/[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 240 by ph10, Tue Sep 11 15:47:20 2007 UTC revision 335 by ph10, Sat Apr 12 14:36:14 2008 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) 1997-2007 University of Cambridge             Copyright (c) 1997-2008 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 140  static const short int escapes[] = { Line 140  static const short int escapes[] = {
140  #endif  #endif
141    
142    
143  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
144  searched linearly. Put all the names into a single string, in order to reduce  searched linearly. Put all the names into a single string, in order to reduce
145  the number of relocations when a shared library is dynamically linked. */  the number of relocations when a shared library is dynamically linked. */
146    
147  typedef struct verbitem {  typedef struct verbitem {
# Line 150  typedef struct verbitem { Line 150  typedef struct verbitem {
150  } verbitem;  } verbitem;
151    
152  static const char verbnames[] =  static const char verbnames[] =
153    "ACCEPT\0"    "ACCEPT\0"
154    "COMMIT\0"    "COMMIT\0"
155    "F\0"    "F\0"
156    "FAIL\0"    "FAIL\0"
157    "PRUNE\0"    "PRUNE\0"
158    "SKIP\0"    "SKIP\0"
159    "THEN";    "THEN";
160    
161  static verbitem verbs[] = {  static const verbitem verbs[] = {
162    { 6, OP_ACCEPT },    { 6, OP_ACCEPT },
163    { 6, OP_COMMIT },    { 6, OP_COMMIT },
164    { 1, OP_FAIL },    { 1, OP_FAIL },
# Line 168  static verbitem verbs[] = { Line 168  static verbitem verbs[] = {
168    { 4, OP_THEN  }    { 4, OP_THEN  }
169  };  };
170    
171  static int verbcount = sizeof(verbs)/sizeof(verbitem);  static const int verbcount = sizeof(verbs)/sizeof(verbitem);
172    
173    
174  /* Tables of names of POSIX character classes and their lengths. The names are  /* Tables of names of POSIX character classes and their lengths. The names are
175  now all in a single string, to reduce the number of relocations when a shared  now all in a single string, to reduce the number of relocations when a shared
176  library is dynamically loaded. The list of lengths is terminated by a zero  library is dynamically loaded. The list of lengths is terminated by a zero
177  length entry. The first three must be alpha, lower, upper, as this is assumed  length entry. The first three must be alpha, lower, upper, as this is assumed
178  for handling case independence. */  for handling case independence. */
179    
180  static const char posix_names[] =  static const char posix_names[] =
181    "alpha\0"  "lower\0"  "upper\0"  "alnum\0"  "ascii\0"  "blank\0"    "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"    "cntrl\0"  "digit\0"  "graph\0"  "print\0"  "punct\0"  "space\0"
183    "word\0"   "xdigit";    "word\0"   "xdigit";
184    
185  static const uschar posix_name_lengths[] = {  static const uschar posix_name_lengths[] = {
# Line 219  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. This used to be a table of strings, but in order to reduce  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,  the number of relocations needed when a shared library is loaded dynamically,
224  it is now one long string. We cannot use a table of offsets, because the  it is now one long string. We cannot use a table of offsets, because the
225  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
226  simply count through to the one we want - this isn't a performance issue  simply count through to the one we want - this isn't a performance issue
227  because these strings are used only when there is a compilation error. */  because these strings are used only when there is a compilation error. */
228    
229  static const char error_texts[] =  static const char error_texts[] =
# Line 241  static const char error_texts[] = Line 241  static const char error_texts[] =
241    /* 10 */    /* 10 */
242    "operand of unlimited repeat could match the empty string\0"  /** DEAD **/    "operand of unlimited repeat could match the empty string\0"  /** DEAD **/
243    "internal error: unexpected repeat\0"    "internal error: unexpected repeat\0"
244    "unrecognized character after (?\0"    "unrecognized character after (? or (?-\0"
245    "POSIX named classes are supported only within a class\0"    "POSIX named classes are supported only within a class\0"
246    "missing )\0"    "missing )\0"
247    /* 15 */    /* 15 */
# Line 295  static const char error_texts[] = Line 295  static const char error_texts[] =
295    /* 55 */    /* 55 */
296    "repeating a DEFINE group is not allowed\0"    "repeating a DEFINE group is not allowed\0"
297    "inconsistent NEWLINE options\0"    "inconsistent NEWLINE options\0"
298    "\\g is not followed by a braced name or an optionally braced non-zero number\0"    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
299    "(?+ or (?- or (?(+ or (?(- must be followed by a non-zero number\0"    "a numbered reference must not be zero\0"
300    "(*VERB) with an argument is not supported\0"    "(*VERB) with an argument is not supported\0"
301    /* 60 */    /* 60 */
302    "(*VERB) not recognized\0"    "(*VERB) not recognized\0"
303    "number is too big";    "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 439  static BOOL Line 441  static BOOL
441  *            Find an error text                  *  *            Find an error text                  *
442  *************************************************/  *************************************************/
443    
444  /* The error texts are now all in one long string, to save on relocations. As  /* 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.  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  Instead, just count through the strings. This is not a performance issue
447  because it happens only when there has been a compilation error.  because it happens only when there has been a compilation error.
