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

Diff of /code/trunk/pcre_compile.c

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

revision 545 by ph10, Wed Jun 16 10:51:15 2010 UTC revision 708 by ph10, Fri Sep 23 11:03:03 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 393  static const char error_texts[] = Line 393  static const char error_texts[] =
393    "internal error: previously-checked referenced subpattern not found\0"    "internal error: previously-checked referenced subpattern not found\0"
394    "DEFINE group contains more than one branch\0"    "DEFINE group contains more than one branch\0"
395    /* 55 */    /* 55 */
396    "repeating a DEFINE group is not allowed\0"    "repeating a DEFINE group is not allowed\0"  /** DEAD **/
397    "inconsistent NEWLINE options\0"    "inconsistent NEWLINE options\0"
398    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
399    "a numbered reference must not be zero\0"    "a numbered reference must not be zero\0"
# Line 408  static const char error_texts[] = Line 408  static const char error_texts[] =
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 *, int *, branch_chain *, compile_data *, int *);
551    
552    
# Line 576  return s; Line 578  return s;
578    
579    
580  /*************************************************  /*************************************************
581    *            Check for counted repeat            *
582    *************************************************/
583    
584    /* This function is called when a '{' is encountered in a place where it might
585    start a quantifier. It looks ahead to see if it really is a quantifier or not.
586    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
587    where the ddds are digits.
588    
589    Arguments:
590      p         pointer to the first char after '{'
591    
592    Returns:    TRUE or FALSE
593    */
594    
595    static BOOL
596    is_counted_repeat(const uschar *p)
597    {
598    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
599    while ((digitab[*p] & ctype_digit) != 0) p++;
600    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
601    
602    if (*p++ != CHAR_COMMA) return FALSE;
603    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
604    
605    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
606    while ((digitab[*p] & ctype_digit) != 0) p++;
607    
608    return (*p == CHAR_RIGHT_CURLY_BRACKET);
609    }
610    
611    
612    
613    /*************************************************
614  *            Handle escapes                      *  *            Handle escapes                      *
615  *************************************************/  *************************************************/
616    
# Line 646  else Line 681  else
681      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
682      break;      break;
683    
684      /* \g must be followed by one of a number of specific things:      /* In a character class, \g is just a literal "g". Outside a character
685        class, \g must be followed by one of a number of specific things:
686    
687      (1) A number, either plain or braced. If positive, it is an absolute      (1) A number, either plain or braced. If positive, it is an absolute
688      backreference. If negative, it is a relative backreference. This is a Perl      backreference. If negative, it is a relative backreference. This is a Perl
# Line 663  else Line 699  else
699      the -ESC_g code (cf \k). */      the -ESC_g code (cf \k). */
700    
701      case CHAR_g:      case CHAR_g:
702        if (isclass) break;
703      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
704        {        {
705        c = -ESC_g;        c = -ESC_g;
# Line 841  else Line 878  else
878      break;      break;
879    
880      /* 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.
881      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
882        coding is ASCII-specific, but then the whole concept of \cx is
883      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
884    
885      case CHAR_c:      case CHAR_c:
# Line 851  else Line 889  else
889        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
890        break;        break;
891        }        }
892    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
893  #ifndef EBCDIC  /* ASCII/UTF-8 coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
894          {
895          *errorcodeptr = ERR68;
896          break;
897          }
898      if (c >= CHAR_a && c <= CHAR_z) c -= 32;      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
899      c ^= 0x40;      c ^= 0x40;
900  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
901      if (c >= CHAR_a && c <= CHAR_z) c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
902      c ^= 0xC0;      c ^= 0xC0;
903  #endif  #endif
# Line 879  else Line 921  else
921    }    }
922    
923  /* Perl supports \N{name} for character names, as well as plain \N for "not  /* Perl supports \N{name} for character names, as well as plain \N for "not
924  newline". PCRE does not support \N{name}. */  newline". PCRE does not support \N{name}. However, it does support
925    quantification such as \N{2,3}. */
926    
927  if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET)  if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
928         !is_counted_repeat(ptr+2))
929    *errorcodeptr = ERR37;    *errorcodeptr = ERR37;
930    
931  /* If PCRE_UCP is set, we change the values for \d etc. */  /* If PCRE_UCP is set, we change the values for \d etc. */
# Line 991  return -1; Line 1035  return -1;
1035    
1036    
1037  /*************************************************  /*************************************************
 *            Check for counted repeat            *  
 *************************************************/  
   
 /* This function is called when a '{' is encountered in a place where it might  
 start a quantifier. It looks ahead to see if it really is a quantifier or not.  
 It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}  
 where the ddds are digits.  
   
 Arguments:  
   p         pointer to the first char after '{'  
   
 Returns:    TRUE or FALSE  
 */  
   
 static BOOL  
 is_counted_repeat(const uschar *p)  
 {  
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
 if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;  
   
 if (*p++ != CHAR_COMMA) return FALSE;  
 if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;  
   
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
   
 return (*p == CHAR_RIGHT_CURLY_BRACKET);  
 }  
   
   
   
