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

Diff of /code/trunk/pcre_compile.c

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

revision 199 by ph10, Tue Jul 31 14:39:09 2007 UTC revision 275 by ph10, Wed Nov 21 15:35:09 2007 UTC
# Line 42  POSSIBILITY OF SUCH DAMAGE. Line 42  POSSIBILITY OF SUCH DAMAGE.
42  supporting internal functions that are not used by other modules. */  supporting internal functions that are not used by other modules. */
43    
44    
45  #ifdef HAVE_CONFIG_H  #ifdef HAVE_CONFIG_H
46  #include <config.h>  #include "config.h"
47  #endif  #endif
48    
49  #define NLBLOCK cd             /* Block containing newline information */  #define NLBLOCK cd             /* Block containing newline information */
50  #define PSSTART start_pattern  /* Field containing processed string start */  #define PSSTART start_pattern  /* Field containing processed string start */
# Line 65  used by pcretest. DEBUG is not defined w Line 65  used by pcretest. DEBUG is not defined w
65    
66  #define SETBIT(a,b) a[b/8] |= (1 << (b%8))  #define SETBIT(a,b) a[b/8] |= (1 << (b%8))
67    
68    /* Maximum length value to check against when making sure that the integer that
69    holds the compiled pattern length does not overflow. We make it a bit less than
70    INT_MAX to allow for adding in group terminating bytes, so that we don't have
71    to check them every time. */
72    
73    #define OFLOW_MAX (INT_MAX - 20)
74    
75    
76  /*************************************************  /*************************************************
77  *      Code parameters and static tables         *  *      Code parameters and static tables         *
# Line 133  static const short int escapes[] = { Line 140  static const short int escapes[] = {
140  #endif  #endif
141    
142    
143  /* Tables of names of POSIX character classes and their lengths. The list is  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
144  terminated by a zero length entry. The first three must be alpha, lower, upper,  searched linearly. Put all the names into a single string, in order to reduce
145  as this is assumed for handling case independence. */  the number of relocations when a shared library is dynamically linked. */
146    
147  static const char *const posix_names[] = {  typedef struct verbitem {
148    "alpha", "lower", "upper",    int   len;
149    "alnum", "ascii", "blank", "cntrl", "digit", "graph",    int   op;
150    "print", "punct", "space", "word",  "xdigit" };  } verbitem;
151    
152    static const char verbnames[] =
153      "ACCEPT\0"
154      "COMMIT\0"
155      "F\0"
156      "FAIL\0"
157      "PRUNE\0"
158      "SKIP\0"
159      "THEN";
160    
161    static verbitem verbs[] = {
162      { 6, OP_ACCEPT },
163      { 6, OP_COMMIT },
164      { 1, OP_FAIL },
165      { 4, OP_FAIL },
166      { 5, OP_PRUNE },
167      { 4, OP_SKIP  },
168      { 4, OP_THEN  }
169    };
170    
171    static int verbcount = sizeof(verbs)/sizeof(verbitem);
172    
173    
174    /* Tables of names of POSIX character classes and their lengths. The names are
175    now all in a single string, to reduce the number of relocations when a shared
176    library is dynamically loaded. The list of lengths is terminated by a zero
177    length entry. The first three must be alpha, lower, upper, as this is assumed
178    for handling case independence. */
179    
180    static const char posix_names[] =
181      "alpha\0"  "lower\0"  "upper\0"  "alnum\0"  "ascii\0"  "blank\0"
182      "cntrl\0"  "digit\0"  "graph\0"  "print\0"  "punct\0"  "space\0"
183      "word\0"   "xdigit";
184    
185  static const uschar posix_name_lengths[] = {  static const uschar posix_name_lengths[] = {
186    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
# Line 179  static const int posix_class_maps[] = { Line 219  static const int posix_class_maps[] = {
219  /* The texts of compile-time error messages. These are "char *" because they  /* The texts of compile-time error messages. These are "char *" because they
220  are passed to the outside world. Do not ever re-use any error number, because  are passed to the outside world. Do not ever re-use any error number, because
221  they are documented. Always add a new error instead. Messages marked DEAD below  they are documented. Always add a new error instead. Messages marked DEAD below
222  are no longer used. */  are no longer used. This used to be a table of strings, but in order to reduce
223    the number of relocations needed when a shared library is loaded dynamically,
224  static const char *error_texts[] = {  it is now one long string. We cannot use a table of offsets, because the
225    "no error",  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
226    "\\ at end of pattern",  simply count through to the one we want - this isn't a performance issue
227    "\\c at end of pattern",  because these strings are used only when there is a compilation error. */
228    "unrecognized character follows \\",  
229    "numbers out of order in {} quantifier",  static const char error_texts[] =
230      "no error\0"
231      "\\ at end of pattern\0"
232      "\\c at end of pattern\0"
233      "unrecognized character follows \\\0"
234      "numbers out of order in {} quantifier\0"
235    /* 5 */    /* 5 */
236    "number too big in {} quantifier",    "number too big in {} quantifier\0"
237    "missing terminating ] for character class",    "missing terminating ] for character class\0"
238    "invalid escape sequence in character class",    "invalid escape sequence in character class\0"
239    "range out of order in character class",    "range out of order in character class\0"
240    "nothing to repeat",    "nothing to repeat\0"
241    /* 10 */    /* 10 */
242    "operand of unlimited repeat could match the empty string",  /** DEAD **/    "operand of unlimited repeat could match the empty string\0"  /** DEAD **/
243    "internal error: unexpected repeat",    "internal error: unexpected repeat\0"
244    "unrecognized character after (?",    "unrecognized character after (? or (?-\0"
245    "POSIX named classes are supported only within a class",    "POSIX named classes are supported only within a class\0"
246    "missing )",    "missing )\0"
247    /* 15 */    /* 15 */
248    "reference to non-existent subpattern",    "reference to non-existent subpattern\0"
249    "erroffset passed as NULL",    "erroffset passed as NULL\0"
250    "unknown option bit(s) set",    "unknown option bit(s) set\0"
251    "missing ) after comment",    "missing ) after comment\0"
252    "parentheses nested too deeply",  /** DEAD **/    "parentheses nested too deeply\0"  /** DEAD **/
253    /* 20 */    /* 20 */
254    "regular expression too large",    "regular expression is too large\0"
255    "failed to get memory",    "failed to get memory\0"
256    "unmatched parentheses",    "unmatched parentheses\0"
257    "internal error: code overflow",    "internal error: code overflow\0"
258    "unrecognized character after (?<",    "unrecognized character after (?<\0"
259    /* 25 */    /* 25 */
260    "lookbehind assertion is not fixed length",    "lookbehind assertion is not fixed length\0"
261    "malformed number or name after (?(",    "malformed number or name after (?(\0"
262    "conditional group contains more than two branches",    "conditional group contains more than two branches\0"
263    "assertion expected after (?(",    "assertion expected after (?(\0"
264    "(?R or (?[+-]digits must be followed by )",    "(?R or (?[+-]digits must be followed by )\0"
265    /* 30 */    /* 30 */
266    "unknown POSIX class name",    "unknown POSIX class name\0"
267    "POSIX collating elements are not supported",    "POSIX collating elements are not supported\0"
268    "this version of PCRE is not compiled with PCRE_UTF8 support",    "this version of PCRE is not compiled with PCRE_UTF8 support\0"
269    "spare error",  /** DEAD **/    "spare error\0"  /** DEAD **/
270    "character value in \\x{...} sequence is too large",    "character value in \\x{...} sequence is too large\0"
271    /* 35 */    /* 35 */
272    "invalid condition (?(0)",    "invalid condition (?(0)\0"
273    "\\C not allowed in lookbehind assertion",    "\\C not allowed in lookbehind assertion\0"
274    "PCRE does not support \\L, \\l, \\N, \\U, or \\u",    "PCRE does not support \\L, \\l, \\N, \\U, or \\u\0"
275    "number after (?C is > 255",    "number after (?C is > 255\0"
276    "closing ) for (?C expected",    "closing ) for (?C expected\0"
277    /* 40 */    /* 40 */
278    "recursive call could loop indefinitely",    "recursive call could loop indefinitely\0"
279    "unrecognized character after (?P",    "unrecognized character after (?P\0"
280    "syntax error in subpattern name (missing terminator)",    "syntax error in subpattern name (missing terminator)\0"
281    "two named subpatterns have the same name",    "two named subpatterns have the same name\0"
282    "invalid UTF-8 string",    "invalid UTF-8 string\0"
283    /* 45 */    /* 45 */
284    "support for \\P, \\p, and \\X has not been compiled",    "support for \\P, \\p, and \\X has not been compiled\0"
285    "malformed \\P or \\p sequence",    "malformed \\P or \\p sequence\0"
286    "unknown property name after \\P or \\p",    "unknown property name after \\P or \\p\0"
287    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)",    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