448    
449  Argument:   the error number  Argument:   the error number
# Line 452  static const char * Line 454  static const char *
454  find_error_text(int n)  find_error_text(int n)
455  {  {
456  const char *s = error_texts;  const char *s = error_texts;
457  for (; n > 0; n--) while (*s++ != 0);  for (; n > 0; n--) while (*s++ != 0);
458  return s;  return s;
459  }  }
460    
# Line 496  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 529  else Line 531  else
531      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
532      break;      break;
533    
534      /* \g must be followed by a number, either plain or braced. If positive, it      /* \g must be followed by one of a number of specific things:
535      is an absolute backreference. If negative, it is a relative backreference.  
536      This is a Perl 5.10 feature. Perl 5.10 also supports \g{name} as a      (1) A number, either plain or braced. If positive, it is an absolute
537      reference to a named group. This is part of Perl's movement towards a      backreference. If negative, it is a relative backreference. This is a Perl
538      unified syntax for back references. As this is synonymous with \k{name}, we      5.10 feature.
539      fudge it up by pretending it really was \k. */  
540        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
541        is part of Perl's movement towards a unified syntax for back references. As
542        this is synonymous with \k{name}, we fudge it up by pretending it really
543        was \k.
544    
545        (3) For Oniguruma compatibility we also support \g followed by a name or a
546        number either in angle brackets or in single quotes. However, these are
547        (possibly recursive) subroutine calls, _not_ backreferences. Just return
548        the -ESC_g code (cf \k). */
549    
550      case 'g':      case 'g':
551        if (ptr[1] == '<' || ptr[1] == '\'')
552          {
553          c = -ESC_g;
554          break;
555          }
556    
557        /* Handle the Perl-compatible cases */
558    
559      if (ptr[1] == '{')      if (ptr[1] == '{')
560        {        {
561        const uschar *p;        const uschar *p;
# Line 563  else Line 582  else
582      while ((digitab[ptr[1]] & ctype_digit) != 0)      while ((digitab[ptr[1]] & ctype_digit) != 0)
583        c = c * 10 + *(++ptr) - '0';        c = c * 10 + *(++ptr) - '0';
584    
585      if (c < 0)      if (c < 0)   /* Integer overflow */
586        {        {
587        *errorcodeptr = ERR61;        *errorcodeptr = ERR61;
588        break;        break;
589        }        }
590    
591      if (c == 0 || (braced && *(++ptr) != '}'))      if (braced && *(++ptr) != '}')
592        {        {
593        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
594        break;        break;
595        }        }
596    
597        if (c == 0)
598          {
599          *errorcodeptr = ERR58;
600          break;
601          }
602    
603      if (negated)      if (negated)
604        {        {
# Line 609  else Line 634  else
634        c -= '0';        c -= '0';
635        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
636          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - '0';
637        if (c < 0)        if (c < 0)    /* Integer overflow */
638          {          {
639          *errorcodeptr = ERR61;          *errorcodeptr = ERR61;
640          break;          break;
# Line 722  else Line 747  else
747      break;      break;
748    
749      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
750      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
751      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
752      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
753      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
754    
755      default:      default:
756      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 1506  for (;;) Line 1531  for (;;)
1531  can match the empty string or not. It is called from could_be_empty()  can match the empty string or not. It is called from could_be_empty()
1532  below and from compile_branch() when checking for an unlimited repeat of a  below and from compile_branch() when checking for an unlimited repeat of a
1533  group that can match nothing. Note that first_significant_code() skips over  group that can match nothing. Note that first_significant_code() skips over
1534  assertions. If we hit an unclosed bracket, we return "empty" - this means we've  backward and negative forward assertions when its final argument is TRUE. If we
1535  struck an inner bracket whose current branch will already have been scanned.  hit an unclosed bracket, we return "empty" - this means we've struck an inner
1536    bracket whose current branch will already have been scanned.
