memmove(bracode + 1 + LINK_SIZE, bracode, IN_UCHARS(nlen));
6252                code += 1 + LINK_SIZE;                code += 1 + LINK_SIZE;
6253                nlen += 1 + LINK_SIZE;                nlen += 1 + LINK_SIZE;
# Line 5729  for (;; ptr++) Line 6294  for (;; ptr++)
6294        goto FAILED;        goto FAILED;
6295        }        }
6296    
6297      /* If the character following a repeat is '+', or if certain optimization      /* If the character following a repeat is '+', possessive_quantifier is
6298      tests above succeeded, possessive_quantifier is TRUE. For some opcodes,      TRUE. For some opcodes, there are special alternative opcodes for this
6299      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
6300      wrap the entire repeated item inside OP_ONCE brackets. Logically, the '+'      brackets. Logically, the '+' notation is just syntactic sugar, taken from
6301      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.  
6302    
6303      Some (but not all) possessively repeated subpatterns have already been      Some (but not all) possessively repeated subpatterns have already been
6304      completely handled in the code just above. For them, possessive_quantifier      completely handled in the code just above. For them, possessive_quantifier
6305      is always FALSE at this stage.      is always FALSE at this stage. Note that the repeated item starts at
6306        tempcode, not at previous, which might be the first part of a string whose
6307      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. */  
6308    
6309      if (possessive_quantifier)      if (possessive_quantifier)
6310        {        {
6311        int len;        int len;
6312    
6313        if (*tempcode == OP_TYPEEXACT)        /* Possessifying an EXACT quantifier has no effect, so we can ignore it.
6314          However, QUERY, STAR, or UPTO may follow (for quantifiers such as {5,6},
6315          {5,}, or {5,10}). We skip over an EXACT item; if the length of what
6316          remains is greater than zero, there's a further opcode that can be
6317          handled. If not, do nothing, leaving the EXACT alone. */
6318    
6319          switch(*tempcode)
6320            {
6321            case OP_TYPEEXACT:
6322          tempcode += PRIV(OP_lengths)[*tempcode] +          tempcode += PRIV(OP_lengths)[*tempcode] +
6323            ((tempcode[1 + IMM2_SIZE] == OP_PROP            ((tempcode[1 + IMM2_SIZE] == OP_PROP
6324            || tempcode[1 + IMM2_SIZE] == OP_NOTPROP)? 2 : 0);            || tempcode[1 + IMM2_SIZE] == OP_NOTPROP)? 2 : 0);
6325            break;
6326    
6327        else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)          /* CHAR opcodes are used for exacts whose count is 1. */
6328          {  
6329            case OP_CHAR:
6330            case OP_CHARI:
6331            case OP_NOT:
6332            case OP_NOTI:
6333            case OP_EXACT:
6334            case OP_EXACTI:
6335            case OP_NOTEXACT:
6336            case OP_NOTEXACTI:
6337          tempcode += PRIV(OP_lengths)[*tempcode];          tempcode += PRIV(OP_lengths)[*tempcode];
6338  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
6339          if (utf && HAS_EXTRALEN(tempcode[-1]))          if (utf && HAS_EXTRALEN(tempcode[-1]))
6340            tempcode += GET_EXTRALEN(tempcode[-1]);            tempcode += GET_EXTRALEN(tempcode[-1]);
6341  #endif  #endif
6342            break;
6343    
6344            /* For the class opcodes, the repeat operator appears at the end;
6345            adjust tempcode to point to it. */
6346    
6347            case OP_CLASS:
6348            case OP_NCLASS:
6349            tempcode += 1 + 32/sizeof(pcre_uchar);
6350            break;
6351    
6352    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
6353            case OP_XCLASS:
6354            tempcode += GET(tempcode, 1);
6355            break;
6356    #endif
6357          }          }
6358    
6359          /* If tempcode is equal to code (which points to the end of the repeated
6360          item), it means we have skipped an EXACT item but there is no following
6361          QUERY, STAR, or UPTO; the value of len will be 0, and we do nothing. In
6362          all other cases, tempcode will be pointing to the repeat opcode, and will
6363          be less than code, so the value of len will be greater than 0. */
6364    
6365        len = (int)(code - tempcode);        len = (int)(code - tempcode);
6366          if (len > 0)
6367            {
6368            unsigned int repcode = *tempcode;
6369    
6370            /* There is a table for possessifying opcodes, all of which are less
6371            than OP_CALLOUT. A zero entry means there is no possessified version.