 /*************************************************  
1038  *         Read repeat counts                     *  *         Read repeat counts                     *
1039  *************************************************/  *************************************************/
1040    
# Line 1099  top-level call starts at the beginning o Line 1110  top-level call starts at the beginning o
1110  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
1111  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
1112  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
1113  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
1114  encountered, the name will be terminated by '>' because that is checked in the  track of subpatterns that reset the capturing group numbers - the (?| feature.
1115  first pass. Recursion is used to keep track of subpatterns that reset the  
1116  capturing group numbers - the (?| feature.  This function was originally called only from the second pass, in which we know
1117    that if (?< or (?' or (?P< is encountered, the name will be correctly
1118    terminated because that is checked in the first pass. There is now one call to
1119    this function in the first pass, to check for a recursive back reference by
1120    name (so that we can make the whole group atomic). In this case, we need check
1121    only up to the current position in the pattern, and that is still OK because
1122    and previous occurrences will have been checked. To make this work, the test
1123    for "end of pattern" is a check against cd->end_pattern in the main loop,
1124    instead of looking for a binary zero. This means that the special first-pass
1125    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1126    processing items within the loop are OK, because afterwards the main loop will
1127    terminate.)
1128    
1129  Arguments:  Arguments:
1130    ptrptr       address of the current character pointer (updated)    ptrptr       address of the current character pointer (updated)
# Line 1110  Arguments: Line 1132  Arguments:
1132    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1133    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1134    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1135      utf8         TRUE if we are in UTF-8 mode
1136    count        pointer to the current capturing subpattern number (updated)    count        pointer to the current capturing subpattern number (updated)
1137    
1138  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 1140  Returns: the number of the named s
1140    
1141  static int  static int
1142  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,
1143    BOOL xmode, int *count)    BOOL xmode, BOOL utf8, int *count)
1144  {  {
1145  uschar *ptr = *ptrptr;  uschar *ptr = *ptrptr;
1146  int start_count = *count;  int start_count = *count;
# Line 1202  if (ptr[0] == CHAR_LEFT_PARENTHESIS) Line 1225  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1225    }    }
1226    
1227  /* Past any initial parenthesis handling, scan for parentheses or vertical  /* Past any initial parenthesis handling, scan for parentheses or vertical
1228  bars. */  bars. Stop if we get to cd->end_pattern. Note that this is important for the
1229    first-pass call when this value is temporarily adjusted to stop at the current
1230    position. So DO NOT change this to a test for binary zero. */
1231    
1232  for (; *ptr != 0; ptr++)  for (; ptr < cd->end_pattern; ptr++)
1233    {    {
1234    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1235    
# Line 1278  for (; *ptr != 0; ptr++) Line 1303  for (; *ptr != 0; ptr++)
1303    
1304    if (xmode && *ptr == CHAR_NUMBER_SIGN)    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1305      {      {
1306      while (*(++ptr) != 0 && *ptr != CHAR_NL) {};      ptr++;
1307        while (*ptr != 0)
1308          {
1309          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1310          ptr++;
1311    #ifdef SUPPORT_UTF8
1312          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
1313    #endif
1314          }
1315      if (*ptr == 0) goto FAIL_EXIT;      if (*ptr == 0) goto FAIL_EXIT;
1316      continue;      continue;
1317      }      }
# Line 1287  for (; *ptr != 0; ptr++) Line 1320  for (; *ptr != 0; ptr++)
1320    
1321    if (*ptr == CHAR_LEFT_PARENTHESIS)    if (*ptr == CHAR_LEFT_PARENTHESIS)
1322      {      {
1323      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, count);      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, count);
1324      if (rc > 0) return rc;      if (rc > 0) return rc;
1325      if (*ptr == 0) goto FAIL_EXIT;      if (*ptr == 0) goto FAIL_EXIT;
1326      }      }
# Line 1333  Arguments: Line 1366  Arguments:
1366    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1367    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1368    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1369      utf8         TRUE if we are in UTF-8 mode
1370    
1371  Returns:       the number of the found subpattern, or -1 if not found  Returns:       the number of the found subpattern, or -1 if not found
1372  */  */
1373    
1374  static int  static int
1375  find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode)  find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode,
1376      BOOL utf8)
1377  {  {
1378  uschar *ptr = (uschar *)cd->start_pattern;  uschar *ptr = (uschar *)cd->start_pattern;
1379  int count = 0;  int count = 0;
# Line 1351  matching closing parens. That is why we Line 1386  matching closing parens. That is why we
1386    
1387  for (;;)  for (;;)
1388    {    {
1389    rc = find_parens_sub(&ptr, cd, name, lorn, xmode, &count);    rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, &count);
1390    if (rc > 0 || *ptr++ == 0) break;    if (rc > 0 || *ptr++ == 0) break;
1391    }    }
1392    
# Line 1367  return rc; Line 1402  return rc;
1402    
1403  /* This is called by several functions that scan a compiled expression looking  /* This is called by several functions that scan a compiled expression looking
1404  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
1405  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
1406  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
1407  assertions, and also the \b assertion; for others it does not.  does not.
1408    
1409  Arguments:  Arguments:
1410    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  
1411    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1412    
1413  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1414  */  */
1415    
1416  static const uschar*  static const uschar*
1417  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const uschar *code, BOOL skipassert)
   BOOL skipassert)  
1418  {  {
1419  for (;;)  for (;;)
1420    {    {
1421    switch ((int)*code)    switch ((int)*code)
1422      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1423      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1424      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1425      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
# Line 1444  and doing the check at the end; a flag s Line 1469  and doing the check at the end; a flag s
1469    
1470  Arguments:  Arguments:
1471    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1472    options  the compiling options    utf8     TRUE in UTF-8 mode
1473    atend    TRUE if called when the pattern is complete    atend    TRUE if called when the pattern is complete
1474    cd       the "compile data" structure    cd       the "compile data" structure
1475    
# Line 1455  Returns: the fixed length, Line 1480  Returns: the fixed length,
1480  */  */
1481    
1482  static int  static int
1483  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)  find_fixedlength(uschar *code, BOOL utf8, BOOL atend, compile_data *cd)
1484  {  {
1485  int length = -1;  int length = -1;
1486    
# Line 1472  for (;;) Line 1497  for (;;)
1497    register int op = *cc;    register int op = *cc;
1498    switch (op)    switch (op)
1499      {      {
1500        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1501        OP_BRA (normal non-capturing bracket) because the other variants of these
1502        opcodes are all concerned with unlimited repeated groups, which of course
1503        are not of fixed length. They will cause a -1 response from the default
1504        case of this switch. */
1505    
1506      case OP_CBRA:      case OP_CBRA:
1507      case OP_BRA:      case OP_BRA:
1508      case OP_ONCE:      case OP_ONCE:
1509      case OP_COND:      case OP_COND:
1510      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), utf8, atend, cd);
1511      if (d < 0) return d;      if (d < 0) return d;
1512      branchlength += d;      branchlength += d;
1513      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1485  for (;;) Line 1516  for (;;)
1516    
1517      /* 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
1518      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
1519      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.
1520        Note that we must not include the OP_KETRxxx opcodes here, because they
1521        all imply an unlimited repeat. */
1522    
1523      case OP_ALT:      case OP_ALT:
1524      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1525      case OP_END:      case OP_END:
1526      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1527        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
# Line 1508  for (;;) Line 1539  for (;;)
1539      cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */      cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1540      do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */      do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1541      if (cc > cs && cc < ce) return -1;                /* Recursion */      if (cc > cs && cc < ce) return -1;                /* Recursion */
1542      d = find_fixedlength(cs + 2, options, atend, cd);      d = find_fixedlength(cs + 2, utf8, atend, cd);
1543      if (d < 0) return d;      if (d < 0) return d;
1544      branchlength += d;      branchlength += d;
1545      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
# Line 1531  for (;;) Line 1562  for (;;)
1562      case OP_RREF:      case OP_RREF:
1563      case OP_NRREF:      case OP_NRREF:
1564      case OP_DEF:      case OP_DEF:
     case OP_OPT:  
1565      case OP_CALLOUT:      case OP_CALLOUT:
1566      case OP_SOD:      case OP_SOD:
1567      case OP_SOM:      case OP_SOM:
# Line 1539  for (;;) Line 1569  for (;;)
1569      case OP_EOD:      case OP_EOD:
1570      case OP_EODN:      case OP_EODN:
1571      case OP_CIRC:      case OP_CIRC:
1572        case OP_CIRCM:
1573      case OP_DOLL:      case OP_DOLL:
1574        case OP_DOLLM:
1575      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1576      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1577      cc += _pcre_OP_lengths[*cc];      cc += _pcre_OP_lengths[*cc];
# Line 1548  for (;;) Line 1580  for (;;)
1580      /* Handle literal characters */      /* Handle literal characters */
1581    
1582      case OP_CHAR:      case OP_CHAR:
1583      case OP_CHARNC:      case OP_CHARI:
1584      case OP_NOT:      case OP_NOT:
1585        case OP_NOTI:
1586      branchlength++;      branchlength++;
1587      cc += 2;      cc += 2;
1588  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1589      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];  
1590  #endif  #endif
1591      break;      break;
1592    
# Line 1565  for (;;) Line 1597  for (;;)
1597      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1598      cc += 4;      cc += 4;
1599  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1600      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];  
1601  #endif  #endif
1602      break;      break;
1603    
# Line 1668  _pcre_find_bracket(const uschar *code, B Line 1699  _pcre_find_bracket(const uschar *code, B
1699  for (;;)  for (;;)
1700    {    {
1701    register int c = *code;    register int c = *code;
1702    
1703    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1704    
1705    /* 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 1686  for (;;) Line 1718  for (;;)
1718    
1719    /* Handle capturing bracket */    /* Handle capturing bracket */
1720    
1721    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
1722               c == OP_CBRAPOS || c == OP_SCBRAPOS)
1723      {      {
1724      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
1725      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
# Line 1724  for (;;) Line 1757  for (;;)
1757        case OP_MARK:        case OP_MARK:
1758        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
1759        case OP_SKIP_ARG:        case OP_SKIP_ARG:
       case OP_THEN_ARG:  
1760        code += code[1];        code += code[1];
1761        break;        break;
1762    
1763          case OP_THEN_ARG:
1764          code += code[1+LINK_SIZE];
1765          break;
1766        }        }
1767    
1768      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1741  for (;;) Line 1777  for (;;)
1777      if (utf8) switch(c)      if (utf8) switch(c)
1778        {        {
1779        case OP_CHAR:        case OP_CHAR:
1780        case OP_CHARNC:        case OP_CHARI:
1781        case OP_EXACT:        case OP_EXACT:
1782          case OP_EXACTI:
1783        case OP_UPTO:        case OP_UPTO:
1784          case OP_UPTOI:
1785        case OP_MINUPTO:        case OP_MINUPTO:
1786          case OP_MINUPTOI:
1787        case OP_POSUPTO:        case OP_POSUPTO:
1788          case OP_POSUPTOI:
1789        case OP_STAR:        case OP_STAR:
1790          case OP_STARI:
1791        case OP_MINSTAR:        case OP_MINSTAR:
1792          case OP_MINSTARI:
1793        case OP_POSSTAR:        case OP_POSSTAR:
1794          case OP_POSSTARI:
1795        case OP_PLUS:        case OP_PLUS:
1796          case OP_PLUSI:
1797        case OP_MINPLUS:        case OP_MINPLUS:
1798          case OP_MINPLUSI:
1799        case OP_POSPLUS:        case OP_POSPLUS:
1800          case OP_POSPLUSI:
1801        case OP_QUERY:        case OP_QUERY:
1802          case OP_QUERYI:
1803        case OP_MINQUERY:        case OP_MINQUERY:
1804          case OP_MINQUERYI:
1805        case OP_POSQUERY:        case OP_POSQUERY:
1806          case OP_POSQUERYI:
1807        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1808        break;        break;
1809        }        }
# Line 1827  for (;;) Line 1876  for (;;)
1876        case OP_MARK:        case OP_MARK:
1877        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
1878        case OP_SKIP_ARG:        case OP_SKIP_ARG:
       case OP_THEN_ARG:  
1879        code += code[1];        code += code[1];
1880        break;        break;
1881    
1882          case OP_THEN_ARG:
1883          code += code[1+LINK_SIZE];
1884          break;
1885        }        }
1886    
1887      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1844  for (;;) Line 1896  for (;;)
1896      if (utf8) switch(c)      if (utf8) switch(c)
1897        {        {
1898        case OP_CHAR:        case OP_CHAR:
1899        case OP_CHARNC:        case OP_CHARI:
1900        case OP_EXACT:        case OP_EXACT:
1901          case OP_EXACTI:
1902        case OP_UPTO:        case OP_UPTO:
1903          case OP_UPTOI:
1904        case OP_MINUPTO:        case OP_MINUPTO:
1905          case OP_MINUPTOI:
1906        case OP_POSUPTO:        case OP_POSUPTO:
1907          case OP_POSUPTOI:
1908        case OP_STAR:        case OP_STAR:
1909          case OP_STARI:
1910        case OP_MINSTAR:        case OP_MINSTAR:
1911          case OP_MINSTARI:
1912        case OP_POSSTAR:        case OP_POSSTAR:
1913          case OP_POSSTARI:
1914        case OP_PLUS:        case OP_PLUS:
1915          case OP_PLUSI:
1916        case OP_MINPLUS:        case OP_MINPLUS:
1917          case OP_MINPLUSI:
1918        case OP_POSPLUS:        case OP_POSPLUS:
1919          case OP_POSPLUSI:
1920        case OP_QUERY:        case OP_QUERY:
1921          case OP_QUERYI:
1922        case OP_MINQUERY:        case OP_MINQUERY:
1923          case OP_MINQUERYI:
1924        case OP_POSQUERY:        case OP_POSQUERY:
1925          case OP_POSQUERYI:
1926        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1927        break;        break;
1928        }        }
# Line 1896  could_be_empty_branch(const uschar *code Line 1961  could_be_empty_branch(const uschar *code
1961    compile_data *cd)    compile_data *cd)
1962  {  {
1963  register int c;  register int c;
1964  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], TRUE);
1965       code < endcode;       code < endcode;
1966       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], TRUE))
1967    {    {
1968    const uschar *ccode;    const uschar *ccode;
1969    
# Line 1914  for (code = first_significant_code(code Line 1979  for (code = first_significant_code(code
1979      continue;      continue;
1980      }      }
1981    
   /* 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;  
     }  
   