1537    
1538  Arguments:  Arguments:
1539    code        points to start of search    code        points to start of search
# Line 1529  for (code = first_significant_code(code Line 1555  for (code = first_significant_code(code
1555    
1556    c = *code;    c = *code;
1557    
1558      /* Skip over forward assertions; the other assertions are skipped by
1559      first_significant_code() with a TRUE final argument. */
1560    
1561      if (c == OP_ASSERT)
1562        {
1563        do code += GET(code, 1); while (*code == OP_ALT);
1564        c = *code;
1565        continue;
1566        }
1567    
1568    /* Groups with zero repeats can of course be empty; skip them. */    /* Groups with zero repeats can of course be empty; skip them. */
1569    
1570    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)
1571      {      {
1572      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1573      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
# Line 1724  return TRUE; Line 1760  return TRUE;
1760  *************************************************/  *************************************************/
1761    
1762  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
1763  encountered in a character class. It checks whether this is followed by an  encountered in a character class. It checks whether this is followed by a
1764  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
1765  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
1766    
1767    Originally, this function only recognized a sequence of letters between the
1768    terminators, but it seems that Perl recognizes any sequence of characters,
1769    though of course unknown POSIX names are subsequently rejected. Perl gives an
1770    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
1771    didn't consider this to be a POSIX class. Likewise for [:1234:].
1772    
1773    The problem in trying to be exactly like Perl is in the handling of escapes. We
1774    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
1775    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
1776    below handles the special case of \], but does not try to do any other escape
1777    processing. This makes it different from Perl for cases such as [:l\ower:]
1778    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
1779    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
1780    I think.
1781    
1782  Argument:  Arguments:
1783    ptr      pointer to the initial [    ptr      pointer to the initial [
1784    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
1785    
1786  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
1787  */  */
1788    
1789  static BOOL  static BOOL
1790  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
1791  {  {
1792  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
1793  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
1794  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
1795    {    {
1796    *endptr = ptr;    if (*ptr == '\\' && ptr[1] == ']') ptr++; else
1797    return TRUE;      {
1798        if (*ptr == ']') return FALSE;
1799        if (*ptr == terminator && ptr[1] == ']')
1800          {
1801          *endptr = ptr;
1802          return TRUE;
1803          }
1804        }
1805    }    }
1806  return FALSE;  return FALSE;
1807  }  }
# Line 1777  while (posix_name_lengths[yield] != 0) Line 1832  while (posix_name_lengths[yield] != 0)
1832    {    {
1833    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
1834      strncmp((const char *)ptr, pn, len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
1835    pn += posix_name_lengths[yield] + 1;    pn += posix_name_lengths[yield] + 1;
1836    yield++;    yield++;
1837    }    }
1838  return -1;  return -1;
# Line 1792  return -1; Line 1847  return -1;
1847  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
1848  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
1849  earlier groups that are outside the current group). However, when a group is  earlier groups that are outside the current group). However, when a group is
1850  optional (i.e. the minimum quantifier is zero), OP_BRAZERO is inserted before  optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is
1851  it, after it has been compiled. This means that any OP_RECURSE items within it  inserted before it, after it has been compiled. This means that any OP_RECURSE
1852  that refer to the group itself or any contained groups have to have their  items within it that refer to the group itself or any contained groups have to
1853  offsets adjusted. That one of the jobs of this function. Before it is called,  have their offsets adjusted. That one of the jobs of this function. Before it