6372            */
6373    
6374            if (repcode < OP_CALLOUT && opcode_possessify[repcode] > 0)
6375              *tempcode = opcode_possessify[repcode];
6376    
6377            /* For opcode without a special possessified version, wrap the item in
6378            ONCE brackets. Because we are moving code along, we must ensure that any
6379            pending recursive references are updated. */
6380    
6381            else
6382              {
6383              *code = OP_END;
6384              adjust_recurse(tempcode, 1 + LINK_SIZE, utf, cd, save_hwm_offset);
6385              memmove(tempcode + 1 + LINK_SIZE, tempcode, IN_UCHARS(len));
6386              code += 1 + LINK_SIZE;
6387              len += 1 + LINK_SIZE;
6388              tempcode[0] = OP_ONCE;
6389              *code++ = OP_KET;
6390              PUTINC(code, 0, len);
6391              PUT(tempcode, 1, len);
6392              }
6393            }
6394    
6395    #ifdef NEVER
6396        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
6397          {          {
6398          case OP_STAR:  *tempcode = OP_POSSTAR; break;          case OP_STAR:  *tempcode = OP_POSSTAR; break;
# Line 5793  for (;; ptr++) Line 6420  for (;; ptr++)
6420          case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;          case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
6421          case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;          case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;
6422    
6423            case OP_CRSTAR:   *tempcode = OP_CRPOSSTAR; break;
6424            case OP_CRPLUS:   *tempcode = OP_CRPOSPLUS; break;
6425            case OP_CRQUERY:  *tempcode = OP_CRPOSQUERY; break;
6426            case OP_CRRANGE:  *tempcode = OP_CRPOSRANGE; break;
6427    
6428          /* Because we are moving code along, we must ensure that any          /* Because we are moving code along, we must ensure that any
6429          pending recursive references are updated. */          pending recursive references are updated. */
6430    
6431          default:          default:
6432          *code = OP_END;          *code = OP_END;
6433          adjust_recurse(tempcode, 1 + LINK_SIZE, utf, cd, save_hwm);          adjust_recurse(tempcode, 1 + LINK_SIZE, utf, cd, save_hwm_offset);
6434          memmove(tempcode + 1 + LINK_SIZE, tempcode, IN_UCHARS(len));          memmove(tempcode + 1 + LINK_SIZE, tempcode, IN_UCHARS(len));
6435          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
6436          len += 1 + LINK_SIZE;          len += 1 + LINK_SIZE;
# Line 5808  for (;; ptr++) Line 6440  for (;; ptr++)
6440          PUT(tempcode, 1, len);          PUT(tempcode, 1, len);
6441          break;          break;
6442          }          }
6443    #endif
6444        }        }
6445    
6446      /* 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 5829  for (;; ptr++) Line 6462  for (;; ptr++)
6462      newoptions = options;      newoptions = options;
6463      skipbytes = 0;      skipbytes = 0;
6464      bravalue = OP_CBRA;      bravalue = OP_CBRA;
6465      save_hwm = cd->hwm;      save_hwm_offset = cd->hwm - cd->start_workspace;
6466      reset_bracount = FALSE;      reset_bracount = FALSE;
6467    
6468      /* First deal with various "verbs" that can be introduced by '*'. */      /* First deal with various "verbs" that can be introduced by '*'. */
# Line 5996  for (;; ptr++) Line 6629  for (;; ptr++)
6629          tempptr = ptr;          tempptr = ptr;
6630    
6631          /* A condition can be an assertion, a number (referring to a numbered          /* A condition can be an assertion, a number (referring to a numbered
6632          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',
6633          recursion. R<digits> and R&name are also permitted for recursion tests.          referring to recursion. R<digits> and R&name are also permitted for
6634            recursion tests.
6635          There are several syntaxes for testing a named group: (?(name)) is used  
6636          by Python; Perl 5.10 onwards uses (?(<name>) or (?('name')).          There are ways of testing a named group: (?(name)) is used by Python;
6637            Perl 5.10 onwards uses (?(<name>) or (?('name')).
6638          There are two unfortunate ambiguities, caused by history. (a) 'R' can  
6639          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
6640          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
6641          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.  
