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

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

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

revision 532 by ph10, Tue Jun 1 16:21:42 2010 UTC revision 613 by ph10, Sat Jul 2 16:59:52 2011 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2010 University of Cambridge             Copyright (c) 1997-2011 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 261  static const int posix_class_maps[] = { Line 261  static const int posix_class_maps[] = {
261    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
262  };  };
263    
264  /* Table of substitutes for \d etc when PCRE_UCP is set. The POSIX class  /* Table of substitutes for \d etc when PCRE_UCP is set. The POSIX class
265  substitutes must be in the order of the names, defined above, and there are  substitutes must be in the order of the names, defined above, and there are
266  both positive and negative cases. NULL means no substitute. */  both positive and negative cases. NULL means no substitute. */
267    
268  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
# Line 272  static const uschar *substitutes[] = { Line 272  static const uschar *substitutes[] = {
272    (uschar *)"\\P{Xsp}",   /* \S */       /* NOTE: Xsp is Perl space */    (uschar *)"\\P{Xsp}",   /* \S */       /* NOTE: Xsp is Perl space */
273    (uschar *)"\\p{Xsp}",   /* \s */    (uschar *)"\\p{Xsp}",   /* \s */
274    (uschar *)"\\P{Xwd}",   /* \W */    (uschar *)"\\P{Xwd}",   /* \W */
275    (uschar *)"\\p{Xwd}"    /* \w */    (uschar *)"\\p{Xwd}"    /* \w */
276  };  };
277    
278  static const uschar *posix_substitutes[] = {  static const uschar *posix_substitutes[] = {
279    (uschar *)"\\p{L}",     /* alpha */    (uschar *)"\\p{L}",     /* alpha */
280    (uschar *)"\\p{Ll}",    /* lower */    (uschar *)"\\p{Ll}",    /* lower */
281    (uschar *)"\\p{Lu}",    /* upper */    (uschar *)"\\p{Lu}",    /* upper */
282    (uschar *)"\\p{Xan}",   /* alnum */    (uschar *)"\\p{Xan}",   /* alnum */
283    NULL,                   /* ascii */    NULL,                   /* ascii */
284    (uschar *)"\\h",        /* blank */    (uschar *)"\\h",        /* blank */
285    NULL,                   /* cntrl */    NULL,                   /* cntrl */
# Line 289  static const uschar *posix_substitutes[] Line 289  static const uschar *posix_substitutes[]
289    NULL,                   /* punct */    NULL,                   /* punct */
290    (uschar *)"\\p{Xps}",   /* space */    /* NOTE: Xps is POSIX space */    (uschar *)"\\p{Xps}",   /* space */    /* NOTE: Xps is POSIX space */
291    (uschar *)"\\p{Xwd}",   /* word */    (uschar *)"\\p{Xwd}",   /* word */
292    NULL,                   /* xdigit */    NULL,                   /* xdigit */
293    /* Negated cases */    /* Negated cases */
294    (uschar *)"\\P{L}",     /* ^alpha */    (uschar *)"\\P{L}",     /* ^alpha */
295    (uschar *)"\\P{Ll}",    /* ^lower */    (uschar *)"\\P{Ll}",    /* ^lower */
296    (uschar *)"\\P{Lu}",    /* ^upper */    (uschar *)"\\P{Lu}",    /* ^upper */
297    (uschar *)"\\P{Xan}",   /* ^alnum */    (uschar *)"\\P{Xan}",   /* ^alnum */
298    NULL,                   /* ^ascii */    NULL,                   /* ^ascii */
299    (uschar *)"\\H",        /* ^blank */    (uschar *)"\\H",        /* ^blank */
300    NULL,                   /* ^cntrl */    NULL,                   /* ^cntrl */
# Line 304  static const uschar *posix_substitutes[] Line 304  static const uschar *posix_substitutes[]
304    NULL,                   /* ^punct */    NULL,                   /* ^punct */
305    (uschar *)"\\P{Xps}",   /* ^space */   /* NOTE: Xps is POSIX space */    (uschar *)"\\P{Xps}",   /* ^space */   /* NOTE: Xps is POSIX space */
306    (uschar *)"\\P{Xwd}",   /* ^word */    (uschar *)"\\P{Xwd}",   /* ^word */
307    NULL                    /* ^xdigit */    NULL                    /* ^xdigit */
308  };  };
309  #define POSIX_SUBSIZE (sizeof(posix_substitutes)/sizeof(uschar *))  #define POSIX_SUBSIZE (sizeof(posix_substitutes)/sizeof(uschar *))
310  #endif  #endif
311    
312  #define STRING(a)  # a  #define STRING(a)  # a
313  #define XSTRING(s) STRING(s)  #define XSTRING(s) STRING(s)
# Line 407  static const char error_texts[] = Line 407  static const char error_texts[] =
407    /* 65 */    /* 65 */
408    "different names for subpatterns of the same number are not allowed\0"    "different names for subpatterns of the same number are not allowed\0"
409    "(*MARK) must have an argument\0"    "(*MARK) must have an argument\0"
410    "this version of PCRE is not compiled with PCRE_UCP support\0"    "this version of PCRE is not compiled with PCRE_UCP support\0"
411      "\\c must be followed by an ASCII character\0"
412    ;    ;
413    
414  /* 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 545  static const unsigned char ebcdic_charta
545  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
546    
547  static BOOL  static BOOL
548    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,    compile_regex(int, uschar **, const uschar **, int *, BOOL, BOOL, int, int *,
549      int *, int *, branch_chain *, compile_data *, int *);      int *, branch_chain *, compile_data *, int *);
550    
551    
552    
# Line 841  else Line 842  else
842      break;      break;
843    
844      /* 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.
845      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
846        coding is ASCII-specific, but then the whole concept of \cx is
847      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
848    
849      case CHAR_c:      case CHAR_c:
# Line 851  else Line 853  else
853        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
854        break;        break;
855        }        }
856    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
857  #ifndef EBCDIC  /* ASCII/UTF-8 coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
858          {
859          *errorcodeptr = ERR68;
860          break;
861          }
862      if (c >= CHAR_a && c <= CHAR_z) c -= 32;      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
863      c ^= 0x40;      c ^= 0x40;
864  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
865      if (c >= CHAR_a && c <= CHAR_z) c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
866      c ^= 0xC0;      c ^= 0xC0;
867  #endif  #endif
# Line 1099  top-level call starts at the beginning o Line 1105  top-level call starts at the beginning o
1105  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
1106  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
1107  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
1108  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
1109  encountered, the name will be terminated by '>' because that is checked in the  track of subpatterns that reset the capturing group numbers - the (?| feature.
1110  first pass. Recursion is used to keep track of subpatterns that reset the  
1111  capturing group numbers - the (?| feature.  This function was originally called only from the second pass, in which we know
1112    that if (?< or (?' or (?P< is encountered, the name will be correctly
1113    terminated because that is checked in the first pass. There is now one call to
1114    this function in the first pass, to check for a recursive back reference by
1115    name (so that we can make the whole group atomic). In this case, we need check
1116    only up to the current position in the pattern, and that is still OK because
1117    and previous occurrences will have been checked. To make this work, the test
1118    for "end of pattern" is a check against cd->end_pattern in the main loop,
1119    instead of looking for a binary zero. This means that the special first-pass
1120    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1121    processing items within the loop are OK, because afterwards the main loop will
1122    terminate.)
1123    
1124  Arguments:  Arguments:
1125    ptrptr       address of the current character pointer (updated)    ptrptr       address of the current character pointer (updated)
# Line 1110  Arguments: Line 1127  Arguments:
1127    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1128    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1129    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1130      utf8         TRUE if we are in UTF-8 mode
1131    count        pointer to the current capturing subpattern number (updated)    count        pointer to the current capturing subpattern number (updated)
1132    
1133  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 1135  Returns:       the number of the named s
1135    
1136  static int  static int
1137  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,
1138    BOOL xmode, int *count)    BOOL xmode, BOOL utf8, int *count)
1139  {  {
1140  uschar *ptr = *ptrptr;  uschar *ptr = *ptrptr;
1141  int start_count = *count;  int start_count = *count;
# Line 1129  dealing with. The very first call may no Line 1147  dealing with. The very first call may no
1147    
1148  if (ptr[0] == CHAR_LEFT_PARENTHESIS)  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1149    {    {
1150    if (ptr[1] == CHAR_QUESTION_MARK &&    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1151        ptr[2] == CHAR_VERTICAL_LINE)  
1152      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1153    
1154      /* Handle a normal, unnamed capturing parenthesis. */
1155    
1156      else if (ptr[1] != CHAR_QUESTION_MARK)
1157        {
1158        *count += 1;
1159        if (name == NULL && *count == lorn) return *count;
1160        ptr++;
1161        }
1162    
1163      /* All cases now have (? at the start. Remember when we are in a group
1164      where the parenthesis numbers are duplicated. */
1165    
1166      else if (ptr[2] == CHAR_VERTICAL_LINE)
1167      {      {
1168      ptr += 3;      ptr += 3;
1169      dup_parens = TRUE;      dup_parens = TRUE;
1170      }      }
1171    
1172    /* Handle a normal, unnamed capturing parenthesis */    /* Handle comments; all characters are allowed until a ket is reached. */
1173    
1174    else if (ptr[1] != CHAR_QUESTION_MARK && ptr[1] != CHAR_ASTERISK)    else if (ptr[2] == CHAR_NUMBER_SIGN)
1175      {      {
1176      *count += 1;      for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1177      if (name == NULL && *count == lorn) return *count;      goto FAIL_EXIT;
     ptr++;  
1178      }      }
1179    
1180    /* Handle a condition. If it is an assertion, just carry on so that it    /* Handle a condition. If it is an assertion, just carry on so that it
1181    is processed as normal. If not, skip to the closing parenthesis of the    is processed as normal. If not, skip to the closing parenthesis of the
1182    condition (there can't be any nested parens. */    condition (there can't be any nested parens). */
1183    
1184    else if (ptr[2] == CHAR_LEFT_PARENTHESIS)    else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1185      {      {
# Line 1159  if (ptr[0] == CHAR_LEFT_PARENTHESIS) Line 1191  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1191        }        }
1192      }      }
1193    
1194    /* We have either (? or (* and not a condition */    /* Start with (? but not a condition. */
1195    
1196    else    else
1197      {      {
# Line 1188  if (ptr[0] == CHAR_LEFT_PARENTHESIS) Line 1220  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1220    }    }
1221    
1222  /* Past any initial parenthesis handling, scan for parentheses or vertical  /* Past any initial parenthesis handling, scan for parentheses or vertical
1223  bars. */  bars. Stop if we get to cd->end_pattern. Note that this is important for the
1224    first-pass call when this value is temporarily adjusted to stop at the current
1225    position. So DO NOT change this to a test for binary zero. */
1226    
1227  for (; *ptr != 0; ptr++)  for (; ptr < cd->end_pattern; ptr++)
1228    {    {
1229    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1230    
# Line 1264  for (; *ptr != 0; ptr++) Line 1298  for (; *ptr != 0; ptr++)
1298    
1299    if (xmode && *ptr == CHAR_NUMBER_SIGN)    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1300      {      {
1301      while (*(++ptr) != 0 && *ptr != CHAR_NL) {};      ptr++;
1302        while (*ptr != 0)
1303          {
1304          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1305          ptr++;
1306    #ifdef SUPPORT_UTF8
1307          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
1308    #endif
1309          }
1310      if (*ptr == 0) goto FAIL_EXIT;      if (*ptr == 0) goto FAIL_EXIT;
1311      continue;      continue;
1312      }      }
# Line 1273  for (; *ptr != 0; ptr++) Line 1315  for (; *ptr != 0; ptr++)
1315    
1316    if (*ptr == CHAR_LEFT_PARENTHESIS)    if (*ptr == CHAR_LEFT_PARENTHESIS)
1317      {      {
1318      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, count);      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, count);
1319      if (rc > 0) return rc;      if (rc > 0) return rc;
1320      if (*ptr == 0) goto FAIL_EXIT;      if (*ptr == 0) goto FAIL_EXIT;
1321      }      }
# Line 1281  for (; *ptr != 0; ptr++) Line 1323  for (; *ptr != 0; ptr++)
1323    else if (*ptr == CHAR_RIGHT_PARENTHESIS)    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1324      {      {
1325      if (dup_parens && *count < hwm_count) *count = hwm_count;      if (dup_parens && *count < hwm_count) *count = hwm_count;
1326      *ptrptr = ptr;      goto FAIL_EXIT;
     return -1;  
1327      }      }
1328    
1329    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
# Line 1320  Arguments: Line 1361  Arguments:
1361    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1362    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1363    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1364      utf8         TRUE if we are in UTF-8 mode
1365    
1366  Returns:       the number of the found subpattern, or -1 if not found  Returns:       the number of the found subpattern, or -1 if not found
1367  */  */
1368    
1369  static int  static int
1370  find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode)  find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode,
1371      BOOL utf8)
1372  {  {
1373  uschar *ptr = (uschar *)cd->start_pattern;  uschar *ptr = (uschar *)cd->start_pattern;
1374  int count = 0;  int count = 0;
# Line 1338  matching closing parens. That is why we Line 1381  matching closing parens. That is why we
1381    
1382  for (;;)  for (;;)
1383    {    {
1384    rc = find_parens_sub(&ptr, cd, name, lorn, xmode, &count);    rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, &count);
1385    if (rc > 0 || *ptr++ == 0) break;    if (rc > 0 || *ptr++ == 0) break;
1386    }    }
1387    
# Line 1354  return rc; Line 1397  return rc;
1397    
1398  /* This is called by several functions that scan a compiled expression looking  /* This is called by several functions that scan a compiled expression looking
1399  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
1400  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
1401  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
1402  assertions, and also the \b assertion; for others it does not.  does not.