1982    /* For a recursion/subroutine call, if its end has been reached, which    /* For a recursion/subroutine call, if its end has been reached, which
1983    implies a subroutine call, we can scan it. */    implies a backward reference subroutine call, we can scan it. If it's a
1984      forward reference subroutine call, we can't. To detect forward reference
1985      we have to scan up the list that is kept in the workspace. This function is
1986      called only when doing the real compile, not during the pre-compile that
1987      measures the size of the compiled pattern. */
1988    
1989    if (c == OP_RECURSE)    if (c == OP_RECURSE)
1990      {      {
1991      BOOL empty_branch = FALSE;      const uschar *scode;
1992      const uschar *scode = cd->start_code + GET(code, 1);      BOOL empty_branch;
1993    
1994        /* Test for forward reference */
1995    
1996        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
1997          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
1998    
1999        /* Not a forward reference, test for completed backward reference */
2000    
2001        empty_branch = FALSE;
2002        scode = cd->start_code + GET(code, 1);
2003      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2004    
2005        /* Completed backwards reference */
2006    
2007      do      do
2008        {        {
2009        if (could_be_empty_branch(scode, endcode, utf8, cd))        if (could_be_empty_branch(scode, endcode, utf8, cd))
# Line 1942  for (code = first_significant_code(code Line 2014  for (code = first_significant_code(code
2014        scode += GET(scode, 1);        scode += GET(scode, 1);
2015        }        }
2016      while (*scode == OP_ALT);      while (*scode == OP_ALT);
2017    
2018      if (!empty_branch) return FALSE;  /* All branches are non-empty */      if (!empty_branch) return FALSE;  /* All branches are non-empty */
2019      continue;      continue;
2020      }      }
2021    
2022      /* Groups with zero repeats can of course be empty; skip them. */
2023    
2024      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2025          c == OP_BRAPOSZERO)
2026        {
2027        code += _pcre_OP_lengths[c];
2028        do code += GET(code, 1); while (*code == OP_ALT);
2029        c = *code;
2030        continue;
2031        }
2032    
2033      /* A nested group that is already marked as "could be empty" can just be
2034      skipped. */
2035    
2036      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2037          c == OP_SCBRA || c == OP_SCBRAPOS)
2038        {
2039        do code += GET(code, 1); while (*code == OP_ALT);
2040        c = *code;
2041        continue;
2042        }
2043    
2044    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
2045    
2046    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_BRA  || c == OP_BRAPOS ||
2047          c == OP_CBRA || c == OP_CBRAPOS ||
2048          c == OP_ONCE || c == OP_COND)
2049      {      {
2050      BOOL empty_branch;      BOOL empty_branch;
2051      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
# Line 2035  for (code = first_significant_code(code Line 2132  for (code = first_significant_code(code
2132      case OP_ALLANY:      case OP_ALLANY:
2133      case OP_ANYBYTE:      case OP_ANYBYTE:
2134      case OP_CHAR:      case OP_CHAR:
2135      case OP_CHARNC:      case OP_CHARI:
2136      case OP_NOT:      case OP_NOT:
2137        case OP_NOTI:
2138      case OP_PLUS:      case OP_PLUS:
2139      case OP_MINPLUS:      case OP_MINPLUS:
2140      case OP_POSPLUS:      case OP_POSPLUS:
# Line 2076  for (code = first_significant_code(code Line 2174  for (code = first_significant_code(code
2174      case OP_KET:      case OP_KET:
2175      case OP_KETRMAX:      case OP_KETRMAX:
2176      case OP_KETRMIN:      case OP_KETRMIN:
2177        case OP_KETRPOS:
2178      case OP_ALT:      case OP_ALT:
2179      return TRUE;      return TRUE;
2180    
# Line 2084  for (code = first_significant_code(code Line 2183  for (code = first_significant_code(code
2183    
2184  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2185      case OP_STAR:      case OP_STAR:
2186        case OP_STARI:
2187      case OP_MINSTAR:      case OP_MINSTAR:
2188        case OP_MINSTARI:
2189      case OP_POSSTAR:      case OP_POSSTAR:
2190        case OP_POSSTARI:
2191      case OP_QUERY:      case OP_QUERY:
2192        case OP_QUERYI:
2193      case OP_MINQUERY:      case OP_MINQUERY:
2194        case OP_MINQUERYI:
2195      case OP_POSQUERY:      case OP_POSQUERY:
2196        case OP_POSQUERYI:
2197      if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];      if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2198      break;      break;
2199    
2200      case OP_UPTO:      case OP_UPTO:
2201        case OP_UPTOI:
2202      case OP_MINUPTO:      case OP_MINUPTO:
2203        case OP_MINUPTOI:
2204      case OP_POSUPTO:      case OP_POSUPTO:
2205        case OP_POSUPTOI:
2206      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2207      break;      break;
2208  #endif  #endif
# Line 2105  for (code = first_significant_code(code Line 2213  for (code = first_significant_code(code
2213      case OP_MARK:      case OP_MARK:
2214      case OP_PRUNE_ARG:      case OP_PRUNE_ARG:
2215      case OP_SKIP_ARG:      case OP_SKIP_ARG:
     case OP_THEN_ARG:  
2216      code += code[1];      code += code[1];
2217      break;      break;
2218    
2219        case OP_THEN_ARG:
2220        code += code[1+LINK_SIZE];
2221        break;
2222    
2223      /* None of the remaining opcodes are required to match a character. */      /* None of the remaining opcodes are required to match a character. */
2224    
2225      default:      default:
# Line 2129  return TRUE; Line 2240  return TRUE;
2240  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
2241  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,
2242  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.
2243    This function is called only during the real compile, not during the
2244    pre-compile.
2245    
2246  Arguments:  Arguments:
2247    code        points to start of the recursion    code        points to start of the recursion
# Line 2179  where Perl recognizes it as the POSIX cl Line 2292  where Perl recognizes it as the POSIX cl
2292  "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,  "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2293  I think.  I think.
2294    
2295    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2296    It seems that the appearance of a nested POSIX class supersedes an apparent
2297    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2298    a digit.
2299    
2300    In Perl, unescaped square brackets may also appear as part of class names. For
2301    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2302    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2303    seem right at all. PCRE does not allow closing square brackets in POSIX class
2304    names.
2305    
2306  Arguments:  Arguments:
2307    ptr      pointer to the initial [    ptr      pointer to the initial [
2308    endptr   where to return the end pointer    endptr   where to return the end pointer
# Line 2193  int terminator; /* Don't combin Line 2317  int terminator; /* Don't combin
2317  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2318  for (++ptr; *ptr != 0; ptr++)  for (++ptr; *ptr != 0; ptr++)
2319    {    {
2320    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2321        ptr++;
2322      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2323      else
2324      {      {
     if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;  
2325      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2326        {        {
2327        *endptr = ptr;        *endptr = ptr;
2328        return TRUE;        return TRUE;
2329        }        }
2330        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2331             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2332              ptr[1] == CHAR_EQUALS_SIGN) &&
2333            check_posix_syntax(ptr, endptr))
2334          return FALSE;
2335      }      }
2336    }    }
2337  return FALSE;  return FALSE;
# Line 2506  if ((options & PCRE_EXTENDED) != 0) Line 2637  if ((options & PCRE_EXTENDED) != 0)
2637      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2638      if (*ptr == CHAR_NUMBER_SIGN)      if (*ptr == CHAR_NUMBER_SIGN)
2639        {        {
2640        while (*(++ptr) != 0)        ptr++;
2641          while (*ptr != 0)
2642            {
2643          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2644            ptr++;
2645    #ifdef SUPPORT_UTF8
2646            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2647    #endif
2648            }
2649        }        }
2650      else break;      else break;
2651      }      }
# Line 2543  if ((options & PCRE_EXTENDED) != 0) Line 2681  if ((options & PCRE_EXTENDED) != 0)
2681      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2682      if (*ptr == CHAR_NUMBER_SIGN)      if (*ptr == CHAR_NUMBER_SIGN)
2683        {        {
2684        while (*(++ptr) != 0)        ptr++;
2685          while (*ptr != 0)
2686            {
2687          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2688            ptr++;
2689    #ifdef SUPPORT_UTF8
2690            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2691    #endif
2692            }
2693        }        }
2694      else break;      else break;
2695      }      }
# Line 2569  if (next >= 0) switch(op_code) Line 2714  if (next >= 0) switch(op_code)
2714  #endif  #endif
2715    return c != next;    return c != next;
2716    
2717    /* 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
2718    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
2719    high-valued characters. */    high-valued characters. */
2720    
2721    case OP_CHARNC:    case OP_CHARI:
2722  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2723    GETCHARTEST(c, previous);    GETCHARTEST(c, previous);
2724  #else  #else
# Line 2596  if (next >= 0) switch(op_code) Line 2741  if (next >= 0) switch(op_code)
2741  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2742    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2743    
2744    /* For OP_NOT, its data is always a single-byte character. */    /* For OP_NOT and OP_NOTI, the data is always a single-byte character. These
2745      opcodes are not used for multi-byte characters, because they are coded using
2746      an XCLASS instead. */
2747    
2748    case OP_NOT:    case OP_NOT:
2749      return (c = *previous) == next;
2750    
2751      case OP_NOTI:
2752    if ((c = *previous) == next) return TRUE;    if ((c = *previous) == next) return TRUE;
   if ((options & PCRE_CASELESS) == 0) return FALSE;  
2753  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2754    if (utf8)    if (utf8)
2755      {      {
# Line 2705  replaced by OP_PROP codes when PCRE_UCP Line 2854  replaced by OP_PROP codes when PCRE_UCP
2854  switch(op_code)  switch(op_code)
2855    {    {
2856    case OP_CHAR:    case OP_CHAR:
2857    case OP_CHARNC:    case OP_CHARI:
2858  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2859    GETCHARTEST(c, previous);    GETCHARTEST(c, previous);
2860  #else  #else
# Line 2892  Arguments: Line 3041  Arguments:
3041    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
3042    reqbyteptr     set to the last literal character required, else < 0    reqbyteptr     set to the last literal character required, else < 0
3043    bcptr          points to current branch chain    bcptr          points to current branch chain
3044      cond_depth     conditional nesting depth
3045    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
3046    lengthptr      NULL during the real compile phase    lengthptr      NULL during the real compile phase
3047                   points to length accumulator during pre-compile phase                   points to length accumulator during pre-compile phase
# Line 2903  Returns: TRUE on success Line 3053  Returns: TRUE on success
3053  static BOOL  static BOOL
3054  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
3055    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
3056    compile_data *cd, int *lengthptr)    int cond_depth, compile_data *cd, int *lengthptr)
3057  {  {
3058  int repeat_type, op_type;  int repeat_type, op_type;
3059  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
# Line 2912  int greedy_default, greedy_non_default; Line 3062  int greedy_default, greedy_non_default;
3062  int firstbyte, reqbyte;  int firstbyte, reqbyte;
3063  int zeroreqbyte, zerofirstbyte;  int zeroreqbyte, zerofirstbyte;
3064  int req_caseopt, reqvary, tempreqvary;  int req_caseopt, reqvary, tempreqvary;
3065  int options = *optionsptr;  int options = *optionsptr;               /* May change dynamically */
3066  int after_manual_callout = 0;  int after_manual_callout = 0;
3067  int length_prevgroup = 0;  int length_prevgroup = 0;
3068  register int c;  register int c;
# Line 2930  uschar *previous_callout = NULL; Line 3080  uschar *previous_callout = NULL;
3080  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
3081  uschar classbits[32];  uschar classbits[32];
3082    
3083    /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3084    must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3085    dynamically as we process the pattern. */
3086    
3087  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3088  BOOL class_utf8;  BOOL class_utf8;
3089  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
# Line 2938  uschar *class_utf8data_base; Line 3092  uschar *class_utf8data_base;
3092  uschar utf8_char[6];  uschar utf8_char[6];
3093  #else  #else
3094  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
 uschar *utf8_char = NULL;  
3095  #endif  #endif
3096    
3097  #ifdef PCRE_DEBUG  #ifdef PCRE_DEBUG
# Line 3110  for (;; ptr++) Line 3263  for (;; ptr++)
3263      previous_callout = NULL;      previous_callout = NULL;
3264      }      }
3265    
3266    /* In extended mode, skip white space and comments */    /* In extended mode, skip white space and comments. */
3267    
3268    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3269      {      {
3270      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
3271      if (c == CHAR_NUMBER_SIGN)      if (c == CHAR_NUMBER_SIGN)
3272        {        {
3273        while (*(++ptr) != 0)        ptr++;
3274          while (*ptr != 0)
3275          {          {
3276          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3277            ptr++;
3278    #ifdef SUPPORT_UTF8
3279            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
3280    #endif
3281          }          }
3282        if (*ptr != 0) continue;        if (*ptr != 0) continue;
3283    
# Line 3164  for (;; ptr++) Line 3322  for (;; ptr++)
3322      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3323    
3324      case CHAR_CIRCUMFLEX_ACCENT:      case CHAR_CIRCUMFLEX_ACCENT:
3325        previous = NULL;
3326      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3327        {        {
3328        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3329          *code++ = OP_CIRCM;
3330        }        }
3331      previous = NULL;      else *code++ = OP_CIRC;
     *code++ = OP_CIRC;  
3332      break;      break;
3333    
3334      case CHAR_DOLLAR_SIGN:      case CHAR_DOLLAR_SIGN:
3335      previous = NULL;      previous = NULL;
3336      *code++ = OP_DOLL;      *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
3337      break;      break;
3338    
3339      /* 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 3494  for (;; ptr++) Line 3653  for (;; ptr++)
3653              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3654              continue;              continue;
3655    
3656                /* Perl 5.004 onwards omits VT from \s, but we must preserve it
3657                if it was previously set by something earlier in the character
3658                class. */
3659    
3660              case ESC_s:              case ESC_s:
3661              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];              classbits[0] |= cbits[cbit_space];
3662              classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= cbits[cbit_space+1] & ~0x08;
3663                for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3664              continue;              continue;
3665    
3666              case ESC_S:              case ESC_S:
# Line 3915  for (;; ptr++) Line 4079  for (;; ptr++)
4079    
4080      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
4081      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
4082      operate on single-bytes only. This is an historical hangover. Maybe one day      operate on single-bytes characters only. This is an historical hangover.
4083      we can tidy these opcodes to handle multi-byte characters.      Maybe one day we can tidy these opcodes to handle multi-byte characters.
4084    
4085      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
4086      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.
4087      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
4088      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
4089      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
4090      reqbyte, save the previous value for reinstating. */      of reqbyte, save the previous value for reinstating. */
4091    
4092  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4093      if (class_charcount == 1 && !class_utf8 &&      if (class_charcount == 1 && !class_utf8 &&
# Line 3934  for (;; ptr++) Line 4098  for (;; ptr++)
4098        {        {
4099        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
4100    
4101        /* The OP_NOT opcode works on one-byte characters only. */        /* The OP_NOT[I] opcodes work on one-byte characters only. */
4102    
4103        if (negate_class)        if (negate_class)
4104          {          {
4105          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
4106          zerofirstbyte = firstbyte;          zerofirstbyte = firstbyte;
4107          *code++ = OP_NOT;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
4108          *code++ = class_lastchar;          *code++ = class_lastchar;
4109          break;          break;
4110          }          }
# Line 4068  for (;; ptr++) Line 4232  for (;; ptr++)
4232      op_type = 0;                    /* Default single-char op codes */      op_type = 0;                    /* Default single-char op codes */
4233      possessive_quantifier = FALSE;  /* Default not possessive quantifier */      possessive_quantifier = FALSE;  /* Default not possessive quantifier */
4234    
4235      /* Save start of previous item, in case we have to move it up to make space      /* Save start of previous item, in case we have to move it up in order to
4236      for an inserted OP_ONCE for the additional '+' extension. */      insert something before it. */
4237    
4238      tempcode = previous;      tempcode = previous;
4239    
# Line 4092  for (;; ptr++) Line 4256  for (;; ptr++)
4256        }        }
4257      else repeat_type = greedy_default;      else repeat_type = greedy_default;
4258    
4259        /* If previous was a recursion call, wrap it in atomic brackets so that
4260        previous becomes the atomic group. All recursions were so wrapped in the
4261        past, but it no longer happens for non-repeated recursions. In fact, the
4262        repeated ones could be re-implemented independently so as not to need this,
4263        but for the moment we rely on the code for repeating groups. */
4264    
4265        if (*previous == OP_RECURSE)
4266          {
4267          memmove(previous + 1 + LINK_SIZE, previous, 1 + LINK_SIZE);
4268          *previous = OP_ONCE;
4269          PUT(previous, 1, 2 + 2*LINK_SIZE);
4270          previous[2 + 2*LINK_SIZE] = OP_KET;
4271          PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
4272          code += 2 + 2 * LINK_SIZE;
4273          length_prevgroup = 3 + 3*LINK_SIZE;
4274    
4275          /* When actually compiling, we need to check whether this was a forward
4276          reference, and if so, adjust the offset. */
4277    
4278          if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
4279            {
4280            int offset = GET(cd->hwm, -LINK_SIZE);
4281            if (offset == previous + 1 - cd->start_code)
4282              PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
4283            }
4284          }
4285    
4286        /* Now handle repetition for the different types of item. */
4287    
4288      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
4289      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
4290      that it is set in reqbyte - it might not be if a sequence such as x{3} is      that it is set in reqbyte - it might not be if a sequence such as x{3} is
4291      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
4292      instead.  */      instead.  */
4293    
4294      if (*previous == OP_CHAR || *previous == OP_CHARNC)      if (*previous == OP_CHAR || *previous == OP_CHARI)
4295        {        {
4296          op_type = (*previous == OP_CHAR)? 0 : OP_STARI - OP_STAR;
4297    
4298        /* 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
4299        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
4300        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 4144  for (;; ptr++) Line 4339  for (;; ptr++)
4339      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
4340      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-
4341      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
4342      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
4343      currently used only for single-byte chars. */      are currently used only for single-byte chars. */
4344    
4345      else if (*previous == OP_NOT)      else if (*previous == OP_NOT || *previous == OP_NOTI)
4346        {        {
4347        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = ((*previous == OP_NOT)? OP_NOTSTAR : OP_NOTSTARI) - OP_STAR;
4348        c = previous[1];        c = previous[1];
4349        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4350            repeat_max < 0 &&            repeat_max < 0 &&
# Line 4346  for (;; ptr++) Line 4541  for (;; ptr++)
4541  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4542               *previous == OP_XCLASS ||               *previous == OP_XCLASS ||
4543  #endif  #endif
4544               *previous == OP_REF)               *previous == OP_REF ||
4545                 *previous == OP_REFI)
4546        {        {
4547        if (repeat_max == 0)        if (repeat_max == 0)
4548          {          {
# Line 4380  for (;; ptr++) Line 4576  for (;; ptr++)
4576        }        }
4577    
4578      /* 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
4579      cases. */      cases. Note that at this point we can encounter only the "basic" bracket
4580        opcodes such as BRA and CBRA, as this is the place where they get converted
4581        into the more special varieties such as BRAPOS and SBRA. A test for >=
4582        OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
4583        ASSERTBACK_NOT, ONCE, BRA, CBRA, and COND. Originally, PCRE did not allow
4584        repetition of assertions, but now it does, for Perl compatibility. */
4585    
4586      else if (*previous == OP_BRA  || *previous == OP_CBRA ||      else if (*previous >= OP_ASSERT && *previous <= OP_COND)
              *previous == OP_ONCE || *previous == OP_COND)  
4587        {        {
4588        register int i;        register int i;
       int ketoffset = 0;  
4589        int len = (int)(code - previous);        int len = (int)(code - previous);
4590        uschar *bralink = NULL;        uschar *bralink = NULL;
4591          uschar *brazeroptr = NULL;
4592    
4593        /* Repeating a DEFINE group is pointless */        /* Repeating a DEFINE group is pointless, but Perl allows the syntax, so
4594          we just ignore the repeat. */
4595    
4596        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
4597          {          goto END_REPEAT;
4598          *errorcodeptr = ERR55;  
4599          goto FAILED;        /* There is no sense in actually repeating assertions. The only potential
4600          }        use of repetition is in cases when the assertion is optional. Therefore,
4601          if the minimum is greater than zero, just ignore the repeat. If the
4602          maximum is not not zero or one, set it to 1. */
4603    
4604        /* If the maximum repeat count is unlimited, find the end of the bracket        if (*previous < OP_ONCE)    /* Assertion */
4605        by scanning through from the start, and compute the offset back to it          {
4606        from the current code pointer. There may be an OP_OPT setting following          if (repeat_min > 0) goto END_REPEAT;
4607        the final KET, so we can't find the end just by going back from the code          if (repeat_max < 0 || repeat_max > 1) repeat_max = 1;
       pointer. */  
   