6642    
6643          For compatibility with auto-callouts, we allow a callout to be          For compatibility with auto-callouts, we allow a callout to be
6644          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 6028  for (;; ptr++) Line 6660  for (;; ptr++)
6660                (tempptr[2] == CHAR_EQUALS_SIGN ||                (tempptr[2] == CHAR_EQUALS_SIGN ||
6661                 tempptr[2] == CHAR_EXCLAMATION_MARK ||                 tempptr[2] == CHAR_EXCLAMATION_MARK ||
6662                 tempptr[2] == CHAR_LESS_THAN_SIGN))                 tempptr[2] == CHAR_LESS_THAN_SIGN))
6663              {
6664              cd->iscondassert = TRUE;
6665            break;            break;
6666              }
6667    
6668          /* 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
6669          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. */
6670    
6671          code[1+LINK_SIZE] = OP_CREF;          code[1+LINK_SIZE] = OP_CREF;
6672          skipbytes = 1+IMM2_SIZE;          skipbytes = 1+IMM2_SIZE;
6673          refsign = -1;          refsign = -1;     /* => not a number */
6674            namelen = -1;     /* => not a name; must set to avoid warning */
6675            name = NULL;      /* Always set to avoid warning */
6676            recno = 0;        /* Always set to avoid warning */
6677    
6678          /* Check for a test for recursion in a named group. */          /* Check for a test for recursion in a named group. */
6679    
6680          if (ptr[1] == CHAR_R && ptr[2] == CHAR_AMPERSAND)          ptr++;
6681            if (*ptr == CHAR_R && ptr[1] == CHAR_AMPERSAND)
6682            {            {
6683            terminator = -1;            terminator = -1;
6684            ptr += 2;            ptr += 2;
# Line 6047  for (;; ptr++) Line 6686  for (;; ptr++)
6686            }            }
6687    
6688          /* 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
6689          syntax (?(<name>) or (?('name') */          syntax (?(<name>) or (?('name'), and also allow for the original PCRE
6690            syntax of (?(name) or for (?(+n), (?(-n), and just (?(n). */
6691    
6692          else if (ptr[1] == CHAR_LESS_THAN_SIGN)          else if (*ptr == CHAR_LESS_THAN_SIGN)
6693            {            {
6694            terminator = CHAR_GREATER_THAN_SIGN;            terminator = CHAR_GREATER_THAN_SIGN;
6695            ptr++;            ptr++;
6696            }            }
6697          else if (ptr[1] == CHAR_APOSTROPHE)          else if (*ptr == CHAR_APOSTROPHE)
6698            {            {
6699            terminator = CHAR_APOSTROPHE;            terminator = CHAR_APOSTROPHE;
6700            ptr++;            ptr++;
# Line 6062  for (;; ptr++) Line 6702  for (;; ptr++)
6702          else          else
6703            {            {
6704            terminator = CHAR_NULL;            terminator = CHAR_NULL;
6705            if (ptr[1] == CHAR_MINUS || ptr[1] == CHAR_PLUS) refsign = *(++ptr);            if (*ptr == CHAR_MINUS || *ptr == CHAR_PLUS) refsign = *ptr++;
6706                else if (IS_DIGIT(*ptr)) refsign = 0;
6707            }            }
6708    
6709          /* We now expect to read a name; any thing else is an error */          /* Handle a number */
6710    
6711          if (!MAX_255(ptr[1]) || (cd->ctypes[ptr[1]] & ctype_word) == 0)          if (refsign >= 0)
6712            {            {
6713            ptr += 1;  /* To get the right offset */            while (IS_DIGIT(*ptr))
6714            *errorcodeptr = ERR28;              {
6715            goto FAILED;              recno = recno * 10 + (int)(*ptr - CHAR_0);
6716                ptr++;
6717                }
6718            }            }
6719    
6720          /* 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
6721            a name is one of a number of duplicates, a different opcode is used and
6722            it needs more memory. Unfortunately we cannot tell whether a name is a
6723            duplicate in the first pass, so we have to allow for more memory. */
6724    
6725          recno = 0;          else
         name = ++ptr;  
         while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_word) != 0)  
6726            {            {
6727            if (recno >= 0)            if (IS_DIGIT(*ptr))
6728              recno = (IS_DIGIT(*ptr))? recno * 10 + (int)(*ptr - CHAR_0) : -1;              {
6729            ptr++;              *errorcodeptr = ERR84;
6730                goto FAILED;
6731                }
6732              if (!MAX_255(*ptr) || (cd->ctypes[*ptr] & ctype_word) == 0)
6733                {
6734                *errorcodeptr = ERR28;   /* Assertion expected */
6735                goto FAILED;
6736                }
6737              name = ptr++;
6738              while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_word) != 0)
6739                {
6740                ptr++;
6741                }
6742              namelen = (int)(ptr - name);
6743              if (lengthptr != NULL) *lengthptr += IMM2_SIZE;
6744            }            }
6745          namelen = (int)(ptr - name);  
6746            /* Check the terminator */
6747    
6748          if ((terminator > 0 && *ptr++ != (pcre_uchar)terminator) ||          if ((terminator > 0 && *ptr++ != (pcre_uchar)terminator) ||
6749              *ptr++ != CHAR_RIGHT_PARENTHESIS)              *ptr++ != CHAR_RIGHT_PARENTHESIS)
6750            {            {
6751            ptr--;      /* Error offset */            ptr--;                  /* Error offset */
6752            *errorcodeptr = ERR26;            *errorcodeptr = ERR26;  /* Malformed number or name */
6753            goto FAILED;            goto FAILED;
6754            }            }
6755    
# Line 6099  for (;; ptr++) Line 6758  for (;; ptr++)
6758          if (lengthptr != NULL) break;          if (lengthptr != NULL) break;
6759    
6760          /* 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
6761          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
6762          be digits, in which case recno will be set. */          recno. */
6763    
6764          if (refsign > 0)          if (refsign >= 0)
6765            {            {
6766            if (recno <= 0)            if (recno <= 0)
6767              {              {
6768              *errorcodeptr = ERR58;              *errorcodeptr = ERR35;
6769              goto FAILED;              goto FAILED;
6770              }              }
6771            recno = (refsign == CHAR_MINUS)?            if (refsign != 0) recno = (refsign == CHAR_MINUS)?
