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

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

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

revision 518 by ph10, Tue May 18 15:47:01 2010 UTC revision 635 by ph10, Sat Jul 23 16:19:50 2011 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) 1997-2010 University of Cambridge             Copyright (c) 1997-2011 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 261  static const int posix_class_maps[] = { Line 261  static const int posix_class_maps[] = {
261    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
262  };  };
263    
264  /* 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. The POSIX class
265  substitutes must be in the order of the names, defined above, and there are  substitutes must be in the order of the names, defined above, and there are
266  both positive and negative cases. NULL means no substitute. */  both positive and negative cases. NULL means no substitute. */
267    
268  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
# Line 272  static const uschar *substitutes[] = { Line 272  static const uschar *substitutes[] = {
272    (uschar *)"\\P{Xsp}",   /* \S */       /* NOTE: Xsp is Perl space */    (uschar *)"\\P{Xsp}",   /* \S */       /* NOTE: Xsp is Perl space */
273    (uschar *)"\\p{Xsp}",   /* \s */    (uschar *)"\\p{Xsp}",   /* \s */
274    (uschar *)"\\P{Xwd}",   /* \W */    (uschar *)"\\P{Xwd}",   /* \W */
275    (uschar *)"\\p{Xwd}"    /* \w */    (uschar *)"\\p{Xwd}"    /* \w */
276  };  };
277    
278  static const uschar *posix_substitutes[] = {  static const uschar *posix_substitutes[] = {
279    (uschar *)"\\p{L}",     /* alpha */    (uschar *)"\\p{L}",     /* alpha */
280    (uschar *)"\\p{Ll}",    /* lower */    (uschar *)"\\p{Ll}",    /* lower */
281    (uschar *)"\\p{Lu}",    /* upper */    (uschar *)"\\p{Lu}",    /* upper */
282    (uschar *)"\\p{Xan}",   /* alnum */    (uschar *)"\\p{Xan}",   /* alnum */
283    NULL,                   /* ascii */    NULL,                   /* ascii */
284    (uschar *)"\\h",        /* blank */    (uschar *)"\\h",        /* blank */
285    NULL,                   /* cntrl */    NULL,                   /* cntrl */
# Line 289  static const uschar *posix_substitutes[] Line 289  static const uschar *posix_substitutes[]
289    NULL,                   /* punct */    NULL,                   /* punct */
290    (uschar *)"\\p{Xps}",   /* space */    /* NOTE: Xps is POSIX space */    (uschar *)"\\p{Xps}",   /* space */    /* NOTE: Xps is POSIX space */
291    (uschar *)"\\p{Xwd}",   /* word */    (uschar *)"\\p{Xwd}",   /* word */
292    NULL,                   /* xdigit */    NULL,                   /* xdigit */
293    /* Negated cases */    /* Negated cases */
294    (uschar *)"\\P{L}",     /* ^alpha */    (uschar *)"\\P{L}",     /* ^alpha */
295    (uschar *)"\\P{Ll}",    /* ^lower */    (uschar *)"\\P{Ll}",    /* ^lower */
296    (uschar *)"\\P{Lu}",    /* ^upper */    (uschar *)"\\P{Lu}",    /* ^upper */
297    (uschar *)"\\P{Xan}",   /* ^alnum */    (uschar *)"\\P{Xan}",   /* ^alnum */
298    NULL,                   /* ^ascii */    NULL,                   /* ^ascii */
299    (uschar *)"\\H",        /* ^blank */    (uschar *)"\\H",        /* ^blank */
300    NULL,                   /* ^cntrl */    NULL,                   /* ^cntrl */
# Line 304  static const uschar *posix_substitutes[] Line 304  static const uschar *posix_substitutes[]
304    NULL,                   /* ^punct */    NULL,                   /* ^punct */
305    (uschar *)"\\P{Xps}",   /* ^space */   /* NOTE: Xps is POSIX space */    (uschar *)"\\P{Xps}",   /* ^space */   /* NOTE: Xps is POSIX space */
306    (uschar *)"\\P{Xwd}",   /* ^word */    (uschar *)"\\P{Xwd}",   /* ^word */
307    NULL                    /* ^xdigit */    NULL                    /* ^xdigit */
308  };  };
309  #define POSIX_SUBSIZE (sizeof(posix_substitutes)/sizeof(uschar *))  #define POSIX_SUBSIZE (sizeof(posix_substitutes)/sizeof(uschar *))
310  #endif  #endif
311    
312  #define STRING(a)  # a  #define STRING(a)  # a
313  #define XSTRING(s) STRING(s)  #define XSTRING(s) STRING(s)
# Line 407  static const char error_texts[] = Line 407  static const char error_texts[] =
407    /* 65 */    /* 65 */
408    "different names for subpatterns of the same number are not allowed\0"    "different names for subpatterns of the same number are not allowed\0"
409    "(*MARK) must have an argument\0"    "(*MARK) must have an argument\0"
410    "this version of PCRE is not compiled with PCRE_UCP support\0"    "this version of PCRE is not compiled with PCRE_UCP support\0"
411      "\\c must be followed by an ASCII character\0"
412      "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
413    ;    ;
414    
415  /* 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 544  static const unsigned char ebcdic_charta Line 546  static const unsigned char ebcdic_charta
546  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
547    
548  static BOOL  static BOOL
549    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,    compile_regex(int, uschar **, const uschar **, int *, BOOL, BOOL, int, int *,
550      int *, int *, branch_chain *, compile_data *, int *);      int *, branch_chain *, compile_data *, int *);
551    
552    
553    
# Line 841  else Line 843  else
843      break;      break;
844    
845      /* 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.
846      This coding is ASCII-specific, but then the whole concept of \cx is      An error is given if the byte following \c is not an ASCII character. This
847        coding is ASCII-specific, but then the whole concept of \cx is
848      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
849    
850      case CHAR_c:      case CHAR_c:
# Line 851  else Line 854  else
854        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
855        break;        break;
856        }        }
857    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
858  #ifndef EBCDIC  /* ASCII/UTF-8 coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
859          {
860          *errorcodeptr = ERR68;
861          break;
862          }
863      if (c >= CHAR_a && c <= CHAR_z) c -= 32;      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
864      c ^= 0x40;      c ^= 0x40;
865  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
866      if (c >= CHAR_a && c <= CHAR_z) c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
867      c ^= 0xC0;      c ^= 0xC0;
868  #endif  #endif
# Line 1099  top-level call starts at the beginning o Line 1106  top-level call starts at the beginning o
1106  start at a parenthesis. It scans along a pattern's text looking for capturing  start at a parenthesis. It scans along a pattern's text looking for capturing
1107  subpatterns, and counting them. If it finds a named pattern that matches the  subpatterns, and counting them. If it finds a named pattern that matches the
1108  name it is given, it returns its number. Alternatively, if the name is NULL, it  name it is given, it returns its number. Alternatively, if the name is NULL, it
1109  returns when it reaches a given numbered subpattern. We know that if (?P< is  returns when it reaches a given numbered subpattern. Recursion is used to keep
1110  encountered, the name will be terminated by '>' because that is checked in the  track of subpatterns that reset the capturing group numbers - the (?| feature.
1111  first pass. Recursion is used to keep track of subpatterns that reset the  
1112  capturing group numbers - the (?| feature.  This function was originally called only from the second pass, in which we know
1113    that if (?< or (?' or (?P< is encountered, the name will be correctly
1114    terminated because that is checked in the first pass. There is now one call to
1115    this function in the first pass, to check for a recursive back reference by
1116    name (so that we can make the whole group atomic). In this case, we need check
1117    only up to the current position in the pattern, and that is still OK because
1118    and previous occurrences will have been checked. To make this work, the test
1119    for "end of pattern" is a check against cd->end_pattern in the main loop,
1120    instead of looking for a binary zero. This means that the special first-pass
1121    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1122    processing items within the loop are OK, because afterwards the main loop will
1123    terminate.)
1124    
1125  Arguments:  Arguments:
1126    ptrptr       address of the current character pointer (updated)    ptrptr       address of the current character pointer (updated)
# Line 1110  Arguments: Line 1128  Arguments:
1128    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1129    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1130    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1131      utf8         TRUE if we are in UTF-8 mode
1132    count        pointer to the current capturing subpattern number (updated)    count        pointer to the current capturing subpattern number (updated)
1133    
1134  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
# Line 1117  Returns: the number of the named s Line 1136  Returns: the number of the named s
1136    
1137  static int  static int
1138  find_parens_sub(uschar **ptrptr, compile_data *cd, const uschar *name, int lorn,  find_parens_sub(uschar **ptrptr, compile_data *cd, const uschar *name, int lorn,
1139    BOOL xmode, int *count)    BOOL xmode, BOOL utf8, int *count)
1140  {  {
1141  uschar *ptr = *ptrptr;  uschar *ptr = *ptrptr;
1142  int start_count = *count;  int start_count = *count;
# Line 1129  dealing with. The very first call may no Line 1148  dealing with. The very first call may no
1148    
1149  if (ptr[0] == CHAR_LEFT_PARENTHESIS)  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1150    {    {
1151    if (ptr[1] == CHAR_QUESTION_MARK &&    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1152        ptr[2] == CHAR_VERTICAL_LINE)  
1153      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1154    
1155      /* Handle a normal, unnamed capturing parenthesis. */
1156    
1157      else if (ptr[1] != CHAR_QUESTION_MARK)
1158        {
1159        *count += 1;
1160        if (name == NULL && *count == lorn) return *count;
1161        ptr++;
1162        }
1163    
1164      /* All cases now have (? at the start. Remember when we are in a group
1165      where the parenthesis numbers are duplicated. */
1166    
1167      else if (ptr[2] == CHAR_VERTICAL_LINE)
1168      {      {
1169      ptr += 3;      ptr += 3;
1170      dup_parens = TRUE;      dup_parens = TRUE;
1171      }      }
1172    
1173    /* Handle a normal, unnamed capturing parenthesis */    /* Handle comments; all characters are allowed until a ket is reached. */
1174    
1175    else if (ptr[1] != CHAR_QUESTION_MARK && ptr[1] != CHAR_ASTERISK)    else if (ptr[2] == CHAR_NUMBER_SIGN)
1176      {      {
1177      *count += 1;      for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1178      if (name == NULL && *count == lorn) return *count;      goto FAIL_EXIT;
     ptr++;  
1179      }      }
1180    
1181    /* Handle a condition. If it is an assertion, just carry on so that it    /* Handle a condition. If it is an assertion, just carry on so that it
1182    is processed as normal. If not, skip to the closing parenthesis of the    is processed as normal. If not, skip to the closing parenthesis of the
1183    condition (there can't be any nested parens. */    condition (there can't be any nested parens). */
1184    
1185    else if (ptr[2] == CHAR_LEFT_PARENTHESIS)    else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1186      {      {
# Line 1159  if (ptr[0] == CHAR_LEFT_PARENTHESIS) Line 1192  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1192        }        }
1193      }      }
1194    
1195    /* We have either (? or (* and not a condition */    /* Start with (? but not a condition. */
1196    
1197    else    else
1198      {      {
# Line 1188  if (ptr[0] == CHAR_LEFT_PARENTHESIS) Line 1221  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1221    }    }
1222    
1223  /* Past any initial parenthesis handling, scan for parentheses or vertical  /* Past any initial parenthesis handling, scan for parentheses or vertical
1224  bars. */  bars. Stop if we get to cd->end_pattern. Note that this is important for the
1225    first-pass call when this value is temporarily adjusted to stop at the current
1226    position. So DO NOT change this to a test for binary zero. */
1227    
1228  for (; *ptr != 0; ptr++)  for (; ptr < cd->end_pattern; ptr++)
1229    {    {
1230    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1231    
# Line 1264  for (; *ptr != 0; ptr++) Line 1299  for (; *ptr != 0; ptr++)
1299    
1300    if (xmode && *ptr == CHAR_NUMBER_SIGN)    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1301      {      {
1302      while (*(++ptr) != 0 && *ptr != CHAR_NL) {};      ptr++;
1303        while (*ptr != 0)
1304          {
1305          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1306          ptr++;
1307    #ifdef SUPPORT_UTF8
1308          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
1309    #endif
1310          }
1311      if (*ptr == 0) goto FAIL_EXIT;      if (*ptr == 0) goto FAIL_EXIT;
1312      continue;      continue;
1313      }      }
# Line 1273  for (; *ptr != 0; ptr++) Line 1316  for (; *ptr != 0; ptr++)
1316    
1317    if (*ptr == CHAR_LEFT_PARENTHESIS)    if (*ptr == CHAR_LEFT_PARENTHESIS)
1318      {      {
1319      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, count);      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, count);
1320      if (rc > 0) return rc;      if (rc > 0) return rc;
1321      if (*ptr == 0) goto FAIL_EXIT;      if (*ptr == 0) goto FAIL_EXIT;
1322      }      }