288    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")",    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
289    /* 50 */    /* 50 */
290    "repeated subpattern is too long",    "repeated subpattern is too long\0"    /** DEAD **/
291    "octal value is greater than \\377 (not in UTF-8 mode)",    "octal value is greater than \\377 (not in UTF-8 mode)\0"
292    "internal error: overran compiling workspace",    "internal error: overran compiling workspace\0"
293    "internal error: previously-checked referenced subpattern not found",    "internal error: previously-checked referenced subpattern not found\0"
294    "DEFINE group contains more than one branch",    "DEFINE group contains more than one branch\0"
295    /* 55 */    /* 55 */
296    "repeating a DEFINE group is not allowed",    "repeating a DEFINE group is not allowed\0"
297    "inconsistent NEWLINE options",    "inconsistent NEWLINE options\0"
298    "\\g is not followed by a braced name or an optionally braced non-zero number",    "\\g is not followed by a braced name or an optionally braced non-zero number\0"
299    "(?+ or (?- or (?(+ or (?(- must be followed by a non-zero number"    "(?+ or (?- or (?(+ or (?(- must be followed by a non-zero number\0"
300  };    "(*VERB) with an argument is not supported\0"
301      /* 60 */
302      "(*VERB) not recognized\0"
303      "number is too big\0"
304      "subpattern name expected\0"
305      "digit expected after (?+";
306    
307    
308  /* 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 388  static BOOL Line 438  static BOOL
438    
439    
440  /*************************************************  /*************************************************
441    *            Find an error text                  *
442    *************************************************/
443    
444    /* The error texts are now all in one long string, to save on relocations. As
445    some of the text is of unknown length, we can't use a table of offsets.
446    Instead, just count through the strings. This is not a performance issue
447    because it happens only when there has been a compilation error.
448    
449    Argument:   the error number
450    Returns:    pointer to the error string
451    */
452    
453    static const char *
454    find_error_text(int n)
455    {
456    const char *s = error_texts;
457    for (; n > 0; n--) while (*s++ != 0);
458    return s;
459    }
460    
461    
462    /*************************************************
463  *            Handle escapes                      *  *            Handle escapes                      *
464  *************************************************/  *************************************************/
465    
# Line 408  Arguments: Line 480  Arguments:
480    
481  Returns:         zero or positive => a data character  Returns:         zero or positive => a data character
482                   negative => a special escape sequence                   negative => a special escape sequence
483                   on error, errorptr is set                   on error, errorcodeptr is set
484  */  */
485    
486  static int  static int
# Line 426  ptr--;                            /* Set Line 498  ptr--;                            /* Set
498    
499  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
500    
501  /* Non-alphamerics are literals. For digits or letters, do an initial lookup in  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
502  a table. A non-zero result is something that can be returned immediately.  in a table. A non-zero result is something that can be returned immediately.
503  Otherwise further processing may be required. */  Otherwise further processing may be required. */
504    
505  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII coding */
506  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  else if (c < '0' || c > 'z') {}                           /* Not alphanumeric */
507  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - '0']) != 0) c = i;
508    
509  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
510  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */
511  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
512  #endif  #endif
513    
# Line 493  else Line 565  else
565      while ((digitab[ptr[1]] & ctype_digit) != 0)      while ((digitab[ptr[1]] & ctype_digit) != 0)
566        c = c * 10 + *(++ptr) - '0';        c = c * 10 + *(++ptr) - '0';
567    
568        if (c < 0)
569          {
570          *errorcodeptr = ERR61;
571          break;
572          }
573    
574      if (c == 0 || (braced && *(++ptr) != '}'))      if (c == 0 || (braced && *(++ptr) != '}'))
575        {        {
576        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
577        return 0;        break;
578        }        }
579    
580      if (negated)      if (negated)
# Line 504  else Line 582  else
582        if (c > bracount)        if (c > bracount)
583          {          {
584          *errorcodeptr = ERR15;          *errorcodeptr = ERR15;
585          return 0;          break;
586          }          }
587        c = bracount - (c - 1);        c = bracount - (c - 1);
588        }        }
# Line 533  else Line 611  else
611        c -= '0';        c -= '0';
612        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
613          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - '0';
614          if (c < 0)
615            {
616            *errorcodeptr = ERR61;
617            break;
618            }
619        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
620          {          {
621          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 628  else Line 711  else
711      if (c == 0)      if (c == 0)
712        {        {
713        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
714        return 0;        break;
715        }        }
716    
717  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII coding */
# Line 641  else Line 724  else
724      break;      break;
725    
726      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
727      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
728      for Perl compatibility, it is a literal. This code looks a bit odd, but      otherwise, for Perl compatibility, it is a literal. This code looks a bit
729      there used to be some cases other than the default, and there may be again      odd, but there used to be some cases other than the default, and there may
730      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
731    
732      default:      default:
733      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 733  top = _pcre_utt_size; Line 816  top = _pcre_utt_size;
816  while (bot < top)  while (bot < top)
817    {    {
818    i = (bot + top) >> 1;    i = (bot + top) >> 1;
819    c = strcmp(name, _pcre_utt[i].name);    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);
820    if (c == 0)    if (c == 0)
821      {      {
822      *dptr = _pcre_utt[i].value;      *dptr = _pcre_utt[i].value;
# Line 907  for (; *ptr != 0; ptr++) Line 990  for (; *ptr != 0; ptr++)
990      {      {
991      while (*(++ptr) != ']')      while (*(++ptr) != ']')
992        {        {
993          if (*ptr == 0) return -1;
994        if (*ptr == '\\')        if (*ptr == '\\')
995          {          {
996          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) return -1;
# Line 934  for (; *ptr != 0; ptr++) Line 1018  for (; *ptr != 0; ptr++)
1018    /* An opening parens must now be a real metacharacter */    /* An opening parens must now be a real metacharacter */
1019    
1020    if (*ptr != '(') continue;    if (*ptr != '(') continue;
1021    if (ptr[1] != '?')    if (ptr[1] != '?' && ptr[1] != '*')
1022      {      {
1023      count++;      count++;
1024      if (name == NULL && count == lorn) return count;      if (name == NULL && count == lorn) return count;
# Line 1062  for (;;) Line 1146  for (;;)
1146    {    {
1147    int d;    int d;
1148    register int op = *cc;    register int op = *cc;
   
1149    switch (op)    switch (op)
1150      {      {
1151      case OP_CBRA:      case OP_CBRA:
# Line 1151  for (;;) Line 1234  for (;;)
1234    
1235      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1236      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1237        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1238      cc += 4;      cc += 4;
1239      break;      break;
1240    
# Line 1259  for (;;) Line 1343  for (;;)
1343      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1344      }      }
1345    
1346    /* In UTF-8 mode, opcodes that are followed by a character may be followed by    /* Otherwise, we can get the item's length from the table, except that for
1347    a multi-byte character. The length in the table is a minimum, so we have to    repeated character types, we have to test for \p and \P, which have an extra
1348    arrange to skip the extra bytes. */    two bytes of parameters. */
1349    
1350    else    else
1351      {      {
1352        switch(c)
1353          {
1354          case OP_TYPESTAR:
1355          case OP_TYPEMINSTAR:
1356          case OP_TYPEPLUS:
1357          case OP_TYPEMINPLUS:
1358          case OP_TYPEQUERY:
1359          case OP_TYPEMINQUERY:
1360          case OP_TYPEPOSSTAR:
1361          case OP_TYPEPOSPLUS:
1362          case OP_TYPEPOSQUERY:
1363          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1364          break;
1365    
1366          case OP_TYPEUPTO:
1367          case OP_TYPEMINUPTO:
1368          case OP_TYPEEXACT:
1369          case OP_TYPEPOSUPTO:
1370          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1371          break;
1372          }
1373    
1374        /* Add in the fixed length from the table */
1375    
1376      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1377    
1378      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1379      a multi-byte character. The length in the table is a minimum, so we have to
1380      arrange to skip the extra bytes. */
1381    
1382  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1383      if (utf8) switch(c)      if (utf8) switch(c)
1384        {        {
# Line 1323  for (;;) Line 1436  for (;;)
1436    
1437    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1438    
1439    /* Otherwise, we get the item's length from the table. In UTF-8 mode, opcodes    /* Otherwise, we can get the item's length from the table, except that for
1440    that are followed by a character may be followed by a multi-byte character.    repeated character types, we have to test for \p and \P, which have an extra
1441    The length in the table is a minimum, so we have to arrange to skip the extra    two bytes of parameters. */
   bytes. */  
1442    
1443    else    else
1444      {      {
1445        switch(c)
1446          {
1447          case OP_TYPESTAR:
1448          case OP_TYPEMINSTAR:
1449          case OP_TYPEPLUS:
1450          case OP_TYPEMINPLUS:
1451          case OP_TYPEQUERY:
1452          case OP_TYPEMINQUERY:
1453          case OP_TYPEPOSSTAR:
1454          case OP_TYPEPOSPLUS:
1455          case OP_TYPEPOSQUERY:
1456          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1457          break;
1458    
1459          case OP_TYPEPOSUPTO:
1460          case OP_TYPEUPTO:
1461          case OP_TYPEMINUPTO:
1462          case OP_TYPEEXACT:
1463          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1464          break;
1465          }
1466    
1467        /* Add in the fixed length from the table */
1468    
1469      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1470    
1471        /* In UTF-8 mode, opcodes that are followed by a character may be followed
1472        by a multi-byte character. The length in the table is a minimum, so we have
1473        to arrange to skip the extra bytes. */
1474    
1475  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1476      if (utf8) switch(c)      if (utf8) switch(c)
1477        {        {
# Line 1402  for (code = first_significant_code(code Line 1543  for (code = first_significant_code(code
1543    
1544    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
1545    
1546    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE)    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)
1547      {      {
1548      BOOL empty_branch;      BOOL empty_branch;
1549      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
# Line 1426  for (code = first_significant_code(code Line 1567  for (code = first_significant_code(code
1567    
1568    switch (c)    switch (c)
1569      {      {
1570      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
1571        cannot be represented just by a bit map. This includes negated single
1572        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
1573        actual length is stored in the compiled code, so we must update "code"
1574        here. */
1575    
1576  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1577      case OP_XCLASS:      case OP_XCLASS:
1578      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
1579      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
1580  #endif  #endif
1581    
# Line 1492  for (code = first_significant_code(code Line 1637  for (code = first_significant_code(code
1637      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1638      return FALSE;      return FALSE;
1639    
1640        /* These are going to continue, as they may be empty, but we have to
1641        fudge the length for the \p and \P cases. */
1642    
1643        case OP_TYPESTAR:
1644        case OP_TYPEMINSTAR:
1645        case OP_TYPEPOSSTAR:
1646        case OP_TYPEQUERY:
1647        case OP_TYPEMINQUERY:
1648        case OP_TYPEPOSQUERY:
1649        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1650        break;
1651    
1652        /* Same for these */
1653    
1654        case OP_TYPEUPTO:
1655        case OP_TYPEMINUPTO:
1656        case OP_TYPEPOSUPTO:
1657        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1658        break;
1659    
1660      /* End of branch */      /* End of branch */
1661    
1662      case OP_KET:      case OP_KET:
# Line 1608  Returns:     a value representing the na Line 1773  Returns:     a value representing the na
1773  static int  static int
1774  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
1775  {  {
1776    const char *pn = posix_names;
1777  register int yield = 0;  register int yield = 0;
1778  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
1779    {    {
1780    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
1781      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
1782      pn += posix_name_lengths[yield] + 1;
1783    yield++;    yield++;
1784    }    }
1785  return -1;  return -1;
# Line 1654  adjust_recurse(uschar *group, int adjust Line 1821  adjust_recurse(uschar *group, int adjust
1821    uschar *save_hwm)    uschar *save_hwm)
1822  {  {
1823  uschar *ptr = group;  uschar *ptr = group;
1824    
1825  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
1826    {    {
1827    int offset;    int offset;
# Line 2217  req_caseopt = ((options & PCRE_CASELESS) Line 2385  req_caseopt = ((options & PCRE_CASELESS)
2385  for (;; ptr++)  for (;; ptr++)
2386    {    {
2387    BOOL negate_class;    BOOL negate_class;
2388      BOOL should_flip_negation;
2389    BOOL possessive_quantifier;    BOOL possessive_quantifier;
2390    BOOL is_quantifier;    BOOL is_quantifier;
2391    BOOL is_recurse;    BOOL is_recurse;
# Line 2258  for (;; ptr++) Line 2427  for (;; ptr++)
2427      */      */
2428    
2429      if (code < last_code) code = last_code;      if (code < last_code) code = last_code;
2430    
2431        /* Paranoid check for integer overflow */
2432    
2433        if (OFLOW_MAX - *lengthptr < code - last_code)
2434          {
2435          *errorcodeptr = ERR20;
2436          goto FAILED;
2437          }
2438    
2439      *lengthptr += code - last_code;      *lengthptr += code - last_code;
2440      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));
2441    
# Line 2370  for (;; ptr++) Line 2548  for (;; ptr++)
2548      *ptrptr = ptr;      *ptrptr = ptr;
2549      if (lengthptr != NULL)      if (lengthptr != NULL)
2550        {        {
2551          if (OFLOW_MAX - *lengthptr < code - last_code)
2552            {
2553            *errorcodeptr = ERR20;
2554            goto FAILED;
2555            }
2556        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += code - last_code;   /* To include callout length */
2557        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
2558        }        }
# Line 2432  for (;; ptr++) Line 2615  for (;; ptr++)
2615        goto FAILED;        goto FAILED;
2616        }        }
2617    
2618      /* If the first character is '^', set the negation flag and skip it. */      /* If the first character is '^', set the negation flag and skip it. Also,
2619        if the first few characters (either before or after ^) are \Q\E or \E we
2620        skip them too. This makes for compatibility with Perl. */
2621    
2622      if ((c = *(++ptr)) == '^')      negate_class = FALSE;
2623        for (;;)
2624        {        {
       negate_class = TRUE;  
2625        c = *(++ptr);        c = *(++ptr);
2626          if (c == '\\')
2627            {
2628            if (ptr[1] == 'E') ptr++;
2629              else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;
2630                else break;
2631            }
2632          else if (!negate_class && c == '^')
2633            negate_class = TRUE;
2634          else break;
2635        }        }
2636      else  
2637        {      /* If a class contains a negative special such as \S, we need to flip the
2638        negate_class = FALSE;      negation flag at the end, so that support for characters > 255 works
2639        }      correctly (they are all included in the class). */
2640    
2641        should_flip_negation = FALSE;
2642    
2643      /* Keep a count of chars with values < 256 so that we can optimize the case      /* Keep a count of chars with values < 256 so that we can optimize the case
2644      of just a single character (as long as it's < 256). However, For higher      of just a single character (as long as it's < 256). However, For higher
# Line 2516  for (;; ptr++) Line 2712  for (;; ptr++)
2712          if (*ptr == '^')          if (*ptr == '^')
2713            {            {
2714            local_negate = TRUE;            local_negate = TRUE;
2715              should_flip_negation = TRUE;  /* Note negative special */
2716            ptr++;            ptr++;
2717            }            }
2718    
# Line 2582  for (;; ptr++) Line 2779  for (;; ptr++)
2779        of the specials, which just set a flag. The sequence \b is a special        of the specials, which just set a flag. The sequence \b is a special
2780        case. Inside a class (and only there) it is treated as backspace.        case. Inside a class (and only there) it is treated as backspace.