1403    
1404  Arguments:  Arguments:
1405    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  
1406    skipassert   TRUE if certain assertions are to be skipped    skipassert   TRUE if certain assertions are to be skipped
1407    
1408  Returns:       pointer to the first significant opcode  Returns:       pointer to the first significant opcode
1409  */  */
1410    
1411  static const uschar*  static const uschar*
1412  first_significant_code(const uschar *code, int *options, int optbit,  first_significant_code(const uschar *code, BOOL skipassert)
   BOOL skipassert)  
1413  {  {
1414  for (;;)  for (;;)
1415    {    {
1416    switch ((int)*code)    switch ((int)*code)
1417      {      {
     case OP_OPT:  
     if (optbit > 0 && ((int)code[1] & optbit) != (*options & optbit))  
       *options = (int)code[1];  
     code += 2;  
     break;  
   
1418      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1419      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1420      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
# Line 1431  and doing the check at the end; a flag s Line 1464  and doing the check at the end; a flag s
1464    
1465  Arguments:  Arguments:
1466    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1467    options  the compiling options    utf8     TRUE in UTF-8 mode
1468    atend    TRUE if called when the pattern is complete    atend    TRUE if called when the pattern is complete
1469    cd       the "compile data" structure    cd       the "compile data" structure
1470    
# Line 1442  Returns:   the fixed length, Line 1475  Returns:   the fixed length,
1475  */  */
1476    
1477  static int  static int
1478  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)  find_fixedlength(uschar *code, BOOL utf8, BOOL atend, compile_data *cd)
1479  {  {
1480  int length = -1;  int length = -1;
1481    
# Line 1459  for (;;) Line 1492  for (;;)
1492    register int op = *cc;    register int op = *cc;
1493    switch (op)    switch (op)
1494      {      {
1495        /* We only need to continue for OP_CBRA (normal capturing bracket) and
1496        OP_BRA (normal non-capturing bracket) because the other variants of these
1497        opcodes are all concerned with unlimited repeated groups, which of course
1498        are not of fixed length. They will cause a -1 response from the default
1499        case of this switch. */
1500    
1501      case OP_CBRA:      case OP_CBRA:
1502      case OP_BRA:      case OP_BRA:
1503      case OP_ONCE:      case OP_ONCE:
1504      case OP_COND:      case OP_COND:
1505      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), utf8, atend, cd);
1506      if (d < 0) return d;      if (d < 0) return d;
1507      branchlength += d;      branchlength += d;
1508      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1472  for (;;) Line 1511  for (;;)
1511    
1512      /* 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
1513      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
1514      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.
1515        Note that we must not include the OP_KETRxxx opcodes here, because they
1516        all imply an unlimited repeat. */
1517    
1518      case OP_ALT:      case OP_ALT:
1519      case OP_KET:      case OP_KET:
     case OP_KETRMAX:  
     case OP_KETRMIN:  
1520      case OP_END:      case OP_END:
1521      if (length < 0) length = branchlength;      if (length < 0) length = branchlength;
1522        else if (length != branchlength) return -1;        else if (length != branchlength) return -1;
# Line 1495  for (;;) Line 1534  for (;;)
1534      cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */      cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1535      do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */      do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1536      if (cc > cs && cc < ce) return -1;                /* Recursion */      if (cc > cs && cc < ce) return -1;                /* Recursion */
1537      d = find_fixedlength(cs + 2, options, atend, cd);      d = find_fixedlength(cs + 2, utf8, atend, cd);
1538      if (d < 0) return d;      if (d < 0) return d;
1539      branchlength += d;      branchlength += d;
1540      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
# Line 1518  for (;;) Line 1557  for (;;)
1557      case OP_RREF:      case OP_RREF:
1558      case OP_NRREF:      case OP_NRREF:
1559      case OP_DEF:      case OP_DEF:
     case OP_OPT:  
1560      case OP_CALLOUT:      case OP_CALLOUT:
1561      case OP_SOD:      case OP_SOD:
1562      case OP_SOM:      case OP_SOM:
# Line 1526  for (;;) Line 1564  for (;;)
1564      case OP_EOD:      case OP_EOD:
1565      case OP_EODN:      case OP_EODN:
1566      case OP_CIRC:      case OP_CIRC:
1567        case OP_CIRCM:
1568      case OP_DOLL:      case OP_DOLL:
1569        case OP_DOLLM:
1570      case OP_NOT_WORD_BOUNDARY:      case OP_NOT_WORD_BOUNDARY:
1571      case OP_WORD_BOUNDARY:      case OP_WORD_BOUNDARY:
1572      cc += _pcre_OP_lengths[*cc];      cc += _pcre_OP_lengths[*cc];
# Line 1535  for (;;) Line 1575  for (;;)
1575      /* Handle literal characters */      /* Handle literal characters */
1576    
1577      case OP_CHAR:      case OP_CHAR:
1578      case OP_CHARNC:      case OP_CHARI:
1579      case OP_NOT:      case OP_NOT:
1580        case OP_NOTI:
1581      branchlength++;      branchlength++;
1582      cc += 2;      cc += 2;
1583  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1584      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];  
1585  #endif  #endif
1586      break;      break;
1587    
# Line 1552  for (;;) Line 1592  for (;;)
1592      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1593      cc += 4;      cc += 4;
1594  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1595      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];  
1596  #endif  #endif
1597      break;      break;
1598    
# Line 1673  for (;;) Line 1712  for (;;)
1712    
1713    /* Handle capturing bracket */    /* Handle capturing bracket */
1714    
1715    else if (c == OP_CBRA)    else if (c == OP_CBRA || c == OP_SCBRA ||
1716               c == OP_CBRAPOS || c == OP_SCBRAPOS)
1717      {      {
1718      int n = GET2(code, 1+LINK_SIZE);      int n = GET2(code, 1+LINK_SIZE);
1719      if (n == number) return (uschar *)code;      if (n == number) return (uschar *)code;
# Line 1711  for (;;) Line 1751  for (;;)
1751        case OP_MARK:        case OP_MARK:
1752        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
1753        case OP_SKIP_ARG:        case OP_SKIP_ARG:
       case OP_THEN_ARG:  
1754        code += code[1];        code += code[1];
1755        break;        break;
1756    
1757          case OP_THEN_ARG:
1758          code += code[1+LINK_SIZE];
1759          break;
1760        }        }
1761    
1762      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1728  for (;;) Line 1771  for (;;)
1771      if (utf8) switch(c)      if (utf8) switch(c)
1772        {        {
1773        case OP_CHAR:        case OP_CHAR:
1774        case OP_CHARNC:        case OP_CHARI:
1775        case OP_EXACT:        case OP_EXACT:
1776          case OP_EXACTI:
1777        case OP_UPTO:        case OP_UPTO:
1778          case OP_UPTOI:
1779        case OP_MINUPTO:        case OP_MINUPTO:
1780          case OP_MINUPTOI:
1781        case OP_POSUPTO:        case OP_POSUPTO:
1782          case OP_POSUPTOI:
1783        case OP_STAR:        case OP_STAR:
1784          case OP_STARI:
1785        case OP_MINSTAR:        case OP_MINSTAR:
1786          case OP_MINSTARI:
1787        case OP_POSSTAR:        case OP_POSSTAR:
1788          case OP_POSSTARI:
1789        case OP_PLUS:        case OP_PLUS:
1790          case OP_PLUSI:
1791        case OP_MINPLUS:        case OP_MINPLUS:
1792          case OP_MINPLUSI:
1793        case OP_POSPLUS:        case OP_POSPLUS:
1794          case OP_POSPLUSI:
1795        case OP_QUERY:        case OP_QUERY:
1796          case OP_QUERYI:
1797        case OP_MINQUERY:        case OP_MINQUERY:
1798          case OP_MINQUERYI:
1799        case OP_POSQUERY:        case OP_POSQUERY:
1800          case OP_POSQUERYI:
1801        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1802        break;        break;
1803        }        }
# Line 1814  for (;;) Line 1870  for (;;)
1870        case OP_MARK:        case OP_MARK:
1871        case OP_PRUNE_ARG:        case OP_PRUNE_ARG:
1872        case OP_SKIP_ARG:        case OP_SKIP_ARG:
       case OP_THEN_ARG:  
1873        code += code[1];        code += code[1];
1874        break;        break;
1875    
1876          case OP_THEN_ARG:
1877          code += code[1+LINK_SIZE];
1878          break;
1879        }        }
1880    
1881      /* Add in the fixed length from the table */      /* Add in the fixed length from the table */
# Line 1831  for (;;) Line 1890  for (;;)
1890      if (utf8) switch(c)      if (utf8) switch(c)
1891        {        {
1892        case OP_CHAR:        case OP_CHAR:
1893        case OP_CHARNC:        case OP_CHARI:
1894        case OP_EXACT:        case OP_EXACT:
1895          case OP_EXACTI:
1896        case OP_UPTO:        case OP_UPTO:
1897          case OP_UPTOI:
1898        case OP_MINUPTO:        case OP_MINUPTO:
1899          case OP_MINUPTOI:
1900        case OP_POSUPTO:        case OP_POSUPTO:
1901          case OP_POSUPTOI:
1902        case OP_STAR:        case OP_STAR:
1903          case OP_STARI:
1904        case OP_MINSTAR:        case OP_MINSTAR:
1905          case OP_MINSTARI:
1906        case OP_POSSTAR:        case OP_POSSTAR:
1907          case OP_POSSTARI:
1908        case OP_PLUS:        case OP_PLUS:
1909          case OP_PLUSI:
1910        case OP_MINPLUS:        case OP_MINPLUS:
1911          case OP_MINPLUSI:
1912        case OP_POSPLUS:        case OP_POSPLUS:
1913          case OP_POSPLUSI:
1914        case OP_QUERY:        case OP_QUERY:
1915          case OP_QUERYI:
1916        case OP_MINQUERY:        case OP_MINQUERY:
1917          case OP_MINQUERYI:
1918        case OP_POSQUERY:        case OP_POSQUERY:
1919          case OP_POSQUERYI:
1920        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1921        break;        break;
1922        }        }
# Line 1883  could_be_empty_branch(const uschar *code Line 1955  could_be_empty_branch(const uschar *code
1955    compile_data *cd)    compile_data *cd)
1956  {  {
1957  register int c;  register int c;
1958  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], TRUE);
1959       code < endcode;       code < endcode;
1960       code = first_significant_code(code + _pcre_OP_lengths[c], NULL, 0, TRUE))       code = first_significant_code(code + _pcre_OP_lengths[c], TRUE))
1961    {    {
1962    const uschar *ccode;    const uschar *ccode;
1963    
# Line 1901  for (code = first_significant_code(code Line 1973  for (code = first_significant_code(code
1973      continue;      continue;
1974      }      }
1975    