1854  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
1855    OP_END.
1856    
1857  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
1858  recursions and subroutine calls. It must also check the list of such references  recursions and subroutine calls. It must also check the list of such references
# Line 1827  while ((ptr = (uschar *)find_recurse(ptr Line 1883  while ((ptr = (uschar *)find_recurse(ptr
1883    
1884    /* See if this recursion is on the forward reference list. If so, adjust the    /* See if this recursion is on the forward reference list. If so, adjust the
1885    reference. */    reference. */
1886    
1887    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
1888      {      {
1889      offset = GET(hc, 0);      offset = GET(hc, 0);
# Line 2081  if (next >= 0) switch(op_code) Line 2137  if (next >= 0) switch(op_code)
2137    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT, "item" must be a single-byte character. */
2138    
2139    case OP_NOT:    case OP_NOT:
   if (next < 0) return FALSE;  /* Not a character */  
2140    if (item == next) return TRUE;    if (item == next) return TRUE;
2141    if ((options & PCRE_CASELESS) == 0) return FALSE;    if ((options & PCRE_CASELESS) == 0) return FALSE;
2142  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 2344  uschar classbits[32]; Line 2399  uschar classbits[32];
2399  BOOL class_utf8;  BOOL class_utf8;
2400  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
2401  uschar *class_utf8data;  uschar *class_utf8data;
2402    uschar *class_utf8data_base;
2403  uschar utf8_char[6];  uschar utf8_char[6];
2404  #else  #else
2405  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
# Line 2383  req_caseopt = ((options & PCRE_CASELESS) Line 2439  req_caseopt = ((options & PCRE_CASELESS)
2439  for (;; ptr++)  for (;; ptr++)
2440    {    {
2441    BOOL negate_class;    BOOL negate_class;
2442      BOOL should_flip_negation;
2443    BOOL possessive_quantifier;    BOOL possessive_quantifier;
2444    BOOL is_quantifier;    BOOL is_quantifier;
2445    BOOL is_recurse;    BOOL is_recurse;
# Line 2402  for (;; ptr++) Line 2459  for (;; ptr++)
2459    /* Get next byte in the pattern */    /* Get next byte in the pattern */
2460    
2461    c = *ptr;    c = *ptr;
2462    
2463    /* If we are in the pre-compile phase, accumulate the length used for the    /* If we are in the pre-compile phase, accumulate the length used for the
2464    previous cycle of this loop. */    previous cycle of this loop. */
2465    
# Line 2606  for (;; ptr++) Line 2663  for (;; ptr++)
2663      they are encountered at the top level, so we'll do that too. */      they are encountered at the top level, so we'll do that too. */
2664    
2665      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&
2666          check_posix_syntax(ptr, &tempptr, cd))          check_posix_syntax(ptr, &tempptr))
2667        {        {
2668        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;
2669        goto FAILED;        goto FAILED;
# Line 2631  for (;; ptr++) Line 2688  for (;; ptr++)
2688        else break;        else break;
2689        }        }
2690    
2691        /* If a class contains a negative special such as \S, we need to flip the
2692        negation flag at the end, so that support for characters > 255 works
2693        correctly (they are all included in the class). */
2694    
2695        should_flip_negation = FALSE;
2696    
2697      /* 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
2698      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
2699      valued UTF-8 characters, we don't yet do any optimization. */      valued UTF-8 characters, we don't yet do any optimization. */
# Line 2648  for (;; ptr++) Line 2711  for (;; ptr++)
2711  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2712      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
2713      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
2714        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
2715  #endif  #endif
2716    
2717      /* Process characters until ] is reached. By writing this as a "do" it      /* Process characters until ] is reached. By writing this as a "do" it
# Line 2663  for (;; ptr++) Line 2727  for (;; ptr++)
2727          {                           /* Braces are required because the */          {                           /* Braces are required because the */
2728          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
2729          }          }
2730    
2731          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
2732          data and reset the pointer. This is so that very large classes that
2733          contain a zillion UTF-8 characters no longer overwrite the work space
2734          (which is on the stack). */
2735    
2736          if (lengthptr != NULL)
2737            {
2738            *lengthptr += class_utf8data - class_utf8data_base;
2739            class_utf8data = class_utf8data_base;
2740            }
2741    
2742  #endif  #endif
2743    
2744        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
# Line 2686  for (;; ptr++) Line 2762  for (;; ptr++)
2762    
2763        if (c == '[' &&        if (c == '[' &&
2764            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&
2765            check_posix_syntax(ptr, &tempptr, cd))            check_posix_syntax(ptr, &tempptr))
2766          {          {
2767          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
2768          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
# Line 2703  for (;; ptr++) Line 2779  for (;; ptr++)
2779          if (*ptr == '^')          if (*ptr == '^')
2780            {            {
2781            local_negate = TRUE;            local_negate = TRUE;
2782              should_flip_negation = TRUE;  /* Note negative special */
2783            ptr++;            ptr++;
2784            }            }
2785    
# Line 2777  for (;; ptr++) Line 2854  for (;; ptr++)
2854          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