# Line 1281  for (; *ptr != 0; ptr++) Line 1324  for (; *ptr != 0; ptr++)
1324    else if (*ptr == CHAR_RIGHT_PARENTHESIS)    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1325      {      {
1326      if (dup_parens && *count < hwm_count) *count = hwm_count;      if (dup_parens && *count < hwm_count) *count = hwm_count;
1327      *ptrptr = ptr;      goto FAIL_EXIT;
     return -1;  
1328      }      }
1329    
1330    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
# Line 1320  Arguments: Line 1362  Arguments:
1362    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1363    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1364    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1365      utf8         TRUE if we are in UTF-8 mode
1366    
1367  Returns:       the number of the found subpattern, or -1 if not found  Returns:       the number of the found subpattern, or -1 if not found
1368  */  */
1369    
1370  static int  static int
1371  find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode)  find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode,
1372      BOOL utf8)
1373  {  {
1374  uschar *ptr = (uschar *)cd->start_pattern;  uschar *ptr = (uschar *)cd->start_pattern;
1375  int count = 0;  int count = 0;
# Line 1338  matching closing parens. That is why we Line 1382  matching closing parens. That is why we
1382    
1383  for (;;)  for (;;)
1384    {    {
1385    rc = find_parens_sub(&ptr, cd, name, lorn, xmode, &count);    rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, &count);
1386    if (rc > 0 || *ptr++ == 0) break;    if (rc > 0 || *ptr++ == 0) break;
1387    }    }
1388    
# Line 1354  return rc; Line 1398  return rc;
1398    
1399  /* This is called by several functions that scan a compiled expression looking  /* This is called by several functions that scan a compiled expression looking
1400  for a fixed first character, or an anchoring op code etc. It skips over things  for a fixed first character, or an anchoring op code etc. It skips over things
1401  that do not influence this. For some calls, a change of option is important.  that do not influence this. For some calls, it makes sense to skip negative
1402  For some calls, it makes sense to skip negative forward and all backward  forward and all backward assertions, and also the \b assertion; for others it
1403  assertions, and also the \b assertion; for others it does not.  does not.
1404    
1405  Arguments:  Arguments:
1406    code         pointer to the start of the group    code         pointer to the start of the group
   options      pointer to external options  
   optbit       the option bit whose changing is significant, or  
                  zero if none are  
1407    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1408    
1409  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1410  */  */
1411    
1412  static const uschar*  static const uschar*
1413  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const uschar *code, BOOL skipassert)
   BOOL skipassert)  
1414  {  {
1415  for (;;)  for (;;)
1416    {    {
1417    switch ((int)*code)    switch ((int)*code)
1418      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1419      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1420      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1421      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
# Line 1431  and doing the check at the end; a flag s Line 1465  and doing the check at the end; a flag s
1465    
1466  Arguments:  Arguments:
1467    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1468    options  the compiling options    utf8     TRUE in UTF-8 mode
1469    atend    TRUE if called when the pattern is complete    atend    TRUE if called when the pattern is complete
1470    cd       the "compile data" structure    cd       the "compile data" structure
1471    
# Line 1442  Returns: the fixed length, Line 1476  Returns: the fixed length,
1476  */  */
1477    
1478  static int  static int
1479  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)  find_fixedlength(uschar *code, BOOL utf8, BOOL atend, compile_data *cd)
1480  {  {
1481  int length = -1;  int length = -1;
1482    
# Line 1459  for (;;) Line 1493  for (;;)
1493    register int op = *cc;    register int op = *cc;
1494    switch (op)    switch (op)
1495      {      {
1496        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1497        OP_BRA (normal non-capturing bracket) because the other variants of these
1498        opcodes are all concerned with unlimited repeated groups, which of course
1499        are not of fixed length. They will cause a -1 response from the default
1500        case of this switch. */
1501    
1502      case OP_CBRA:      case OP_CBRA:
1503      case OP_BRA:      case OP_BRA:
1504      case OP_ONCE:      case OP_ONCE:
1505      case OP_COND:      case OP_COND:
1506      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), utf8, atend, cd);
1507      if (d < 0) return d;      if (d < 0) return d;
1508      branchlength += d;      branchlength += d;
1509      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1472  for (;;) Line 1512  for (;;)
1512    
1513      /* Reached end of a branch; if it's a ket it is the end of a nested      /* Reached end of a branch; if it's a ket it is the end of a nested
1514      call. If it's ALT it is an alternation in a nested call. If it is      call. If it's ALT it is an alternation in a nested call. If it is
1515      END it's the end of the outer call. All can be handled by the same code. */      END it's the end of the outer call. All can be handled by the same code.
1516        Note that we must not include the OP_KETRxxx opcodes here, because they
1517        all imply an unlimited repeat. */
1518    
1519      case OP_ALT:      case OP_ALT:
1520      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1521      case OP_END:      case OP_END:
1522      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1523        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
# Line 1495  for (;;) Line 1535  for (;;)
1535      cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */      cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1536      do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */      do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1537      if (cc > cs && cc < ce) return -1;                /* Recursion */      if (cc > cs && cc < ce) return -1;                /* Recursion */
1538      d = find_fixedlength(cs + 2, options, atend, cd);      d = find_fixedlength(cs + 2, utf8, atend, cd);
1539      if (d < 0) return d;      if (d < 0) return d;
1540      branchlength += d;      branchlength += d;
1541      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
# Line 1518  for (;;) Line 1558  for (;;)
1558      case OP_RREF:      case OP_RREF:
1559      case OP_NRREF:      case OP_NRREF:
1560      case OP_DEF:      case OP_DEF:
     case OP_OPT:  
1561      case OP_CALLOUT:      case OP_CALLOUT:
1562      case OP_SOD:      case OP_SOD:
1563      case OP_SOM:      case OP_SOM:
# Line 1526  for (;;) Line 1565  for (;;)
1565      case OP_EOD:      case OP_EOD:
1566      case OP_EODN:      case OP_EODN:
1567      case OP_CIRC:      case OP_CIRC:
1568        case OP_CIRCM:
1569      case OP_DOLL:      case OP_DOLL:
1570        case OP_DOLLM:
1571      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1572      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1573      cc += _pcre_OP_lengths[*cc];      cc += _pcre_OP_lengths[*cc];
# Line 1535  for (;;) Line 1576  for (;;)
1576      /* Handle literal characters */      /* Handle literal characters */
1577    
1578      case OP_CHAR:      case OP_CHAR:
1579      case OP_CHARNC:      case OP_CHARI:
1580      case OP_NOT:      case OP_NOT:
1581        case OP_NOTI:
1582      branchlength++;      branchlength++;
1583      cc += 2;      cc += 2;
1584  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1585      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       cc += _pcre_utf8_table4[cc[-1] & 0x3f];  
1586  #endif  #endif
1587      break;      break;
1588    
# Line 1552  for (;;) Line 1593  for (;;)
1593      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1594      cc += 4;      cc += 4;
1595  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1596      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)      if (utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       cc += _pcre_utf8_table4[cc[-1] & 0x3f];  
1597  #endif  #endif
1598      break;      break;
1599    
# Line 1655  _pcre_find_bracket(const uschar *code, B Line 1695  _pcre_find_bracket(const uschar *code, B
1695  for (;;)  for (;;)
1696    {    {
1697    register int c = *code;    register int c = *code;
1698    
1699    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1700    
1701    /* XCLASS is used for classes that cannot be represented just by a bit    /* XCLASS is used for classes that cannot be represented just by a bit
# Line 1673  for (;;) Line 1714  for (;;)
1714    
1715    /* Handle capturing bracket */    /* Handle capturing bracket */
1716    
1717    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
1718               c == OP_CBRAPOS || c == OP_SCBRAPOS)
1719      {      {
1720      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
1721      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
# Line 1711  for (;;) Line 1753  for (;;)
1753        case OP_MARK:        case OP_MARK:
1754        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
1755        case OP_SKIP_ARG:        case OP_SKIP_ARG:
       case OP_THEN_ARG:  
1756        code += code[1];        code += code[1];
1757        break;        break;
1758    
1759          case OP_THEN_ARG:
1760          code += code[1+LINK_SIZE];
1761          break;
1762        }        }
1763    
1764      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1728  for (;;) Line 1773  for (;;)
1773      if (utf8) switch(c)      if (utf8) switch(c)
1774        {        {
1775        case OP_CHAR:        case OP_CHAR:
1776        case OP_CHARNC:        case OP_CHARI:
1777        case OP_EXACT:        case OP_EXACT:
1778          case OP_EXACTI:
1779        case OP_UPTO:        case OP_UPTO:
1780          case OP_UPTOI:
1781        case OP_MINUPTO:        case OP_MINUPTO:
1782          case OP_MINUPTOI:
1783        case OP_POSUPTO:        case OP_POSUPTO:
1784          case OP_POSUPTOI:
1785        case OP_STAR:        case OP_STAR:
1786          case OP_STARI:
1787        case OP_MINSTAR:        case OP_MINSTAR:
1788          case OP_MINSTARI:
1789        case OP_POSSTAR:        case OP_POSSTAR:
1790          case OP_POSSTARI:
1791        case OP_PLUS:        case OP_PLUS:
1792          case OP_PLUSI:
1793        case OP_MINPLUS:        case OP_MINPLUS:
1794          case OP_MINPLUSI:
1795        case OP_POSPLUS:        case OP_POSPLUS:
1796          case OP_POSPLUSI:
1797        case OP_QUERY:        case OP_QUERY:
1798          case OP_QUERYI:
1799        case OP_MINQUERY:        case OP_MINQUERY:
1800          case OP_MINQUERYI:
1801        case OP_POSQUERY:        case OP_POSQUERY:
1802          case OP_POSQUERYI:
1803        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1804        break;        break;
1805        }        }
# Line 1814  for (;;) Line 1872  for (;;)
1872        case OP_MARK:        case OP_MARK:
1873        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
1874        case OP_SKIP_ARG:        case OP_SKIP_ARG:
       case OP_THEN_ARG:  
1875        code += code[1];        code += code[1];
1876        break;        break;
1877    
1878          case OP_THEN_ARG:
1879          code += code[1+LINK_SIZE];
1880          break;
1881        }        }
1882    
1883      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1831  for (;;) Line 1892  for (;;)
1892      if (utf8) switch(c)      if (utf8) switch(c)
1893        {        {
1894        case OP_CHAR:        case OP_CHAR:
1895        case OP_CHARNC:        case OP_CHARI:
1896        case OP_EXACT:        case OP_EXACT:
1897          case OP_EXACTI:
1898        case OP_UPTO:        case OP_UPTO:
1899          case OP_UPTOI:
1900        case OP_MINUPTO:        case OP_MINUPTO:
1901          case OP_MINUPTOI:
1902        case OP_POSUPTO:        case OP_POSUPTO:
1903          case OP_POSUPTOI:
1904        case OP_STAR:        case OP_STAR:
1905          case OP_STARI:
1906        case OP_MINSTAR:        case OP_MINSTAR:
1907          case OP_MINSTARI:
1908        case OP_POSSTAR:        case OP_POSSTAR:
1909          case OP_POSSTARI:
1910        case OP_PLUS:        case OP_PLUS:
1911          case OP_PLUSI:
1912        case OP_MINPLUS:        case OP_MINPLUS:
1913          case OP_MINPLUSI:
1914        case OP_POSPLUS:        case OP_POSPLUS:
1915          case OP_POSPLUSI:
1916        case OP_QUERY:        case OP_QUERY:
1917          case OP_QUERYI:
1918        case OP_MINQUERY:        case OP_MINQUERY:
1919          case OP_MINQUERYI:
1920        case OP_POSQUERY:        case OP_POSQUERY:
1921          case OP_POSQUERYI:
1922        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1923        break;        break;
1924        }        }
# Line 1883  could_be_empty_branch(const uschar *code Line 1957  could_be_empty_branch(const uschar *code
1957    compile_data *cd)    compile_data *cd)
1958  {  {
1959  register int c;  register int c;
1960  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], TRUE);
1961       code < endcode;       code < endcode;
1962       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], TRUE))
1963    {    {
1964    const uschar *ccode;    const uschar *ccode;
1965    
# Line 1901  for (code = first_significant_code(code Line 1975  for (code = first_significant_code(code
1975      continue;      continue;
1976      }      }
1977    
   /* Groups with zero repeats can of course be empty; skip them. */  
   