   /* 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;  
     }  
   
1976    /* For a recursion/subroutine call, if its end has been reached, which    /* For a recursion/subroutine call, if its end has been reached, which
1977    implies a subroutine call, we can scan it. */    implies a subroutine call, we can scan it. */
1978    
# Line 1933  for (code = first_significant_code(code Line 1995  for (code = first_significant_code(code
1995      continue;      continue;
1996      }      }
1997    
1998      /* Groups with zero repeats can of course be empty; skip them. */
1999    
2000      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2001          c == OP_BRAPOSZERO)
2002        {
2003        code += _pcre_OP_lengths[c];
2004        do code += GET(code, 1); while (*code == OP_ALT);
2005        c = *code;
2006        continue;
2007        }
2008    
2009      /* A nested group that is already marked as "could be empty" can just be
2010      skipped. */
2011    
2012      if (c == OP_SBRA  || c == OP_SBRAPOS ||
2013          c == OP_SCBRA || c == OP_SCBRAPOS)
2014        {
2015        do code += GET(code, 1); while (*code == OP_ALT);
2016        c = *code;
2017        continue;
2018        }
2019    
2020    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
2021    
2022    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_BRA  || c == OP_BRAPOS ||
2023          c == OP_CBRA || c == OP_CBRAPOS ||
2024          c == OP_ONCE || c == OP_COND)
2025      {      {
2026      BOOL empty_branch;      BOOL empty_branch;
2027      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
# Line 2022  for (code = first_significant_code(code Line 2108  for (code = first_significant_code(code
2108      case OP_ALLANY:      case OP_ALLANY:
2109      case OP_ANYBYTE:      case OP_ANYBYTE:
2110      case OP_CHAR:      case OP_CHAR:
2111      case OP_CHARNC:      case OP_CHARI:
2112      case OP_NOT:      case OP_NOT:
2113        case OP_NOTI:
2114      case OP_PLUS:      case OP_PLUS:
2115      case OP_MINPLUS:      case OP_MINPLUS:
2116      case OP_POSPLUS:      case OP_POSPLUS:
# Line 2063  for (code = first_significant_code(code Line 2150  for (code = first_significant_code(code
2150      case OP_KET:      case OP_KET:
2151      case OP_KETRMAX:      case OP_KETRMAX:
2152      case OP_KETRMIN:      case OP_KETRMIN:
2153        case OP_KETRPOS:
2154      case OP_ALT:      case OP_ALT:
2155      return TRUE;      return TRUE;
2156    
# Line 2071  for (code = first_significant_code(code Line 2159  for (code = first_significant_code(code
2159    
2160  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2161      case OP_STAR:      case OP_STAR:
2162        case OP_STARI:
2163      case OP_MINSTAR:      case OP_MINSTAR:
2164        case OP_MINSTARI:
2165      case OP_POSSTAR:      case OP_POSSTAR:
2166        case OP_POSSTARI:
2167      case OP_QUERY:      case OP_QUERY:
2168        case OP_QUERYI:
2169      case OP_MINQUERY:      case OP_MINQUERY:
2170        case OP_MINQUERYI:
2171      case OP_POSQUERY:      case OP_POSQUERY:
2172        case OP_POSQUERYI:
2173      if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];      if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2174      break;      break;
2175    
2176      case OP_UPTO:      case OP_UPTO:
2177        case OP_UPTOI:
2178      case OP_MINUPTO:      case OP_MINUPTO:
2179        case OP_MINUPTOI:
2180      case OP_POSUPTO:      case OP_POSUPTO:
2181        case OP_POSUPTOI:
2182      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2183      break;      break;
2184  #endif  #endif
# Line 2092  for (code = first_significant_code(code Line 2189  for (code = first_significant_code(code
2189      case OP_MARK:      case OP_MARK:
2190      case OP_PRUNE_ARG:      case OP_PRUNE_ARG:
2191      case OP_SKIP_ARG:      case OP_SKIP_ARG:
     case OP_THEN_ARG:  
2192      code += code[1];      code += code[1];
2193      break;      break;
2194    
2195        case OP_THEN_ARG:
2196        code += code[1+LINK_SIZE];
2197        break;
2198    
2199      /* None of the remaining opcodes are required to match a character. */      /* None of the remaining opcodes are required to match a character. */
2200    
2201      default:      default:
# Line 2407  Arguments: Line 2507  Arguments:
2507    ptype        the property type    ptype        the property type
2508    pdata        the data for the type    pdata        the data for the type
2509    negated      TRUE if it's a negated property (\P or \p{^)    negated      TRUE if it's a negated property (\P or \p{^)
2510    
2511  Returns:       TRUE if auto-possessifying is OK  Returns:       TRUE if auto-possessifying is OK
2512  */  */
2513    
2514  static BOOL  static BOOL
2515  check_char_prop(int c, int ptype, int pdata, BOOL negated)  check_char_prop(int c, int ptype, int pdata, BOOL negated)
# Line 2453  switch(ptype) Line 2553  switch(ptype)
2553            _pcre_ucp_gentype[prop->chartype] == ucp_N ||            _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2554            c == CHAR_UNDERSCORE) == negated;            c == CHAR_UNDERSCORE) == negated;
2555    }    }
2556  return FALSE;  return FALSE;
2557  }  }
2558  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2559    
# Line 2478  Returns:        TRUE if possessifying is Line 2578  Returns:        TRUE if possessifying is
2578  */  */
2579    
2580  static BOOL  static BOOL
2581  check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,  check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
2582    int options, compile_data *cd)    int options, compile_data *cd)
2583  {  {
2584  int c, next;  int c, next;
# Line 2493  if ((options & PCRE_EXTENDED) != 0) Line 2593  if ((options & PCRE_EXTENDED) != 0)
2593      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2594      if (*ptr == CHAR_NUMBER_SIGN)      if (*ptr == CHAR_NUMBER_SIGN)
2595        {        {
2596        while (*(++ptr) != 0)        ptr++;
2597          while (*ptr != 0)
2598            {
2599          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2600            ptr++;
2601    #ifdef SUPPORT_UTF8
2602            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2603    #endif
2604            }
2605        }        }
2606      else break;      else break;
2607      }      }
# Line 2530  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 2549  the next item is a character. */ Line 2663  the next item is a character. */
2663  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
2664    {    {
2665    case OP_CHAR:    case OP_CHAR:
2666  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2667    GETCHARTEST(c, previous);    GETCHARTEST(c, previous);
2668  #else  #else
2669    c = *previous;    c = *previous;
2670  #endif  #endif
2671    return c != next;    return c != next;
2672    
2673    /* 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
2674    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
2675    high-valued characters. */    high-valued characters. */
2676    
2677    case OP_CHARNC:    case OP_CHARI:
2678  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2679    GETCHARTEST(c, previous);    GETCHARTEST(c, previous);
2680  #else  #else
2681    c = *previous;    c = *previous;
2682  #endif  #endif
2683    if (c == next) return FALSE;    if (c == next) return FALSE;
2684  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2685    if (utf8)    if (utf8)
# Line 2583  if (next >= 0) switch(op_code) Line 2697  if (next >= 0) switch(op_code)
2697  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2698    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2699    
2700    /* 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
2701      opcodes are not used for multi-byte characters, because they are coded using
2702      an XCLASS instead. */
2703    
2704    case OP_NOT:    case OP_NOT:
2705      return (c = *previous) == next;
2706    
2707      case OP_NOTI:
2708    if ((c = *previous) == next) return TRUE;    if ((c = *previous) == next) return TRUE;
   if ((options & PCRE_CASELESS) == 0) return FALSE;  
2709  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2710    if (utf8)    if (utf8)
2711      {      {
# Line 2603  if (next >= 0) switch(op_code) Line 2721  if (next >= 0) switch(op_code)
2721    else    else
2722  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2723    return (c == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2724    
2725    /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.    /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2726    When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */    When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
2727    
2728    case OP_DIGIT:    case OP_DIGIT:
2729    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
# Line 2673  if (next >= 0) switch(op_code) Line 2791  if (next >= 0) switch(op_code)
2791  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2792    case OP_PROP:    case OP_PROP:
2793    return check_char_prop(next, previous[0], previous[1], FALSE);    return check_char_prop(next, previous[0], previous[1], FALSE);
2794    
2795    case OP_NOTPROP:    case OP_NOTPROP:
2796    return check_char_prop(next, previous[0], previous[1], TRUE);    return check_char_prop(next, previous[0], previous[1], TRUE);
2797  #endif  #endif
# Line 2683  if (next >= 0) switch(op_code) Line 2801  if (next >= 0) switch(op_code)
2801    }    }
2802    
2803    
2804  /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP  /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP
2805  is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are  is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
2806  generated only when PCRE_UCP is *not* set, that is, when only ASCII  generated only when PCRE_UCP is *not* set, that is, when only ASCII
2807  characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are  characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
2808  replaced by OP_PROP codes when PCRE_UCP is set. */  replaced by OP_PROP codes when PCRE_UCP is set. */
2809    
2810  switch(op_code)  switch(op_code)
2811    {    {
2812    case OP_CHAR:    case OP_CHAR:
2813    case OP_CHARNC:    case OP_CHARI:
2814  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2815    GETCHARTEST(c, previous);    GETCHARTEST(c, previous);
2816  #else  #else
2817    c = *previous;    c = *previous;
2818  #endif  #endif
2819    switch(-next)    switch(-next)
2820      {      {
2821      case ESC_d:      case ESC_d:
# Line 2761  switch(op_code) Line 2879  switch(op_code)
2879        default:        default:
2880        return -next == ESC_v;        return -next == ESC_v;
2881        }        }
2882    
2883      /* When PCRE_UCP is set, these values get generated for \d etc. Find      /* When PCRE_UCP is set, these values get generated for \d etc. Find
2884      their substitutions and process them. The result will always be either      their substitutions and process them. The result will always be either