       if (repeat_max == -1)  
         {  
         register uschar *ket = previous;  
         do ket += GET(ket, 1); while (*ket != OP_KET);  
         ketoffset = (int)(code - ket);  
4608          }          }
4609    
4610        /* 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
# Line 4429  for (;; ptr++) Line 4625  for (;; ptr++)
4625          **   goto END_REPEAT;          **   goto END_REPEAT;
4626          **   }          **   }
4627    
4628          However, that fails when a group is referenced as a subroutine from          However, that fails when a group or a subgroup within it is referenced
4629          elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it          as a subroutine from elsewhere in the pattern, so now we stick in
4630          so that it is skipped on execution. As we don't have a list of which          OP_SKIPZERO in front of it so that it is skipped on execution. As we
4631          groups are referenced, we cannot do this selectively.          don't have a list of which groups are referenced, we cannot do this
4632            selectively.
4633    
4634          If the maximum is 1 or unlimited, we just have to stick in the BRAZERO          If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4635          and do no more at this point. However, we do need to adjust any          and do no more at this point. However, we do need to adjust any
# Line 4452  for (;; ptr++) Line 4649  for (;; ptr++)
4649              *previous++ = OP_SKIPZERO;              *previous++ = OP_SKIPZERO;
4650              goto END_REPEAT;              goto END_REPEAT;
4651              }              }
4652              brazeroptr = previous;    /* Save for possessive optimizing */
4653            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4654            }            }
4655    
# Line 4616  for (;; ptr++) Line 4814  for (;; ptr++)
4814            }            }
4815          }          }
4816    
4817        /* 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
4818        can't just offset backwards from the current code point, because we        ONCE brackets, that's all we need to do. However, possessively repeated
4819        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
4820        correct offset was computed above.        behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to
4821          deal with possessive ONCEs specially.
4822    
4823          Otherwise, if the quantifier was possessive, we convert the BRA code to
4824          the POS form, and the KET code to KETRPOS. (It turns out to be convenient
4825          at runtime to detect this kind of subpattern at both the start and at the
4826          end.) The use of special opcodes makes it possible to reduce greatly the
4827          stack usage in pcre_exec(). If the group is preceded by OP_BRAZERO,
4828          convert this to OP_BRAPOSZERO. Then cancel the possessive flag so that
4829          the default action below, of wrapping everything inside atomic brackets,
4830          does not happen.
4831    
4832        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
4833        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
4834        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
4835        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
4836        atomic groups at runtime, but in a different way.] */        runtime, but in a different way.] */
4837    
4838        else        else
4839          {          {
4840          uschar *ketcode = code - ketoffset;          uschar *ketcode = code - 1 - LINK_SIZE;
4841          uschar *bracode = ketcode - GET(ketcode, 1);          uschar *bracode = ketcode - GET(ketcode, 1);
4842          *ketcode = OP_KETRMAX + repeat_type;  
4843          if (lengthptr == NULL && *bracode != OP_ONCE)          if (*bracode == OP_ONCE && possessive_quantifier) *bracode = OP_BRA;
4844            if (*bracode == OP_ONCE)
4845              *ketcode = OP_KETRMAX + repeat_type;
4846            else
4847            {            {
4848            uschar *scode = bracode;            if (possessive_quantifier)
           do  
4849              {              {
4850              if (could_be_empty_branch(scode, ketcode, utf8, cd))              *bracode += 1;                   /* Switch to xxxPOS opcodes */
4851                *ketcode = OP_KETRPOS;
4852                if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
4853                possessive_quantifier = FALSE;
4854                }
4855              else *ketcode = OP_KETRMAX + repeat_type;
4856    
4857              if (lengthptr == NULL)
4858                {
4859                uschar *scode = bracode;
4860                do
4861                {                {
4862                *bracode += OP_SBRA - OP_BRA;                if (could_be_empty_branch(scode, ketcode, utf8, cd))
4863                break;                  {
4864                    *bracode += OP_SBRA - OP_BRA;
4865                    break;
4866                    }
4867                  scode += GET(scode, 1);
4868                }                }
4869              scode += GET(scode, 1);              while (*scode == OP_ALT);
4870              }              }
           while (*scode == OP_ALT);  
4871            }            }
4872          }          }
4873        }        }
# Line 4665  for (;; ptr++) Line 4888  for (;; ptr++)
4888        }        }
4889    
4890      /* If the character following a repeat is '+', or if certain optimization      /* If the character following a repeat is '+', or if certain optimization
4891      tests above succeeded, possessive_quantifier is TRUE. For some of the      tests above succeeded, possessive_quantifier is TRUE. For some opcodes,
4892      simpler opcodes, there is an special alternative opcode for this. For      there are special alternative opcodes for this case. For anything else, we
4893      anything else, we wrap the entire repeated item inside OP_ONCE brackets.      wrap the entire repeated item inside OP_ONCE brackets. Logically, the '+'
4894      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
4895      but the special opcodes can optimize it a bit. The repeated item starts at      special opcodes can optimize it.
4896      tempcode, not at previous, which might be the first part of a string whose  
4897      (former) last char we repeated.      Possessively repeated subpatterns have already been handled in the code
4898        just above, so possessive_quantifier is always FALSE for them at this
4899        stage.
4900    
4901        Note that the repeated item starts at tempcode, not at previous, which
4902        might be the first part of a string whose (former) last char we repeated.
4903    
4904      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
4905      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 4702  for (;; ptr++) Line 4930  for (;; ptr++)
4930          case OP_QUERY: *tempcode = OP_POSQUERY; break;          case OP_QUERY: *tempcode = OP_POSQUERY; break;
4931          case OP_UPTO:  *tempcode = OP_POSUPTO; break;          case OP_UPTO:  *tempcode = OP_POSUPTO; break;
4932    
4933          case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;          case OP_STARI:  *tempcode = OP_POSSTARI; break;
4934          case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;          case OP_PLUSI:  *tempcode = OP_POSPLUSI; break;
4935          case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;          case OP_QUERYI: *tempcode = OP_POSQUERYI; break;
4936          case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;          case OP_UPTOI:  *tempcode = OP_POSUPTOI; break;
4937    
4938          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;
4939          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;
4940          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4941          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4942    
4943            case OP_NOTSTARI:  *tempcode = OP_NOTPOSSTARI; break;
4944            case OP_NOTPLUSI:  *tempcode = OP_NOTPOSPLUSI; break;
4945            case OP_NOTQUERYI: *tempcode = OP_NOTPOSQUERYI; break;
4946            case OP_NOTUPTOI:  *tempcode = OP_NOTPOSUPTOI; break;
4947    
4948            case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;
4949            case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;
4950            case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
4951            case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;
4952    
4953          /* Because we are moving code along, we must ensure that any          /* Because we are moving code along, we must ensure that any
4954          pending recursive references are updated. */          pending recursive references are updated. */
4955    
# Line 4765  for (;; ptr++) Line 5003  for (;; ptr++)
5003        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
5004        namelen = (int)(ptr - name);        namelen = (int)(ptr - name);
5005    
5006          /* It appears that Perl allows any characters whatsoever, other than
5007          a closing parenthesis, to appear in arguments, so we no longer insist on
5008          letters, digits, and underscores. */
5009    
5010        if (*ptr == CHAR_COLON)        if (*ptr == CHAR_COLON)
5011          {          {
5012          arg = ++ptr;          arg = ++ptr;
5013          while ((cd->ctypes[*ptr] & (ctype_letter|ctype_digit)) != 0          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
           || *ptr == '_') ptr++;  
5014          arglen = (int)(ptr - arg);          arglen = (int)(ptr - arg);
5015          }          }
5016    
# Line 4786  for (;; ptr++) Line 5027  for (;; ptr++)
5027          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
5028              strncmp((char *)name, vn, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
5029            {            {
5030            /* Check for open captures before ACCEPT */            /* Check for open captures before ACCEPT and convert it to
5031              ASSERT_ACCEPT if in an assertion. */
5032    
5033            if (verbs[i].op == OP_ACCEPT)            if (verbs[i].op == OP_ACCEPT)
5034              {              {
5035              open_capitem *oc;              open_capitem *oc;
5036                if (arglen != 0)
5037                  {
5038                  *errorcodeptr = ERR59;
5039                  goto FAILED;
5040                  }
5041              cd->had_accept = TRUE;              cd->had_accept = TRUE;
5042              for (oc = cd->open_caps; oc != NULL; oc = oc->next)              for (oc = cd->open_caps; oc != NULL; oc = oc->next)
5043                {                {
5044                *code++ = OP_CLOSE;                *code++ = OP_CLOSE;
5045                PUT2INC(code, 0, oc->number);                PUT2INC(code, 0, oc->number);
5046                }                }
5047                *code++ = (cd->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT;
5048    
5049                /* Do not set firstbyte after *ACCEPT */
5050                if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
5051              }              }
5052    
5053            /* Handle the cases with/without an argument */            /* Handle other cases with/without an argument */
5054    
5055            if (arglen == 0)            else if (arglen == 0)
5056              {              {
5057              if (verbs[i].op < 0)   /* Argument is mandatory */              if (verbs[i].op < 0)   /* Argument is mandatory */
5058                {                {
5059                *errorcodeptr = ERR66;                *errorcodeptr = ERR66;
5060                goto FAILED;                goto FAILED;
5061                }                }
5062              *code++ = verbs[i].op;              *code = verbs[i].op;
5063                if (*code++ == OP_THEN)
5064                  {
5065                  PUT(code, 0, code - bcptr->current_branch - 1);
5066                  code += LINK_SIZE;
5067                  cd->external_flags |= PCRE_HASTHEN;
5068                  }
5069              }              }
5070    
5071            else            else
# Line 4818  for (;; ptr++) Line 5075  for (;; ptr++)
5075                *errorcodeptr = ERR59;                *errorcodeptr = ERR59;
5076                goto FAILED;                goto FAILED;
5077                }                }
5078              *code++ = verbs[i].op_arg;              *code = verbs[i].op_arg;
5079                if (*code++ == OP_THEN_ARG)
5080                  {
5081                  PUT(code, 0, code - bcptr->current_branch - 1);
5082                  code += LINK_SIZE;
5083                  }
5084              *code++ = arglen;              *code++ = arglen;
5085              memcpy(code, arg, arglen);              memcpy(code, arg, arglen);
5086              code += arglen;              code += arglen;
# Line 5012  for (;; ptr++) Line 5274  for (;; ptr++)
5274          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
5275    
5276          else if ((i = find_parens(cd, name, namelen,          else if ((i = find_parens(cd, name, namelen,
5277                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0, utf8)) > 0)
5278            {            {
5279            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
5280            code[1+LINK_SIZE]++;            code[1+LINK_SIZE]++;
# Line 5080  for (;; ptr++) Line 5342  for (;; ptr++)
5342          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5343          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
5344          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
5345            cd->assert_depth += 1;
5346          ptr++;          ptr++;
5347          break;          break;
5348    
# Line 5094  for (;; ptr++) Line 5357  for (;; ptr++)
5357            continue;            continue;
5358            }            }
5359          bravalue = OP_ASSERT_NOT;          bravalue = OP_ASSERT_NOT;
5360            cd->assert_depth += 1;
5361          break;          break;
5362    
5363    
# Line 5103  for (;; ptr++) Line 5367  for (;; ptr++)
5367            {            {
5368            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
5369            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
5370              cd->assert_depth += 1;
5371            ptr += 2;            ptr += 2;
5372            break;            break;
5373    
5374            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
5375            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
5376              cd->assert_depth += 1;
5377            ptr += 2;            ptr += 2;
5378            break;            break;
5379    
# Line 5313  for (;; ptr++) Line 5579  for (;; ptr++)
5579          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
5580          namelen = (int)(ptr - name);          namelen = (int)(ptr - name);
5581    