   if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)  
     {  
     code += _pcre_OP_lengths[c];  
     do code += GET(code, 1); while (*code == OP_ALT);  
     c = *code;  
     continue;  
     }  
   
1978    /* For a recursion/subroutine call, if its end has been reached, which    /* For a recursion/subroutine call, if its end has been reached, which
1979    implies a subroutine call, we can scan it. */    implies a backward reference subroutine call, we can scan it. If it's a
1980      forward reference subroutine call, we can't. To detect forward reference
1981      we have to scan up the list that is kept in the workspace. This function is
1982      called only when doing the real compile, not during the pre-compile that
1983      measures the size of the compiled pattern. */
1984    
1985    if (c == OP_RECURSE)    if (c == OP_RECURSE)
1986      {      {
1987      BOOL empty_branch = FALSE;      const uschar *scode;
1988      const uschar *scode = cd->start_code + GET(code, 1);      BOOL empty_branch;
1989    
1990        /* Test for forward reference */
1991    
1992        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
1993          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
1994    
1995        /* Not a forward reference, test for completed backward reference */
1996    
1997        empty_branch = FALSE;
1998        scode = cd->start_code + GET(code, 1);
1999      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2000    
2001        /* Completed backwards reference */
2002    
2003      do      do
2004        {        {
2005        if (could_be_empty_branch(scode, endcode, utf8, cd))        if (could_be_empty_branch(scode, endcode, utf8, cd))
# Line 1929  for (code = first_significant_code(code Line 2010  for (code = first_significant_code(code
2010        scode += GET(scode, 1);        scode += GET(scode, 1);
2011        }        }
2012      while (*scode == OP_ALT);      while (*scode == OP_ALT);
2013    
2014      if (!empty_branch) return FALSE;  /* All branches are non-empty */      if (!empty_branch) return FALSE;  /* All branches are non-empty */
2015      continue;      continue;
2016      }      }
2017    
2018      /* Groups with zero repeats can of course be empty; skip them. */
2019    
2020      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2021          c == OP_BRAPOSZERO)
2022        {
2023        code += _pcre_OP_lengths[c];
2024        do code += GET(code, 1); while (*code == OP_ALT);
2025        c = *code;
2026        continue;
2027        }
2028    
2029      /* A nested group that is already marked as "could be empty" can just be
2030      skipped. */
2031    
2032      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2033          c == OP_SCBRA || c == OP_SCBRAPOS)
2034        {
2035        do code += GET(code, 1); while (*code == OP_ALT);
2036        c = *code;
2037        continue;
2038        }
2039    
2040    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
2041    
2042    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_BRA  || c == OP_BRAPOS ||
2043          c == OP_CBRA || c == OP_CBRAPOS ||
2044          c == OP_ONCE || c == OP_COND)
2045      {      {
2046      BOOL empty_branch;      BOOL empty_branch;
2047      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
# Line 2022  for (code = first_significant_code(code Line 2128  for (code = first_significant_code(code
2128      case OP_ALLANY:      case OP_ALLANY:
2129      case OP_ANYBYTE:      case OP_ANYBYTE:
2130      case OP_CHAR:      case OP_CHAR:
2131      case OP_CHARNC:      case OP_CHARI:
2132      case OP_NOT:      case OP_NOT:
2133        case OP_NOTI:
2134      case OP_PLUS:      case OP_PLUS:
2135      case OP_MINPLUS:      case OP_MINPLUS:
2136      case OP_POSPLUS:      case OP_POSPLUS:
# Line 2063  for (code = first_significant_code(code Line 2170  for (code = first_significant_code(code
2170      case OP_KET:      case OP_KET:
2171      case OP_KETRMAX:      case OP_KETRMAX:
2172      case OP_KETRMIN:      case OP_KETRMIN:
2173        case OP_KETRPOS:
2174      case OP_ALT:      case OP_ALT:
2175      return TRUE;      return TRUE;
2176    
# Line 2071  for (code = first_significant_code(code Line 2179  for (code = first_significant_code(code
2179    
2180  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2181      case OP_STAR:      case OP_STAR:
2182        case OP_STARI:
2183      case OP_MINSTAR:      case OP_MINSTAR:
2184        case OP_MINSTARI:
2185      case OP_POSSTAR:      case OP_POSSTAR:
2186        case OP_POSSTARI:
2187      case OP_QUERY:      case OP_QUERY:
2188        case OP_QUERYI:
2189      case OP_MINQUERY:      case OP_MINQUERY:
2190        case OP_MINQUERYI:
2191      case OP_POSQUERY:      case OP_POSQUERY:
2192        case OP_POSQUERYI:
2193      if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];      if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2194      break;      break;
2195    
2196      case OP_UPTO:      case OP_UPTO:
2197        case OP_UPTOI:
2198      case OP_MINUPTO:      case OP_MINUPTO:
2199        case OP_MINUPTOI:
2200      case OP_POSUPTO:      case OP_POSUPTO:
2201        case OP_POSUPTOI:
2202      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2203      break;      break;
2204  #endif  #endif
# Line 2092  for (code = first_significant_code(code Line 2209  for (code = first_significant_code(code
2209      case OP_MARK:      case OP_MARK:
2210      case OP_PRUNE_ARG:      case OP_PRUNE_ARG:
2211      case OP_SKIP_ARG:      case OP_SKIP_ARG:
     case OP_THEN_ARG:  
2212      code += code[1];      code += code[1];
2213      break;      break;
2214    
2215        case OP_THEN_ARG:
2216        code += code[1+LINK_SIZE];
2217        break;
2218    
2219      /* None of the remaining opcodes are required to match a character. */      /* None of the remaining opcodes are required to match a character. */
2220    
2221      default:      default:
# Line 2116  return TRUE; Line 2236  return TRUE;
2236  the current branch of the current pattern to see if it could match the empty  the current branch of the current pattern to see if it could match the empty
2237  string. If it could, we must look outwards for branches at other levels,  string. If it could, we must look outwards for branches at other levels,
2238  stopping when we pass beyond the bracket which is the subject of the recursion.  stopping when we pass beyond the bracket which is the subject of the recursion.
2239    This function is called only during the real compile, not during the
2240    pre-compile.
2241    
2242  Arguments:  Arguments:
2243    code        points to start of the recursion    code        points to start of the recursion
# Line 2316  auto_callout(uschar *code, const uschar Line 2438  auto_callout(uschar *code, const uschar
2438  {  {
2439  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2440  *code++ = 255;  *code++ = 255;
2441  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2442  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2443  return code + 2*LINK_SIZE;  return code + 2*LINK_SIZE;
2444  }  }
2445    
# Line 2342  Returns: nothing Line 2464  Returns: nothing
2464  static void  static void
2465  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)
2466  {  {
2467  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2468  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2469  }  }
2470    
# Line 2392  for (++c; c <= d; c++) Line 2514  for (++c; c <= d; c++)
2514    
2515  return TRUE;  return TRUE;
2516  }  }
2517    
2518    
2519    
2520    /*************************************************
2521    *        Check a character and a property        *
2522    *************************************************/
2523    
2524    /* This function is called by check_auto_possessive() when a property item
2525    is adjacent to a fixed character.
2526    
2527    Arguments:
2528      c            the character
2529      ptype        the property type
2530      pdata        the data for the type
2531      negated      TRUE if it's a negated property (\P or \p{^)
2532    
2533    Returns:       TRUE if auto-possessifying is OK
2534    */
2535    
2536    static BOOL
2537    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2538    {
2539    const ucd_record *prop = GET_UCD(c);
2540    switch(ptype)
2541      {
2542      case PT_LAMP:
2543      return (prop->chartype == ucp_Lu ||
2544              prop->chartype == ucp_Ll ||
2545              prop->chartype == ucp_Lt) == negated;
2546    
2547      case PT_GC:
2548      return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;
2549    
2550      case PT_PC:
2551      return (pdata == prop->chartype) == negated;
2552    
2553      case PT_SC:
2554      return (pdata == prop->script) == negated;
2555    
2556      /* These are specials */
2557    
2558      case PT_ALNUM:
2559      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2560              _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;
2561    
2562      case PT_SPACE:    /* Perl space */
2563      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2564              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2565              == negated;
2566    
2567      case PT_PXSPACE:  /* POSIX space */
2568      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2569              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2570              c == CHAR_FF || c == CHAR_CR)
2571              == negated;
2572    
2573      case PT_WORD:
2574      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2575              _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2576              c == CHAR_UNDERSCORE) == negated;
2577      }
2578    return FALSE;
2579    }
2580  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2581    
2582    
# Line 2405  whether the next thing could possibly ma Line 2590  whether the next thing could possibly ma
2590  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
2591    
2592  Arguments:  Arguments:
2593    op_code       the repeated op code    previous      pointer to the repeated opcode
   this          data for this item, depends on the opcode  
2594    utf8          TRUE in UTF-8 mode    utf8          TRUE in UTF-8 mode
   utf8_char     used for utf8 character bytes, NULL if not relevant  
2595    ptr           next character in pattern    ptr           next character in pattern
2596    options       options bits    options       options bits
2597    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 2417  Returns: TRUE if possessifying is Line 2600  Returns: TRUE if possessifying is
2600  */  */
2601    
2602  static BOOL  static BOOL
2603  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
2604    const uschar *ptr, int options, compile_data *cd)    int options, compile_data *cd)
2605  {  {
2606  int next;  int c, next;
2607    int op_code = *previous++;
2608    
2609  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
2610    
# Line 2431  if ((options & PCRE_EXTENDED) != 0) Line 2615  if ((options & PCRE_EXTENDED) != 0)
2615      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2616      if (*ptr == CHAR_NUMBER_SIGN)      if (*ptr == CHAR_NUMBER_SIGN)
2617        {        {
2618        while (*(++ptr) != 0)        ptr++;
2619          while (*ptr != 0)
2620            {
2621          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2622            ptr++;
2623    #ifdef SUPPORT_UTF8
2624            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2625    #endif
2626            }
2627        }        }
2628      else break;      else break;
2629      }      }
# Line 2468  if ((options & PCRE_EXTENDED) != 0) Line 2659  if ((options & PCRE_EXTENDED) != 0)
2659      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2660      if (*ptr == CHAR_NUMBER_SIGN)      if (*ptr == CHAR_NUMBER_SIGN)
2661        {        {
2662        while (*(++ptr) != 0)        ptr++;
2663          while (*ptr != 0)
2664            {
2665          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2666            ptr++;
2667    #ifdef SUPPORT_UTF8
2668            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2669    #endif
2670            }
2671        }        }
2672      else break;      else break;
2673      }      }
# Line 2481  if (*ptr == CHAR_ASTERISK || *ptr == CHA Line 2679  if (*ptr == CHAR_ASTERISK || *ptr == CHA
2679    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2680      return FALSE;      return FALSE;
2681    
2682  /* Now compare the next item with the previous opcode. If the previous is a  /* Now compare the next item with the previous opcode. First, handle cases when
2683  positive single character match, "item" either contains the character or, if  the next item is a character. */
 "item" is greater than 127 in utf8 mode, the character's bytes are in  
 utf8_char. */  
   