2885      -ESC_p or -ESC_P. Then fall through to process those values. */      -ESC_p or -ESC_P. Then fall through to process those values. */
2886    
2887  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2888      case ESC_du:      case ESC_du:
2889      case ESC_DU:      case ESC_DU:
# Line 2780  switch(op_code) Line 2898  switch(op_code)
2898        if (temperrorcode != 0) return FALSE;        if (temperrorcode != 0) return FALSE;
2899        ptr++;    /* For compatibility */        ptr++;    /* For compatibility */
2900        }        }
2901      /* Fall through */      /* Fall through */
2902    
2903      case ESC_p:      case ESC_p:
2904      case ESC_P:      case ESC_P:
2905        {        {
2906        int ptype, pdata, errorcodeptr;        int ptype, pdata, errorcodeptr;
2907        BOOL negated;        BOOL negated;
2908    
2909        ptr--;      /* Make ptr point at the p or P */        ptr--;      /* Make ptr point at the p or P */
2910        ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);        ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
2911        if (ptype < 0) return FALSE;        if (ptype < 0) return FALSE;
2912        ptr++;      /* Point past the final curly ket */        ptr++;      /* Point past the final curly ket */
2913    
2914        /* If the property item is optional, we have to give up. (When generated        /* If the property item is optional, we have to give up. (When generated
2915        from \d etc by PCRE_UCP, this test will have been applied much earlier,        from \d etc by PCRE_UCP, this test will have been applied much earlier,
2916        to the original \d etc. At this point, ptr will point to a zero byte. */        to the original \d etc. At this point, ptr will point to a zero byte. */
2917    
2918        if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||        if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2919          strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)          strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2920            return FALSE;            return FALSE;
2921    
2922        /* Do the property check. */        /* Do the property check. */
2923    
2924        return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);        return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
2925        }        }
2926  #endif  #endif
2927    
2928      default:      default:
2929      return FALSE;      return FALSE;
2930      }      }
2931    
2932    /* In principle, support for Unicode properties should be integrated here as    /* In principle, support for Unicode properties should be integrated here as
2933    well. It means re-organizing the above code so as to get hold of the property    well. It means re-organizing the above code so as to get hold of the property
2934    values before switching on the op-code. However, I wonder how many patterns    values before switching on the op-code. However, I wonder how many patterns
2935    combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,    combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
2936    these op-codes are never generated.) */    these op-codes are never generated.) */
2937    
2938    case OP_DIGIT:    case OP_DIGIT:
2939    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
# Line 2831  switch(op_code) Line 2949  switch(op_code)
2949    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2950    
2951    case OP_HSPACE:    case OP_HSPACE:
2952    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
2953           next == -ESC_w || next == -ESC_v || next == -ESC_R;           next == -ESC_w || next == -ESC_v || next == -ESC_R;
2954    
2955    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
2956    return next == -ESC_h;    return next == -ESC_h;
2957    
2958    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
2959    case OP_ANYNL:    case OP_ANYNL:
2960    case OP_VSPACE:    case OP_VSPACE:
2961    return next == -ESC_V || next == -ESC_d || next == -ESC_w;    return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2962    
# Line 2846  switch(op_code) Line 2964  switch(op_code)
2964    return next == -ESC_v || next == -ESC_R;    return next == -ESC_v || next == -ESC_R;
2965    
2966    case OP_WORDCHAR:    case OP_WORDCHAR:
2967    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
2968           next == -ESC_v || next == -ESC_R;           next == -ESC_v || next == -ESC_R;
2969    
2970    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
# Line 2982  for (;; ptr++) Line 3100  for (;; ptr++)
3100    
3101    c = *ptr;    c = *ptr;
3102    
3103    /* If we are at the end of a nested substitution, revert to the outer level    /* If we are at the end of a nested substitution, revert to the outer level
3104    string. Nesting only happens one level deep. */    string. Nesting only happens one level deep. */
3105    
3106    if (c == 0 && nestptr != NULL)    if (c == 0 && nestptr != NULL)
# Line 3104  for (;; ptr++) Line 3222  for (;; ptr++)
3222      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
3223      if (c == CHAR_NUMBER_SIGN)      if (c == CHAR_NUMBER_SIGN)
3224        {        {
3225        while (*(++ptr) != 0)        ptr++;
3226          while (*ptr != 0)
3227          {          {
3228          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3229            ptr++;
3230    #ifdef SUPPORT_UTF8
3231            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
3232    #endif
3233          }          }
3234        if (*ptr != 0) continue;        if (*ptr != 0) continue;
3235    
# Line 3151  for (;; ptr++) Line 3274  for (;; ptr++)
3274      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3275    
3276      case CHAR_CIRCUMFLEX_ACCENT:      case CHAR_CIRCUMFLEX_ACCENT:
3277        previous = NULL;
3278      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3279        {        {
3280        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3281          *code++ = OP_CIRCM;
3282        }        }
3283      previous = NULL;      else *code++ = OP_CIRC;
     *code++ = OP_CIRC;  
3284      break;      break;
3285    
3286      case CHAR_DOLLAR_SIGN:      case CHAR_DOLLAR_SIGN:
3287      previous = NULL;      previous = NULL;
3288      *code++ = OP_DOLL;      *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
3289      break;      break;
3290    
3291      /* 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 3289  for (;; ptr++) Line 3413  for (;; ptr++)
3413          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3414          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3415          }          }
3416    
3417        /* In the pre-compile phase, accumulate the length of any UTF-8 extra        /* In the pre-compile phase, accumulate the length of any UTF-8 extra
3418        data and reset the pointer. This is so that very large classes that        data and reset the pointer. This is so that very large classes that
3419        contain a zillion UTF-8 characters no longer overwrite the work space        contain a zillion UTF-8 characters no longer overwrite the work space
# Line 3358  for (;; ptr++) Line 3482  for (;; ptr++)
3482    
3483          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3484            posix_class = 0;            posix_class = 0;
3485    
3486          /* When PCRE_UCP is set, some of the POSIX classes are converted to          /* When PCRE_UCP is set, some of the POSIX classes are converted to
3487          different escape sequences that use Unicode properties. */          different escape sequences that use Unicode properties. */
3488    
3489  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3490          if ((options & PCRE_UCP) != 0)          if ((options & PCRE_UCP) != 0)
3491            {            {
3492            int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);            int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3493            if (posix_substitutes[pc] != NULL)            if (posix_substitutes[pc] != NULL)
3494              {              {
3495              nestptr = tempptr + 1;              nestptr = tempptr + 1;
3496              ptr = posix_substitutes[pc] - 1;              ptr = posix_substitutes[pc] - 1;
3497              continue;              continue;
3498              }              }
3499            }            }
3500  #endif  #endif
3501          /* In the non-UCP case, we build the bit map for the POSIX class in a          /* In the non-UCP case, we build the bit map for the POSIX class in a
3502          chunk of local store because we may be adding and subtracting from it,          chunk of local store because we may be adding and subtracting from it,
3503          and we don't want to subtract bits that may be in the main map already.          and we don't want to subtract bits that may be in the main map already.
# Line 3460  for (;; ptr++) Line 3584  for (;; ptr++)
3584              case ESC_SU:              case ESC_SU:
3585              nestptr = ptr;              nestptr = ptr;
3586              ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */              ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3587              class_charcount -= 2;                /* Undo! */              class_charcount -= 2;                /* Undo! */
3588              continue;              continue;
3589  #endif  #endif
3590              case ESC_d:              case ESC_d:
# Line 3481  for (;; ptr++) Line 3605  for (;; ptr++)
3605              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3606              continue;              continue;
3607    
3608                /* Perl 5.004 onwards omits VT from \s, but we must preserve it
3609                if it was previously set by something earlier in the character
3610                class. */
3611    
3612              case ESC_s:              case ESC_s:
3613              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];              classbits[0] |= cbits[cbit_space];
3614              classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= cbits[cbit_space+1] & ~0x08;
3615                for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3616              continue;              continue;
3617    
3618              case ESC_S:              case ESC_S:
# Line 3902  for (;; ptr++) Line 4031  for (;; ptr++)
4031    
4032      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
4033      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR      characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
4034      operate on single-bytes only. This is an historical hangover. Maybe one day      operate on single-bytes characters only. This is an historical hangover.
4035      we can tidy these opcodes to handle multi-byte characters.      Maybe one day we can tidy these opcodes to handle multi-byte characters.
4036    
4037      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
4038      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.