2781        Elsewhere it marks a word boundary. Other escapes have preset maps ready        Elsewhere it marks a word boundary. Other escapes have preset maps ready
2782        to or into the one we are building. We assume they have more than one        to 'or' into the one we are building. We assume they have more than one
2783        character in them, so set class_charcount bigger than one. */        character in them, so set class_charcount bigger than one. */
2784    
2785        if (c == '\\')        if (c == '\\')
# Line 2590  for (;; ptr++) Line 2787  for (;; ptr++)
2787          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
2788          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
2789    
2790          if (-c == ESC_b) c = '\b';       /* \b is backslash in a class */          if (-c == ESC_b) c = '\b';       /* \b is backspace in a class */
2791          else if (-c == ESC_X) c = 'X';   /* \X is literal X in a class */          else if (-c == ESC_X) c = 'X';   /* \X is literal X in a class */
2792          else if (-c == ESC_R) c = 'R';   /* \R is literal R in a class */          else if (-c == ESC_R) c = 'R';   /* \R is literal R in a class */
2793          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
# Line 2602  for (;; ptr++) Line 2799  for (;; ptr++)
2799            else inescq = TRUE;            else inescq = TRUE;
2800            continue;            continue;
2801            }            }
2802            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
2803    
2804          if (c < 0)          if (c < 0)
2805            {            {
# Line 2617  for (;; ptr++) Line 2815  for (;; ptr++)
2815              continue;              continue;
2816    
2817              case ESC_D:              case ESC_D:
2818                should_flip_negation = TRUE;
2819              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
2820              continue;              continue;
2821    
# Line 2625  for (;; ptr++) Line 2824  for (;; ptr++)
2824              continue;              continue;
2825    
2826              case ESC_W:              case ESC_W:
2827                should_flip_negation = TRUE;
2828              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
2829              continue;              continue;
2830    
# Line 2634  for (;; ptr++) Line 2834  for (;; ptr++)
2834              continue;              continue;
2835    
2836              case ESC_S:              case ESC_S:
2837                should_flip_negation = TRUE;
2838              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
2839              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
2840              continue;              continue;
# Line 2828  for (;; ptr++) Line 3029  for (;; ptr++)
3029    
3030        oldptr = ptr;        oldptr = ptr;
3031    
3032          /* Remember \r or \n */
3033    
3034          if (c == '\r' || c == '\n') cd->external_flags |= PCRE_HASCRORLF;
3035    
3036          /* Check for range */
3037    
3038        if (!inescq && ptr[1] == '-')        if (!inescq && ptr[1] == '-')
3039          {          {
3040          int d;          int d;
# Line 2869  for (;; ptr++) Line 3076  for (;; ptr++)
3076            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3077            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3078    
3079            /* \b is backslash; \X is literal X; \R is literal R; any other            /* \b is backspace; \X is literal X; \R is literal R; any other
3080            special means the '-' was literal */            special means the '-' was literal */
3081    
3082            if (d < 0)            if (d < 0)
# Line 2895  for (;; ptr++) Line 3102  for (;; ptr++)
3102    
3103          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3104    
3105            /* Remember \r or \n */
3106    
3107            if (d == '\r' || d == '\n') cd->external_flags |= PCRE_HASCRORLF;
3108    
3109          /* In UTF-8 mode, if the upper limit is > 255, or > 127 for caseless          /* In UTF-8 mode, if the upper limit is > 255, or > 127 for caseless
3110          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
3111          matching for characters > 127 is available only if UCP support is          matching for characters > 127 is available only if UCP support is
# Line 3048  for (;; ptr++) Line 3259  for (;; ptr++)
3259        goto FAILED;        goto FAILED;
3260        }        }
3261    
3262    
3263    /* This code has been disabled because it would mean that \s counts as
3264    an explicit \r or \n reference, and that's not really what is wanted. Now
3265    we set the flag only if there is a literal "\r" or "\n" in the class. */
3266    
3267    #if 0
3268        /* Remember whether \r or \n are in this class */
3269    
3270        if (negate_class)
3271          {
3272          if ((classbits[1] & 0x24) != 0x24) cd->external_flags |= PCRE_HASCRORLF;
3273          }
3274        else
3275          {
3276          if ((classbits[1] & 0x24) != 0) cd->external_flags |= PCRE_HASCRORLF;
3277          }
3278    #endif
3279    
3280    
3281      /* If class_charcount is 1, we saw precisely one character whose value is      /* If class_charcount is 1, we saw precisely one character whose value is
3282      less than 256. In non-UTF-8 mode we can always optimize. In UTF-8 mode, we      less than 256. As long as there were no characters >= 128 and there was no
3283      can optimize the negative case only if there were no characters >= 128      use of \p or \P, in other words, no use of any XCLASS features, we can
3284      because OP_NOT and the related opcodes like OP_NOTSTAR operate on      optimize.
3285      single-bytes only. This is an historical hangover. Maybe one day we can  
3286      tidy these opcodes to handle multi-byte characters.      In UTF-8 mode, we can optimize the negative case only if there were no
3287        characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
3288        operate on single-bytes only. This is an historical hangover. Maybe one day
3289        we can tidy these opcodes to handle multi-byte characters.