6772              cd->bracount - recno + 1 : recno +cd->bracount;              cd->bracount - recno + 1 : recno + cd->bracount;
6773            if (recno <= 0 || recno > cd->final_bracount)            if (recno <= 0 || recno > cd->final_bracount)
6774              {              {
6775              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
6776              goto FAILED;              goto FAILED;
6777              }              }
6778            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
6779              if (recno > cd->top_backref) cd->top_backref = recno;
6780            break;            break;
6781            }            }
6782    
6783          /* 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. */  
6784    
6785          slot = cd->name_table;          slot = cd->name_table;
6786          for (i = 0; i < cd->names_found; i++)          for (i = 0; i < cd->names_found; i++)
# Line 6133  for (;; ptr++) Line 6789  for (;; ptr++)
6789            slot += cd->name_entry_size;            slot += cd->name_entry_size;
6790            }            }
6791    
6792          /* Found the named subpattern */          /* Found the named subpattern. If the name is duplicated, add one to
6793            the opcode to change CREF/RREF into DNCREF/DNRREF and insert
6794            appropriate data values. Otherwise, just insert the unique subpattern
6795            number. */
6796    
6797          if (i < cd->names_found)          if (i < cd->names_found)
6798            {            {
6799            recno = GET2(slot, 0);            int offset = i++;
6800            PUT2(code, 2+LINK_SIZE, recno);            int count = 1;
6801            code[1+LINK_SIZE]++;            recno = GET2(slot, 0);   /* Number from first found */
6802              if (recno > cd->top_backref) cd->top_backref = recno;
6803              for (; i < cd->names_found; i++)
6804                {
6805                slot += cd->name_entry_size;
6806                if (STRNCMP_UC_UC(name, slot+IMM2_SIZE, namelen) != 0 ||
6807                  (slot+IMM2_SIZE)[namelen] != 0) break;
6808                count++;
6809                }
6810    
6811              if (count > 1)
6812                {
6813                PUT2(code, 2+LINK_SIZE, offset);
6814                PUT2(code, 2+LINK_SIZE+IMM2_SIZE, count);
6815                skipbytes += IMM2_SIZE;
6816                code[1+LINK_SIZE]++;
6817                }
6818              else  /* Not a duplicated name */
6819                {
6820                PUT2(code, 2+LINK_SIZE, recno);
6821                }
6822            }            }
6823    
6824          /* If terminator == CHAR_NULL it means that the name followed directly          /* If terminator == CHAR_NULL it means that the name followed directly
6825          after the opening parenthesis [e.g. (?(abc)...] and in this case there          after the opening parenthesis [e.g. (?(abc)...] and in this case there
6826          are some further alternatives to try. For the cases where terminator !=          are some further alternatives to try. For the cases where terminator !=
6827          0 [things like (?(<name>... or (?('name')... or (?(R&name)... ] we have          CHAR_NULL [things like (?(<name>... or (?('name')... or (?(R&name)... ]
6828          now checked all the possibilities, so give an error. */          we have now checked all the possibilities, so give an error. */
6829    
6830          else if (terminator != CHAR_NULL)          else if (terminator != CHAR_NULL)
6831            {            {
# Line 6183  for (;; ptr++) Line 6862  for (;; ptr++)
6862            skipbytes = 1;            skipbytes = 1;
6863            }            }
6864    
6865          /* 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) */  
6866    
6867          else          else
6868            {            {
6869            *errorcodeptr = (recno == 0)? ERR35: ERR15;            *errorcodeptr = ERR15;
6870            goto FAILED;            goto FAILED;
6871            }            }
6872          break;          break;
# Line 6208  for (;; ptr++) Line 6879  for (;; ptr++)
6879          ptr++;          ptr++;
6880          break;          break;
6881    
6882            /* Optimize (?!) to (*FAIL) unless it is quantified - which is a weird
6883            thing to do, but Perl allows all assertions to be quantified, and when
6884            they contain capturing parentheses there may be a potential use for
6885            this feature. Not that that applies to a quantified (?!) but we allow
6886            it for uniformity. */
6887    
6888          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
6889          case CHAR_EXCLAMATION_MARK:            /* Negative lookahead */          case CHAR_EXCLAMATION_MARK:            /* Negative lookahead */
6890          ptr++;          ptr++;
6891          if (*ptr == CHAR_RIGHT_PARENTHESIS)    /* Optimize (?!) */          if (*ptr == CHAR_RIGHT_PARENTHESIS && ptr[1] != CHAR_ASTERISK &&