4039      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
4040      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
4041      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
4042      reqbyte, save the previous value for reinstating. */      of reqbyte, save the previous value for reinstating. */
4043    
4044  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4045      if (class_charcount == 1 && !class_utf8 &&      if (class_charcount == 1 && !class_utf8 &&
4046        (!utf8 || !negate_class || class_lastchar < 128))        (!utf8 || !negate_class || class_lastchar < 128))
# Line 3921  for (;; ptr++) Line 4050  for (;; ptr++)
4050        {        {
4051        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
4052    
4053        /* The OP_NOT opcode works on one-byte characters only. */        /* The OP_NOT[I] opcodes work on one-byte characters only. */
4054    
4055        if (negate_class)        if (negate_class)
4056          {          {
4057          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
4058          zerofirstbyte = firstbyte;          zerofirstbyte = firstbyte;
4059          *code++ = OP_NOT;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
4060          *code++ = class_lastchar;          *code++ = class_lastchar;
4061          break;          break;
4062          }          }
# Line 3991  for (;; ptr++) Line 4120  for (;; ptr++)
4120        }        }
4121  #endif  #endif
4122    
4123      /* If there are no characters > 255, or they are all to be included or      /* If there are no characters > 255, or they are all to be included or
4124      excluded, set the opcode to OP_CLASS or OP_NCLASS, depending on whether the      excluded, set the opcode to OP_CLASS or OP_NCLASS, depending on whether the
4125      whole class was negated and whether there were negative specials such as \S      whole class was negated and whether there were negative specials such as \S
4126      (non-UCP) in the class. Then copy the 32-byte map into the code vector,      (non-UCP) in the class. Then copy the 32-byte map into the code vector,
# Line 4085  for (;; ptr++) Line 4214  for (;; ptr++)
4214      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
4215      instead.  */      instead.  */
4216    
4217      if (*previous == OP_CHAR || *previous == OP_CHARNC)      if (*previous == OP_CHAR || *previous == OP_CHARI)
4218        {        {
4219          op_type = (*previous == OP_CHAR)? 0 : OP_STARI - OP_STAR;
4220    
4221        /* 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
4222        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
4223        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 4131  for (;; ptr++) Line 4262  for (;; ptr++)
4262      /* If previous was a single negated character ([^a] or similar), we use      /* If previous was a single negated character ([^a] or similar), we use
4263      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-
4264      character repeats by setting opt_type to add a suitable offset into      character repeats by setting opt_type to add a suitable offset into
4265      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
4266      currently used only for single-byte chars. */      are currently used only for single-byte chars. */
4267    
4268      else if (*previous == OP_NOT)      else if (*previous == OP_NOT || *previous == OP_NOTI)
4269        {        {
4270        op_type = OP_NOTSTAR - OP_STAR;  /* Use "not" opcodes */        op_type = ((*previous == OP_NOT)? OP_NOTSTAR : OP_NOTSTARI) - OP_STAR;
4271        c = previous[1];        c = previous[1];
4272        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4273            repeat_max < 0 &&            repeat_max < 0 &&
# Line 4333  for (;; ptr++) Line 4464  for (;; ptr++)
4464  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4465               *previous == OP_XCLASS ||               *previous == OP_XCLASS ||
4466  #endif  #endif
4467               *previous == OP_REF)               *previous == OP_REF ||
4468                 *previous == OP_REFI)
4469        {        {
4470        if (repeat_max == 0)        if (repeat_max == 0)
4471          {          {
# Line 4367  for (;; ptr++) Line 4499  for (;; ptr++)
4499        }        }
4500    
4501      /* 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
4502      cases. */      cases. Note that at this point we can encounter only the "basic" BRA and
4503        KET opcodes, as this is the place where they get converted into the more
4504        special varieties. */
4505    
4506      else if (*previous == OP_BRA  || *previous == OP_CBRA ||      else if (*previous == OP_BRA  || *previous == OP_CBRA ||
4507               *previous == OP_ONCE || *previous == OP_COND)               *previous == OP_ONCE || *previous == OP_COND)
4508        {        {
4509        register int i;        register int i;
       int ketoffset = 0;  
4510        int len = (int)(code - previous);        int len = (int)(code - previous);
4511        uschar *bralink = NULL;        uschar *bralink = NULL;
4512          uschar *brazeroptr = NULL;
4513    
4514        /* Repeating a DEFINE group is pointless */        /* Repeating a DEFINE group is pointless */
4515    
4516        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
# Line 4385  for (;; ptr++) Line 4519  for (;; ptr++)
4519          goto FAILED;          goto FAILED;
4520          }          }
4521    
       /* If the maximum repeat count is unlimited, find the end of the bracket  
       by scanning through from the start, and compute the offset back to it  
       from the current code pointer. There may be an OP_OPT setting following  
       the final KET, so we can't find the end just by going back from the code  
       pointer. */  
   
       if (repeat_max == -1)  
         {  
         register uschar *ket = previous;  
         do ket += GET(ket, 1); while (*ket != OP_KET);  
         ketoffset = (int)(code - ket);  
         }  
   
4522        /* 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
4523        OP_BRAZERO in front of it, and because the group appears once in the        OP_BRAZERO in front of it, and because the group appears once in the
4524        data, whereas in other cases it appears the minimum number of times. For        data, whereas in other cases it appears the minimum number of times. For
# Line 4439  for (;; ptr++) Line 4560  for (;; ptr++)
4560              *previous++ = OP_SKIPZERO;              *previous++ = OP_SKIPZERO;
4561              goto END_REPEAT;              goto END_REPEAT;
4562              }              }
4563              brazeroptr = previous;    /* Save for possessive optimizing */
4564            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4565            }            }
4566    
# Line 4603  for (;; ptr++) Line 4725  for (;; ptr++)
4725            }            }
4726          }          }
4727    
4728        /* 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
4729        can't just offset backwards from the current code point, because we        ONCE brackets, that's all we need to do.
4730        don't know if there's been an options resetting after the ket. The  
4731        correct offset was computed above.        Otherwise, if the quantifier was possessive, we convert the BRA code to
4732          the POS form, and the KET code to KETRPOS. (It turns out to be convenient
4733          at runtime to detect this kind of subpattern at both the start and at the
4734          end.) The use of special opcodes makes it possible to reduce greatly the
4735          stack usage in pcre_exec(). If the group is preceded by OP_BRAZERO,
4736          convert this to OP_BRAPOSZERO. Then cancel the possessive flag so that
4737          the default action below, of wrapping everything inside atomic brackets,
4738          does not happen.
4739    
4740        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
4741        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
4742        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
4743        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
4744        atomic groups at runtime, but in a different way.] */        runtime, but in a different way.] */
4745    
4746        else        else
4747          {          {
4748          uschar *ketcode = code - ketoffset;          uschar *ketcode = code - 1 - LINK_SIZE;
4749          uschar *bracode = ketcode - GET(ketcode, 1);          uschar *bracode = ketcode - GET(ketcode, 1);
4750          *ketcode = OP_KETRMAX + repeat_type;  
4751          if (lengthptr == NULL && *bracode != OP_ONCE)          if (*bracode == OP_ONCE)
4752              *ketcode = OP_KETRMAX + repeat_type;
4753            else
4754            {            {
4755            uschar *scode = bracode;            if (possessive_quantifier)
4756            do              {
4757                *bracode += 1;                   /* Switch to xxxPOS opcodes */
4758                *ketcode = OP_KETRPOS;
4759                if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
4760                possessive_quantifier = FALSE;
4761                }
4762              else *ketcode = OP_KETRMAX + repeat_type;
4763    
4764              if (lengthptr == NULL)
4765              {              {
4766              if (could_be_empty_branch(scode, ketcode, utf8, cd))              uschar *scode = bracode;
4767                do
4768                {                {
4769                *bracode += OP_SBRA - OP_BRA;                if (could_be_empty_branch(scode, ketcode, utf8, cd))
4770                break;                  {
4771                    *bracode += OP_SBRA - OP_BRA;
4772                    break;
4773                    }
4774                  scode += GET(scode, 1);
4775                }                }
4776              scode += GET(scode, 1);              while (*scode == OP_ALT);
4777              }              }
           while (*scode == OP_ALT);  
4778            }            }
4779          }          }
4780        }        }
# Line 4652  for (;; ptr++) Line 4795  for (;; ptr++)
4795        }        }
4796    
4797      /* If the character following a repeat is '+', or if certain optimization      /* If the character following a repeat is '+', or if certain optimization
4798      tests above succeeded, possessive_quantifier is TRUE. For some of the      tests above succeeded, possessive_quantifier is TRUE. For some opcodes,
4799      simpler opcodes, there is an special alternative opcode for this. For      there are special alternative opcodes for this case. For anything else, we
4800      anything else, we wrap the entire repeated item inside OP_ONCE brackets.      wrap the entire repeated item inside OP_ONCE brackets. Logically, the '+'
4801      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
4802      but the special opcodes can optimize it a bit. The repeated item starts at      special opcodes can optimize it.
4803      tempcode, not at previous, which might be the first part of a string whose  
4804      (former) last char we repeated.      Possessively repeated subpatterns have already been handled in the code
4805        just above, so possessive_quantifier is always FALSE for them at this
4806        stage.
4807    
4808        Note that the repeated item starts at tempcode, not at previous, which
4809        might be the first part of a string whose (former) last char we repeated.