5582          /* 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
5583          reference number. */          a dummy reference number, because it was not used in the first pass.
5584            However, with the change of recursive back references to be atomic,
5585            we have to look for the number so that this state can be identified, as
5586            otherwise the incorrect length is computed. If it's not a backwards
5587            reference, the dummy number will do. */
5588    
5589          if (lengthptr != NULL)          if (lengthptr != NULL)
5590            {            {
5591              const uschar *temp;
5592    
5593            if (namelen == 0)            if (namelen == 0)
5594              {              {
5595              *errorcodeptr = ERR62;              *errorcodeptr = ERR62;
# Line 5333  for (;; ptr++) Line 5605  for (;; ptr++)
5605              *errorcodeptr = ERR48;              *errorcodeptr = ERR48;
5606              goto FAILED;              goto FAILED;
5607              }              }
5608            recno = 0;  
5609              /* The name table does not exist in the first pass, so we cannot
5610              do a simple search as in the code below. Instead, we have to scan the
5611              pattern to find the number. It is important that we scan it only as
5612              far as we have got because the syntax of named subpatterns has not
5613              been checked for the rest of the pattern, and find_parens() assumes
5614              correct syntax. In any case, it's a waste of resources to scan
5615              further. We stop the scan at the current point by temporarily
5616              adjusting the value of cd->endpattern. */
5617    
5618              temp = cd->end_pattern;
5619              cd->end_pattern = ptr;
5620              recno = find_parens(cd, name, namelen,
5621                (options & PCRE_EXTENDED) != 0, utf8);
5622              cd->end_pattern = temp;
5623              if (recno < 0) recno = 0;    /* Forward ref; set dummy number */
5624            }            }
5625    
5626          /* 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 5358  for (;; ptr++) Line 5645  for (;; ptr++)
5645              }              }
5646            else if ((recno =                /* Forward back reference */            else if ((recno =                /* Forward back reference */
5647                      find_parens(cd, name, namelen,                      find_parens(cd, name, namelen,
5648                        (options & PCRE_EXTENDED) != 0)) <= 0)                        (options & PCRE_EXTENDED) != 0, utf8)) <= 0)
5649              {              {
5650              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
5651              goto FAILED;              goto FAILED;
# Line 5469  for (;; ptr++) Line 5756  for (;; ptr++)
5756              if (called == NULL)              if (called == NULL)
5757                {                {
5758                if (find_parens(cd, NULL, recno,                if (find_parens(cd, NULL, recno,
5759                      (options & PCRE_EXTENDED) != 0) < 0)                      (options & PCRE_EXTENDED) != 0, utf8) < 0)
5760                  {                  {
5761                  *errorcodeptr = ERR15;                  *errorcodeptr = ERR15;
5762                  goto FAILED;                  goto FAILED;
# Line 5477  for (;; ptr++) Line 5764  for (;; ptr++)
5764    
5765                /* 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
5766                offset below, what it actually inserted is the reference number                offset below, what it actually inserted is the reference number
5767                of the group. */                of the group. Then remember the forward reference. */
5768    
5769                called = cd->start_code + recno;                called = cd->start_code + recno;
5770                PUTINC(cd->hwm, 0, (int)(code + 2 + LINK_SIZE - cd->start_code));                PUTINC(cd->hwm, 0, (int)(code + 1 - cd->start_code));
5771                }                }
5772    
5773              /* If not a forward reference, and the subpattern is still open,              /* If not a forward reference, and the subpattern is still open,
5774              this is a recursive call. We check to see if this is a left              this is a recursive call. We check to see if this is a left
5775              recursion that could loop for ever, and diagnose that case. */              recursion that could loop for ever, and diagnose that case. We
5776                must not, however, do this check if we are in a conditional
5777                subpattern because the condition might be testing for recursion in
5778                a pattern such as /(?(R)a+|(?R)b)/, which is perfectly valid.
5779                Forever loops are also detected at runtime, so those that occur in
5780                conditional subpatterns will be picked up then. */
5781    
5782              else if (GET(called, 1) == 0 &&              else if (GET(called, 1) == 0 && cond_depth <= 0 &&
5783                       could_be_empty(called, code, bcptr, utf8, cd))                       could_be_empty(called, code, bcptr, utf8, cd))
5784                {                {
5785                *errorcodeptr = ERR40;                *errorcodeptr = ERR40;
# Line 5495  for (;; ptr++) Line 5787  for (;; ptr++)
5787                }                }
5788              }              }
5789    
5790            /* Insert the recursion/subroutine item, automatically wrapped inside            /* Insert the recursion/subroutine item. */
           "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;  
5791    
5792            *code = OP_RECURSE;            *code = OP_RECURSE;
5793            PUT(code, 1, (int)(called - cd->start_code));            PUT(code, 1, (int)(called - cd->start_code));
5794            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;  
5795            }            }
5796    
5797          /* Can't determine a first byte now */          /* Can't determine a first byte now */
# Line 5572  for (;; ptr++) Line 5852  for (;; ptr++)
5852          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
5853          both phases.          both phases.
5854    
5855          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
5856          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. */  
5857    
5858          if (*ptr == CHAR_RIGHT_PARENTHESIS)          if (*ptr == CHAR_RIGHT_PARENTHESIS)
5859            {            {
# Line 5585  for (;; ptr++) Line 5864  for (;; ptr++)
5864              }              }
5865            else            else
5866              {              {
             if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))  
               {  
               *code++ = OP_OPT;  
               *code++ = newoptions & PCRE_IMS;  
               }  
5867              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);
5868              greedy_non_default = greedy_default ^ 1;              greedy_non_default = greedy_default ^ 1;
5869              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
5870              }              }
5871    
5872            /* Change options at this level, and pass them back for use            /* Change options at this level, and pass them back for use
5873            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). */  
5874    
5875            *optionsptr = options = newoptions;            *optionsptr = options = newoptions;
5876            previous = NULL;       /* This item can't be repeated */            previous = NULL;       /* This item can't be repeated */
# Line 5634  for (;; ptr++) Line 5906  for (;; ptr++)
5906        skipbytes = 2;        skipbytes = 2;
5907        }        }
5908    
5909      /* Process nested bracketed regex. Assertions may not be repeated, but      /* Process nested bracketed regex. Assertions used not to be repeatable,
5910      other kinds can be. All their opcodes are >= OP_ONCE. We copy code into a      but this was changed for Perl compatibility, so all kinds can now be
5911      non-register variable in order to be able to pass its address because some      repeated. We copy code into a non-register variable (tempcode) in order to
5912      compilers complain otherwise. Pass in a new setting for the ims options if      be able to pass its address because some compilers complain otherwise. */
     they have changed. */  
5913    
5914      previous = (bravalue >= OP_ONCE)? code : NULL;      previous = code;                      /* For handling repetition */
5915      *code = bravalue;      *code = bravalue;
5916      tempcode = code;      tempcode = code;
5917      tempreqvary = cd->req_varyopt;     /* Save value before bracket */      tempreqvary = cd->req_varyopt;        /* Save value before bracket */
5918      length_prevgroup = 0;              /* Initialize for pre-compile phase */      length_prevgroup = 0;                 /* Initialize for pre-compile phase */
5919    
5920      if (!compile_regex(      if (!compile_regex(
5921           newoptions,                   /* The complete new option state */           newoptions,                      /* The complete new option state */
5922           options & PCRE_IMS,           /* The previous ims option state */           &tempcode,                       /* Where to put code (updated) */
5923           &tempcode,                    /* Where to put code (updated) */           &ptr,                            /* Input pointer (updated) */
5924           &ptr,                         /* Input pointer (updated) */           errorcodeptr,                    /* Where to put an error message */
          errorcodeptr,                 /* Where to put an error message */  
5925           (bravalue == OP_ASSERTBACK ||           (bravalue == OP_ASSERTBACK ||
5926            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */
5927           reset_bracount,               /* True if (?| group */           reset_bracount,                  /* True if (?| group */
5928           skipbytes,                    /* Skip over bracket number */           skipbytes,                       /* Skip over bracket number */
5929           &subfirstbyte,                /* For possible first char */           cond_depth +
5930           &subreqbyte,                  /* For possible last char */             ((bravalue == OP_COND)?1:0),   /* Depth of condition subpatterns */
5931           bcptr,                        /* Current branch chain */           &subfirstbyte,                   /* For possible first char */
5932           cd,                           /* Tables block */           &subreqbyte,                     /* For possible last char */
5933           (lengthptr == NULL)? NULL :   /* Actual compile phase */           bcptr,                           /* Current branch chain */
5934             &length_prevgroup           /* Pre-compile phase */           cd,                              /* Tables block */
5935             (lengthptr == NULL)? NULL :      /* Actual compile phase */
5936               &length_prevgroup              /* Pre-compile phase */
5937           ))           ))
5938        goto FAILED;        goto FAILED;
5939    
5940        if (bravalue >= OP_ASSERT && bravalue <= OP_ASSERTBACK_NOT)
5941          cd->assert_depth -= 1;
5942    
5943      /* 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
5944      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
5945      and any option resetting that may follow it. The pattern pointer (ptr)      and any option resetting that may follow it. The pattern pointer (ptr)
# Line 5735  for (;; ptr++) Line 6010  for (;; ptr++)
6010          goto FAILED;          goto FAILED;
6011          }          }
6012        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;
6013        *code++ = OP_BRA;        code++;   /* This already contains bravalue */
6014        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
6015        *code++ = OP_KET;        *code++ = OP_KET;
6016        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
# Line 5903  for (;; ptr++) Line 6178  for (;; ptr++)
6178          }          }
6179    
6180        /* \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).
6181        We also support \k{name} (.NET syntax) */        We also support \k{name} (.NET syntax).  */
6182    
6183        if (-c == ESC_k && (ptr[1] == CHAR_LESS_THAN_SIGN ||        if (-c == ESC_k)
           ptr[1] == CHAR_APOSTROPHE || ptr[1] == CHAR_LEFT_CURLY_BRACKET))  
6184          {          {
6185            if ((ptr[1] != CHAR_LESS_THAN_SIGN &&
6186              ptr[1] != CHAR_APOSTROPHE && ptr[1] != CHAR_LEFT_CURLY_BRACKET))
6187              {
6188              *errorcodeptr = ERR69;
6189              break;
6190              }
6191          is_recurse = FALSE;          is_recurse = FALSE;
6192          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
6193            CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?            CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?
# Line 5927  for (;; ptr++) Line 6207  for (;; ptr++)
6207          HANDLE_REFERENCE:    /* Come here from named backref handling */          HANDLE_REFERENCE:    /* Come here from named backref handling */
6208          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
6209          previous = code;          previous = code;
6210          *code++ = OP_REF;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_REFI : OP_REF;
6211          PUT2INC(code, 0, recno);          PUT2INC(code, 0, recno);
6212          cd->backref_map |= (recno < 32)? (1 << recno) : 1;          cd->backref_map |= (recno < 32)? (1 << recno) : 1;
6213          if (recno > cd->top_backref) cd->top_backref = recno;          if (recno > cd->top_backref) cd->top_backref = recno;
# Line 6035  for (;; ptr++) Line 6315  for (;; ptr++)
6315    
6316      ONE_CHAR:      ONE_CHAR:
6317      previous = code;      previous = code;
6318      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARNC : OP_CHAR;      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARI : OP_CHAR;
6319      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];
6320    
6321      /* Remember if \r or \n were seen */      /* Remember if \r or \n were seen */
# Line 6064  for (;; ptr++) Line 6344  for (;; ptr++)
6344        else firstbyte = reqbyte = REQ_NONE;        else firstbyte = reqbyte = REQ_NONE;
6345        }        }
6346    
6347      /* firstbyte was previously set; we can set reqbyte only the length is      /* firstbyte was previously set; we can set reqbyte only if the length is
6348      1 or the matching is caseful. */      1 or the matching is caseful. */
6349    
6350      else      else
# Line 6099  return FALSE; Line 6379  return FALSE;
6379  /* 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
6380  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
6381  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.  
   