3290    
3291      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
3292      1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note      1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note
# Line 3063  for (;; ptr++) Line 3296  for (;; ptr++)
3296      reqbyte, save the previous value for reinstating. */      reqbyte, save the previous value for reinstating. */
3297    
3298  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3299      if (class_charcount == 1 &&      if (class_charcount == 1 && !class_utf8 &&
3300            (!utf8 ||        (!utf8 || !negate_class || class_lastchar < 128))
           (!class_utf8 && (!negate_class || class_lastchar < 128))))  
   
3301  #else  #else
3302      if (class_charcount == 1)      if (class_charcount == 1)
3303  #endif  #endif
# Line 3109  for (;; ptr++) Line 3340  for (;; ptr++)
3340      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3341    
3342      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
3343      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode, unless there was a negated special
3344      we can omit the bitmap in the actual compiled code. */      such as \S in the class, because in that case all characters > 255 are in
3345        the class, so any that were explicitly given as well can be ignored. If
3346        (when there are explicit characters > 255 that must be listed) there are no
3347        characters < 256, we can omit the bitmap in the actual compiled code. */
3348    
3349  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3350      if (class_utf8)      if (class_utf8 && !should_flip_negation)
3351        {        {
3352        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
3353        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 3139  for (;; ptr++) Line 3373  for (;; ptr++)
3373        }        }
3374  #endif  #endif
3375    
3376      /* If there are no characters > 255, negate the 32-byte map if necessary,      /* If there are no characters > 255, set the opcode to OP_CLASS or
3377      and copy it into the code vector. If this is the first thing in the branch,      OP_NCLASS, depending on whether the whole class was negated and whether
3378      there can be no first char setting, whatever the repeat count. Any reqbyte      there were negative specials such as \S in the class. Then copy the 32-byte
3379      setting must remain unchanged after any kind of repeat. */      map into the code vector, negating it if necessary. */
3380    
3381        *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
3382      if (negate_class)      if (negate_class)
3383        {        {
       *code++ = OP_NCLASS;  
3384        if (lengthptr == NULL)    /* Save time in the pre-compile phase */        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
3385          for (c = 0; c < 32; c++) code[c] = ~classbits[c];          for (c = 0; c < 32; c++) code[c] = ~classbits[c];
3386        }        }
3387      else      else
3388        {        {
       *code++ = OP_CLASS;  
3389        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
3390        }        }
3391      code += 32;      code += 32;
# Line 3338  for (;; ptr++) Line 3571  for (;; ptr++)
3571        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
3572        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
3573    
3574        if (repeat_max != 1) cd->nopartial = TRUE;        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;
3575    
3576        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
3577    
# Line 3488  for (;; ptr++) Line 3721  for (;; ptr++)
3721        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
3722        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
3723    
3724        if (repeat_max != 1) cd->nopartial = TRUE;        if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL;
3725    
3726        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
3727          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 3524  for (;; ptr++) Line 3757  for (;; ptr++)
3757          goto FAILED;          goto FAILED;
3758          }          }
3759    
       /* This is a paranoid check to stop integer overflow later on */  
   
       if (len > MAX_DUPLENGTH)  
         {  
         *errorcodeptr = ERR50;  
         goto FAILED;  
         }  
   
3760        /* If the maximum repeat count is unlimited, find the end of the bracket        /* If the maximum repeat count is unlimited, find the end of the bracket
3761        by scanning through from the start, and compute the offset back to it        by scanning through from the start, and compute the offset back to it
3762        from the current code pointer. There may be an OP_OPT setting following        from the current code pointer. There may be an OP_OPT setting following
# Line 3620  for (;; ptr++) Line 3845  for (;; ptr++)
3845          if (repeat_min > 1)          if (repeat_min > 1)
3846            {            {
3847            /* In the pre-compile phase, we don't actually do the replication. We            /* In the pre-compile phase, we don't actually do the replication. We
3848            just adjust the length as if we had. */            just adjust the length as if we had. Do some paranoid checks for
3849              potential integer overflow. */
3850    
3851            if (lengthptr != NULL)            if (lengthptr != NULL)
3852              *lengthptr += (repeat_min - 1)*length_prevgroup;              {
3853                int delta = (repeat_min - 1)*length_prevgroup;
3854                if ((double)(repeat_min - 1)*(double)length_prevgroup >
3855                                                                (double)INT_MAX ||
3856                    OFLOW_MAX - *lengthptr < delta)
3857                  {
3858                  *errorcodeptr = ERR20;
3859                  goto FAILED;
3860                  }
3861                *lengthptr += delta;
3862                }
3863    
3864            /* This is compiling for real */            /* This is compiling for real */
3865    
# Line 3661  for (;; ptr++) Line 3897  for (;; ptr++)
3897          /* In the pre-compile phase, we don't actually do the replication. We          /* In the pre-compile phase, we don't actually do the replication. We
3898          just adjust the length as if we had. For each repetition we must add 1          just adjust the length as if we had. For each repetition we must add 1
3899          to the length for BRAZERO and for all but the last repetition we must          to the length for BRAZERO and for all but the last repetition we must
3900          add 2 + 2*LINKSIZE to allow for the nesting that occurs. */          add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
3901            paranoid checks to avoid integer overflow. */
3902    
3903          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
3904            *lengthptr += repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            {
3905              2 - 2*LINK_SIZE;  /* Last one doesn't nest */            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
3906                          2 - 2*LINK_SIZE;   /* Last one doesn't nest */
3907              if ((double)repeat_max *
3908                    (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
3909                      > (double)INT_MAX ||
3910                  OFLOW_MAX - *lengthptr < delta)
3911                {
3912                *errorcodeptr = ERR20;
3913                goto FAILED;
3914                }
3915              *lengthptr += delta;
3916              }
3917    
3918          /* This is compiling for real */          /* This is compiling for real */
3919    
# Line 3817  for (;; ptr++) Line 4065  for (;; ptr++)
4065      /* ===================================================================*/      /* ===================================================================*/
4066      /* Start of nested parenthesized sub-expression, or comment or lookahead or      /* Start of nested parenthesized sub-expression, or comment or lookahead or
4067      lookbehind or option setting or condition or all the other extended      lookbehind or option setting or condition or all the other extended
4068      parenthesis forms. First deal with the specials; all are introduced by ?,      parenthesis forms.  */
     and the appearance of any of them means that this is not a capturing  
     group. */  
4069    
4070      case '(':      case '(':
4071      newoptions = options;      newoptions = options;
# Line 3828  for (;; ptr++) Line 4074  for (;; ptr++)
4074      save_hwm = cd->hwm;      save_hwm = cd->hwm;
4075      reset_bracount = FALSE;      reset_bracount = FALSE;
4076    
4077      if (*(++ptr) == '?')      /* First deal with various "verbs" that can be introduced by '*'. */
4078    
4079        if (*(++ptr) == '*' && (cd->ctypes[ptr[1]] & ctype_letter) != 0)
4080          {
4081          int i, namelen;
4082          const char *vn = verbnames;
4083          const uschar *name = ++ptr;
4084          previous = NULL;
4085          while ((cd->ctypes[*++ptr] & ctype_letter) != 0);
4086          if (*ptr == ':')
4087            {
4088            *errorcodeptr = ERR59;   /* Not supported */
4089            goto FAILED;
4090            }
4091          if (*ptr != ')')
4092            {
4093            *errorcodeptr = ERR60;
4094            goto FAILED;
4095            }
4096          namelen = ptr - name;
4097          for (i = 0; i < verbcount; i++)
4098            {
4099            if (namelen == verbs[i].len &&
4100                strncmp((char *)name, vn, namelen) == 0)
4101              {
4102              *code = verbs[i].op;
4103              if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;
4104              break;
4105              }
4106            vn += verbs[i].len + 1;
4107            }
4108          if (i < verbcount) continue;
4109          *errorcodeptr = ERR60;
4110          goto FAILED;
4111          }
4112    
4113        /* Deal with the extended parentheses; all are introduced by '?', and the
4114        appearance of any of them means that this is not a capturing group. */
4115    
4116        else if (*ptr == '?')
4117        {        {
4118        int i, set, unset, namelen;        int i, set, unset, namelen;
4119        int *optset;        int *optset;
# Line 3963  for (;; ptr++) Line 4248  for (;; ptr++)
4248              *errorcodeptr = ERR58;              *errorcodeptr = ERR58;
4249              goto FAILED;              goto FAILED;
4250              }              }
4251            if (refsign == '-')            recno = (refsign == '-')?