4810    
4811      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
4812      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 4689  for (;; ptr++) Line 4837  for (;; ptr++)
4837          case OP_QUERY: *tempcode = OP_POSQUERY; break;          case OP_QUERY: *tempcode = OP_POSQUERY; break;
4838          case OP_UPTO:  *tempcode = OP_POSUPTO; break;          case OP_UPTO:  *tempcode = OP_POSUPTO; break;
4839    
4840          case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;          case OP_STARI:  *tempcode = OP_POSSTARI; break;
4841          case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;          case OP_PLUSI:  *tempcode = OP_POSPLUSI; break;
4842          case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;          case OP_QUERYI: *tempcode = OP_POSQUERYI; break;
4843          case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;          case OP_UPTOI:  *tempcode = OP_POSUPTOI; break;
4844    
4845          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;          case OP_NOTSTAR:  *tempcode = OP_NOTPOSSTAR; break;
4846          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;          case OP_NOTPLUS:  *tempcode = OP_NOTPOSPLUS; break;
4847          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4848          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4849    
4850            case OP_NOTSTARI:  *tempcode = OP_NOTPOSSTARI; break;
4851            case OP_NOTPLUSI:  *tempcode = OP_NOTPOSPLUSI; break;
4852            case OP_NOTQUERYI: *tempcode = OP_NOTPOSQUERYI; break;
4853            case OP_NOTUPTOI:  *tempcode = OP_NOTPOSUPTOI; break;
4854    
4855            case OP_TYPESTAR:  *tempcode = OP_TYPEPOSSTAR; break;
4856            case OP_TYPEPLUS:  *tempcode = OP_TYPEPOSPLUS; break;
4857            case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
4858            case OP_TYPEUPTO:  *tempcode = OP_TYPEPOSUPTO; break;
4859    
4860          /* Because we are moving code along, we must ensure that any          /* Because we are moving code along, we must ensure that any
4861          pending recursive references are updated. */          pending recursive references are updated. */
4862    
# Line 4757  for (;; ptr++) Line 4915  for (;; ptr++)
4915          arg = ++ptr;          arg = ++ptr;
4916          while ((cd->ctypes[*ptr] & (ctype_letter|ctype_digit)) != 0          while ((cd->ctypes[*ptr] & (ctype_letter|ctype_digit)) != 0
4917            || *ptr == '_') ptr++;            || *ptr == '_') ptr++;
4918          arglen = ptr - arg;          arglen = (int)(ptr - arg);
4919          }          }
4920    
4921        if (*ptr != CHAR_RIGHT_PARENTHESIS)        if (*ptr != CHAR_RIGHT_PARENTHESIS)
# Line 4773  for (;; ptr++) Line 4931  for (;; ptr++)
4931          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
4932              strncmp((char *)name, vn, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
4933            {            {
4934            /* Check for open captures before ACCEPT */            /* Check for open captures before ACCEPT and convert it to
4935              ASSERT_ACCEPT if in an assertion. */
4936    
4937            if (verbs[i].op == OP_ACCEPT)            if (verbs[i].op == OP_ACCEPT)
4938              {              {
4939              open_capitem *oc;              open_capitem *oc;
4940                if (arglen != 0)
4941                  {
4942                  *errorcodeptr = ERR59;
4943                  goto FAILED;
4944                  }
4945              cd->had_accept = TRUE;              cd->had_accept = TRUE;
4946              for (oc = cd->open_caps; oc != NULL; oc = oc->next)              for (oc = cd->open_caps; oc != NULL; oc = oc->next)
4947                {                {
4948                *code++ = OP_CLOSE;                *code++ = OP_CLOSE;
4949                PUT2INC(code, 0, oc->number);                PUT2INC(code, 0, oc->number);
4950                }                }
4951                *code++ = (cd->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT;
4952              }              }
4953    
4954            /* Handle the cases with/without an argument */            /* Handle other cases with/without an argument */
4955    
4956            if (arglen == 0)            else if (arglen == 0)
4957              {              {
4958              if (verbs[i].op < 0)   /* Argument is mandatory */              if (verbs[i].op < 0)   /* Argument is mandatory */
4959                {                {
4960                *errorcodeptr = ERR66;                *errorcodeptr = ERR66;
4961                goto FAILED;                goto FAILED;
4962                }                }
4963              *code++ = verbs[i].op;              *code = verbs[i].op;
4964                if (*code++ == OP_THEN)
4965                  {
4966                  PUT(code, 0, code - bcptr->current_branch - 1);
4967                  code += LINK_SIZE;
4968                  }
4969              }              }
4970    
4971            else            else
# Line 4805  for (;; ptr++) Line 4975  for (;; ptr++)
4975                *errorcodeptr = ERR59;                *errorcodeptr = ERR59;
4976                goto FAILED;                goto FAILED;
4977                }                }
4978              *code++ = verbs[i].op_arg;              *code = verbs[i].op_arg;
4979                if (*code++ == OP_THEN_ARG)
4980                  {
4981                  PUT(code, 0, code - bcptr->current_branch - 1);
4982                  code += LINK_SIZE;
4983                  }
4984              *code++ = arglen;              *code++ = arglen;
4985              memcpy(code, arg, arglen);              memcpy(code, arg, arglen);
4986              code += arglen;              code += arglen;
# Line 4999  for (;; ptr++) Line 5174  for (;; ptr++)
5174          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
5175    
5176          else if ((i = find_parens(cd, name, namelen,          else if ((i = find_parens(cd, name, namelen,
5177                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0, utf8)) > 0)
5178            {            {
5179            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
5180            code[1+LINK_SIZE]++;            code[1+LINK_SIZE]++;
# Line 5067  for (;; ptr++) Line 5242  for (;; ptr++)
5242          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5243          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
5244          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
5245            cd->assert_depth += 1;
5246          ptr++;          ptr++;
5247          break;          break;
5248    
# Line 5081  for (;; ptr++) Line 5257  for (;; ptr++)
5257            continue;            continue;
5258            }            }
5259          bravalue = OP_ASSERT_NOT;          bravalue = OP_ASSERT_NOT;
5260            cd->assert_depth += 1;
5261          break;          break;
5262    
5263    
# Line 5090  for (;; ptr++) Line 5267  for (;; ptr++)
5267            {            {
5268            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
5269            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
5270              cd->assert_depth += 1;
5271            ptr += 2;            ptr += 2;
5272            break;            break;
5273    
5274            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
5275            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
5276              cd->assert_depth += 1;
5277            ptr += 2;            ptr += 2;
5278            break;            break;
5279    
# Line 5300  for (;; ptr++) Line 5479  for (;; ptr++)
5479          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
5480          namelen = (int)(ptr - name);          namelen = (int)(ptr - name);
5481    
5482          /* 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
5483          reference number. */          a dummy reference number, because it was not used in the first pass.
5484            However, with the change of recursive back references to be atomic,
5485            we have to look for the number so that this state can be identified, as
5486            otherwise the incorrect length is computed. If it's not a backwards
5487            reference, the dummy number will do. */
5488    
5489          if (lengthptr != NULL)          if (lengthptr != NULL)
5490            {            {
5491              const uschar *temp;
5492    
5493            if (namelen == 0)            if (namelen == 0)
5494              {              {
5495              *errorcodeptr = ERR62;              *errorcodeptr = ERR62;
# Line 5320  for (;; ptr++) Line 5505  for (;; ptr++)
5505              *errorcodeptr = ERR48;              *errorcodeptr = ERR48;
5506              goto FAILED;              goto FAILED;
5507              }              }
5508            recno = 0;  
5509              /* The name table does not exist in the first pass, so we cannot
5510              do a simple search as in the code below. Instead, we have to scan the
5511              pattern to find the number. It is important that we scan it only as
5512              far as we have got because the syntax of named subpatterns has not
5513              been checked for the rest of the pattern, and find_parens() assumes
5514              correct syntax. In any case, it's a waste of resources to scan
5515              further. We stop the scan at the current point by temporarily
5516              adjusting the value of cd->endpattern. */
5517    
5518              temp = cd->end_pattern;
5519              cd->end_pattern = ptr;
5520              recno = find_parens(cd, name, namelen,
5521                (options & PCRE_EXTENDED) != 0, utf8);
5522              cd->end_pattern = temp;
5523              if (recno < 0) recno = 0;    /* Forward ref; set dummy number */
5524            }            }
5525    
5526          /* 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 5345  for (;; ptr++) Line 5545  for (;; ptr++)
5545              }              }
5546            else if ((recno =                /* Forward back reference */            else if ((recno =                /* Forward back reference */
5547                      find_parens(cd, name, namelen,                      find_parens(cd, name, namelen,
5548                        (options & PCRE_EXTENDED) != 0)) <= 0)                        (options & PCRE_EXTENDED) != 0, utf8)) <= 0)
5549              {              {
5550              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
5551              goto FAILED;              goto FAILED;
# Line 5456  for (;; ptr++) Line 5656  for (;; ptr++)
5656              if (called == NULL)              if (called == NULL)
5657                {                {
5658                if (find_parens(cd, NULL, recno,                if (find_parens(cd, NULL, recno,
5659                      (options & PCRE_EXTENDED) != 0) < 0)                      (options & PCRE_EXTENDED) != 0, utf8) < 0)
5660                  {                  {
5661                  *errorcodeptr = ERR15;                  *errorcodeptr = ERR15;
5662                  goto FAILED;                  goto FAILED;
# Line 5559  for (;; ptr++) Line 5759  for (;; ptr++)
5759          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
5760          both phases.          both phases.