6382  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
6383  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
6384  value of lengthptr distinguishes the two phases.  value of lengthptr distinguishes the two phases.
6385    
6386  Arguments:  Arguments:
6387    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  
6388    codeptr        -> the address of the current code pointer    codeptr        -> the address of the current code pointer
6389    ptrptr         -> the address of the current pattern pointer    ptrptr         -> the address of the current pattern pointer
6390    errorcodeptr   -> pointer to error code variable    errorcodeptr   -> pointer to error code variable
6391    lookbehind     TRUE if this is a lookbehind assertion    lookbehind     TRUE if this is a lookbehind assertion
6392    reset_bracount TRUE to reset the count for each branch    reset_bracount TRUE to reset the count for each branch
6393    skipbytes      skip this many bytes at start (for brackets and OP_COND)    skipbytes      skip this many bytes at start (for brackets and OP_COND)
6394      cond_depth     depth of nesting for conditional subpatterns
6395    firstbyteptr   place to put the first required character, or a negative number    firstbyteptr   place to put the first required character, or a negative number
6396    reqbyteptr     place to put the last required character, or a negative number    reqbyteptr     place to put the last required character, or a negative number
6397    bcptr          pointer to the chain of currently open branches    bcptr          pointer to the chain of currently open branches
# Line 6128  Returns: TRUE on success Line 6403  Returns: TRUE on success
6403  */  */
6404    
6405  static BOOL  static BOOL
6406  compile_regex(int options, int oldims, uschar **codeptr, const uschar **ptrptr,  compile_regex(int options, uschar **codeptr, const uschar **ptrptr,
6407    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,
6408    int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr, compile_data *cd,    int cond_depth, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
6409    int *lengthptr)    compile_data *cd, int *lengthptr)
6410  {  {
6411  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
6412  uschar *code = *codeptr;  uschar *code = *codeptr;
# Line 6145  int branchfirstbyte, branchreqbyte; Line 6420  int branchfirstbyte, branchreqbyte;
6420  int length;  int length;
6421  int orig_bracount;  int orig_bracount;
6422  int max_bracount;  int max_bracount;
 int old_external_options = cd->external_options;  
6423  branch_chain bc;  branch_chain bc;
6424    
6425  bc.outer = bcptr;  bc.outer = bcptr;
# Line 6169  pre-compile phase to find out whether an Line 6443  pre-compile phase to find out whether an
6443    
6444  /* 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
6445  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
6446  detect groups that contain recursive back references to themselves. */  detect groups that contain recursive back references to themselves. Note that
6447    only OP_CBRA need be tested here; changing this opcode to one of its variants,
6448    e.g. OP_SCBRAPOS, happens later, after the group has been compiled. */
6449    
6450  if (*code == OP_CBRA)  if (*code == OP_CBRA)
6451    {    {
# Line 6195  for (;;) Line 6471  for (;;)
6471    
6472    if (reset_bracount) cd->bracount = orig_bracount;    if (reset_bracount) cd->bracount = orig_bracount;
6473    
   /* 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;  
     }  
   