2855          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
2856    
2857          if (-c == ESC_b) c = '\b';       /* \b is backslash in a class */          if (-c == ESC_b) c = '\b';       /* \b is backspace in a class */
2858          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 */
2859          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 */
2860          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
# Line 2805  for (;; ptr++) Line 2882  for (;; ptr++)
2882              continue;              continue;
2883    
2884              case ESC_D:              case ESC_D:
2885                should_flip_negation = TRUE;
2886              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
2887              continue;              continue;
2888    
# Line 2813  for (;; ptr++) Line 2891  for (;; ptr++)
2891              continue;              continue;
2892    
2893              case ESC_W:              case ESC_W:
2894                should_flip_negation = TRUE;
2895              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
2896              continue;              continue;
2897    
# Line 2822  for (;; ptr++) Line 2901  for (;; ptr++)
2901              continue;              continue;
2902    
2903              case ESC_S:              case ESC_S:
2904                should_flip_negation = TRUE;
2905              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
2906              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
2907              continue;              continue;
2908    
             case ESC_E: /* Perl ignores an orphan \E */  
             continue;  
   
2909              default:    /* Not recognized; fall through */              default:    /* Not recognized; fall through */
2910              break;      /* Need "default" setting to stop compiler warning. */              break;      /* Need "default" setting to stop compiler warning. */
2911              }              }
# Line 3063  for (;; ptr++) Line 3140  for (;; ptr++)
3140            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3141            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3142    
3143            /* \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
3144            special means the '-' was literal */            special means the '-' was literal */
3145    
3146            if (d < 0)            if (d < 0)
# Line 3327  we set the flag only if there is a liter Line 3404  we set the flag only if there is a liter
3404      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3405    
3406      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
3407      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode, unless there was a negated special
3408      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
3409        the class, so any that were explicitly given as well can be ignored. If
3410        (when there are explicit characters > 255 that must be listed) there are no
3411        characters < 256, we can omit the bitmap in the actual compiled code. */
3412    
3413  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3414      if (class_utf8)      if (class_utf8 && !should_flip_negation)
3415        {        {
3416        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
3417        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 3357  we set the flag only if there is a liter Line 3437  we set the flag only if there is a liter
3437        }        }
3438  #endif  #endif
3439    
3440      /* 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
3441      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
3442      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
3443      setting must remain unchanged after any kind of repeat. */      map into the code vector, negating it if necessary. */
3444    
3445        *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
3446      if (negate_class)      if (negate_class)
3447        {        {
       *code++ = OP_NCLASS;  
3448        if (lengthptr == NULL)    /* Save time in the pre-compile phase */        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
3449          for (c = 0; c < 32; c++) code[c] = ~classbits[c];          for (c = 0; c < 32; c++) code[c] = ~classbits[c];
3450        }        }
3451      else      else
3452        {        {
       *code++ = OP_CLASS;  
3453        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
3454        }        }
3455      code += 32;      code += 32;
# Line 3764  we set the flag only if there is a liter Line 3843  we set the flag only if there is a liter
3843    
3844        if (repeat_min == 0)        if (repeat_min == 0)
3845          {          {
3846          /* If the maximum is also zero, we just omit the group from the output          /* If the maximum is also zero, we used to just omit the group from the
3847          altogether. */          output altogether, like this:
   
         if (repeat_max == 0)  
           {  
           code = previous;  
           goto END_REPEAT;  
           }  
3848    
3849          /* If the maximum is 1 or unlimited, we just have to stick in the          ** if (repeat_max == 0)
3850          BRAZERO and do no more at this point. However, we do need to adjust          **   {
3851          any OP_RECURSE calls inside the group that refer to the group itself or          **   code = previous;
3852          any internal or forward referenced group, because the offset is from          **   goto END_REPEAT;
3853          the start of the whole regex. Temporarily terminate the pattern while          **   }
3854          doing this. */  
3855            However, that fails when a group is referenced as a subroutine from
3856            elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it
3857            so that it is skipped on execution. As we don't have a list of which
3858            groups are referenced, we cannot do this selectively.