   
 /* Handle cases when the next item is a character. */  
2684    
2685  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
2686    {    {
2687    case OP_CHAR:    case OP_CHAR:
2688  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2689    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2690  #else  #else
2691    (void)(utf8_char);  /* Keep compiler happy by referencing function argument */    c = *previous;
2692  #endif  #endif
2693    return item != next;    return c != next;
2694    
2695    /* For CHARNC (caseless character) we must check the other case. If we have    /* For CHARI (caseless character) we must check the other case. If we have
2696    Unicode property support, we can use it to test the other case of    Unicode property support, we can use it to test the other case of
2697    high-valued characters. */    high-valued characters. */
2698    
2699    case OP_CHARNC:    case OP_CHARI:
2700  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2701    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2702    #else
2703      c = *previous;
2704  #endif  #endif
2705    if (item == next) return FALSE;    if (c == next) return FALSE;
2706  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2707    if (utf8)    if (utf8)
2708      {      {
# Line 2518  if (next >= 0) switch(op_code) Line 2713  if (next >= 0) switch(op_code)
2713  #else  #else
2714      othercase = NOTACHAR;      othercase = NOTACHAR;
2715  #endif  #endif
2716      return (unsigned int)item != othercase;      return (unsigned int)c != othercase;
2717      }      }
2718    else    else
2719  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2720    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2721    
2722    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT and OP_NOTI, the data is always a single-byte character. These
2723      opcodes are not used for multi-byte characters, because they are coded using
2724      an XCLASS instead. */
2725    
2726    case OP_NOT:    case OP_NOT:
2727    if (item == next) return TRUE;    return (c = *previous) == next;
2728    if ((options & PCRE_CASELESS) == 0) return FALSE;  
2729      case OP_NOTI:
2730      if ((c = *previous) == next) return TRUE;
2731  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2732    if (utf8)    if (utf8)
2733      {      {
# Line 2539  if (next >= 0) switch(op_code) Line 2738  if (next >= 0) switch(op_code)
2738  #else  #else
2739      othercase = NOTACHAR;      othercase = NOTACHAR;
2740  #endif  #endif
2741      return (unsigned int)item == othercase;      return (unsigned int)c == othercase;
2742      }      }
2743    else    else
2744  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2745    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2746    
2747      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2748      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
2749    
2750    case OP_DIGIT:    case OP_DIGIT:
2751    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
# Line 2586  if (next >= 0) switch(op_code) Line 2788  if (next >= 0) switch(op_code)
2788      case 0x202f:      case 0x202f:
2789      case 0x205f:      case 0x205f:
2790      case 0x3000:      case 0x3000:
2791      return op_code != OP_HSPACE;      return op_code == OP_NOT_HSPACE;
2792      default:      default:
2793      return op_code == OP_HSPACE;      return op_code != OP_NOT_HSPACE;
2794      }      }
2795    
2796      case OP_ANYNL:
2797    case OP_VSPACE:    case OP_VSPACE:
2798    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
2799    switch(next)    switch(next)
# Line 2602  if (next >= 0) switch(op_code) Line 2805  if (next >= 0) switch(op_code)
2805      case 0x85:      case 0x85:
2806      case 0x2028:      case 0x2028:
2807      case 0x2029:      case 0x2029:
2808      return op_code != OP_VSPACE;      return op_code == OP_NOT_VSPACE;
2809      default:      default:
2810      return op_code == OP_VSPACE;      return op_code != OP_NOT_VSPACE;
2811      }      }
2812    
2813    #ifdef SUPPORT_UCP
2814      case OP_PROP:
2815      return check_char_prop(next, previous[0], previous[1], FALSE);
2816    
2817      case OP_NOTPROP:
2818      return check_char_prop(next, previous[0], previous[1], TRUE);
2819    #endif
2820    
2821    default:    default:
2822    return FALSE;    return FALSE;
2823    }    }
2824    
2825    
2826  /* Handle the case when the next item is \d, \s, etc. */  /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP
2827    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
2828    generated only when PCRE_UCP is *not* set, that is, when only ASCII
2829    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
2830    replaced by OP_PROP codes when PCRE_UCP is set. */
2831    
2832  switch(op_code)  switch(op_code)
2833    {    {
2834    case OP_CHAR:    case OP_CHAR:
2835    case OP_CHARNC:    case OP_CHARI:
2836  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2837    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2838    #else
2839      c = *previous;
2840  #endif  #endif
2841    switch(-next)    switch(-next)
2842      {      {
2843      case ESC_d:      case ESC_d:
2844      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
2845    
2846      case ESC_D:      case ESC_D:
2847      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
2848    
2849      case ESC_s:      case ESC_s:
2850      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
2851    
2852      case ESC_S:      case ESC_S:
2853      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
2854    
2855      case ESC_w:      case ESC_w:
2856      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
2857    
2858      case ESC_W:      case ESC_W:
2859      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
2860    
2861      case ESC_h:      case ESC_h:
2862      case ESC_H:      case ESC_H:
2863      switch(item)      switch(c)
2864        {        {
2865        case 0x09:        case 0x09:
2866        case 0x20:        case 0x20:
# Line 2671  switch(op_code) Line 2888  switch(op_code)
2888    
2889      case ESC_v:      case ESC_v:
2890      case ESC_V:      case ESC_V:
2891      switch(item)      switch(c)
2892        {        {
2893        case 0x0a:        case 0x0a:
2894        case 0x0b:        case 0x0b:
# Line 2685  switch(op_code) Line 2902  switch(op_code)
2902        return -next == ESC_v;        return -next == ESC_v;
2903        }        }
2904    
2905        /* When PCRE_UCP is set, these values get generated for \d etc. Find
2906        their substitutions and process them. The result will always be either
2907        -ESC_p or -ESC_P. Then fall through to process those values. */
2908    
2909    #ifdef SUPPORT_UCP
2910        case ESC_du:
2911        case ESC_DU:
2912        case ESC_wu:
2913        case ESC_WU:
2914        case ESC_su:
2915        case ESC_SU:
2916          {
2917          int temperrorcode = 0;
2918          ptr = substitutes[-next - ESC_DU];
2919          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
2920          if (temperrorcode != 0) return FALSE;
2921          ptr++;    /* For compatibility */
2922          }
2923        /* Fall through */
2924    
2925        case ESC_p:
2926        case ESC_P:
2927          {
2928          int ptype, pdata, errorcodeptr;
2929          BOOL negated;
2930    
2931          ptr--;      /* Make ptr point at the p or P */
2932          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
2933          if (ptype < 0) return FALSE;
2934          ptr++;      /* Point past the final curly ket */
2935    
2936          /* If the property item is optional, we have to give up. (When generated
2937          from \d etc by PCRE_UCP, this test will have been applied much earlier,
2938          to the original \d etc. At this point, ptr will point to a zero byte. */
2939    
2940          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2941            strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2942              return FALSE;
2943    
2944          /* Do the property check. */
2945    
2946          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
2947          }
2948    #endif
2949    
2950      default:      default:
2951      return FALSE;      return FALSE;
2952      }      }
2953    
2954      /* In principle, support for Unicode properties should be integrated here as
2955      well. It means re-organizing the above code so as to get hold of the property
2956      values before switching on the op-code. However, I wonder how many patterns
2957      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
2958      these op-codes are never generated.) */
2959    
2960    case OP_DIGIT:    case OP_DIGIT:
2961    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2962           next == -ESC_h || next == -ESC_v;           next == -ESC_h || next == -ESC_v || next == -ESC_R;
2963    
2964    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
2965    return next == -ESC_d;    return next == -ESC_d;
2966    
2967    case OP_WHITESPACE:    case OP_WHITESPACE:
2968    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
2969    
2970    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
2971    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2972    
2973    case OP_HSPACE:    case OP_HSPACE:
2974    return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
2975             next == -ESC_w || next == -ESC_v || next == -ESC_R;
2976    
2977    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
2978    return next == -ESC_h;    return next == -ESC_h;
2979    
2980    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
2981      case OP_ANYNL:
2982    case OP_VSPACE:    case OP_VSPACE:
2983    return next == -ESC_V || next == -ESC_d || next == -ESC_w;    return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2984    
2985    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
2986    return next == -ESC_v;    return next == -ESC_v || next == -ESC_R;
2987    
2988    case OP_WORDCHAR:    case OP_WORDCHAR:
2989    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
2990             next == -ESC_v || next == -ESC_R;
2991    
2992    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
2993    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2768  int greedy_default, greedy_non_default; Line 3039  int greedy_default, greedy_non_default;
3039  int firstbyte, reqbyte;  int firstbyte, reqbyte;
3040  int zeroreqbyte, zerofirstbyte;  int zeroreqbyte, zerofirstbyte;
3041  int req_caseopt, reqvary, tempreqvary;  int req_caseopt, reqvary, tempreqvary;
3042  int options = *optionsptr;  int options = *optionsptr;               /* May change dynamically */
3043  int after_manual_callout = 0;  int after_manual_callout = 0;
3044  int length_prevgroup = 0;  int length_prevgroup = 0;
3045  register int c;  register int c;
# Line 2786  uschar *previous_callout = NULL; Line 3057  uschar *previous_callout = NULL;
3057  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
3058  uschar classbits[32];  uschar classbits[32];
3059    
3060    /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3061    must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3062    dynamically as we process the pattern. */
3063    
3064  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3065  BOOL class_utf8;  BOOL class_utf8;
3066  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
# Line 2851  for (;; ptr++) Line 3126  for (;; ptr++)
3126    
3127    c = *ptr;    c = *ptr;
3128    
3129    /* If we are at the end of a nested substitution, revert to the outer level    /* If we are at the end of a nested substitution, revert to the outer level
3130    string. Nesting only happens one level deep. */    string. Nesting only happens one level deep. */
3131    
3132    if (c == 0 && nestptr != NULL)    if (c == 0 && nestptr != NULL)
# Line 2891  for (;; ptr++) Line 3166  for (;; ptr++)
3166        goto FAILED;        goto FAILED;
3167        }        }
3168    
3169      *lengthptr += code - last_code;      *lengthptr += (int)(code - last_code);
3170      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));
3171    
3172      /* If "previous" is set and it is not at the start of the work space, move      /* If "previous" is set and it is not at the start of the work space, move
# Line 2966  for (;; ptr++) Line 3241  for (;; ptr++)
3241      previous_callout = NULL;      previous_callout = NULL;
3242      }      }
3243    
3244    /* In extended mode, skip white space and comments */    /* In extended mode, skip white space and comments. */
3245    
3246    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3247      {      {
3248      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
3249      if (c == CHAR_NUMBER_SIGN)      if (c == CHAR_NUMBER_SIGN)
3250        {        {
3251        while (*(++ptr) != 0)        ptr++;
3252          while (*ptr != 0)
3253          {          {
3254          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3255            ptr++;
3256    #ifdef SUPPORT_UTF8
3257            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
3258    #endif
3259          }          }
3260        if (*ptr != 0) continue;        if (*ptr != 0) continue;
3261    
# Line 3009  for (;; ptr++) Line 3289  for (;; ptr++)
3289          *errorcodeptr = ERR20;          *errorcodeptr = ERR20;
3290          goto FAILED;          goto FAILED;
3291          }          }
3292        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += (int)(code - last_code);   /* To include callout length */
3293        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
3294        }        }
3295      return TRUE;      return TRUE;
# Line 3020  for (;; ptr++) Line 3300  for (;; ptr++)
3300      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3301    
3302      case CHAR_CIRCUMFLEX_ACCENT:      case CHAR_CIRCUMFLEX_ACCENT:
3303        previous = NULL;
3304      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3305        {        {
3306        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3307          *code++ = OP_CIRCM;
3308        }        }
3309      previous = NULL;      else *code++ = OP_CIRC;
     *code++ = OP_CIRC;  
3310      break;      break;
3311    
3312      case CHAR_DOLLAR_SIGN:      case CHAR_DOLLAR_SIGN:
3313      previous = NULL;      previous = NULL;
3314      *code++ = OP_DOLL;      *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
3315      break;      break;
3316    
3317      /* There can never be a first char if '.' is first, whatever happens about      /* There can never be a first char if '.' is first, whatever happens about
# Line 3158  for (;; ptr++) Line 3439  for (;; ptr++)
3439          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3440          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3441          }          }
3442    
3443        /* In the pre-compile phase, accumulate the length of any UTF-8 extra        /* In the pre-compile phase, accumulate the length of any UTF-8 extra
3444        data and reset the pointer. This is so that very large classes that        data and reset the pointer. This is so that very large classes that
3445        contain a zillion UTF-8 characters no longer overwrite the work space        contain a zillion UTF-8 characters no longer overwrite the work space
# Line 3214  for (;; ptr++) Line 3495  for (;; ptr++)
3495            ptr++;            ptr++;
3496            }            }
3497    
3498          posix_class = check_posix_name(ptr, tempptr - ptr);          posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3499          if (posix_class < 0)          if (posix_class < 0)
3500            {            {
3501            *errorcodeptr = ERR30;            *errorcodeptr = ERR30;
# Line 3227  for (;; ptr++) Line 3508  for (;; ptr++)
3508    
3509          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3510            posix_class = 0;            posix_class = 0;
3511    
3512          /* 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
3513          different escape sequences that use Unicode properties. */          different escape sequences that use Unicode properties. */
3514    
3515  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3516          if ((options & PCRE_UCP) != 0)          if ((options & PCRE_UCP) != 0)
3517            {            {
3518            int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);            int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3519            if (posix_substitutes[pc] != NULL)            if (posix_substitutes[pc] != NULL)
3520              {              {
3521              nestptr = tempptr + 1;              nestptr = tempptr + 1;
3522              ptr = posix_substitutes[pc] - 1;              ptr = posix_substitutes[pc] - 1;
3523              continue;              continue;
3524              }              }
3525            }            }
3526  #endif  #endif
3527          /* In the non-UCP case, we build the bit map for the POSIX class in a          /* In the non-UCP case, we build the bit map for the POSIX class in a
3528          chunk of local store because we may be adding and subtracting from it,          chunk of local store because we may be adding and subtracting from it,
3529          and we don't want to subtract bits that may be in the main map already.          and we don't want to subtract bits that may be in the main map already.
# Line 3329  for (;; ptr++) Line 3610  for (;; ptr++)
3610              case ESC_SU:              case ESC_SU:
3611              nestptr = ptr;              nestptr = ptr;
3612              ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */              ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3613              class_charcount -= 2;                /* Undo! */              class_charcount -= 2;                /* Undo! */
3614              continue;              continue;
3615  #endif  #endif
3616              case ESC_d:              case ESC_d:
# Line 3350  for (;; ptr++) Line 3631  for (;; ptr++)
3631              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3632              continue;              continue;
3633    
3634                /* Perl 5.004 onwards omits VT from \s, but we must preserve it
3635                if it was previously set by something earlier in the character
3636                class. */
3637    
3638              case ESC_s:              case ESC_s:
3639              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];              classbits[0] |= cbits[cbit_space];
3640              classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= cbits[cbit_space+1] & ~0x08;
3641                for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3642              continue;              continue;
3643    
3644              case ESC_S:              case ESC_S:
# Line 3771  for (;; ptr++) Line 4057  for (;; ptr++)
4057    
4058      In UTF-8 mode, we can optimize the negative case only if there were no      In UTF-8 mode, we can optimize the negative case only if there were no
4059      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
4060      operate on single-bytes only. This is an historical hangover. Maybe one day      operate on single-bytes characters only. This is an historical hangover.
4061      we can tidy these opcodes to handle multi-byte characters.      Maybe one day we can tidy these opcodes to handle multi-byte characters.
4062    
4063      The optimization throws away the bit map. We turn the item into a      The optimization throws away the bit map. We turn the item into a
4064      1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note      1-character OP_CHAR[I] if it's positive, or OP_NOT[I] if it's negative.
4065      that OP_NOT does not support multibyte characters. In the positive case, it      Note that OP_NOT[I] does not support multibyte characters. In the positive
4066      can cause firstbyte to be set. Otherwise, there can be no first char if      case, it can cause firstbyte to be set. Otherwise, there can be no first
4067      this item is first, whatever repeat count may follow. In the case of      char if this item is first, whatever repeat count may follow. In the case
4068      reqbyte, save the previous value for reinstating. */      of reqbyte, save the previous value for reinstating. */
4069    
4070  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4071      if (class_charcount == 1 && !class_utf8 &&      if (class_charcount == 1 && !class_utf8 &&
4072        (!utf8 || !negate_class || class_lastchar < 128))        (!utf8 || !negate_class || class_lastchar < 128))
# Line 3790  for (;; ptr++) Line 4076  for (;; ptr++)
4076        {        {
4077        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
4078    
4079        /* The OP_NOT opcode works on one-byte characters only. */        /* The OP_NOT[I] opcodes work on one-byte characters only. */
4080    
4081        if (negate_class)        if (negate_class)
4082          {          {
4083          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
4084          zerofirstbyte = firstbyte;          zerofirstbyte = firstbyte;
4085          *code++ = OP_NOT;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
4086          *code++ = class_lastchar;          *code++ = class_lastchar;
4087          break;          break;
4088          }          }
# Line 3860  for (;; ptr++) Line 4146  for (;; ptr++)
4146        }        }
4147  #endif  #endif
4148    
4149      /* If there are no characters > 255, or they are all to be included or      /* If there are no characters > 255, or they are all to be included or
4150      excluded, set the opcode to OP_CLASS or OP_NCLASS, depending on whether the      excluded, set the opcode to OP_CLASS or OP_NCLASS, depending on whether the
4151      whole class was negated and whether there were negative specials such as \S      whole class was negated and whether there were negative specials such as \S
4152      (non-UCP) in the class. Then copy the 32-byte map into the code vector,      (non-UCP) in the class. Then copy the 32-byte map into the code vector,
# Line 3947  for (;; ptr++) Line 4233  for (;; ptr++)
4233        ptr++;        ptr++;
4234        }        }
4235      else repeat_type = greedy_default;      else repeat_type = greedy_default;
4236    
4237        /* If previous was a recursion call, wrap it in atomic brackets so that
4238        previous becomes the atomic group. All recursions were so wrapped in the
4239        past, but it no longer happens for non-repeated recursions. In fact, the
4240        repeated ones could be re-implemented independently so as not to need this,
4241        but for the moment we rely on the code for repeating groups. */
4242    
4243        if (*previous == OP_RECURSE)
4244          {
4245          memmove(previous + 1 + LINK_SIZE, previous, 1 + LINK_SIZE);
4246          *previous = OP_ONCE;
4247          PUT(previous, 1, 2 + 2*LINK_SIZE);
4248          previous[2 + 2*LINK_SIZE] = OP_KET;
4249          PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
4250          code += 2 + 2 * LINK_SIZE;
4251          length_prevgroup = 3 + 3*LINK_SIZE;
4252    
4253          /* When actually compiling, we need to check whether this was a forward
4254          reference, and if so, adjust the offset. */
4255    
4256          if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
4257            {
4258            int offset = GET(cd->hwm, -LINK_SIZE);
4259            if (offset == previous + 1 - cd->start_code)
4260              PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
4261            }
4262          }
4263    
4264        /* Now handle repetition for the different types of item. */
4265    
4266      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
4267      repeat item instead. If a char item has a minumum of more than one, ensure      repeat item instead. If a char item has a minumum of more than one, ensure
# Line 3954  for (;; ptr++) Line 4269  for (;; ptr++)
4269      the first thing in a branch because the x will have gone into firstbyte      the first thing in a branch because the x will have gone into firstbyte
4270      instead.  */      instead.  */
4271    
4272      if (*previous == OP_CHAR || *previous == OP_CHARNC)      if (*previous == OP_CHAR || *previous == OP_CHARI)
4273        {        {
4274          op_type = (*previous == OP_CHAR)? 0 : OP_STARI - OP_STAR;
4275    
4276        /* Deal with UTF-8 characters that take up more than one byte. It's        /* Deal with UTF-8 characters that take up more than one byte. It's
4277        easier to write this out separately than try to macrify it. Use c to        easier to write this out separately than try to macrify it. Use c to
4278        hold the length of the character in bytes, plus 0x80 to flag that it's a        hold the length of the character in bytes, plus 0x80 to flag that it's a
# Line 3988  for (;; ptr++) Line 4305  for (;; ptr++)
4305    
4306        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4307            repeat_max < 0 &&            repeat_max < 0 &&
4308            check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
             options, cd))  
4309          {          {
4310          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4311          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 4001  for (;; ptr++) Line 4317  for (;; ptr++)
4317      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
4318      one of the special opcodes, replacing it. The code is shared with single-      one of the special opcodes, replacing it. The code is shared with single-
4319      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
4320      repeat_type. We can also test for auto-possessification. OP_NOT is      repeat_type. We can also test for auto-possessification. OP_NOT and OP_NOTI
4321      currently used only for single-byte chars. */      are currently used only for single-byte chars. */
4322    
4323      else if (*previous == OP_NOT)      else if (*previous == OP_NOT || *previous == OP_NOTI)
4324        {        {
4325        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = ((*previous == OP_NOT)? OP_NOTSTAR : OP_NOTSTARI) - OP_STAR;
4326        c = previous[1];        c = previous[1];
4327        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4328            repeat_max < 0 &&            repeat_max < 0 &&
4329            check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4330          {          {
4331          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4332          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 4034  for (;; ptr++) Line 4350  for (;; ptr++)
4350    
4351        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4352            repeat_max < 0 &&            repeat_max < 0 &&
4353            check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4354          {          {
4355          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4356          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 4203  for (;; ptr++) Line 4519  for (;; ptr++)
4519  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4520               *previous == OP_XCLASS ||               *previous == OP_XCLASS ||
4521  #endif  #endif
4522               *previous == OP_REF)               *previous == OP_REF ||
4523                 *previous == OP_REFI)
4524        {        {
4525        if (repeat_max == 0)        if (repeat_max == 0)
4526          {          {
# Line 4237  for (;; ptr++) Line 4554  for (;; ptr++)
4554        }        }
4555    
4556      /* If previous was a bracket group, we may have to replicate it in certain      /* If previous was a bracket group, we may have to replicate it in certain
4557      cases. */      cases. Note that at this point we can encounter only the "basic" BRA and
4558        KET opcodes, as this is the place where they get converted into the more
4559        special varieties. */
4560    
4561      else if (*previous == OP_BRA  || *previous == OP_CBRA ||      else if (*previous == OP_BRA  || *previous == OP_CBRA ||
4562               *previous == OP_ONCE || *previous == OP_COND)               *previous == OP_ONCE || *previous == OP_COND)
4563        {        {
4564        register int i;        register int i;
4565        int ketoffset = 0;        int len = (int)(code - previous);
       int len = code - previous;  
4566        uschar *bralink = NULL;        uschar *bralink = NULL;
4567          uschar *brazeroptr = NULL;
4568    
4569        /* Repeating a DEFINE group is pointless */        /* Repeating a DEFINE group is pointless */
4570    
4571        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
# Line 4255  for (;; ptr++) Line 4574  for (;; ptr++)
4574          goto FAILED;          goto FAILED;
4575          }          }
4576    
       /* If the maximum repeat count is unlimited, find the end of the bracket  
       by scanning through from the start, and compute the offset back to it  
       from the current code pointer. There may be an OP_OPT setting following  
       the final KET, so we can't find the end just by going back from the code  
       pointer. */  
   