6474    /* Set up dummy OP_REVERSE if lookbehind assertion */    /* Set up dummy OP_REVERSE if lookbehind assertion */
6475    
6476    if (lookbehind)    if (lookbehind)
# Line 6218  for (;;) Line 6485  for (;;)
6485    into the length. */    into the length. */
6486    
6487    if (!compile_branch(&options, &code, &ptr, errorcodeptr, &branchfirstbyte,    if (!compile_branch(&options, &code, &ptr, errorcodeptr, &branchfirstbyte,
6488          &branchreqbyte, &bc, cd, (lengthptr == NULL)? NULL : &length))          &branchreqbyte, &bc, cond_depth, cd,
6489            (lengthptr == NULL)? NULL : &length))
6490      {      {
6491      *ptrptr = ptr;      *ptrptr = ptr;
6492      return FALSE;      return FALSE;
6493      }      }
6494    
   /* 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;  
   
6495    /* Keep the highest bracket count in case (?| was used and some branch    /* Keep the highest bracket count in case (?| was used and some branch
6496    has fewer than the rest. */    has fewer than the rest. */
6497    
# Line 6293  for (;;) Line 6552  for (;;)
6552        {        {
6553        int fixed_length;        int fixed_length;
6554        *code = OP_END;        *code = OP_END;
6555        fixed_length = find_fixedlength(last_branch, options, FALSE, cd);        fixed_length = find_fixedlength(last_branch,  (options & PCRE_UTF8) != 0,
6556            FALSE, cd);
6557        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
6558        if (fixed_length == -3)        if (fixed_length == -3)
6559          {          {
# Line 6314  for (;;) Line 6574  for (;;)
6574    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
6575    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
6576    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
6577    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. */  
6578    
6579    if (*ptr != CHAR_VERTICAL_LINE)    if (*ptr != CHAR_VERTICAL_LINE)
6580      {      {
# Line 6360  for (;;) Line 6618  for (;;)
6618        cd->open_caps = cd->open_caps->next;        cd->open_caps = cd->open_caps->next;
6619        }        }
6620    
     /* Reset options if needed. */  
   
     if ((options & PCRE_IMS) != oldims && *ptr == CHAR_RIGHT_PARENTHESIS)  
       {  
       *code++ = OP_OPT;  
       *code++ = oldims;  
       length += 2;  
       }  
   