6892                 ptr[1] != CHAR_PLUS && ptr[1] != CHAR_QUESTION_MARK &&
6893                (ptr[1] != CHAR_LEFT_CURLY_BRACKET || !is_counted_repeat(ptr+2)))
6894            {            {
6895            *code++ = OP_FAIL;            *code++ = OP_FAIL;
6896            previous = NULL;            previous = NULL;
# Line 6308  for (;; ptr++) Line 6986  for (;; ptr++)
6986          terminator = (*ptr == CHAR_LESS_THAN_SIGN)?          terminator = (*ptr == CHAR_LESS_THAN_SIGN)?
6987            CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;            CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
6988          name = ++ptr;          name = ++ptr;
6989            if (IS_DIGIT(*ptr))
6990              {
6991              *errorcodeptr = ERR84;   /* Group name must start with non-digit */
6992              goto FAILED;
6993              }
6994          while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_word) != 0) ptr++;          while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
6995          namelen = (int)(ptr - name);          namelen = (int)(ptr - name);
6996    
# Line 6422  for (;; ptr++) Line 7104  for (;; ptr++)
7104    
7105          NAMED_REF_OR_RECURSE:          NAMED_REF_OR_RECURSE:
7106          name = ++ptr;          name = ++ptr;
7107            if (IS_DIGIT(*ptr))
7108              {
7109              *errorcodeptr = ERR84;   /* Group name must start with non-digit */
7110              goto FAILED;
7111              }
7112          while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_word) != 0) ptr++;          while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
7113          namelen = (int)(ptr - name);          namelen = (int)(ptr - name);
7114    
# Line 6469  for (;; ptr++) Line 7156  for (;; ptr++)
7156            /* Count named back references. */            /* Count named back references. */
7157    
7158            if (!is_recurse) cd->namedrefcount++;            if (!is_recurse) cd->namedrefcount++;
7159    
7160              /* We have to allow for a named reference to a duplicated name (this
7161              cannot be determined until the second pass). This needs an extra
7162              16-bit data item. */
7163    
7164              *lengthptr += IMM2_SIZE;
7165            }            }
7166    
7167          /* In the real compile, search the name table. We check the name          /* In the real compile, search the name table. We check the name
# Line 6515  for (;; ptr++) Line 7208  for (;; ptr++)
7208            for (i++; i < cd->names_found; i++)            for (i++; i < cd->names_found; i++)
7209              {              {
7210              if (STRCMP_UC_UC(slot + IMM2_SIZE, cslot + IMM2_SIZE) != 0) break;              if (STRCMP_UC_UC(slot + IMM2_SIZE, cslot + IMM2_SIZE) != 0) break;
7211    
7212    
7213              count++;              count++;
7214              cslot += cd->name_entry_size;              cslot += cd->name_entry_size;
7215              }              }
# Line 6814  for (;; ptr++) Line 7509  for (;; ptr++)
7509        skipbytes = IMM2_SIZE;        skipbytes = IMM2_SIZE;
7510        }        }
7511    
7512      /* Process nested bracketed regex. Assertions used not to be repeatable,      /* Process nested bracketed regex. First check for parentheses nested too
7513      but this was changed for Perl compatibility, so all kinds can now be      deeply. */
7514      repeated. We copy code into a non-register variable (tempcode) in order to  
7515      be able to pass its address because some compilers complain otherwise. */      if ((cd->parens_depth += 1) > PARENS_NEST_LIMIT)
7516          {
7517          *errorcodeptr = ERR82;
7518          goto FAILED;
7519          }
7520    
7521        /* All assertions used not to be repeatable, but this was changed for Perl
7522        compatibility. All kinds can now be repeated except for assertions that are
7523        conditions (Perl also forbids these to be repeated). We copy code into a
7524        non-register variable (tempcode) in order to be able to pass its address
7525        because some compilers complain otherwise. At the start of a conditional
7526        group whose condition is an assertion, cd->iscondassert is set. We unset it
7527        here so as to allow assertions later in the group to be quantified. */
7528    
7529        if (bravalue >= OP_ASSERT && bravalue <= OP_ASSERTBACK_NOT &&
7530            cd->iscondassert)
7531          {
7532          previous = NULL;
7533          cd->iscondassert = FALSE;
7534          }
7535        else previous = code;
7536    
     previous = code;                      /* For handling repetition */  
7537      *code = bravalue;      *code = bravalue;
7538      tempcode = code;      tempcode = code;
7539      tempreqvary = cd->req_varyopt;        /* Save value before bracket */      tempreqvary = cd->req_varyopt;        /* Save value before bracket */