       if (repeat_max == -1)  
         {  
         register uschar *ket = previous;  
         do ket += GET(ket, 1); while (*ket != OP_KET);  
         ketoffset = code - ket;  
         }  
   
4577        /* The case of a zero minimum is special because of the need to stick        /* The case of a zero minimum is special because of the need to stick
4578        OP_BRAZERO in front of it, and because the group appears once in the        OP_BRAZERO in front of it, and because the group appears once in the
4579        data, whereas in other cases it appears the minimum number of times. For        data, whereas in other cases it appears the minimum number of times. For
# Line 4309  for (;; ptr++) Line 4615  for (;; ptr++)
4615              *previous++ = OP_SKIPZERO;              *previous++ = OP_SKIPZERO;
4616              goto END_REPEAT;              goto END_REPEAT;
4617              }              }
4618              brazeroptr = previous;    /* Save for possessive optimizing */
4619            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4620            }            }
4621    
# Line 4333  for (;; ptr++) Line 4640  for (;; ptr++)
4640            /* We chain together the bracket offset fields that have to be            /* We chain together the bracket offset fields that have to be
4641            filled in later when the ends of the brackets are reached. */            filled in later when the ends of the brackets are reached. */
4642    
4643            offset = (bralink == NULL)? 0 : previous - bralink;            offset = (bralink == NULL)? 0 : (int)(previous - bralink);
4644            bralink = previous;            bralink = previous;
4645            PUTINC(previous, 0, offset);            PUTINC(previous, 0, offset);
4646            }            }
# Line 4442  for (;; ptr++) Line 4749  for (;; ptr++)
4749              {              {
4750              int offset;              int offset;
4751              *code++ = OP_BRA;              *code++ = OP_BRA;
4752              offset = (bralink == NULL)? 0 : code - bralink;              offset = (bralink == NULL)? 0 : (int)(code - bralink);
4753              bralink = code;              bralink = code;
4754              PUTINC(code, 0, offset);              PUTINC(code, 0, offset);
4755              }              }
# Line 4463  for (;; ptr++) Line 4770  for (;; ptr++)
4770          while (bralink != NULL)          while (bralink != NULL)
4771            {            {
4772            int oldlinkoffset;            int oldlinkoffset;
4773            int offset = code - bralink + 1;            int offset = (int)(code - bralink + 1);
4774            uschar *bra = code - offset;            uschar *bra = code - offset;
4775            oldlinkoffset = GET(bra, 1);            oldlinkoffset = GET(bra, 1);
4776            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;
# Line 4473  for (;; ptr++) Line 4780  for (;; ptr++)
4780            }            }
4781          }          }
4782    
4783        /* If the maximum is unlimited, set a repeater in the final copy. We        /* If the maximum is unlimited, set a repeater in the final copy. For
4784        can't just offset backwards from the current code point, because we        ONCE brackets, that's all we need to do. However, possessively repeated
4785        don't know if there's been an options resetting after the ket. The        ONCE brackets can be converted into non-capturing brackets, as the
4786        correct offset was computed above.        behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to
4787          deal with possessive ONCEs specially.
4788    
4789          Otherwise, if the quantifier was possessive, we convert the BRA code to
4790          the POS form, and the KET code to KETRPOS. (It turns out to be convenient
4791          at runtime to detect this kind of subpattern at both the start and at the
4792          end.) The use of special opcodes makes it possible to reduce greatly the
4793          stack usage in pcre_exec(). If the group is preceded by OP_BRAZERO,
4794          convert this to OP_BRAPOSZERO. Then cancel the possessive flag so that
4795          the default action below, of wrapping everything inside atomic brackets,
4796          does not happen.
4797    
4798        Then, when we are doing the actual compile phase, check to see whether        Then, when we are doing the actual compile phase, check to see whether
4799        this group is a non-atomic one that could match an empty string. If so,        this group is one that could match an empty string. If so, convert the
4800        convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so        initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so that runtime
4801        that runtime checking can be done. [This check is also applied to        checking can be done. [This check is also applied to ONCE groups at
4802        atomic groups at runtime, but in a different way.] */        runtime, but in a different way.] */
4803    
4804        else        else
4805          {          {
4806          uschar *ketcode = code - ketoffset;          uschar *ketcode = code - 1 - LINK_SIZE;
4807          uschar *bracode = ketcode - GET(ketcode, 1);          uschar *bracode = ketcode - GET(ketcode, 1);
4808          *ketcode = OP_KETRMAX + repeat_type;  
4809          if (lengthptr == NULL && *bracode != OP_ONCE)          if (*bracode == OP_ONCE && possessive_quantifier) *bracode = OP_BRA;
4810            if (*bracode == OP_ONCE)
4811              *ketcode = OP_KETRMAX + repeat_type;
4812            else
4813            {            {
4814            uschar *scode = bracode;            if (possessive_quantifier)
4815            do              {
4816                *bracode += 1;                   /* Switch to xxxPOS opcodes */
4817                *ketcode = OP_KETRPOS;
4818                if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
4819                possessive_quantifier = FALSE;
4820                }
4821              else *ketcode = OP_KETRMAX + repeat_type;
4822    
4823              if (lengthptr == NULL)
4824              {              {
4825              if (could_be_empty_branch(scode, ketcode, utf8, cd))              uschar *scode = bracode;
4826                do
4827                {                {
4828                *bracode += OP_SBRA - OP_BRA;                if (could_be_empty_branch(scode, ketcode, utf8, cd))
4829                break;                  {
4830                    *bracode += OP_SBRA - OP_BRA;
4831                    break;
4832                    }
4833                  scode += GET(scode, 1);
4834                }                }
4835              scode += GET(scode, 1);              while (*scode == OP_ALT);
4836              }              }
           while (*scode == OP_ALT);  
4837            }            }
4838          }          }
4839        }        }
# Line 4522  for (;; ptr++) Line 4854  for (;; ptr++)
4854        }        }
4855    
4856      /* If the character following a repeat is '+', or if certain optimization      /* If the character following a repeat is '+', or if certain optimization
4857      tests above succeeded, possessive_quantifier is TRUE. For some of the      tests above succeeded, possessive_quantifier is TRUE. For some opcodes,
4858      simpler opcodes, there is an special alternative opcode for this. For      there are special alternative opcodes for this case. For anything else, we
4859      anything else, we wrap the entire repeated item inside OP_ONCE brackets.      wrap the entire repeated item inside OP_ONCE brackets. Logically, the '+'
4860      The '+' notation is just syntactic sugar, taken from Sun's Java package,      notation is just syntactic sugar, taken from Sun's Java package, but the
4861      but the special opcodes can optimize it a bit. The repeated item starts at      special opcodes can optimize it.
4862      tempcode, not at previous, which might be the first part of a string whose  
4863      (former) last char we repeated.      Possessively repeated subpatterns have already been handled in the code
4864        just above, so possessive_quantifier is always FALSE for them at this
4865        stage.
4866    
4867        Note that the repeated item starts at tempcode, not at previous, which
4868        might be the first part of a string whose (former) last char we repeated.
4869    
4870      Possessifying an 'exact' quantifier has no effect, so we can ignore it. But      Possessifying an 'exact' quantifier has no effect, so we can ignore it. But
4871      an 'upto' may follow. We skip over an 'exact' item, and then test the      an 'upto' may follow. We skip over an 'exact' item, and then test the
# Line 4551  for (;; ptr++) Line 4888  for (;; ptr++)
4888  #endif  #endif
4889          }          }
4890    
4891        len = code - tempcode;        len = (int)(code - tempcode);
4892        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4893          {          {
4894          case OP_STAR:  *tempcode = OP_POSSTAR; break;          case OP_STAR:  *tempcode = OP_POSSTAR; break;
# Line 4559  for (;; ptr++) Line 4896  for (;; ptr++)
4896          case OP_QUERY: *tempcode = OP_POSQUERY; break;          case OP_QUERY: *tempcode = OP_POSQUERY; break;
4897          case OP_UPTO:  *tempcode = OP_POSUPTO; break;          case OP_UPTO:  *tempcode = OP_POSUPTO; break;
4898    
4899          case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;          case OP_STARI:  *tempcode = OP_POSSTARI; break;
4900          case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;          case OP_PLUSI:  *tempcode = OP_POSPLUSI; break;
4901          case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;          case OP_QUERYI: *tempcode = OP_POSQUERYI; break;
4902          case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;          case OP_UPTOI:  *tempcode = OP_POSUPTOI; break;
4903    
4904          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;
4905          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;
4906          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4907          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4908    
4909            case OP_NOTSTARI:  *tempcode = OP_NOTPOSSTARI; break;
4910            case OP_NOTPLUSI:  *tempcode = OP_NOTPOSPLUSI; break;
4911            case OP_NOTQUERYI: *tempcode = OP_NOTPOSQUERYI; break;
4912            case OP_NOTUPTOI:  *tempcode = OP_NOTPOSUPTOI; break;
4913    
4914            case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;
4915            case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;
4916            case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
4917            case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;
4918    
4919          /* Because we are moving code along, we must ensure that any          /* Because we are moving code along, we must ensure that any
4920          pending recursive references are updated. */          pending recursive references are updated. */
4921    
# Line 4620  for (;; ptr++) Line 4967  for (;; ptr++)
4967        const uschar *arg = NULL;        const uschar *arg = NULL;
4968        previous = NULL;        previous = NULL;
4969        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
4970        namelen = ptr - name;        namelen = (int)(ptr - name);
4971    
4972        if (*ptr == CHAR_COLON)        if (*ptr == CHAR_COLON)
4973          {          {
4974          arg = ++ptr;          arg = ++ptr;
4975          while ((cd->ctypes[*ptr] & (ctype_letter|ctype_digit)) != 0          while ((cd->ctypes[*ptr] & (ctype_letter|ctype_digit)) != 0
4976            || *ptr == '_') ptr++;            || *ptr == '_') ptr++;
4977          arglen = ptr - arg;          arglen = (int)(ptr - arg);
4978          }          }
4979    
4980        if (*ptr != CHAR_RIGHT_PARENTHESIS)        if (*ptr != CHAR_RIGHT_PARENTHESIS)
# Line 4643  for (;; ptr++) Line 4990  for (;; ptr++)
4990          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
4991              strncmp((char *)name, vn, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
4992            {            {
4993            /* Check for open captures before ACCEPT */            /* Check for open captures before ACCEPT and convert it to
4994              ASSERT_ACCEPT if in an assertion. */
4995    
4996            if (verbs[i].op == OP_ACCEPT)            if (verbs[i].op == OP_ACCEPT)
4997              {              {
4998              open_capitem *oc;              open_capitem *oc;
4999                if (arglen != 0)
5000                  {
5001                  *errorcodeptr = ERR59;
5002                  goto FAILED;
5003                  }
5004              cd->had_accept = TRUE;              cd->had_accept = TRUE;
5005              for (oc = cd->open_caps; oc != NULL; oc = oc->next)              for (oc = cd->open_caps; oc != NULL; oc = oc->next)
5006                {                {
5007                *code++ = OP_CLOSE;                *code++ = OP_CLOSE;
5008                PUT2INC(code, 0, oc->number);                PUT2INC(code, 0, oc->number);
5009                }                }
5010                *code++ = (cd->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT;
5011              }              }
5012    
5013            /* Handle the cases with/without an argument */            /* Handle other cases with/without an argument */
5014    
5015            if (arglen == 0)            else if (arglen == 0)
5016              {              {
5017              if (verbs[i].op < 0)   /* Argument is mandatory */              if (verbs[i].op < 0)   /* Argument is mandatory */
5018                {                {
5019                *errorcodeptr = ERR66;                *errorcodeptr = ERR66;
5020                goto FAILED;                goto FAILED;
5021                }                }
5022              *code++ = verbs[i].op;              *code = verbs[i].op;
5023                if (*code++ == OP_THEN)
5024                  {
5025                  PUT(code, 0, code - bcptr->current_branch - 1);
5026                  code += LINK_SIZE;
5027                  }
5028              }              }
5029    
5030            else            else
# Line 4675  for (;; ptr++) Line 5034  for (;; ptr++)
5034                *errorcodeptr = ERR59;                *errorcodeptr = ERR59;
5035                goto FAILED;                goto FAILED;
5036                }                }
5037              *code++ = verbs[i].op_arg;              *code = verbs[i].op_arg;
5038                if (*code++ == OP_THEN_ARG)
5039                  {
5040                  PUT(code, 0, code - bcptr->current_branch - 1);
5041                  code += LINK_SIZE;
5042                  }
5043              *code++ = arglen;              *code++ = arglen;
5044              memcpy(code, arg, arglen);              memcpy(code, arg, arglen);
5045              code += arglen;              code += arglen;
# Line 4808  for (;; ptr++) Line 5172  for (;; ptr++)
5172                recno * 10 + *ptr - CHAR_0 : -1;                recno * 10 + *ptr - CHAR_0 : -1;
5173            ptr++;            ptr++;
5174            }            }
5175          namelen = ptr - name;          namelen = (int)(ptr - name);
5176    
5177          if ((terminator > 0 && *ptr++ != terminator) ||          if ((terminator > 0 && *ptr++ != terminator) ||
5178              *ptr++ != CHAR_RIGHT_PARENTHESIS)              *ptr++ != CHAR_RIGHT_PARENTHESIS)
# Line 4869  for (;; ptr++) Line 5233  for (;; ptr++)
5233          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
5234    
5235          else if ((i = find_parens(cd, name, namelen,          else if ((i = find_parens(cd, name, namelen,
5236                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0, utf8)) > 0)
5237            {            {
5238            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
5239            code[1+LINK_SIZE]++;            code[1+LINK_SIZE]++;
# Line 4937  for (;; ptr++) Line 5301  for (;; ptr++)
5301          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5302          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
5303          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
5304            cd->assert_depth += 1;
5305          ptr++;          ptr++;
5306          break;          break;
5307    
# Line 4951  for (;; ptr++) Line 5316  for (;; ptr++)
5316            continue;            continue;
5317            }            }
5318          bravalue = OP_ASSERT_NOT;          bravalue = OP_ASSERT_NOT;
5319            cd->assert_depth += 1;
5320          break;          break;
5321    
5322    
# Line 4960  for (;; ptr++) Line 5326  for (;; ptr++)
5326            {            {
5327            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
5328            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
5329              cd->assert_depth += 1;
5330            ptr += 2;            ptr += 2;
5331            break;            break;
5332    
5333            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
5334            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
5335              cd->assert_depth += 1;
5336            ptr += 2;            ptr += 2;
5337            break;            break;
5338    
# Line 5004  for (;; ptr++) Line 5372  for (;; ptr++)
5372              goto FAILED;              goto FAILED;
5373              }              }
5374            *code++ = n;            *code++ = n;
5375            PUT(code, 0, ptr - cd->start_pattern + 1);  /* Pattern offset */            PUT(code, 0, (int)(ptr - cd->start_pattern + 1)); /* Pattern offset */
5376            PUT(code, LINK_SIZE, 0);                    /* Default length */            PUT(code, LINK_SIZE, 0);                          /* Default length */
5377            code += 2 * LINK_SIZE;            code += 2 * LINK_SIZE;
5378            }            }
5379          previous = NULL;          previous = NULL;
# Line 5038  for (;; ptr++) Line 5406  for (;; ptr++)
5406            name = ++ptr;            name = ++ptr;
5407    
5408            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
5409            namelen = ptr - name;            namelen = (int)(ptr - name);
5410    
5411            /* In the pre-compile phase, just do a syntax check. */            /* In the pre-compile phase, just do a syntax check. */
5412    
# Line 5168  for (;; ptr++) Line 5536  for (;; ptr++)
5536          NAMED_REF_OR_RECURSE:          NAMED_REF_OR_RECURSE:
5537          name = ++ptr;          name = ++ptr;
5538          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
5539          namelen = ptr - name;          namelen = (int)(ptr - name);
5540    
5541          /* In the pre-compile phase, do a syntax check and set a dummy          /* In the pre-compile phase, do a syntax check. We used to just set
5542          reference number. */          a dummy reference number, because it was not used in the first pass.
5543            However, with the change of recursive back references to be atomic,
5544            we have to look for the number so that this state can be identified, as
5545            otherwise the incorrect length is computed. If it's not a backwards
5546            reference, the dummy number will do. */
5547    
5548          if (lengthptr != NULL)          if (lengthptr != NULL)
5549            {            {
5550              const uschar *temp;
5551    
5552            if (namelen == 0)            if (namelen == 0)
5553              {              {
5554              *errorcodeptr = ERR62;              *errorcodeptr = ERR62;
# Line 5190  for (;; ptr++) Line 5564  for (;; ptr++)
5564              *errorcodeptr = ERR48;              *errorcodeptr = ERR48;
5565              goto FAILED;              goto FAILED;
5566              }              }
5567            recno = 0;  
5568              /* The name table does not exist in the first pass, so we cannot
5569              do a simple search as in the code below. Instead, we have to scan the
5570              pattern to find the number. It is important that we scan it only as
5571              far as we have got because the syntax of named subpatterns has not
5572              been checked for the rest of the pattern, and find_parens() assumes
5573              correct syntax. In any case, it's a waste of resources to scan
5574              further. We stop the scan at the current point by temporarily
5575              adjusting the value of cd->endpattern. */
5576    
5577              temp = cd->end_pattern;
5578              cd->end_pattern = ptr;
5579              recno = find_parens(cd, name, namelen,
5580                (options & PCRE_EXTENDED) != 0, utf8);
5581              cd->end_pattern = temp;
5582              if (recno < 0) recno = 0;    /* Forward ref; set dummy number */
5583            }            }
5584    
5585          /* In the real compile, seek the name in the table. We check the name          /* In the real compile, seek the name in the table. We check the name
# Line 5215  for (;; ptr++) Line 5604  for (;; ptr++)
5604              }              }
5605            else if ((recno =                /* Forward back reference */            else if ((recno =                /* Forward back reference */
5606                      find_parens(cd, name, namelen,                      find_parens(cd, name, namelen,
5607                        (options & PCRE_EXTENDED) != 0)) <= 0)                        (options & PCRE_EXTENDED) != 0, utf8)) <= 0)
5608              {              {
5609              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
5610              goto FAILED;              goto FAILED;
# Line 5326  for (;; ptr++) Line 5715  for (;; ptr++)
5715              if (called == NULL)              if (called == NULL)
5716                {                {
5717                if (find_parens(cd, NULL, recno,                if (find_parens(cd, NULL, recno,
5718                      (options & PCRE_EXTENDED) != 0) < 0)                      (options & PCRE_EXTENDED) != 0, utf8) < 0)
5719                  {                  {
5720                  *errorcodeptr = ERR15;                  *errorcodeptr = ERR15;
5721                  goto FAILED;                  goto FAILED;
# Line 5334  for (;; ptr++) Line 5723  for (;; ptr++)
5723    
5724                /* Fudge the value of "called" so that when it is inserted as an                /* Fudge the value of "called" so that when it is inserted as an
5725                offset below, what it actually inserted is the reference number                offset below, what it actually inserted is the reference number
5726                of the group. */                of the group. Then remember the forward reference. */
5727    
5728                called = cd->start_code + recno;                called = cd->start_code + recno;
5729                PUTINC(cd->hwm, 0, code + 2 + LINK_SIZE - cd->start_code);                PUTINC(cd->hwm, 0, (int)(code + 1 - cd->start_code));
5730                }                }
5731    
5732              /* If not a forward reference, and the subpattern is still open,              /* If not a forward reference, and the subpattern is still open,
# Line 5352  for (;; ptr++) Line 5741  for (;; ptr++)
5741                }                }
5742              }              }
5743    
5744            /* Insert the recursion/subroutine item, automatically wrapped inside            /* Insert the recursion/subroutine item. */
5745            "once" brackets. Set up a "previous group" length so that a  
           subsequent quantifier will work. */  
   