6621      /* Retain the highest bracket number, in case resetting was used. */      /* Retain the highest bracket number, in case resetting was used. */
6622    
6623      cd->bracount = max_bracount;      cd->bracount = max_bracount;
# Line 6428  for (;;) Line 6677  for (;;)
6677  /* 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
6678  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
6679  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
6680  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
6681  counts, since OP_CIRC can match in the middle.  be found, because ^ generates OP_CIRCM in that mode.
6682    
6683  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.
6684  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 6450  of the more common cases more precisely. Line 6699  of the more common cases more precisely.
6699    
6700  Arguments:  Arguments:
6701    code           points to start of expression (the bracket)    code           points to start of expression (the bracket)
   options        points to the options setting  
6702    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
6703                    handles up to substring 31; after that we just have to take                    handles up to substring 31; after that we just have to take
6704                    the less precise approach                    the less precise approach
# Line 6460  Returns: TRUE or FALSE Line 6708  Returns: TRUE or FALSE
6708  */  */
6709    
6710  static BOOL  static BOOL
6711  is_anchored(register const uschar *code, int *options, unsigned int bracket_map,  is_anchored(register const uschar *code, unsigned int bracket_map,
6712    unsigned int backref_map)    unsigned int backref_map)
6713  {  {
6714  do {  do {
6715     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],
6716       options, PCRE_MULTILINE, FALSE);       FALSE);
6717     register int op = *scode;     register int op = *scode;
6718    
6719     /* Non-capturing brackets */     /* Non-capturing brackets */
6720    
6721     if (op == OP_BRA)     if (op == OP_BRA  || op == OP_BRAPOS ||
6722           op == OP_SBRA || op == OP_SBRAPOS)
6723       {       {
6724       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, bracket_map, backref_map)) return FALSE;
6725       }       }
6726    
6727     /* Capturing brackets */     /* Capturing brackets */
6728    
6729     else if (op == OP_CBRA)     else if (op == OP_CBRA  || op == OP_CBRAPOS ||
6730                op == OP_SCBRA || op == OP_SCBRAPOS)
6731       {       {
6732       int n = GET2(scode, 1+LINK_SIZE);       int n = GET2(scode, 1+LINK_SIZE);
6733       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);
6734       if (!is_anchored(scode, options, new_map, backref_map)) return FALSE;       if (!is_anchored(scode, new_map, backref_map)) return FALSE;
6735       }       }
6736    
6737     /* Other brackets */     /* Other brackets */
6738    
6739     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_COND)     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_COND)
6740       {       {
6741       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, bracket_map, backref_map)) return FALSE;
6742       }       }
6743    
6744     /* .* 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 6503  do { Line 6753  do {
6753    
6754     /* Check for explicit anchoring */     /* Check for explicit anchoring */
6755    
6756     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;  
6757     code += GET(code, 1);     code += GET(code, 1);
6758     }     }
6759  while (*code == OP_ALT);   /* Loop for each alternative */  while (*code == OP_ALT);   /* Loop for each alternative */
# Line 6541  is_startline(const uschar *code, unsigne Line 6789  is_startline(const uschar *code, unsigne
6789  {  {
6790  do {  do {
6791     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],
6792       NULL, 0, FALSE);       FALSE);
6793     register int op = *scode;     register int op = *scode;
6794    
6795     /* 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 6568  do { Line 6816  do {
6816         scode += 1 + LINK_SIZE;         scode += 1 + LINK_SIZE;
6817         break;         break;
6818         }         }
6819       scode = first_significant_code(scode, NULL, 0, FALSE);       scode = first_significant_code(scode, FALSE);
6820       op = *scode;       op = *scode;
6821       }       }
6822    
6823     /* Non-capturing brackets */     /* Non-capturing brackets */
6824    
6825     if (op == OP_BRA)     if (op == OP_BRA  || op == OP_BRAPOS ||
6826           op == OP_SBRA || op == OP_SBRAPOS)
6827       {       {
6828       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
6829       }       }
6830    
6831     /* Capturing brackets */     /* Capturing brackets */
6832    
6833     else if (op == OP_CBRA)     else if (op == OP_CBRA  || op == OP_CBRAPOS ||
6834                op == OP_SCBRA || op == OP_SCBRAPOS)
6835       {       {
6836       int n = GET2(scode, 1+LINK_SIZE);       int n = GET2(scode, 1+LINK_SIZE);
6837       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);
# Line 6605  do { Line 6855  do {
6855    
6856     /* Check for explicit circumflex */     /* Check for explicit circumflex */
6857    
6858     else if (op != OP_CIRC) return FALSE;     else if (op != OP_CIRC && op != OP_CIRCM) return FALSE;
6859    
6860     /* Move on to the next alternative */     /* Move on to the next alternative */
6861    
# Line 6631  we return that char, otherwise -1. Line 6881  we return that char, otherwise -1.
6881    
6882  Arguments:  Arguments:
6883    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)  
6884    inassert   TRUE if in an assertion    inassert   TRUE if in an assertion
6885    
6886  Returns:     -1 or the fixed first char  Returns:     -1 or the fixed first char
6887  */  */
6888    
6889  static int  static int
6890  find_firstassertedchar(const uschar *code, int *options, BOOL inassert)  find_firstassertedchar(const uschar *code, BOOL inassert)
6891  {  {
6892  register int c = -1;  register int c = -1;
6893  do {  do {
6894     int d;     int d;
6895     const uschar *scode =     int xl = (*code == OP_CBRA || *code == OP_SCBRA ||
6896       first_significant_code(code + 1+LINK_SIZE, options, PCRE_CASELESS, TRUE);               *code == OP_CBRAPOS || *code == OP_SCBRAPOS)? 2:0;
6897       const uschar *scode = first_significant_code(code + 1+LINK_SIZE + xl, TRUE);
6898     register int op = *scode;     register int op = *scode;
6899    
6900     switch(op)     switch(op)
# Line 6653  do { Line 6903  do {
6903       return -1;       return -1;
6904    
6905       case OP_BRA:       case OP_BRA:
6906         case OP_BRAPOS:
6907       case OP_CBRA:       case OP_CBRA:
6908         case OP_SCBRA:
6909         case OP_CBRAPOS:
6910         case OP_SCBRAPOS:
6911       case OP_ASSERT:       case OP_ASSERT:
6912       case OP_ONCE:       case OP_ONCE:
6913       case OP_COND:       case OP_COND:
6914       if ((d = find_firstassertedchar(scode, options, op == OP_ASSERT)) < 0)       if ((d = find_firstassertedchar(scode, op == OP_ASSERT)) < 0)
6915         return -1;         return -1;
6916       if (c < 0) c = d; else if (c != d) return -1;       if (c < 0) c = d; else if (c != d) return -1;
6917       break;       break;
6918    
6919       case OP_EXACT:       /* Fall through */       case OP_EXACT:
6920       scode += 2;       scode += 2;
6921         /* Fall through */
6922    
6923       case OP_CHAR:       case OP_CHAR:
      case OP_CHARNC:  
6924       case OP_PLUS:       case OP_PLUS:
6925       case OP_MINPLUS:       case OP_MINPLUS:
6926       case OP_POSPLUS:       case OP_POSPLUS:
6927       if (!inassert) return -1;       if (!inassert) return -1;
6928       if (c < 0)       if (c < 0) c = scode[1];
6929         {         else if (c != scode[1]) return -1;
6930         c = scode[1];       break;
6931         if ((*options & PCRE_CASELESS) != 0) c |= REQ_CASELESS;  
6932         }       case OP_EXACTI:
6933       else if (c != scode[1]) return -1;       scode += 2;
6934         /* Fall through */
6935    
6936         case OP_CHARI:
6937         case OP_PLUSI:
6938         case OP_MINPLUSI:
6939         case OP_POSPLUSI:
6940         if (!inassert) return -1;
6941         if (c < 0) c = scode[1] | REQ_CASELESS;
6942           else if (c != scode[1]) return -1;
6943       break;       break;
6944       }       }
6945    
# Line 6799  while (ptr[skipatstart] == CHAR_LEFT_PAR Line 7062  while (ptr[skipatstart] == CHAR_LEFT_PAR
7062      { skipatstart += 7; options |= PCRE_UTF8; continue; }      { skipatstart += 7; options |= PCRE_UTF8; continue; }
7063    else if (strncmp((char *)(ptr+skipatstart+2), STRING_UCP_RIGHTPAR, 4) == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_UCP_RIGHTPAR, 4) == 0)
7064      { skipatstart += 6; options |= PCRE_UCP; continue; }      { skipatstart += 6; options |= PCRE_UCP; continue; }
7065      else if (strncmp((char *)(ptr+skipatstart+2), STRING_NO_START_OPT_RIGHTPAR, 13) == 0)
7066        { skipatstart += 15; options |= PCRE_NO_START_OPTIMIZE; continue; }
7067    
7068    if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)    if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)
7069      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }
# Line 6825  while (ptr[skipatstart] == CHAR_LEFT_PAR Line 7090  while (ptr[skipatstart] == CHAR_LEFT_PAR
7090    
7091  utf8 = (options & PCRE_UTF8) != 0;  utf8 = (options & PCRE_UTF8) != 0;
7092    
7093  /* 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
7094    return of an error code from _pcre_valid_utf8() is a new feature, introduced in
7095    release 8.13. It is passed back from pcre_[dfa_]exec(), but at the moment is
7096    not used here. */
7097    
7098  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
7099  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&
7100       (*erroroffset = _pcre_valid_utf8((USPTR)pattern, -1)) >= 0)       (errorcode = _pcre_valid_utf8((USPTR)pattern, -1, erroroffset)) != 0)
7101    {    {
7102    errorcode = ERR44;    errorcode = ERR44;
7103    goto PCRE_EARLY_ERROR_RETURN2;    goto PCRE_EARLY_ERROR_RETURN2;
# Line 6854  if ((options & PCRE_UCP) != 0) Line 7122  if ((options & PCRE_UCP) != 0)
7122    
7123  /* Check validity of \R options. */  /* Check validity of \R options. */
7124    
7125  switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))  if ((options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) ==
7126         (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))
7127    {    {
7128    case 0:    errorcode = ERR56;
7129    case PCRE_BSR_ANYCRLF:    goto PCRE_EARLY_ERROR_RETURN;
   case PCRE_BSR_UNICODE:  
   break;  
   default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;  
7130    }    }
7131    
7132  /* Handle different types of newline. The three bits give seven cases. The  /* Handle different types of newline. The three bits give seven cases. The
# Line 6945  outside can help speed up starting point Line 7211  outside can help speed up starting point
7211  ptr += skipatstart;  ptr += skipatstart;
7212  code = cworkspace;  code = cworkspace;
7213  *code = OP_BRA;  *code = OP_BRA;
7214  (void)compile_regex(cd->external_options, cd->external_options & PCRE_IMS,  (void)compile_regex(cd->external_options, &code, &ptr, &errorcode, FALSE,
7215    &code, &ptr, &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd,    FALSE, 0, 0, &firstbyte, &reqbyte, NULL, cd, &length);
   &length);  
7216  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;
7217    
7218  DPRINTF(("end pre-compile: length=%d workspace=%d\n", length,  DPRINTF(("end pre-compile: length=%d workspace=%d\n", length,
# Line 7001  field; this time it's used for rememberi Line 7266  field; this time it's used for rememberi
7266  */  */
7267    
7268  cd->final_bracount = cd->bracount;  /* Save for checking forward references */  cd->final_bracount = cd->bracount;  /* Save for checking forward references */
7269    cd->assert_depth = 0;
7270  cd->bracount = 0;  cd->bracount = 0;
7271  cd->names_found = 0;  cd->names_found = 0;
7272  cd->name_table = (uschar *)re + re->name_table_offset;  cd->name_table = (uschar *)re + re->name_table_offset;
# Line 7019  of the function here. */ Line 7285  of the function here. */
7285  ptr = (const uschar *)pattern + skipatstart;  ptr = (const uschar *)pattern + skipatstart;
7286  code = (uschar *)codestart;  code = (uschar *)codestart;
7287  *code = OP_BRA;  *code = OP_BRA;
7288  (void)compile_regex(re->options, re->options & PCRE_IMS, &code, &ptr,  (void)compile_regex(re->options, &code, &ptr, &errorcode, FALSE, FALSE, 0, 0,
7289    &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd, NULL);    &firstbyte, &reqbyte, NULL, cd, NULL);
7290  re->top_bracket = cd->bracount;  re->top_bracket = cd->bracount;
7291  re->top_backref = cd->top_backref;  re->top_backref = cd->top_backref;
7292  re->flags = cd->external_flags;  re->flags = cd->external_flags;
7293    
7294  if (cd->had_accept) reqbyte = -1;   /* Must disable after (*ACCEPT) */  if (cd->had_accept) reqbyte = REQ_NONE;   /* Must disable after (*ACCEPT) */
7295    
7296  /* If not reached end of pattern on success, there's an excess bracket. */  /* If not reached end of pattern on success, there's an excess bracket. */
7297    
# Line 7086  if (cd->check_lookbehind) Line 7352  if (cd->check_lookbehind)
7352        uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);        uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);
7353        int end_op = *be;        int end_op = *be;
7354        *be = OP_END;        *be = OP_END;
7355        fixed_length = find_fixedlength(cc, re->options, TRUE, cd);        fixed_length = find_fixedlength(cc, (re->options & PCRE_UTF8) != 0, TRUE,
7356            cd);
7357        *be = end_op;        *be = end_op;
7358        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
7359        if (fixed_length < 0)        if (fixed_length < 0)
# Line 7125  start with ^. and also when all branches Line 7392  start with ^. and also when all branches
7392    
7393  if ((re->options & PCRE_ANCHORED) == 0)  if ((re->options & PCRE_ANCHORED) == 0)
7394    {    {
7395    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))  
7396      re->options |= PCRE_ANCHORED;      re->options |= PCRE_ANCHORED;
7397    else    else
7398      {      {
7399      if (firstbyte < 0)      if (firstbyte < 0)
7400        firstbyte = find_firstassertedchar(codestart, &temp_options, FALSE);        firstbyte = find_firstassertedchar(codestart, FALSE);
7401      if (firstbyte >= 0)   /* Remove caseless flag for non-caseable chars */      if (firstbyte >= 0)   /* Remove caseless flag for non-caseable chars */
7402        {        {
7403        int ch = firstbyte & 255;        int ch = firstbyte & 255;

Legend:
Removed from v.545  
changed lines
  Added in v.708

webmaster@exim.org
ViewVC Help
Powered by ViewVC 1.1.12