5761    
5762          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
5763          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. */  
5764    
5765          if (*ptr == CHAR_RIGHT_PARENTHESIS)          if (*ptr == CHAR_RIGHT_PARENTHESIS)
5766            {            {
# Line 5572  for (;; ptr++) Line 5771  for (;; ptr++)
5771              }              }
5772            else            else
5773              {              {
             if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))  
               {  
               *code++ = OP_OPT;  
               *code++ = newoptions & PCRE_IMS;  
               }  
5774              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);
5775              greedy_non_default = greedy_default ^ 1;              greedy_non_default = greedy_default ^ 1;
5776              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
5777              }              }
5778    
5779            /* Change options at this level, and pass them back for use            /* Change options at this level, and pass them back for use
5780            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). */  
5781    
5782            *optionsptr = options = newoptions;            *optionsptr = options = newoptions;
5783            previous = NULL;       /* This item can't be repeated */            previous = NULL;       /* This item can't be repeated */
# Line 5623  for (;; ptr++) Line 5815  for (;; ptr++)
5815    
5816      /* Process nested bracketed regex. Assertions may not be repeated, but      /* Process nested bracketed regex. Assertions may not be repeated, but
5817      other kinds can be. All their opcodes are >= OP_ONCE. We copy code into a      other kinds can be. All their opcodes are >= OP_ONCE. We copy code into a
5818      non-register variable in order to be able to pass its address because some      non-register variable (tempcode) in order to be able to pass its address
5819      compilers complain otherwise. Pass in a new setting for the ims options if      because some compilers complain otherwise. */
     they have changed. */  
5820    
5821      previous = (bravalue >= OP_ONCE)? code : NULL;      previous = (bravalue >= OP_ONCE)? code : NULL;
5822      *code = bravalue;      *code = bravalue;
# Line 5635  for (;; ptr++) Line 5826  for (;; ptr++)
5826    
5827      if (!compile_regex(      if (!compile_regex(
5828           newoptions,                   /* The complete new option state */           newoptions,                   /* The complete new option state */
          options & PCRE_IMS,           /* The previous ims option state */  
5829           &tempcode,                    /* Where to put code (updated) */           &tempcode,                    /* Where to put code (updated) */
5830           &ptr,                         /* Input pointer (updated) */           &ptr,                         /* Input pointer (updated) */
5831           errorcodeptr,                 /* Where to put an error message */           errorcodeptr,                 /* Where to put an error message */
# Line 5651  for (;; ptr++) Line 5841  for (;; ptr++)
5841             &length_prevgroup           /* Pre-compile phase */             &length_prevgroup           /* Pre-compile phase */
5842           ))           ))
5843        goto FAILED;        goto FAILED;
5844    
5845        if (bravalue >= OP_ASSERT && bravalue <= OP_ASSERTBACK_NOT)
5846          cd->assert_depth -= 1;
5847    
5848      /* 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
5849      group, while tempcode has been updated to point past the end of the group      group, while tempcode has been updated to point past the end of the group
# Line 5722  for (;; ptr++) Line 5915  for (;; ptr++)
5915          goto FAILED;          goto FAILED;
5916          }          }
5917        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;
5918        *code++ = OP_BRA;        code++;   /* This already contains bravalue */
5919        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
5920        *code++ = OP_KET;        *code++ = OP_KET;
5921        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
# Line 5795  for (;; ptr++) Line 5988  for (;; ptr++)
5988    
5989      /* ===================================================================*/      /* ===================================================================*/
5990      /* Handle metasequences introduced by \. For ones like \d, the ESC_ values      /* Handle metasequences introduced by \. For ones like \d, the ESC_ values
5991      are arranged to be the negation of the corresponding OP_values in the      are arranged to be the negation of the corresponding OP_values in the
5992      default case when PCRE_UCP is not set. For the back references, the values      default case when PCRE_UCP is not set. For the back references, the values
5993      are ESC_REF plus the reference number. Only back references and those types      are ESC_REF plus the reference number. Only back references and those types
5994      that consume a character may be repeated. We can test for values between      that consume a character may be repeated. We can test for values between
# Line 5914  for (;; ptr++) Line 6107  for (;; ptr++)
6107          HANDLE_REFERENCE:    /* Come here from named backref handling */          HANDLE_REFERENCE:    /* Come here from named backref handling */
6108          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
6109          previous = code;          previous = code;
6110          *code++ = OP_REF;          *code++ = ((options & PCRE_CASELESS) != 0)? OP_REFI : OP_REF;
6111          PUT2INC(code, 0, recno);          PUT2INC(code, 0, recno);
6112          cd->backref_map |= (recno < 32)? (1 << recno) : 1;          cd->backref_map |= (recno < 32)? (1 << recno) : 1;
6113          if (recno > cd->top_backref) cd->top_backref = recno;          if (recno > cd->top_backref) cd->top_backref = recno;
# Line 5973  for (;; ptr++) Line 6166  for (;; ptr++)
6166            ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */            ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
6167            }            }
6168          else          else
6169  #endif  #endif
6170            {            {
6171            previous = (-c > ESC_b && -c < ESC_Z)? code : NULL;            previous = (-c > ESC_b && -c < ESC_Z)? code : NULL;
6172            *code++ = -c;            *code++ = -c;
6173            }            }
6174          }          }
6175        continue;        continue;
6176        }        }
# Line 6022  for (;; ptr++) Line 6215  for (;; ptr++)
6215    
6216      ONE_CHAR:      ONE_CHAR:
6217      previous = code;      previous = code;
6218      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARNC : OP_CHAR;      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARI : OP_CHAR;
6219      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];
6220    
6221      /* Remember if \r or \n were seen */      /* Remember if \r or \n were seen */
# Line 6086  return FALSE; Line 6279  return FALSE;
6279  /* 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
6280  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
6281  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.  
   
6282  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
6283  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
6284  value of lengthptr distinguishes the two phases.  value of lengthptr distinguishes the two phases.
6285    
6286  Arguments:  Arguments:
6287    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  
6288    codeptr        -> the address of the current code pointer    codeptr        -> the address of the current code pointer
6289    ptrptr         -> the address of the current pattern pointer    ptrptr         -> the address of the current pattern pointer
6290    errorcodeptr   -> pointer to error code variable    errorcodeptr   -> pointer to error code variable
# Line 6115  Returns:         TRUE on success Line 6302  Returns:         TRUE on success
6302  */  */
6303    
6304  static BOOL  static BOOL
6305  compile_regex(int options, int oldims, uschar **codeptr, const uschar **ptrptr,  compile_regex(int options, uschar **codeptr, const uschar **ptrptr,
6306    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,
6307    int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr, compile_data *cd,    int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr, compile_data *cd,
6308    int *lengthptr)    int *lengthptr)
# Line 6132  int branchfirstbyte, branchreqbyte; Line 6319  int branchfirstbyte, branchreqbyte;
6319  int length;  int length;
6320  int orig_bracount;  int orig_bracount;
6321  int max_bracount;  int max_bracount;
 int old_external_options = cd->external_options;  
6322  branch_chain bc;  branch_chain bc;
6323    
6324  bc.outer = bcptr;  bc.outer = bcptr;
# Line 6156  pre-compile phase to find out whether an Line 6342  pre-compile phase to find out whether an
6342    
6343  /* 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
6344  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
6345  detect groups that contain recursive back references to themselves. */  detect groups that contain recursive back references to themselves. Note that
6346    only OP_CBRA need be tested here; changing this opcode to one of its variants,
6347    e.g. OP_SCBRAPOS, happens later, after the group has been compiled. */
6348    
6349  if (*code == OP_CBRA)  if (*code == OP_CBRA)
6350    {    {
# Line 6182  for (;;) Line 6370  for (;;)
6370    
6371    if (reset_bracount) cd->bracount = orig_bracount;    if (reset_bracount) cd->bracount = orig_bracount;
6372    
   /* 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;  
     }  
   
6373    /* Set up dummy OP_REVERSE if lookbehind assertion */    /* Set up dummy OP_REVERSE if lookbehind assertion */
6374    
6375    if (lookbehind)    if (lookbehind)
# Line 6211  for (;;) Line 6390  for (;;)
6390      return FALSE;      return FALSE;
6391      }      }
6392    
   /* 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;  
   
6393    /* Keep the highest bracket count in case (?| was used and some branch    /* Keep the highest bracket count in case (?| was used and some branch
6394    has fewer than the rest. */    has fewer than the rest. */
6395    
# Line 6280  for (;;) Line 6450  for (;;)
6450        {        {
6451        int fixed_length;        int fixed_length;
6452        *code = OP_END;        *code = OP_END;
6453        fixed_length = find_fixedlength(last_branch, options, FALSE, cd);        fixed_length = find_fixedlength(last_branch,  (options & PCRE_UTF8) != 0,
6454            FALSE, cd);
6455        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
6456        if (fixed_length == -3)        if (fixed_length == -3)
6457          {          {
# Line 6301  for (;;) Line 6472  for (;;)
6472    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
6473    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
6474    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
6475    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. */  
6476    
6477    if (*ptr != CHAR_VERTICAL_LINE)    if (*ptr != CHAR_VERTICAL_LINE)
6478      {      {
# Line 6347  for (;;) Line 6516  for (;;)
6516        cd->open_caps = cd->open_caps->next;        cd->open_caps = cd->open_caps->next;
6517        }        }
6518    
     /* Reset options if needed. */  
   
     if ((options & PCRE_IMS) != oldims && *ptr == CHAR_RIGHT_PARENTHESIS)  
       {  
       *code++ = OP_OPT;  
       *code++ = oldims;  
       length += 2;  
       }  
   
6519      /* Retain the highest bracket number, in case resetting was used. */      /* Retain the highest bracket number, in case resetting was used. */
6520    
6521      cd->bracount = max_bracount;      cd->bracount = max_bracount;
# Line 6415  for (;;) Line 6575  for (;;)
6575  /* 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
6576  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
6577  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
6578  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
6579  counts, since OP_CIRC can match in the middle.  be found, because ^ generates OP_CIRCM in that mode.
6580    
6581  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.
6582  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 6437  of the more common cases more precisely. Line 6597  of the more common cases more precisely.
6597    
6598  Arguments:  Arguments:
6599    code           points to start of expression (the bracket)    code           points to start of expression (the bracket)
   options        points to the options setting  
6600    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
6601                    handles up to substring 31; after that we just have to take                    handles up to substring 31; after that we just have to take
6602                    the less precise approach                    the less precise approach
# Line 6447  Returns:     TRUE or FALSE Line 6606  Returns:     TRUE or FALSE
6606  */  */
6607    
6608  static BOOL  static BOOL
6609  is_anchored(register const uschar *code, int *options, unsigned int bracket_map,  is_anchored(register const uschar *code, unsigned int bracket_map,
6610    unsigned int backref_map)    unsigned int backref_map)
6611  {  {
6612  do {  do {
6613     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],
6614       options, PCRE_MULTILINE, FALSE);       FALSE);
6615     register int op = *scode;     register int op = *scode;
6616    
6617     /* Non-capturing brackets */     /* Non-capturing brackets */
6618    
6619     if (op == OP_BRA)     if (op == OP_BRA  || op == OP_BRAPOS ||
6620           op == OP_SBRA || op == OP_SBRAPOS)
6621       {       {
6622       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, bracket_map, backref_map)) return FALSE;
6623       }       }
6624    
6625     /* Capturing brackets */     /* Capturing brackets */
6626    
6627     else if (op == OP_CBRA)     else if (op == OP_CBRA  || op == OP_CBRAPOS ||
6628                op == OP_SCBRA || op == OP_SCBRAPOS)
6629       {       {
6630       int n = GET2(scode, 1+LINK_SIZE);       int n = GET2(scode, 1+LINK_SIZE);
6631       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);
6632       if (!is_anchored(scode, options, new_map, backref_map)) return FALSE;       if (!is_anchored(scode, new_map, backref_map)) return FALSE;
6633       }       }
6634    
6635     /* Other brackets */     /* Other brackets */
6636    
6637     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_COND)     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_COND)
6638       {       {
6639       if (!is_anchored(scode, options, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, bracket_map, backref_map)) return FALSE;
6640       }       }
6641    
6642     /* .* 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 6490  do { Line 6651  do {
6651    
6652     /* Check for explicit anchoring */     /* Check for explicit anchoring */
6653    
6654     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;  
6655     code += GET(code, 1);     code += GET(code, 1);
6656     }     }
6657  while (*code == OP_ALT);   /* Loop for each alternative */  while (*code == OP_ALT);   /* Loop for each alternative */
# Line 6528  is_startline(const uschar *code, unsigne Line 6687  is_startline(const uschar *code, unsigne
6687  {  {
6688  do {  do {
6689     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],     const uschar *scode = first_significant_code(code + _pcre_OP_lengths[*code],
6690       NULL, 0, FALSE);       FALSE);
6691     register int op = *scode;     register int op = *scode;
6692    
6693     /* 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 6555  do { Line 6714  do {
6714         scode += 1 + LINK_SIZE;         scode += 1 + LINK_SIZE;
6715         break;         break;
6716         }         }
6717       scode = first_significant_code(scode, NULL, 0, FALSE);       scode = first_significant_code(scode, FALSE);
6718       op = *scode;       op = *scode;
6719       }       }
6720    
6721     /* Non-capturing brackets */     /* Non-capturing brackets */
6722    
6723     if (op == OP_BRA)     if (op == OP_BRA  || op == OP_BRAPOS ||
6724           op == OP_SBRA || op == OP_SBRAPOS)
6725       {       {
6726       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
6727       }       }
6728    
6729     /* Capturing brackets */     /* Capturing brackets */
6730    
6731     else if (op == OP_CBRA)     else if (op == OP_CBRA  || op == OP_CBRAPOS ||
6732                op == OP_SCBRA || op == OP_SCBRAPOS)
6733       {       {
6734       int n = GET2(scode, 1+LINK_SIZE);       int n = GET2(scode, 1+LINK_SIZE);
6735       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);       int new_map = bracket_map | ((n < 32)? (1 << n) : 1);
# Line 6592  do { Line 6753  do {
6753    
6754     /* Check for explicit circumflex */     /* Check for explicit circumflex */
6755    
6756     else if (op != OP_CIRC) return FALSE;     else if (op != OP_CIRC && op != OP_CIRCM) return FALSE;
6757    
6758     /* Move on to the next alternative */     /* Move on to the next alternative */