# Line 6848  for (;; ptr++) Line 7562  for (;; ptr++)
7562           ))           ))
7563        goto FAILED;        goto FAILED;
7564    
7565        cd->parens_depth -= 1;
7566    
7567      /* 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,
7568      generate OP_ONCE_NC instead of OP_ONCE. */      generate OP_ONCE_NC instead of OP_ONCE. */
7569    
# Line 7062  for (;; ptr++) Line 7778  for (;; ptr++)
7778        if (escape == ESC_g)        if (escape == ESC_g)
7779          {          {
7780          const pcre_uchar *p;          const pcre_uchar *p;
7781          save_hwm = cd->hwm;   /* Normally this is set when '(' is read */          pcre_uint32 cf;
7782    
7783            save_hwm_offset = cd->hwm - cd->start_workspace;   /* Normally this is set when '(' is read */
7784          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
7785            CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;            CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
7786    
7787          /* These two statements stop the compiler for warning about possibly          /* These two statements stop the compiler for warning about possibly
7788          unset variables caused by the jump to HANDLE_NUMERICAL_RECURSION. In          unset variables caused by the jump to HANDLE_NUMERICAL_RECURSION. In
7789          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
7790          would actually be in error are never taken. */          would actually be in error are never taken. */
7791    
7792          skipbytes = 0;          skipbytes = 0;
7793          reset_bracount = FALSE;          reset_bracount = FALSE;
7794    
7795          /* Test for a name */          /* If it's not a signed or unsigned number, treat it as a name. */
7796    
7797          if (ptr[1] != CHAR_PLUS && ptr[1] != CHAR_MINUS)          cf = ptr[1];
7798            if (cf != CHAR_PLUS && cf != CHAR_MINUS && !IS_DIGIT(cf))
7799            {            {
           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;  
             }  
7800            is_recurse = TRUE;            is_recurse = TRUE;
7801            goto NAMED_REF_OR_RECURSE;            goto NAMED_REF_OR_RECURSE;
7802            }            }
7803    
7804          /* Test a signed number in angle brackets or quotes. */          /* Signed or unsigned number (cf = ptr[1]) is known to be plus or minus
7805            or a digit. */
7806    
7807          p = ptr + 2;          p = ptr + 2;
7808          while (IS_DIGIT(*p)) p++;          while (IS_DIGIT(*p)) p++;
# Line 7239  for (;; ptr++) Line 7942  for (;; ptr++)
7942    
7943      /* ===================================================================*/      /* ===================================================================*/
7944      /* Handle a literal character. It is guaranteed not to be whitespace or #      /* Handle a literal character. It is guaranteed not to be whitespace or #
7945      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
7946      multi-byte literal character. */      multi-unit literal character. */
7947    
7948      default:      default:
7949      NORMAL_CHAR:      NORMAL_CHAR:
# Line 7404  int length; Line 8107  int length;
8107  unsigned int orig_bracount;  unsigned int orig_bracount;
8108  unsigned int max_bracount;  unsigned int max_bracount;
8109  branch_chain bc;  branch_chain bc;
8110    size_t save_hwm_offset;
8111    
8112    /* If set, call the external function that checks for stack availability. */
8113    
8114    if (PUBL(stack_guard) != NULL && PUBL(stack_guard)())
8115      {
8116      *errorcodeptr= ERR85;
8117      return FALSE;
8118      }
8119    
8120    /* Miscellaneous initialization */
8121    
8122  bc.outer = bcptr;  bc.outer = bcptr;
8123  bc.current_branch = code;  bc.current_branch = code;
# Line 7411  bc.current_branch = code; Line 8125  bc.current_branch = code;
8125  firstchar = reqchar = 0;  firstchar = reqchar = 0;
8126  firstcharflags = reqcharflags = REQ_UNSET;  firstcharflags = reqcharflags = REQ_UNSET;
8127    
8128    save_hwm_offset = cd->hwm - cd->start_workspace;
8129    
8130  /* Accumulate the length for use in the pre-compile phase. Start with the  /* Accumulate the length for use in the pre-compile phase. Start with the
8131  length of the BRA and KET and any extra bytes that are required at the  length of the BRA and KET and any extra bytes that are required at the
8132  beginning. We accumulate in a local variable to save frequent testing of  beginning. We accumulate in a local variable to save frequent testing of
# Line 7604  for (;;) Line 8320  for (;;)
8320    
8321      /* If it was a capturing subpattern, check to see if it contained any      /* If it was a capturing subpattern, check to see if it contained any
8322      recursive back references. If so, we must wrap it in atomic brackets.      recursive back references. If so, we must wrap it in atomic brackets.