4252                cd->bracount - recno + 1 : recno +cd->bracount;
4253              if (recno <= 0 || recno > cd->final_bracount)
4254              {              {
4255              recno = cd->bracount - recno + 1;              *errorcodeptr = ERR15;
4256              if (recno <= 0)              goto FAILED;
               {  
               *errorcodeptr = ERR15;  
               goto FAILED;  
               }  
4257              }              }
           else recno += cd->bracount;  
4258            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4259            break;            break;
4260            }            }
# Line 4044  for (;; ptr++) Line 4326  for (;; ptr++)
4326            skipbytes = 1;            skipbytes = 1;
4327            }            }
4328    
4329          /* Check for the "name" actually being a subpattern number. */          /* Check for the "name" actually being a subpattern number. We are
4330            in the second pass here, so final_bracount is set. */
4331    
4332          else if (recno > 0)          else if (recno > 0 && recno <= cd->final_bracount)
4333            {            {
4334            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4335            }            }
# Line 4070  for (;; ptr++) Line 4353  for (;; ptr++)
4353    
4354          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4355          case '!':                 /* Negative lookahead */          case '!':                 /* Negative lookahead */
         bravalue = OP_ASSERT_NOT;  
4356          ptr++;          ptr++;
4357            if (*ptr == ')')          /* Optimize (?!) */
4358              {
4359              *code++ = OP_FAIL;
4360              previous = NULL;
4361              continue;
4362              }
4363            bravalue = OP_ASSERT_NOT;
4364          break;          break;
4365    
4366    
# Line 4234  for (;; ptr++) Line 4523  for (;; ptr++)
4523    
4524          /* We come here from the Python syntax above that handles both          /* We come here from the Python syntax above that handles both
4525          references (?P=name) and recursion (?P>name), as well as falling          references (?P=name) and recursion (?P>name), as well as falling
4526          through from the Perl recursion syntax (?&name). */          through from the Perl recursion syntax (?&name). We also come here from
4527            the Perl \k<name> or \k'name' back reference syntax and the \k{name}
4528            .NET syntax. */
4529    
4530          NAMED_REF_OR_RECURSE:          NAMED_REF_OR_RECURSE:
4531          name = ++ptr;          name = ++ptr;
# Line 4246  for (;; ptr++) Line 4537  for (;; ptr++)
4537    
4538          if (lengthptr != NULL)          if (lengthptr != NULL)
4539            {            {
4540              if (namelen == 0)
4541                {
4542                *errorcodeptr = ERR62;
4543                goto FAILED;
4544                }
4545            if (*ptr != terminator)            if (*ptr != terminator)
4546              {              {
4547              *errorcodeptr = ERR42;              *errorcodeptr = ERR42;
# Line 4259  for (;; ptr++) Line 4555  for (;; ptr++)
4555            recno = 0;            recno = 0;
4556            }            }
4557    
4558          /* In the real compile, seek the name in the table */          /* In the real compile, seek the name in the table. We check the name
4559            first, and then check that we have reached the end of the name in the
4560            table. That way, if the name that is longer than any in the table,
4561            the comparison will fail without reading beyond the table entry. */
4562    
4563          else          else
4564            {            {
4565            slot = cd->name_table;            slot = cd->name_table;
4566            for (i = 0; i < cd->names_found; i++)            for (i = 0; i < cd->names_found; i++)
4567              {              {
4568              if (strncmp((char *)name, (char *)slot+2, namelen) == 0) break;              if (strncmp((char *)name, (char *)slot+2, namelen) == 0 &&
4569                    slot[2+namelen] == 0)
4570                  break;
4571              slot += cd->name_entry_size;              slot += cd->name_entry_size;
4572              }              }
4573    
# Line 4303  for (;; ptr++) Line 4604  for (;; ptr++)
4604            {            {
4605            const uschar *called;            const uschar *called;
4606    
4607            if ((refsign = *ptr) == '+') ptr++;            if ((refsign = *ptr) == '+')
4608                {
4609                ptr++;
4610                if ((digitab[*ptr] & ctype_digit) == 0)
4611                  {
4612                  *errorcodeptr = ERR63;
4613                  goto FAILED;
4614                  }
4615                }
4616            else if (refsign == '-')            else if (refsign == '-')
4617              {              {
4618              if ((digitab[ptr[1]] & ctype_digit) == 0)              if ((digitab[ptr[1]] & ctype_digit) == 0)
# Line 4429  for (;; ptr++) Line 4738  for (;; ptr++)
4738    
4739              case 'J':    /* Record that it changed in the external options */              case 'J':    /* Record that it changed in the external options */
4740              *optset |= PCRE_DUPNAMES;              *optset |= PCRE_DUPNAMES;
4741              cd->external_options |= PCRE_JCHANGED;              cd->external_flags |= PCRE_JCHANGED;
4742              break;              break;
4743    
4744              case 'i': *optset |= PCRE_CASELESS; break;              case 'i': *optset |= PCRE_CASELESS; break;
# Line 4620  for (;; ptr++) Line 4929  for (;; ptr++)
4929        goto FAILED;        goto FAILED;
4930        }        }
4931    
4932      /* In the pre-compile phase, update the length by the length of the nested      /* In the pre-compile phase, update the length by the length of the group,
4933      group, less the brackets at either end. Then reduce the compiled code to      less the brackets at either end. Then reduce the compiled code to just a
4934      just the brackets so that it doesn't use much memory if it is duplicated by      set of non-capturing brackets so that it doesn't use much memory if it is
4935      a quantifier. */      duplicated by a quantifier.*/
4936    
4937      if (lengthptr != NULL)      if (lengthptr != NULL)
4938        {        {
4939          if (OFLOW_MAX - *lengthptr < length_prevgroup - 2 - 2*LINK_SIZE)
4940            {
4941            *errorcodeptr = ERR20;
4942            goto FAILED;
4943            }
4944        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;
4945        code++;        *code++ = OP_BRA;
4946        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
4947        *code++ = OP_KET;        *code++ = OP_KET;
4948        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
4949          break;    /* No need to waste time with special character handling */
4950        }        }
4951    
4952      /* Otherwise update the main code pointer to the end of the group. */      /* Otherwise update the main code pointer to the end of the group. */
4953    
4954      else code = tempcode;      code = tempcode;
4955    
4956      /* For a DEFINE group, required and first character settings are not      /* For a DEFINE group, required and first character settings are not
4957      relevant. */      relevant. */
# Line 4840  for (;; ptr++) Line 5155  for (;; ptr++)
5155      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARNC : OP_CHAR;      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARNC : OP_CHAR;
5156      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];
5157    
5158        /* Remember if \r or \n were seen */
5159    
5160        if (mcbuffer[0] == '\r' || mcbuffer[0] == '\n')
5161          cd->external_flags |= PCRE_HASCRORLF;
5162    
5163      /* Set the first and required bytes appropriately. If no previous first      /* Set the first and required bytes appropriately. If no previous first
5164      byte, set it from this character, but revert to none on a zero repeat.      byte, set it from this character, but revert to none on a zero repeat.