6759    
# Line 6618  we return that char, otherwise -1. Line 6779  we return that char, otherwise -1.
6779    
6780  Arguments:  Arguments:
6781    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)  
6782    inassert   TRUE if in an assertion    inassert   TRUE if in an assertion
6783    
6784  Returns:     -1 or the fixed first char  Returns:     -1 or the fixed first char
6785  */  */
6786    
6787  static int  static int
6788  find_firstassertedchar(const uschar *code, int *options, BOOL inassert)  find_firstassertedchar(const uschar *code, BOOL inassert)
6789  {  {
6790  register int c = -1;  register int c = -1;
6791  do {  do {
6792     int d;     int d;
6793     const uschar *scode =     int xl = (*code == OP_CBRA || *code == OP_SCBRA ||
6794       first_significant_code(code + 1+LINK_SIZE, options, PCRE_CASELESS, TRUE);               *code == OP_CBRAPOS || *code == OP_SCBRAPOS)? 2:0;
6795       const uschar *scode = first_significant_code(code + 1+LINK_SIZE + xl, TRUE);
6796     register int op = *scode;     register int op = *scode;
6797    
6798     switch(op)     switch(op)
# Line 6640  do { Line 6801  do {
6801       return -1;       return -1;
6802    
6803       case OP_BRA:       case OP_BRA:
6804         case OP_BRAPOS:
6805       case OP_CBRA:       case OP_CBRA:
6806         case OP_SCBRA:
6807         case OP_CBRAPOS:
6808         case OP_SCBRAPOS:
6809       case OP_ASSERT:       case OP_ASSERT:
6810       case OP_ONCE:       case OP_ONCE:
6811       case OP_COND:       case OP_COND:
6812       if ((d = find_firstassertedchar(scode, options, op == OP_ASSERT)) < 0)       if ((d = find_firstassertedchar(scode, op == OP_ASSERT)) < 0)
6813         return -1;         return -1;
6814       if (c < 0) c = d; else if (c != d) return -1;       if (c < 0) c = d; else if (c != d) return -1;
6815       break;       break;
6816    
6817       case OP_EXACT:       /* Fall through */       case OP_EXACT:
6818       scode += 2;       scode += 2;
6819         /* Fall through */
6820    
6821       case OP_CHAR:       case OP_CHAR:
      case OP_CHARNC:  
6822       case OP_PLUS:       case OP_PLUS:
6823       case OP_MINPLUS:       case OP_MINPLUS:
6824       case OP_POSPLUS:       case OP_POSPLUS:
6825       if (!inassert) return -1;       if (!inassert) return -1;
6826       if (c < 0)       if (c < 0) c = scode[1];
6827         {         else if (c != scode[1]) return -1;
6828         c = scode[1];       break;
6829         if ((*options & PCRE_CASELESS) != 0) c |= REQ_CASELESS;  
6830         }       case OP_EXACTI:
6831       else if (c != scode[1]) return -1;       scode += 2;
6832         /* Fall through */
6833    
6834         case OP_CHARI:
6835         case OP_PLUSI:
6836         case OP_MINPLUSI:
6837         case OP_POSPLUSI:
6838         if (!inassert) return -1;
6839         if (c < 0) c = scode[1] | REQ_CASELESS;
6840           else if (c != scode[1]) return -1;
6841       break;       break;
6842       }       }
6843    
# Line 6786  while (ptr[skipatstart] == CHAR_LEFT_PAR Line 6960  while (ptr[skipatstart] == CHAR_LEFT_PAR
6960      { skipatstart += 7; options |= PCRE_UTF8; continue; }      { skipatstart += 7; options |= PCRE_UTF8; continue; }
6961    else if (strncmp((char *)(ptr+skipatstart+2), STRING_UCP_RIGHTPAR, 4) == 0)    else if (strncmp((char *)(ptr+skipatstart+2), STRING_UCP_RIGHTPAR, 4) == 0)
6962      { skipatstart += 6; options |= PCRE_UCP; continue; }      { skipatstart += 6; options |= PCRE_UCP; continue; }
6963      else if (strncmp((char *)(ptr+skipatstart+2), STRING_NO_START_OPT_RIGHTPAR, 13) == 0)
6964        { skipatstart += 15; options |= PCRE_NO_START_OPTIMIZE; continue; }
6965    
6966    if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)    if (strncmp((char *)(ptr+skipatstart+2), STRING_CR_RIGHTPAR, 3) == 0)
6967      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }      { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }
# Line 6809  while (ptr[skipatstart] == CHAR_LEFT_PAR Line 6985  while (ptr[skipatstart] == CHAR_LEFT_PAR
6985      options = (options & ~(PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) | newbsr;      options = (options & ~(PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) | newbsr;
6986    else break;    else break;
6987    }    }
6988    
6989  utf8 = (options & PCRE_UTF8) != 0;  utf8 = (options & PCRE_UTF8) != 0;
6990    
6991  /* 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
6992    return of an error code from _pcre_valid_utf8() is a new feature, introduced in
6993    release 8.13. It is passed back from pcre_[dfa_]exec(), but at the moment is
6994    not used here. */
6995    
6996  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
6997  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&  if (utf8 && (options & PCRE_NO_UTF8_CHECK) == 0 &&
6998       (*erroroffset = _pcre_valid_utf8((USPTR)pattern, -1)) >= 0)       (errorcode = _pcre_valid_utf8((USPTR)pattern, -1, erroroffset)) != 0)
6999    {    {
7000    errorcode = ERR44;    errorcode = ERR44;
7001    goto PCRE_EARLY_ERROR_RETURN2;    goto PCRE_EARLY_ERROR_RETURN2;
# Line 6835  if (utf8) Line 7014  if (utf8)
7014  if ((options & PCRE_UCP) != 0)  if ((options & PCRE_UCP) != 0)
7015    {    {
7016    errorcode = ERR67;    errorcode = ERR67;
7017    goto PCRE_EARLY_ERROR_RETURN;    goto PCRE_EARLY_ERROR_RETURN;
7018    }    }
7019  #endif  #endif
7020    
7021  /* Check validity of \R options. */  /* Check validity of \R options. */
# Line 6932  outside can help speed up starting point Line 7111  outside can help speed up starting point
7111  ptr += skipatstart;  ptr += skipatstart;
7112  code = cworkspace;  code = cworkspace;
7113  *code = OP_BRA;  *code = OP_BRA;
7114  (void)compile_regex(cd->external_options, cd->external_options & PCRE_IMS,  (void)compile_regex(cd->external_options, &code, &ptr, &errorcode, FALSE,
7115    &code, &ptr, &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd,    FALSE, 0, &firstbyte, &reqbyte, NULL, cd, &length);
   &length);  
7116  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;
7117    
7118  DPRINTF(("end pre-compile: length=%d workspace=%d\n", length,  DPRINTF(("end pre-compile: length=%d workspace=%d\n", length,
# Line 6988  field; this time it's used for rememberi Line 7166  field; this time it's used for rememberi
7166  */  */
7167    
7168  cd->final_bracount = cd->bracount;  /* Save for checking forward references */  cd->final_bracount = cd->bracount;  /* Save for checking forward references */
7169    cd->assert_depth = 0;
7170  cd->bracount = 0;  cd->bracount = 0;
7171  cd->names_found = 0;  cd->names_found = 0;
7172  cd->name_table = (uschar *)re + re->name_table_offset;  cd->name_table = (uschar *)re + re->name_table_offset;
# Line 7006  of the function here. */ Line 7185  of the function here. */
7185  ptr = (const uschar *)pattern + skipatstart;  ptr = (const uschar *)pattern + skipatstart;
7186  code = (uschar *)codestart;  code = (uschar *)codestart;
7187  *code = OP_BRA;  *code = OP_BRA;
7188  (void)compile_regex(re->options, re->options & PCRE_IMS, &code, &ptr,  (void)compile_regex(re->options, &code, &ptr, &errorcode, FALSE, FALSE, 0,
7189    &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd, NULL);    &firstbyte, &reqbyte, NULL, cd, NULL);
7190  re->top_bracket = cd->bracount;  re->top_bracket = cd->bracount;
7191  re->top_backref = cd->top_backref;  re->top_backref = cd->top_backref;
7192  re->flags = cd->external_flags;  re->flags = cd->external_flags;
# Line 7073  if (cd->check_lookbehind) Line 7252  if (cd->check_lookbehind)
7252        uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);        uschar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);
7253        int end_op = *be;        int end_op = *be;
7254        *be = OP_END;        *be = OP_END;
7255        fixed_length = find_fixedlength(cc, re->options, TRUE, cd);        fixed_length = find_fixedlength(cc, (re->options & PCRE_UTF8) != 0, TRUE,
7256            cd);
7257        *be = end_op;        *be = end_op;
7258        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
7259        if (fixed_length < 0)        if (fixed_length < 0)
# Line 7112  start with ^. and also when all branches Line 7292  start with ^. and also when all branches
7292    
7293  if ((re->options & PCRE_ANCHORED) == 0)  if ((re->options & PCRE_ANCHORED) == 0)
7294    {    {
7295    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))  
7296      re->options |= PCRE_ANCHORED;      re->options |= PCRE_ANCHORED;
7297    else    else
7298      {      {
7299      if (firstbyte < 0)      if (firstbyte < 0)
7300        firstbyte = find_firstassertedchar(codestart, &temp_options, FALSE);        firstbyte = find_firstassertedchar(codestart, FALSE);
7301      if (firstbyte >= 0)   /* Remove caseless flag for non-caseable chars */      if (firstbyte >= 0)   /* Remove caseless flag for non-caseable chars */
7302        {        {
7303        int ch = firstbyte & 255;        int ch = firstbyte & 255;

Legend:
Removed from v.532  
changed lines
  Added in v.613

  ViewVC Help
Powered by ViewVC 1.1.5