8323      In any event, remove the block from the chain. */      Because we are moving code along, we must ensure that any pending recursive
8324        references are updated. In any event, remove the block from the chain. */
8325    
8326      if (capnumber > 0)      if (capnumber > 0)
8327        {        {
8328        if (cd->open_caps->flag)        if (cd->open_caps->flag)
8329          {          {
8330            *code = OP_END;
8331            adjust_recurse(start_bracket, 1 + LINK_SIZE,
8332              (options & PCRE_UTF8) != 0, cd, save_hwm_offset);
8333          memmove(start_bracket + 1 + LINK_SIZE, start_bracket,          memmove(start_bracket + 1 + LINK_SIZE, start_bracket,
8334            IN_UCHARS(code - start_bracket));            IN_UCHARS(code - start_bracket));
8335          *start_bracket = OP_ONCE;          *start_bracket = OP_ONCE;
# Line 7829  do { Line 8549  do {
8549       switch (*scode)       switch (*scode)
8550         {         {
8551         case OP_CREF:         case OP_CREF:
8552         case OP_NCREF:         case OP_DNCREF:
8553         case OP_RREF:         case OP_RREF:
8554         case OP_NRREF:         case OP_DNRREF:
8555         case OP_DEF:         case OP_DEF:
8556           case OP_FAIL:
8557         return FALSE;         return FALSE;
8558    
8559         default:     /* Assertion */         default:     /* Assertion */
# Line 8228  PCRE_UTF8 == PCRE_UTF16 == PCRE_UTF32. * Line 8949  PCRE_UTF8 == PCRE_UTF16 == PCRE_UTF32. *
8949      { skipatstart += 6; options |= PCRE_UTF8; continue; }      { skipatstart += 6; options |= PCRE_UTF8; continue; }
8950    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)
8951      { skipatstart += 6; options |= PCRE_UCP; continue; }      { skipatstart += 6; options |= PCRE_UCP; continue; }
8952      else if (STRNCMP_UC_C8(ptr+skipatstart+2, STRING_NO_AUTO_POSSESS_RIGHTPAR, 16) == 0)
8953        { skipatstart += 18; options |= PCRE_NO_AUTO_POSSESS; continue; }
8954    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)
8955      { skipatstart += 15; options |= PCRE_NO_START_OPTIMIZE; continue; }      { skipatstart += 15; options |= PCRE_NO_START_OPTIMIZE; continue; }
8956    
# Line 8415  cd->dupnames = FALSE; Line 9138  cd->dupnames = FALSE;
9138  cd->namedrefcount = 0;  cd->namedrefcount = 0;
9139  cd->start_code = cworkspace;  cd->start_code = cworkspace;
9140  cd->hwm = cworkspace;  cd->hwm = cworkspace;
9141    cd->iscondassert = FALSE;
9142  cd->start_workspace = cworkspace;  cd->start_workspace = cworkspace;
9143  cd->workspace_size = COMPILE_WORK_SIZE;  cd->workspace_size = COMPILE_WORK_SIZE;
9144  cd->named_groups = named_groups;  cd->named_groups = named_groups;
# Line 8422  cd->named_group_list_size = NAMED_GROUP_ Line 9146  cd->named_group_list_size = NAMED_GROUP_
9146  cd->start_pattern = (const pcre_uchar *)pattern;  cd->start_pattern = (const pcre_uchar *)pattern;
9147  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));
9148  cd->req_varyopt = 0;  cd->req_varyopt = 0;
9149    cd->parens_depth = 0;
9150  cd->assert_depth = 0;  cd->assert_depth = 0;
9151  cd->max_lookbehind = 0;  cd->max_lookbehind = 0;
9152  cd->external_options = options;  cd->external_options = options;
# Line 8436  outside can help speed up starting point Line 9161  outside can help speed up starting point
9161  ptr += skipatstart;  ptr += skipatstart;
9162  code = cworkspace;  code = cworkspace;
9163  *code = OP_BRA;  *code = OP_BRA;
9164    
9165  (void)compile_regex(cd->external_options, &code, &ptr, &errorcode, FALSE,  (void)compile_regex(cd->external_options, &code, &ptr, &errorcode, FALSE,
9166    FALSE, 0, 0, &firstchar, &firstcharflags, &reqchar, &reqcharflags, NULL,    FALSE, 0, 0, &firstchar, &firstcharflags, &reqchar, &reqcharflags, NULL,