5165      Otherwise, leave the firstbyte value alone, and don't change it on a zero      Otherwise, leave the firstbyte value alone, and don't change it on a zero
# Line 5122  for (;;) Line 5442  for (;;)
5442      *ptrptr = ptr;      *ptrptr = ptr;
5443      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
5444      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
5445      if (lengthptr != NULL) *lengthptr += length;      if (lengthptr != NULL)
5446          {
5447          if (OFLOW_MAX - *lengthptr < length)
5448            {
5449            *errorcodeptr = ERR20;
5450            return FALSE;
5451            }
5452          *lengthptr += length;
5453          }
5454      return TRUE;      return TRUE;
5455      }      }
5456    
# Line 5431  real_pcre *re; Line 5759  real_pcre *re;
5759  int length = 1;  /* For final END opcode */  int length = 1;  /* For final END opcode */
5760  int firstbyte, reqbyte, newline;  int firstbyte, reqbyte, newline;
5761  int errorcode = 0;  int errorcode = 0;
5762    int skipatstart = 0;
5763  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
5764  BOOL utf8;  BOOL utf8;
5765  #endif  #endif
# Line 5509  cd->fcc = tables + fcc_offset; Line 5838  cd->fcc = tables + fcc_offset;
5838  cd->cbits = tables + cbits_offset;  cd->cbits = tables + cbits_offset;
5839  cd->ctypes = tables + ctypes_offset;  cd->ctypes = tables + ctypes_offset;
5840    
5841    /* Check for global one-time settings at the start of the pattern, and remember
5842    the offset for later. */
5843    
5844    while (ptr[skipatstart] == '(' && ptr[skipatstart+1] == '*')
5845      {
5846      int newnl = 0;
5847      int newbsr = 0;
5848    
5849      if (strncmp((char *)(ptr+skipatstart+2), "CR)", 3) == 0)
5850        { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }
5851      else if (strncmp((char *)(ptr+skipatstart+2), "LF)", 3)  == 0)
5852        { skipatstart += 5; newnl = PCRE_NEWLINE_LF; }
5853      else if (strncmp((char *)(ptr+skipatstart+2), "CRLF)", 5)  == 0)
5854        { skipatstart += 7; newnl = PCRE_NEWLINE_CR + PCRE_NEWLINE_LF; }
5855      else if (strncmp((char *)(ptr+skipatstart+2), "ANY)", 4) == 0)
5856        { skipatstart += 6; newnl = PCRE_NEWLINE_ANY; }
5857      else if (strncmp((char *)(ptr+skipatstart+2), "ANYCRLF)", 8)  == 0)
5858        { skipatstart += 10; newnl = PCRE_NEWLINE_ANYCRLF; }
5859    
5860      else if (strncmp((char *)(ptr+skipatstart+2), "BSR_ANYCRLF)", 12) == 0)
5861        { skipatstart += 14; newbsr = PCRE_BSR_ANYCRLF; }
5862      else if (strncmp((char *)(ptr+skipatstart+2), "BSR_UNICODE)", 12) == 0)
5863        { skipatstart += 14; newbsr = PCRE_BSR_UNICODE; }
5864    
5865      if (newnl != 0)
5866        options = (options & ~PCRE_NEWLINE_BITS) | newnl;
5867      else if (newbsr != 0)
5868        options = (options & ~(PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) | newbsr;
5869      else break;
5870      }
5871    
5872    /* Check validity of \R options. */
5873    
5874    switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))
5875      {
5876      case 0:
5877      case PCRE_BSR_ANYCRLF:
5878      case PCRE_BSR_UNICODE:
5879      break;
5880      default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;
5881      }
5882    
5883  /* Handle different types of newline. The three bits give seven cases. The  /* Handle different types of newline. The three bits give seven cases. The
5884  current code allows for fixed one- or two-byte sequences, plus "any" and  current code allows for fixed one- or two-byte sequences, plus "any" and
5885  "anycrlf". */  "anycrlf". */
5886    
5887  switch (options & (PCRE_NEWLINE_CRLF | PCRE_NEWLINE_ANY))  switch (options & PCRE_NEWLINE_BITS)
5888    {    {
5889    case 0: newline = NEWLINE; break;   /* Compile-time default */    case 0: newline = NEWLINE; break;   /* Build-time default */
5890    case PCRE_NEWLINE_CR: newline = '\r'; break;    case PCRE_NEWLINE_CR: newline = '\r'; break;
5891    case PCRE_NEWLINE_LF: newline = '\n'; break;    case PCRE_NEWLINE_LF: newline = '\n'; break;
5892    case PCRE_NEWLINE_CR+    case PCRE_NEWLINE_CR+
# Line 5568  to compile parts of the pattern into; th Line 5939  to compile parts of the pattern into; th
5939  no longer needed, so hopefully this workspace will never overflow, though there  no longer needed, so hopefully this workspace will never overflow, though there
5940  is a test for its doing so. */  is a test for its doing so. */
5941    
5942  cd->bracount = 0;  cd->bracount = cd->final_bracount = 0;
5943  cd->names_found = 0;  cd->names_found = 0;
5944  cd->name_entry_size = 0;  cd->name_entry_size = 0;
5945  cd->name_table = NULL;  cd->name_table = NULL;
# Line 5578  cd->hwm = cworkspace; Line 5949  cd->hwm = cworkspace;
5949  cd->start_pattern = (const uschar *)pattern;  cd->start_pattern = (const uschar *)pattern;
5950  cd->end_pattern = (const uschar *)(pattern + strlen(pattern));  cd->end_pattern = (const uschar *)(pattern + strlen(pattern));
5951  cd->req_varyopt = 0;  cd->req_varyopt = 0;
 cd->nopartial = FALSE;  
5952  cd->external_options = options;  cd->external_options = options;
5953    cd->external_flags = 0;
5954    
5955  /* Now do the pre-compile. On error, errorcode will be set non-zero, so we  /* Now do the pre-compile. On error, errorcode will be set non-zero, so we
5956  don't need to look at the result of the function here. The initial options have  don't need to look at the result of the function here. The initial options have
# Line 5587  been put into the cd block so that they Line 5958  been put into the cd block so that they
5958  found within the regex right at the beginning. Bringing initial option settings  found within the regex right at the beginning. Bringing initial option settings
5959  outside can help speed up starting point checks. */  outside can help speed up starting point checks. */
5960    
5961    ptr += skipatstart;
5962  code = cworkspace;  code = cworkspace;
5963  *code = OP_BRA;  *code = OP_BRA;
5964  (void)compile_regex(cd->external_options, cd->external_options & PCRE_IMS,  (void)compile_regex(cd->external_options, cd->external_options & PCRE_IMS,
# Line 5617  if (re == NULL) Line 5989  if (re == NULL)
5989    goto PCRE_EARLY_ERROR_RETURN;    goto PCRE_EARLY_ERROR_RETURN;
5990    }    }
5991    
5992  /* Put in the magic number, and save the sizes, initial options, and character  /* Put in the magic number, and save the sizes, initial options, internal
5993  table pointer. NULL is used for the default character tables. The nullpad field  flags, and character table pointer. NULL is used for the default character
5994  is at the end; it's there to help in the case when a regex compiled on a system  tables. The nullpad field is at the end; it's there to help in the case when a
5995  with 4-byte pointers is run on another with 8-byte pointers. */  regex compiled on a system with 4-byte pointers is run on another with 8-byte
5996    pointers. */
5997    
5998  re->magic_number = MAGIC_NUMBER;  re->magic_number = MAGIC_NUMBER;
5999  re->size = size;  re->size = size;
6000  re->options = cd->external_options;  re->options = cd->external_options;
6001    re->flags = cd->external_flags;
6002  re->dummy1 = 0;  re->dummy1 = 0;
6003  re->first_byte = 0;  re->first_byte = 0;
6004  re->req_byte = 0;  re->req_byte = 0;
# Line 5642  field. Reset the bracket count and the n Line 6016  field. Reset the bracket count and the n
6016  field; this time it's used for remembering forward references to subpatterns.  field; this time it's used for remembering forward references to subpatterns.