           *code = OP_ONCE;  
           PUT(code, 1, 2 + 2*LINK_SIZE);  
           code += 1 + LINK_SIZE;  
   
5746            *code = OP_RECURSE;            *code = OP_RECURSE;
5747            PUT(code, 1, called - cd->start_code);            PUT(code, 1, (int)(called - cd->start_code));
5748            code += 1 + LINK_SIZE;            code += 1 + LINK_SIZE;
   
           *code = OP_KET;  
           PUT(code, 1, 2 + 2*LINK_SIZE);  
           code += 1 + LINK_SIZE;  
   
           length_prevgroup = 3 + 3*LINK_SIZE;  
5749            }            }
5750    
5751          /* Can't determine a first byte now */          /* Can't determine a first byte now */
# Line 5429  for (;; ptr++) Line 5806  for (;; ptr++)
5806          is necessary to ensure we correctly detect the start of the pattern in          is necessary to ensure we correctly detect the start of the pattern in
5807          both phases.          both phases.
5808    
5809          If we are not at the pattern start, compile code to change the ims          If we are not at the pattern start, reset the greedy defaults and the
5810          options if this setting actually changes any of them, and reset the          case value for firstbyte and reqbyte. */
         greedy defaults and the case value for firstbyte and reqbyte. */  
5811    
5812          if (*ptr == CHAR_RIGHT_PARENTHESIS)          if (*ptr == CHAR_RIGHT_PARENTHESIS)
5813            {            {
# Line 5442  for (;; ptr++) Line 5818  for (;; ptr++)
5818              }              }
5819            else            else
5820              {              {
             if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))  
               {  
               *code++ = OP_OPT;  
               *code++ = newoptions & PCRE_IMS;  
               }  
5821              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);
5822              greedy_non_default = greedy_default ^ 1;              greedy_non_default = greedy_default ^ 1;
5823              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
5824              }              }
5825    
5826            /* Change options at this level, and pass them back for use            /* Change options at this level, and pass them back for use
5827            in subsequent branches. When not at the start of the pattern, this            in subsequent branches. */
           information is also necessary so that a resetting item can be  
           compiled at the end of a group (if we are in a group). */  
5828    
5829            *optionsptr = options = newoptions;            *optionsptr = options = newoptions;
5830            previous = NULL;       /* This item can't be repeated */            previous = NULL;       /* This item can't be repeated */
# Line 5493  for (;; ptr++) Line 5862  for (;; ptr++)
5862    
5863      /* Process nested bracketed regex. Assertions may not be repeated, but      /* Process nested bracketed regex. Assertions may not be repeated, but
5864      other kinds can be. All their opcodes are >= OP_ONCE. We copy code into a      other kinds can be. All their opcodes are >= OP_ONCE. We copy code into a
5865      non-register variable in order to be able to pass its address because some      non-register variable (tempcode) in order to be able to pass its address
5866      compilers complain otherwise. Pass in a new setting for the ims options if      because some compilers complain otherwise. */
     they have changed. */  
5867    
5868      previous = (bravalue >= OP_ONCE)? code : NULL;      previous = (bravalue >= OP_ONCE)? code : NULL;
5869      *code = bravalue;      *code = bravalue;
# Line 5505  for (;; ptr++) Line 5873  for (;; ptr++)
5873    
5874      if (!compile_regex(      if (!compile_regex(
5875           newoptions,                   /* The complete new option state */           newoptions,                   /* The complete new option state */
          options & PCRE_IMS,           /* The previous ims option state */  
5876           &tempcode,                    /* Where to put code (updated) */           &tempcode,                    /* Where to put code (updated) */
5877           &ptr,                         /* Input pointer (updated) */           &ptr,                         /* Input pointer (updated) */
5878           errorcodeptr,                 /* Where to put an error message */           errorcodeptr,                 /* Where to put an error message */
# Line 5521  for (;; ptr++) Line 5888  for (;; ptr++)
5888             &length_prevgroup           /* Pre-compile phase */             &length_prevgroup           /* Pre-compile phase */
5889           ))           ))
5890        goto FAILED;        goto FAILED;
5891    
5892        if (bravalue >= OP_ASSERT && bravalue <= OP_ASSERTBACK_NOT)
5893          cd->assert_depth -= 1;
5894    
5895      /* At the end of compiling, code is still pointing to the start of the      /* At the end of compiling, code is still pointing to the start of the
5896      group, while tempcode has been updated to point past the end of the group      group, while tempcode has been updated to point past the end of the group
# Line 5592  for (;; ptr++) Line 5962  for (;; ptr++)
5962          goto FAILED;          goto FAILED;
5963          }          }
5964        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;
5965        *code++ = OP_BRA;        code++;   /* This already contains bravalue */
5966        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
5967        *code++ = OP_KET;        *code++ = OP_KET;
5968        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
# Line 5665  for (;; ptr++) Line 6035  for (;; ptr++)
6035    
6036      /* ===================================================================*/      /* ===================================================================*/
6037      /* Handle metasequences introduced by \. For ones like \d, the ESC_ values      /* Handle metasequences introduced by \. For ones like \d, the ESC_ values
6038      are arranged to be the negation of the corresponding OP_values in the      are arranged to be the negation of the corresponding OP_values in the
6039      default case when PCRE_UCP is not set. For the back references, the values      default case when PCRE_UCP is not set. For the back references, the values
6040      are ESC_REF plus the reference number. Only back references and those types      are ESC_REF plus the reference number. Only back references and those types
6041      that consume a character may be repeated. We can test for values between      that consume a character may be repeated. We can test for values between
# Line 5760  for (;; ptr++) Line 6130  for (;; ptr++)
6130          }          }
6131    
6132        /* \k<name> or \k'name' is a back reference by name (Perl syntax).        /* \k<name> or \k'name' is a back reference by name (Perl syntax).
6133        We also support \k{name} (.NET syntax) */        We also support \k{name} (.NET syntax).  */
6134    
6135        if (-c == ESC_k && (ptr[1] == CHAR_LESS_THAN_SIGN ||        if (-c == ESC_k)
           ptr[1] == CHAR_APOSTROPHE || ptr[1] == CHAR_LEFT_CURLY_BRACKET))  
6136          {          {
6137            if ((ptr[1] != CHAR_LESS_THAN_SIGN &&
6138              ptr[1] != CHAR_APOSTROPHE && ptr[1] != CHAR_LEFT_CURLY_BRACKET))
6139              {
6140              *errorcodeptr = ERR69;
6141              break;
6142              }
6143          is_recurse = FALSE;          is_recurse = FALSE;
6144          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
6145            CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?            CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?
6146            CHAR_APOSTROPHE : CHAR_RIGHT_CURLY_BRACKET;            CHAR_APOSTROPHE : CHAR_RIGHT_CURLY_BRACKET;
6147          goto NAMED_REF_OR_RECURSE;          goto NAMED_REF_OR_RECURSE;
6148          }          }
6149    
6150        /* Back references are handled specially; must disable firstbyte if        /* Back references are handled specially; must disable firstbyte if
6151        not set to cope with cases like (?=(\w+))\1: which would otherwise set        not set to cope with cases like (?=(\w+))\1: which would otherwise set
# Line 5784  for (;; ptr++) Line 6159  for (;; ptr++)
6159          HANDLE_REFERENCE:    /* Come here from named backref handling */          HANDLE_REFERENCE:    /* Come here from named backref handling */
6160          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
6161          previous = code;          previous = code;
6162          *code++ = OP_REF;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_REFI : OP_REF;
6163          PUT2INC(code, 0, recno);          PUT2INC(code, 0, recno);
6164          cd->backref_map |= (recno < 32)? (1 << recno) : 1;          cd->backref_map |= (recno < 32)? (1 << recno) : 1;
6165          if (recno > cd->top_backref) cd->top_backref = recno;          if (recno > cd->top_backref) cd->top_backref = recno;
# Line 5843  for (;; ptr++) Line 6218  for (;; ptr++)
6218            ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */            ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
6219            }            }
6220          else          else
6221  #endif  #endif
6222            {            {
6223            previous = (-c > ESC_b && -c < ESC_Z)? code : NULL;            previous = (-c > ESC_b && -c < ESC_Z)? code : NULL;
6224            *code++ = -c;            *code++ = -c;
6225            }            }
6226          }          }
6227        continue;        continue;
6228        }        }
# Line 5892  for (;; ptr++) Line 6267  for (;; ptr++)
6267    
6268      ONE_CHAR:      ONE_CHAR:
6269      previous = code;      previous = code;
6270      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARNC : OP_CHAR;      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARI : OP_CHAR;
6271      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];
6272    
6273      /* Remember if \r or \n were seen */      /* Remember if \r or \n were seen */
# Line 5956  return FALSE; Line 6331  return FALSE;
6331  /* On entry, ptr is pointing past the bracket character, but on return it  /* On entry, ptr is pointing past the bracket character, but on return it
6332  points to the closing bracket, or vertical bar, or end of string. The code  points to the closing bracket, or vertical bar, or end of string. The code
6333  variable is pointing at the byte into which the BRA operator has been stored.  variable is pointing at the byte into which the BRA operator has been stored.
 If the ims options are changed at the start (for a (?ims: group) or during any  
 branch, we need to insert an OP_OPT item at the start of every following branch  
 to ensure they get set correctly at run time, and also pass the new options  
 into every subsequent branch compile.  
   