9167    cd, &length);    cd, &length);
# Line 8450  if (length > MAX_PATTERN_SIZE) Line 9176  if (length > MAX_PATTERN_SIZE)
9176    goto PCRE_EARLY_ERROR_RETURN;    goto PCRE_EARLY_ERROR_RETURN;
9177    }    }
9178    
 /* If there are groups with duplicate names and there are also references by  
 name, we must allow for the possibility of named references to duplicated  
 groups. These require an extra data item each. */  
   
 if (cd->dupnames && cd->namedrefcount > 0)  
   length += cd->namedrefcount * IMM2_SIZE * sizeof(pcre_uchar);  
   
9179  /* Compute the size of the data block for storing the compiled pattern. Integer  /* Compute the size of the data block for storing the compiled pattern. Integer
9180  overflow should no longer be possible because nowadays we limit the maximum  overflow should no longer be possible because nowadays we limit the maximum
9181  value of cd->names_found and cd->name_entry_size. */  value of cd->names_found and cd->name_entry_size. */
# Line 8507  field; this time it's used for rememberi Line 9226  field; this time it's used for rememberi
9226  */  */
9227    
9228  cd->final_bracount = cd->bracount;  /* Save for checking forward references */  cd->final_bracount = cd->bracount;  /* Save for checking forward references */
9229    cd->parens_depth = 0;
9230  cd->assert_depth = 0;  cd->assert_depth = 0;
9231  cd->bracount = 0;  cd->bracount = 0;
9232  cd->max_lookbehind = 0;  cd->max_lookbehind = 0;
# Line 8514  cd->name_table = (pcre_uchar *)re + re-> Line 9234  cd->name_table = (pcre_uchar *)re + re->
9234  codestart = cd->name_table + re->name_entry_size * re->name_count;  codestart = cd->name_table + re->name_entry_size * re->name_count;
9235  cd->start_code = codestart;  cd->start_code = codestart;
9236  cd->hwm = (pcre_uchar *)(cd->start_workspace);  cd->hwm = (pcre_uchar *)(cd->start_workspace);
9237    cd->iscondassert = FALSE;
9238  cd->req_varyopt = 0;  cd->req_varyopt = 0;
9239  cd->had_accept = FALSE;  cd->had_accept = FALSE;
9240  cd->had_pruneorskip = FALSE;  cd->had_pruneorskip = FALSE;
# Line 8609  subpattern. */ Line 9330  subpattern. */
9330    
9331  if (errorcode == 0 && re->top_backref > re->top_bracket) errorcode = ERR15;  if (errorcode == 0 && re->top_backref > re->top_bracket) errorcode = ERR15;
9332    
9333  /* Unless disabled, check whether single character iterators can be  /* Unless disabled, check whether any single character iterators can be
9334  auto-possessified. The function overwrites the appropriate opcode values. */  auto-possessified. The function overwrites the appropriate opcode values, so
9335    the type of the pointer must be cast. NOTE: the intermediate variable "temp" is
9336    used in this code because at least one compiler gives a warning about loss of
9337    "const" attribute if the cast (pcre_uchar *)codestart is used directly in the
9338    function call. */
9339    
9340  if ((options & PCRE_NO_AUTO_POSSESSIFY) == 0)  if ((options & PCRE_NO_AUTO_POSSESS) == 0)
9341    auto_possessify((pcre_uchar *)codestart, utf, cd);    {
9342      pcre_uchar *temp = (pcre_uchar *)codestart;
9343      auto_possessify(temp, utf, cd);
9344      }
9345    
9346  /* If there were any lookbehind assertions that contained OP_RECURSE  /* If there were any lookbehind assertions that contained OP_RECURSE
9347  (recursions or subroutine calls), a flag is set for them to be checked here,  (recursions or subroutine calls), a flag is set for them to be checked here,
# Line 8836  return (pcre32 *)re; Line 9564  return (pcre32 *)re;
9564  }  }
9565    
9566  /* End of pcre_compile.c */  /* End of pcre_compile.c */
9567    

Legend:
Removed from v.1364  
changed lines
  Added in v.1534

  ViewVC Help
Powered by ViewVC 1.1.5