3859    
3860            If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
3861            and do no more at this point. However, we do need to adjust any
3862            OP_RECURSE calls inside the group that refer to the group itself or any
3863            internal or forward referenced group, because the offset is from the
3864            start of the whole regex. Temporarily terminate the pattern while doing
3865            this. */
3866    
3867          if (repeat_max <= 1)          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
3868            {            {
3869            *code = OP_END;            *code = OP_END;
3870            adjust_recurse(previous, 1, utf8, cd, save_hwm);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
3871            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
3872            code++;            code++;
3873              if (repeat_max == 0)
3874                {
3875                *previous++ = OP_SKIPZERO;
3876                goto END_REPEAT;
3877                }
3878            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
3879            }            }
3880    
# Line 4006  we set the flag only if there is a liter Line 4095  we set the flag only if there is a liter
4095        int len;        int len;
4096        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||
4097            *tempcode == OP_NOTEXACT)            *tempcode == OP_NOTEXACT)
4098          tempcode += _pcre_OP_lengths[*tempcode];          tempcode += _pcre_OP_lengths[*tempcode] +
4099              ((*tempcode == OP_TYPEEXACT &&
4100                 (tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP))? 2:0);
4101        len = code - tempcode;        len = code - tempcode;
4102        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4103          {          {
# Line 4058  we set the flag only if there is a liter Line 4149  we set the flag only if there is a liter
4149      bravalue = OP_CBRA;      bravalue = OP_CBRA;
4150      save_hwm = cd->hwm;      save_hwm = cd->hwm;
4151      reset_bracount = FALSE;      reset_bracount = FALSE;
4152    
4153      /* First deal with various "verbs" that can be introduced by '*'. */      /* First deal with various "verbs" that can be introduced by '*'. */
4154    
4155      if (*(++ptr) == '*' && (cd->ctypes[ptr[1]] & ctype_letter) != 0)      if (*(++ptr) == '*' && (cd->ctypes[ptr[1]] & ctype_letter) != 0)
4156        {        {
4157        int i, namelen;        int i, namelen;
4158        const char *vn = verbnames;        const char *vn = verbnames;
4159        const uschar *name = ++ptr;        const uschar *name = ++ptr;
4160        previous = NULL;        previous = NULL;
4161        while ((cd->ctypes[*++ptr] & ctype_letter) != 0);        while ((cd->ctypes[*++ptr] & ctype_letter) != 0);
# Line 4088  we set the flag only if there is a liter Line 4179  we set the flag only if there is a liter
4179            if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;            if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;
4180            break;            break;
4181            }            }
4182          vn += verbs[i].len + 1;          vn += verbs[i].len + 1;
4183          }          }
4184        if (i < verbcount) continue;        if (i < verbcount) continue;
4185        *errorcodeptr = ERR60;        *errorcodeptr = ERR60;
# Line 4233  we set the flag only if there is a liter Line 4324  we set the flag only if there is a liter
4324              *errorcodeptr = ERR58;              *errorcodeptr = ERR58;
4325              goto FAILED;              goto FAILED;
4326              }              }
4327            if (refsign == '-')            recno = (refsign == '-')?