6334  This function is used during the pre-compile phase when we are trying to find  This function is used during the pre-compile phase when we are trying to find
6335  out the amount of memory needed, as well as during the real compile phase. The  out the amount of memory needed, as well as during the real compile phase. The
6336  value of lengthptr distinguishes the two phases.  value of lengthptr distinguishes the two phases.
6337    
6338  Arguments:  Arguments:
6339    options        option bits, including any changes for this subpattern    options        option bits, including any changes for this subpattern
   oldims         previous settings of ims option bits  
6340    codeptr        -> the address of the current code pointer    codeptr        -> the address of the current code pointer
6341    ptrptr         -> the address of the current pattern pointer    ptrptr         -> the address of the current pattern pointer
6342    errorcodeptr   -> pointer to error code variable    errorcodeptr   -> pointer to error code variable
# Line 5985  Returns: TRUE on success Line 6354  Returns: TRUE on success
6354  */  */
6355    
6356  static BOOL  static BOOL
6357  compile_regex(int options, int oldims, uschar **codeptr, const uschar **ptrptr,  compile_regex(int options, uschar **codeptr, const uschar **ptrptr,
6358    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,
6359    int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr, compile_data *cd,    int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr, compile_data *cd,
6360    int *lengthptr)    int *lengthptr)
# Line 6002  int branchfirstbyte, branchreqbyte; Line 6371  int branchfirstbyte, branchreqbyte;
6371  int length;  int length;
6372  int orig_bracount;  int orig_bracount;
6373  int max_bracount;  int max_bracount;
 int old_external_options = cd->external_options;  
6374  branch_chain bc;  branch_chain bc;
6375    
6376  bc.outer = bcptr;  bc.outer = bcptr;
# Line 6026  pre-compile phase to find out whether an Line 6394  pre-compile phase to find out whether an
6394    
6395  /* If this is a capturing subpattern, add to the chain of open capturing items  /* If this is a capturing subpattern, add to the chain of open capturing items
6396  so that we can detect them if (*ACCEPT) is encountered. This is also used to  so that we can detect them if (*ACCEPT) is encountered. This is also used to
6397  detect groups that contain recursive back references to themselves. */  detect groups that contain recursive back references to themselves. Note that
6398    only OP_CBRA need be tested here; changing this opcode to one of its variants,
6399    e.g. OP_SCBRAPOS, happens later, after the group has been compiled. */
6400    
6401  if (*code == OP_CBRA)  if (*code == OP_CBRA)
6402    {    {
# Line 6052  for (;;) Line 6422  for (;;)
6422    
6423    if (reset_bracount) cd->bracount = orig_bracount;    if (reset_bracount) cd->bracount = orig_bracount;
6424    
   /* Handle a change of ims options at the start of the branch */  
   
   if ((options & PCRE_IMS) != oldims)  
     {  
     *code++ = OP_OPT;  
     *code++ = options & PCRE_IMS;  
     length += 2;  
     }  
   
6425    /* Set up dummy OP_REVERSE if lookbehind assertion */    /* Set up dummy OP_REVERSE if lookbehind assertion */
6426    
6427    if (lookbehind)    if (lookbehind)
# Line 6081  for (;;) Line 6442  for (;;)
6442      return FALSE;      return FALSE;
6443      }      }
6444    
   /* If the external options have changed during this branch, it means that we  
   are at the top level, and a leading option setting has been encountered. We  
   need to re-set the original option values to take account of this so that,  
   during the pre-compile phase, we know to allow for a re-set at the start of  
   subsequent branches. */  
   
   if (old_external_options != cd->external_options)  
     oldims = cd->external_options & PCRE_IMS;  
   
6445    /* Keep the highest bracket count in case (?| was used and some branch    /* Keep the highest bracket count in case (?| was used and some branch
6446    has fewer than the rest. */    has fewer than the rest. */
6447    
# Line 6150  for (;;) Line 6502  for (;;)
6502        {        {
6503        int fixed_length;        int fixed_length;
6504        *code = OP_END;        *code = OP_END;
6505        fixed_length = find_fixedlength(last_branch, options, FALSE, cd);        fixed_length = find_fixedlength(last_branch,  (options & PCRE_UTF8) != 0,
6506            FALSE, cd);
6507        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
6508        if (fixed_length == -3)        if (fixed_length == -3)
6509          {          {
# Line 6171  for (;;) Line 6524  for (;;)
6524    of offsets, with the field in the BRA item now becoming an offset to the    of offsets, with the field in the BRA item now becoming an offset to the
6525    first alternative. If there are no alternatives, it points to the end of the    first alternative. If there are no alternatives, it points to the end of the
6526    group. The length in the terminating ket is always the length of the whole    group. The length in the terminating ket is always the length of the whole
6527    bracketed item. If any of the ims options were changed inside the group,    bracketed item. Return leaving the pointer at the terminating char. */
   compile a resetting op-code following, except at the very end of the pattern.  
   Return leaving the pointer at the terminating char. */  
6528    
6529    if (*ptr != CHAR_VERTICAL_LINE)    if (*ptr != CHAR_VERTICAL_LINE)
6530      {      {
6531      if (lengthptr == NULL)      if (lengthptr == NULL)
6532        {        {
6533        int branch_length = code - last_branch;        int branch_length = (int)(code - last_branch);
6534        do        do
6535          {          {
6536          int prev_length = GET(last_branch, 1);          int prev_length = GET(last_branch, 1);
# Line 6193  for (;;) Line 6544  for (;;)
6544      /* Fill in the ket */      /* Fill in the ket */
6545    
6546      *code = OP_KET;      *code = OP_KET;
6547      PUT(code, 1, code - start_bracket);      PUT(code, 1, (int)(code - start_bracket));
6548      code += 1 + LINK_SIZE;      code += 1 + LINK_SIZE;
6549    
6550      /* 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
# Line 6208  for (;;) Line 6559  for (;;)
6559            code - start_bracket);            code - start_bracket);
6560          *start_bracket = OP_ONCE;          *start_bracket = OP_ONCE;
6561          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
6562          PUT(start_bracket, 1, code - start_bracket);          PUT(start_bracket, 1, (int)(code - start_bracket));
6563          *code = OP_KET;          *code = OP_KET;
6564          PUT(code, 1, code - start_bracket);          PUT(code, 1, (int)(code - start_bracket));
6565          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
6566          length += 2 + 2*LINK_SIZE;          length += 2 + 2*LINK_SIZE;
6567          }          }
6568        cd->open_caps = cd->open_caps->next;        cd->open_caps = cd->open_caps->next;
6569        }        }
6570    
     /* Reset options if needed. */  
   
     if ((options & PCRE_IMS) != oldims && *ptr == CHAR_RIGHT_PARENTHESIS)  
       {  
       *code++ = OP_OPT;  
       *code++ = oldims;  
       length += 2;  
       }  
   