6017  */  */
6018    
6019    cd->final_bracount = cd->bracount;  /* Save for checking forward references */
6020  cd->bracount = 0;  cd->bracount = 0;
6021  cd->names_found = 0;  cd->names_found = 0;
6022  cd->name_table = (uschar *)re + re->name_table_offset;  cd->name_table = (uschar *)re + re->name_table_offset;
# Line 5649  codestart = cd->name_table + re->name_en Line 6024  codestart = cd->name_table + re->name_en
6024  cd->start_code = codestart;  cd->start_code = codestart;
6025  cd->hwm = cworkspace;  cd->hwm = cworkspace;
6026  cd->req_varyopt = 0;  cd->req_varyopt = 0;
6027  cd->nopartial = FALSE;  cd->had_accept = FALSE;
6028    
6029  /* Set up a starting, non-extracting bracket, then compile the expression. On  /* Set up a starting, non-extracting bracket, then compile the expression. On
6030  error, errorcode will be set non-zero, so we don't need to look at the result  error, errorcode will be set non-zero, so we don't need to look at the result
6031  of the function here. */  of the function here. */
6032    
6033  ptr = (const uschar *)pattern;  ptr = (const uschar *)pattern + skipatstart;
6034  code = (uschar *)codestart;  code = (uschar *)codestart;
6035  *code = OP_BRA;  *code = OP_BRA;
6036  (void)compile_regex(re->options, re->options & PCRE_IMS, &code, &ptr,  (void)compile_regex(re->options, re->options & PCRE_IMS, &code, &ptr,
6037    &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd, NULL);    &errorcode, FALSE, FALSE, 0, &firstbyte, &reqbyte, NULL, cd, NULL);
6038  re->top_bracket = cd->bracount;  re->top_bracket = cd->bracount;
6039  re->top_backref = cd->top_backref;  re->top_backref = cd->top_backref;
6040    re->flags = cd->external_flags;
6041    
6042  if (cd->nopartial) re->options |= PCRE_NOPARTIAL;  if (cd->had_accept) reqbyte = -1;   /* Must disable after (*ACCEPT) */
6043    
6044  /* If not reached end of pattern on success, there's an excess bracket. */  /* If not reached end of pattern on success, there's an excess bracket. */
6045    
# Line 5705  if (errorcode != 0) Line 6081  if (errorcode != 0)
6081    PCRE_EARLY_ERROR_RETURN:    PCRE_EARLY_ERROR_RETURN:
6082    *erroroffset = ptr - (const uschar *)pattern;    *erroroffset = ptr - (const uschar *)pattern;
6083    PCRE_EARLY_ERROR_RETURN2:    PCRE_EARLY_ERROR_RETURN2:
6084    *errorptr = error_texts[errorcode];    *errorptr = find_error_text(errorcode);
6085    if (errorcodeptr != NULL) *errorcodeptr = errorcode;    if (errorcodeptr != NULL) *errorcodeptr = errorcode;
6086    return NULL;    return NULL;
6087    }    }
# Line 5734  if ((re->options & PCRE_ANCHORED) == 0) Line 6110  if ((re->options & PCRE_ANCHORED) == 0)
6110        int ch = firstbyte & 255;        int ch = firstbyte & 255;
6111        re->first_byte = ((firstbyte & REQ_CASELESS) != 0 &&        re->first_byte = ((firstbyte & REQ_CASELESS) != 0 &&
6112           cd->fcc[ch] == ch)? ch : firstbyte;           cd->fcc[ch] == ch)? ch : firstbyte;
6113        re->options |= PCRE_FIRSTSET;        re->flags |= PCRE_FIRSTSET;
6114        }        }
6115      else if (is_startline(codestart, 0, cd->backref_map))      else if (is_startline(codestart, 0, cd->backref_map))
6116        re->options |= PCRE_STARTLINE;        re->flags |= PCRE_STARTLINE;
6117      }      }
6118    }    }
6119    
# Line 5751  if (reqbyte >= 0 && Line 6127  if (reqbyte >= 0 &&
6127    int ch = reqbyte & 255;    int ch = reqbyte & 255;
6128    re->req_byte = ((reqbyte & REQ_CASELESS) != 0 &&    re->req_byte = ((reqbyte & REQ_CASELESS) != 0 &&
6129      cd->fcc[ch] == ch)? (reqbyte & ~REQ_CASELESS) : reqbyte;      cd->fcc[ch] == ch)? (reqbyte & ~REQ_CASELESS) : reqbyte;
6130    re->options |= PCRE_REQCHSET;    re->flags |= PCRE_REQCHSET;
6131    }    }
6132    
6133  /* Print out the compiled data if debugging is enabled. This is never the  /* Print out the compiled data if debugging is enabled. This is never the
# Line 5762  case when building a production library. Line 6138  case when building a production library.
6138  printf("Length = %d top_bracket = %d top_backref = %d\n",  printf("Length = %d top_bracket = %d top_backref = %d\n",
6139    length, re->top_bracket, re->top_backref);    length, re->top_bracket, re->top_backref);
6140    
6141  if (re->options != 0)  printf("Options=%08x\n", re->options);
   {  
   printf("%s%s%s%s%s%s%s%s%s\n",  
     ((re->options & PCRE_NOPARTIAL) != 0)? "nopartial " : "",  
     ((re->options & PCRE_ANCHORED) != 0)? "anchored " : "",  
     ((re->options & PCRE_CASELESS) != 0)? "caseless " : "",  
     ((re->options & PCRE_EXTENDED) != 0)? "extended " : "",  
     ((re->options & PCRE_MULTILINE) != 0)? "multiline " : "",  
     ((re->options & PCRE_DOTALL) != 0)? "dotall " : "",  
     ((re->options & PCRE_DOLLAR_ENDONLY) != 0)? "endonly " : "",  
     ((re->options & PCRE_EXTRA) != 0)? "extra " : "",  
     ((re->options & PCRE_UNGREEDY) != 0)? "ungreedy " : "");  
   }  
6142    
6143  if ((re->options & PCRE_FIRSTSET) != 0)  if ((re->flags & PCRE_FIRSTSET) != 0)
6144    {    {
6145    int ch = re->first_byte & 255;    int ch = re->first_byte & 255;
6146    const char *caseless = ((re->first_byte & REQ_CASELESS) == 0)?    const char *caseless = ((re->first_byte & REQ_CASELESS) == 0)?
# Line 5785  if ((re->options & PCRE_FIRSTSET) != 0) Line 6149  if ((re->options & PCRE_FIRSTSET) != 0)
6149      else printf("First char = \\x%02x%s\n", ch, caseless);      else printf("First char = \\x%02x%s\n", ch, caseless);
6150    }    }
6151    
6152  if ((re->options & PCRE_REQCHSET) != 0)  if ((re->flags & PCRE_REQCHSET) != 0)
6153    {    {
6154    int ch = re->req_byte & 255;    int ch = re->req_byte & 255;
6155    const char *caseless = ((re->req_byte & REQ_CASELESS) == 0)?    const char *caseless = ((re->req_byte & REQ_CASELESS) == 0)?
# Line 5802  was compiled can be seen. */ Line 6166  was compiled can be seen. */
6166  if (code - codestart > length)  if (code - codestart > length)
6167    {    {
6168    (pcre_free)(re);    (pcre_free)(re);
6169    *errorptr = error_texts[ERR23];    *errorptr = find_error_text(ERR23);
6170    *erroroffset = ptr - (uschar *)pattern;    *erroroffset = ptr - (uschar *)pattern;
6171    if (errorcodeptr != NULL) *errorcodeptr = ERR23;    if (errorcodeptr != NULL) *errorcodeptr = ERR23;
6172    return NULL;    return NULL;

Legend:
Removed from v.199  
changed lines
  Added in v.275

  ViewVC Help
Powered by ViewVC 1.1.5