4328                cd->bracount - recno + 1 : recno +cd->bracount;
4329              if (recno <= 0 || recno > cd->final_bracount)
4330              {              {
4331              recno = cd->bracount - recno + 1;              *errorcodeptr = ERR15;
4332              if (recno <= 0)              goto FAILED;
               {  
               *errorcodeptr = ERR15;  
               goto FAILED;  
               }  
4333              }              }
           else recno += cd->bracount;  
4334            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4335            break;            break;
4336            }            }
# Line 4314  we set the flag only if there is a liter Line 4402  we set the flag only if there is a liter
4402            skipbytes = 1;            skipbytes = 1;
4403            }            }
4404    
4405          /* Check for the "name" actually being a subpattern number. */          /* Check for the "name" actually being a subpattern number. We are
4406            in the second pass here, so final_bracount is set. */
4407    
4408          else if (recno > 0)          else if (recno > 0 && recno <= cd->final_bracount)
4409            {            {
4410            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4411            }            }
# Line 4510  we set the flag only if there is a liter Line 4599  we set the flag only if there is a liter
4599    
4600          /* We come here from the Python syntax above that handles both          /* We come here from the Python syntax above that handles both
4601          references (?P=name) and recursion (?P>name), as well as falling          references (?P=name) and recursion (?P>name), as well as falling
4602          through from the Perl recursion syntax (?&name). */          through from the Perl recursion syntax (?&name). We also come here from
4603            the Perl \k<name> or \k'name' back reference syntax and the \k{name}
4604            .NET syntax, and the Oniguruma \g<...> and \g'...' subroutine syntax. */
4605    
4606          NAMED_REF_OR_RECURSE:          NAMED_REF_OR_RECURSE:
4607          name = ++ptr;          name = ++ptr;
# Line 4522  we set the flag only if there is a liter Line 4613  we set the flag only if there is a liter
4613    
4614          if (lengthptr != NULL)          if (lengthptr != NULL)
4615            {            {
4616              if (namelen == 0)
4617                {
4618                *errorcodeptr = ERR62;
4619                goto FAILED;
4620                }
4621            if (*ptr != terminator)            if (*ptr != terminator)
4622              {              {
4623              *errorcodeptr = ERR42;              *errorcodeptr = ERR42;
# Line 4535  we set the flag only if there is a liter Line 4631  we set the flag only if there is a liter
4631            recno = 0;            recno = 0;
4632            }            }
4633    
4634          /* In the real compile, seek the name in the table */          /* In the real compile, seek the name in the table. We check the name
4635            first, and then check that we have reached the end of the name in the
4636            table. That way, if the name that is longer than any in the table,
4637            the comparison will fail without reading beyond the table entry. */
4638    
4639          else          else
4640            {            {
4641            slot = cd->name_table;            slot = cd->name_table;
4642            for (i = 0; i < cd->names_found; i++)            for (i = 0; i < cd->names_found; i++)
4643              {              {
4644              if (strncmp((char *)name, (char *)slot+2, namelen) == 0) break;              if (strncmp((char *)name, (char *)slot+2, namelen) == 0 &&
4645                    slot[2+namelen] == 0)
4646                  break;
4647              slot += cd->name_entry_size;              slot += cd->name_entry_size;
4648              }              }
4649    
# Line 4578  we set the flag only if there is a liter Line 4679  we set the flag only if there is a liter
4679          case '5': case '6': case '7': case '8': case '9':   /* subroutine */          case '5': case '6': case '7': case '8': case '9':   /* subroutine */
4680            {            {
4681            const uschar *called;            const uschar *called;
4682              terminator = ')';
4683    
4684              /* Come here from the \g<...> and \g'...' code (Oniguruma
4685              compatibility). However, the syntax has been checked to ensure that
4686              the ... are a (signed) number, so that neither ERR63 nor ERR29 will
4687              be called on this path, nor with the jump to OTHER_CHAR_AFTER_QUERY
4688              ever be taken. */
4689    
4690              HANDLE_NUMERICAL_RECURSION:
4691    
4692            if ((refsign = *ptr) == '+') ptr++;            if ((refsign = *ptr) == '+')
4693                {
4694                ptr++;
4695                if ((digitab[*ptr] & ctype_digit) == 0)
4696                  {
4697                  *errorcodeptr = ERR63;
4698                  goto FAILED;
4699                  }
4700                }
4701            else if (refsign == '-')            else if (refsign == '-')
4702              {              {
4703              if ((digitab[ptr[1]] & ctype_digit) == 0)              if ((digitab[ptr[1]] & ctype_digit) == 0)
# Line 4591  we set the flag only if there is a liter Line 4709  we set the flag only if there is a liter
4709            while((digitab[*ptr] & ctype_digit) != 0)            while((digitab[*ptr] & ctype_digit) != 0)
4710              recno = recno * 10 + *ptr++ - '0';              recno = recno * 10 + *ptr++ - '0';
4711    
4712            if (*ptr != ')')            if (*ptr != terminator)
4713              {              {
4714              *errorcodeptr = ERR29;              *errorcodeptr = ERR29;
4715              goto FAILED;              goto FAILED;
# Line 4987  we set the flag only if there is a liter Line 5105  we set the flag only if there is a liter
5105      back references and those types that consume a character may be repeated.      back references and those types that consume a character may be repeated.