6571      /* Retain the highest bracket number, in case resetting was used. */      /* Retain the highest bracket number, in case resetting was used. */
6572    
6573      cd->bracount = max_bracount;      cd->bracount = max_bracount;
# Line 6265  for (;;) Line 6607  for (;;)
6607    else    else
6608      {      {
6609      *code = OP_ALT;      *code = OP_ALT;
6610      PUT(code, 1, code - last_branch);      PUT(code, 1, (int)(code - last_branch));
6611      bc.current_branch = last_branch = code;      bc.current_branch = last_branch = code;
6612      code += 1 + LINK_SIZE;      code += 1 + LINK_SIZE;
6613      }      }
# Line 6285  for (;;) Line 6627  for (;;)
6627  /* Try to find out if this is an anchored regular expression. Consider each  /* Try to find out if this is an anchored regular expression. Consider each
6628  alternative branch. If they all start with OP_SOD or OP_CIRC, or with a bracket  alternative branch. If they all start with OP_SOD or OP_CIRC, or with a bracket
6629  all of whose alternatives start with OP_SOD or OP_CIRC (recurse ad lib), then  all of whose alternatives start with OP_SOD or OP_CIRC (recurse ad lib), then
6630  it's anchored. However, if this is a multiline pattern, then only OP_SOD  it's anchored. However, if this is a multiline pattern, then only OP_SOD will
6631  counts, since OP_CIRC can match in the middle.  be found, because ^ generates OP_CIRCM in that mode.
6632    
6633  We can also consider a regex to be anchored if OP_SOM starts all its branches.  We can also consider a regex to be anchored if OP_SOM starts all its branches.
6634  This is the code for \G, which means "match at start of match position, taking  This is the code for \G, which means "match at start of match position, taking
# Line 6307  of the more common cases more precisely. Line 6649  of the more common cases more precisely.
6649    
6650  Arguments:  Arguments:
6651    code           points to start of expression (the bracket)    code           points to start of expression (the bracket)
   options        points to the options setting  
6652    bracket_map    a bitmap of which brackets we are inside while testing; this    bracket_map    a bitmap of which brackets we are inside while testing; this
6653                    handles up to substring 31; after that we just have to take                    handles up to substring 31; after that we just have to take
6654                    the less precise approach                    the less precise approach
# Line 6317  Returns: TRUE or FALSE Line 6658  Returns: TRUE or FALSE
6658  */  */
6659    
6660  static BOOL  static BOOL
6661  is_anchored(register const uschar *code, int *options, unsigned int bracket_map,  is_anchored(register const uschar *code, unsigned int bracket_map,
6662    unsigned int backref_map)    unsigned int backref_map)
6663  {  {
6664  do {  do {
6665     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],
6666       options, PCRE_MULTILINE, FALSE);       FALSE);
6667     register int op = *scode;     register int op = *scode;
6668    
6669     /* Non-capturing brackets */     /* Non-capturing brackets */
6670    
6671     if (op == OP_BRA)     if (op == OP_BRA  || op == OP_BRAPOS ||
6672           op == OP_SBRA || op == OP_SBRAPOS)
6673       {       {
6674       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, bracket_map, backref_map)) return FALSE;
6675       }       }
6676    
6677     /* Capturing brackets */     /* Capturing brackets */
6678    
6679     else if (op == OP_CBRA)     else if (op == OP_CBRA  || op == OP_CBRAPOS ||
6680                op == OP_SCBRA || op == OP_SCBRAPOS)
6681       {       {
6682       int n = GET2(scode, 1+LINK_SIZE);       int n = GET2(scode, 1+LINK_SIZE);
6683       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);
6684       if (!is_anchored(scode, options, new_map, backref_map)) return FALSE;       if (!is_anchored(scode, new_map, backref_map)) return FALSE;
6685       }       }
6686    
6687     /* Other brackets */     /* Other brackets */
6688    
6689     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_COND)     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_COND)
6690       {       {
6691       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, bracket_map, backref_map)) return FALSE;
6692       }       }
6693    
6694     /* .* is not anchored unless DOTALL is set (which generates OP_ALLANY) and     /* .* is not anchored unless DOTALL is set (which generates OP_ALLANY) and
# Line 6360  do { Line 6703  do {
6703    
6704     /* Check for explicit anchoring */     /* Check for explicit anchoring */
6705    
6706     else if (op != OP_SOD && op != OP_SOM &&     else if (op != OP_SOD && op != OP_SOM && op != OP_CIRC) return FALSE;
            ((*options & PCRE_MULTILINE) != 0 || op != OP_CIRC))  
      return FALSE;  
6707     code += GET(code, 1);     code += GET(code, 1);
6708     }     }
6709  while (*code == OP_ALT);   /* Loop for each alternative */  while (*code == OP_ALT);   /* Loop for each alternative */
# Line 6398  is_startline(const uschar *code, unsigne Line 6739  is_startline(const uschar *code, unsigne
6739  {  {
6740  do {  do {
6741     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],
6742       NULL, 0, FALSE);       FALSE);
6743     register int op = *scode;     register int op = *scode;
6744    
6745     /* If we are at the start of a conditional assertion group, *both* the     /* If we are at the start of a conditional assertion group, *both* the
# Line 6425  do { Line 6766  do {
6766         scode += 1 + LINK_SIZE;         scode += 1 + LINK_SIZE;
6767         break;         break;
6768         }         }
6769       scode = first_significant_code(scode, NULL, 0, FALSE);       scode = first_significant_code(scode, FALSE);
6770       op = *scode;       op = *scode;
6771       }       }
6772    
6773     /* Non-capturing brackets */     /* Non-capturing brackets */
6774    
6775     if (op == OP_BRA)     if (op == OP_BRA  || op == OP_BRAPOS ||
6776           op == OP_SBRA || op == OP_SBRAPOS)
6777       {       {
6778       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
6779       }       }
6780    
6781     /* Capturing brackets */     /* Capturing brackets */
6782    
6783     else if (op == OP_CBRA)     else if (op == OP_CBRA  || op == OP_CBRAPOS ||
6784                op == OP_SCBRA || op == OP_SCBRAPOS)
6785       {       {
6786       int n = GET2(scode, 1+LINK_SIZE);       int n = GET2(scode, 1+LINK_SIZE);
6787       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);
# Line 6462  do { Line 6805  do {
6805    
6806     /* Check for explicit circumflex */     /* Check for explicit circumflex */
6807    
6808     else if (op != OP_CIRC) return FALSE;     else if (op != OP_CIRC && op != OP_CIRCM) return FALSE;
6809    
6810     /* Move on to the next alternative */     /* Move on to the next alternative */
6811    
# Line 6488  we return that char, otherwise -1. Line 6831  we return that char, otherwise -1.
6831    
6832  Arguments:  Arguments:
6833    code       points to start of expression (the bracket)    code       points to start of expression (the bracket)
   options    pointer to the options (used to check casing changes)  
6834    inassert   TRUE if in an assertion    inassert   TRUE if in an assertion
6835    
6836  Returns:     -1 or the fixed first char  Returns:     -1 or the fixed first char
6837  */  */
6838    
6839  static int  static int
6840  find_firstassertedchar(const uschar *code, int *options, BOOL inassert)  find_firstassertedchar(const uschar *code, BOOL inassert)
6841  {  {
6842  register int c = -1;  register int c = -1;
6843  do {  do {
6844     int d;     int d;
6845     const uschar *scode =     int xl = (*code == OP_CBRA || *code == OP_SCBRA ||
6846       first_significant_code(code + 1+LINK_SIZE, options, PCRE_CASELESS, TRUE);               *code == OP_CBRAPOS || *code == OP_SCBRAPOS)? 2:0;
6847       const uschar *scode = first_significant_code(code + 1+LINK_SIZE + xl, TRUE);
6848     register int op = *scode;     register int op = *scode;
6849    
6850     switch(op)     switch(op)
# Line 6510  do { Line 6853  do {
6853       return -1;       return -1;
6854    
6855       case OP_BRA:       case OP_BRA:
6856         case OP_BRAPOS:
6857       case OP_CBRA:       case OP_CBRA:
6858         case OP_SCBRA:
6859         case OP_CBRAPOS:
6860         case OP_SCBRAPOS:
6861       case OP_ASSERT:       case OP_ASSERT:
6862       case OP_ONCE:       case OP_ONCE:
6863       case OP_COND:       case OP_COND:
6864       if ((d = find_firstassertedchar(scode, options, op == OP_ASSERT)) < 0)       if ((d = find_firstassertedchar(scode, op == OP_ASSERT)) < 0)
6865         return -1;         return -1;
6866       if (c < 0) c = d; else if (c != d) return -1;       if (c < 0) c = d; else if (c != d) return -1;
6867       break;       break;
6868    
6869       case OP_EXACT:       /* Fall through */       case OP_EXACT:
6870       scode += 2;       scode += 2;
6871         /* Fall through */
6872    
6873       case OP_CHAR:       case OP_CHAR:
      case OP_CHARNC:  
6874       case OP_PLUS:       case OP_PLUS:
6875       case OP_MINPLUS:       case OP_MINPLUS:
6876       case OP_POSPLUS:       case OP_POSPLUS:
6877       if (!inassert) return -1;       if (!inassert) return -1;
6878       if (c < 0)       if (c < 0) c = scode[1];
6879         {         else if (c != scode[1]) return -1;
6880         c = scode[1];       break;
6881         if ((*options & PCRE_CASELESS) != 0) c |= REQ_CASELESS;  
6882         }       case OP_EXACTI:
6883       else if (c != scode[1]) return -1;       scode += 2;
6884         /* Fall through */
6885    
6886         case OP_CHARI:
6887         case OP_PLUSI:
6888         case OP_MINPLUSI:
6889         case OP_POSPLUSI:
6890         if (!inassert) return -1;
6891         if (c < 0) c = scode[1] | REQ_CASELESS;
6892           else if (c != scode[1]) return -1;
6893       break;       break;
6894       }       }
6895    
# Line 6656  while (ptr[skipatstart] == CHAR_LEFT_PAR Line 7012  while (ptr[skipatstart] == CHAR_LEFT_PAR
7012      { skipatstart += 7; options |= PCRE_UTF8; continue; }      { skipatstart += 7; options |= PCRE_UTF8; continue; }
7013    else if (strncmp((char *)(ptr+skipatstart+2), STRING_UCP_RIGHTPAR, 4) == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_UCP_RIGHTPAR, 4) == 0)
7014      { skipatstart += 6; options |= PCRE_UCP; continue; }      { skipatstart += 6; options |= PCRE_UCP; continue; }
7015      else if (strncmp((char *)(ptr+skipatstart+2), STRING_NO_START_OPT_RIGHTPAR, 13) == 0)
7016        { skipatstart += 15; options |= PCRE_NO_START_OPTIMIZE; continue; }
7017    
7018    if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)    if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)
7019      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }
# Line 6679  while (ptr[skipatstart] == CHAR_LEFT_PAR Line 7037  while (ptr[skipatstart] == CHAR_LEFT_PAR
7037      options = (options & ~(PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) | newbsr;      options = (options & ~(PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) | newbsr;
7038    else break;    else break;
7039    }    }
7040    
7041  utf8 = (options & PCRE_UTF8) != 0;  utf8 = (options & PCRE_UTF8) != 0;
7042    
7043  /* Can't support UTF8 unless PCRE has been compiled to include the code. */  /* Can't support UTF8 unless PCRE has been compiled to include the code. The
7044    return of an error code from _pcre_valid_utf8() is a new feature, introduced in
7045    release 8.13. It is passed back from pcre_[dfa_]exec(), but at the moment is
7046    not used here. */
7047    
7048  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
7049  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&
7050       (*erroroffset = _pcre_valid_utf8((USPTR)pattern, -1)) >= 0)       (errorcode = _pcre_valid_utf8((USPTR)pattern, -1, erroroffset)) != 0)
7051    {    {
7052    errorcode = ERR44;    errorcode = ERR44;
7053    goto PCRE_EARLY_ERROR_RETURN2;    goto PCRE_EARLY_ERROR_RETURN2;
# Line 6705  if (utf8) Line 7066  if (utf8)
7066  if ((options & PCRE_UCP) != 0)  if ((options & PCRE_UCP) != 0)
7067    {    {
7068    errorcode = ERR67;    errorcode = ERR67;
7069    goto PCRE_EARLY_ERROR_RETURN;    goto PCRE_EARLY_ERROR_RETURN;
7070    }    }
7071  #endif  #endif
7072    
7073  /* Check validity of \R options. */  /* Check validity of \R options. */
# Line 6802  outside can help speed up starting point Line 7163  outside can help speed up starting point
7163  ptr += skipatstart;  ptr += skipatstart;
7164  code = cworkspace;  code = cworkspace;
7165  *code = OP_BRA;  *code = OP_BRA;
7166  (void)compile_regex(cd->external_options, cd->external_options & PCRE_IMS,  (void)compile_regex(cd->external_options, &code, &ptr, &errorcode, FALSE,
7167    &code, &ptr, &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd,    FALSE, 0, &firstbyte, &reqbyte, NULL, cd, &length);
   &length);  
7168  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;
7169    
7170  DPRINTF(("end pre-compile: length=%d workspace=%d\n", length,  DPRINTF(("end pre-compile: length=%d workspace=%d\n", length,
# Line 6837  regex compiled on a system with 4-byte p Line 7197  regex compiled on a system with 4-byte p
7197  pointers. */  pointers. */
7198    
7199  re->magic_number = MAGIC_NUMBER;  re->magic_number = MAGIC_NUMBER;
7200  re->size = size;  re->size = (int)size;
7201  re->options = cd->external_options;  re->options = cd->external_options;
7202  re->flags = cd->external_flags;  re->flags = cd->external_flags;
7203  re->dummy1 = 0;  re->dummy1 = 0;
# Line 6858  field; this time it's used for rememberi Line 7218  field; this time it's used for rememberi
7218  */  */
7219    
7220  cd->final_bracount = cd->bracount;  /* Save for checking forward references */  cd->final_bracount = cd->bracount;  /* Save for checking forward references */
7221    cd->assert_depth = 0;
7222  cd->bracount = 0;  cd->bracount = 0;
7223  cd->names_found = 0;  cd->names_found = 0;
7224  cd->name_table = (uschar *)re + re->name_table_offset;  cd->name_table = (uschar *)re + re->name_table_offset;
# Line 6876  of the function here. */ Line 7237  of the function here. */
7237  ptr = (const uschar *)pattern + skipatstart;  ptr = (const uschar *)pattern + skipatstart;
7238  code = (uschar *)codestart;  code = (uschar *)codestart;
7239  *code = OP_BRA;  *code = OP_BRA;
7240  (void)compile_regex(re->options, re->options & PCRE_IMS, &code, &ptr,  (void)compile_regex(re->options, &code, &ptr, &errorcode, FALSE, FALSE, 0,
7241    &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd, NULL);    &firstbyte, &reqbyte, NULL, cd, NULL);
7242  re->top_bracket = cd->bracount;  re->top_bracket = cd->bracount;
7243  re->top_backref = cd->top_backref;  re->top_backref = cd->top_backref;
7244  re->flags = cd->external_flags;  re->flags = cd->external_flags;
# Line 6908  while (errorcode == 0 && cd->hwm > cwork Line 7269  while (errorcode == 0 && cd->hwm > cwork
7269    recno = GET(codestart, offset);    recno = GET(codestart, offset);
7270    groupptr = _pcre_find_bracket(codestart, utf8, recno);    groupptr = _pcre_find_bracket(codestart, utf8, recno);
7271    if (groupptr == NULL) errorcode = ERR53;    if (groupptr == NULL) errorcode = ERR53;
7272      else PUT(((uschar *)codestart), offset, groupptr - codestart);      else PUT(((uschar *)codestart), offset, (int)(groupptr - codestart));
7273    }    }
7274    
7275  /* Give an error if there's back reference to a non-existent capturing  /* Give an error if there's back reference to a non-existent capturing
# Line 6943  if (cd->check_lookbehind) Line 7304  if (cd->check_lookbehind)
7304        uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);        uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);
7305        int end_op = *be;        int end_op = *be;
7306        *be = OP_END;        *be = OP_END;
7307        fixed_length = find_fixedlength(cc, re->options, TRUE, cd);        fixed_length = find_fixedlength(cc, (re->options & PCRE_UTF8) != 0, TRUE,
7308            cd);
7309        *be = end_op;        *be = end_op;
7310        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
7311        if (fixed_length < 0)        if (fixed_length < 0)
# Line 6963  if (errorcode != 0) Line 7325  if (errorcode != 0)
7325    {    {
7326    (pcre_free)(re);    (pcre_free)(re);
7327    PCRE_EARLY_ERROR_RETURN:    PCRE_EARLY_ERROR_RETURN:
7328    *erroroffset = ptr - (const uschar *)pattern;    *erroroffset = (int)(ptr - (const uschar *)pattern);
7329    PCRE_EARLY_ERROR_RETURN2:    PCRE_EARLY_ERROR_RETURN2:
7330    *errorptr = find_error_text(errorcode);    *errorptr = find_error_text(errorcode);
7331    if (errorcodeptr != NULL) *errorcodeptr = errorcode;    if (errorcodeptr != NULL) *errorcodeptr = errorcode;
# Line 6982  start with ^. and also when all branches Line 7344  start with ^. and also when all branches
7344    
7345  if ((re->options & PCRE_ANCHORED) == 0)  if ((re->options & PCRE_ANCHORED) == 0)
7346    {    {
7347    int temp_options = re->options;   /* May get changed during these scans */    if (is_anchored(codestart, 0, cd->backref_map))
   if (is_anchored(codestart, &temp_options, 0, cd->backref_map))  
7348      re->options |= PCRE_ANCHORED;      re->options |= PCRE_ANCHORED;
7349    else    else
7350      {      {
7351      if (firstbyte < 0)      if (firstbyte < 0)
7352        firstbyte = find_firstassertedchar(codestart, &temp_options, FALSE);        firstbyte = find_firstassertedchar(codestart, FALSE);
7353      if (firstbyte >= 0)   /* Remove caseless flag for non-caseable chars */      if (firstbyte >= 0)   /* Remove caseless flag for non-caseable chars */
7354        {        {
7355        int ch = firstbyte & 255;        int ch = firstbyte & 255;

Legend:
Removed from v.518  
changed lines
  Added in v.635

webmaster@exim.org
ViewVC Help
Powered by ViewVC 1.1.12