5106      We can test for values between ESC_b and ESC_Z for the latter; this may      We can test for values between ESC_b and ESC_Z for the latter; this may
5107      have to change if any new ones are ever created. */      have to change if any new ones are ever created. */
5108    
5109      case '\\':      case '\\':
5110      tempptr = ptr;      tempptr = ptr;
5111      c = check_escape(&ptr, errorcodeptr, cd->bracount, options, FALSE);      c = check_escape(&ptr, errorcodeptr, cd->bracount, options, FALSE);
# Line 5014  we set the flag only if there is a liter Line 5132  we set the flag only if there is a liter
5132    
5133        zerofirstbyte = firstbyte;        zerofirstbyte = firstbyte;
5134        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
5135    
5136          /* \g<name> or \g'name' is a subroutine call by name and \g<n> or \g'n'
5137          is a subroutine call by number (Oniguruma syntax). In fact, the value
5138          -ESC_g is returned only for these cases. So we don't need to check for <
5139          or ' if the value is -ESC_g. For the Perl syntax \g{n} the value is
5140          -ESC_REF+n, and for the Perl syntax \g{name} the result is -ESC_k (as
5141          that is a synonym for a named back reference). */
5142    
5143          if (-c == ESC_g)
5144            {
5145            const uschar *p;
5146            save_hwm = cd->hwm;   /* Normally this is set when '(' is read */
5147            terminator = (*(++ptr) == '<')? '>' : '\'';
5148    
5149            /* These two statements stop the compiler for warning about possibly
5150            unset variables caused by the jump to HANDLE_NUMERICAL_RECURSION. In
5151            fact, because we actually check for a number below, the paths that
5152            would actually be in error are never taken. */
5153    
5154            skipbytes = 0;
5155            reset_bracount = FALSE;
5156    
5157            /* Test for a name */
5158    
5159            if (ptr[1] != '+' && ptr[1] != '-')
5160              {
5161              BOOL isnumber = TRUE;
5162              for (p = ptr + 1; *p != 0 && *p != terminator; p++)
5163                {
5164                if ((cd->ctypes[*p] & ctype_digit) == 0) isnumber = FALSE;
5165                if ((cd->ctypes[*p] & ctype_word) == 0) break;
5166                }
5167              if (*p != terminator)
5168                {
5169                *errorcodeptr = ERR57;
5170                break;
5171                }
5172              if (isnumber)
5173                {
5174                ptr++;
5175                goto HANDLE_NUMERICAL_RECURSION;
5176                }
5177              is_recurse = TRUE;
5178              goto NAMED_REF_OR_RECURSE;
5179              }
5180    
5181            /* Test a signed number in angle brackets or quotes. */
5182    
5183            p = ptr + 2;
5184            while ((digitab[*p] & ctype_digit) != 0) p++;
5185            if (*p != terminator)
5186              {
5187              *errorcodeptr = ERR57;
5188              break;
5189              }
5190            ptr++;
5191            goto HANDLE_NUMERICAL_RECURSION;
5192            }
5193    
5194        /* \k<name> or \k'name' is a back reference by name (Perl syntax).        /* \k<name> or \k'name' is a back reference by name (Perl syntax).
5195        We also support \k{name} (.NET syntax) */        We also support \k{name} (.NET syntax) */
# Line 5745  to fill in forward references to subpatt Line 5921  to fill in forward references to subpatt
5921    
5922  uschar cworkspace[COMPILE_WORK_SIZE];  uschar cworkspace[COMPILE_WORK_SIZE];
5923    
   
5924  /* Set this early so that early errors get offset 0. */  /* Set this early so that early errors get offset 0. */
5925    
5926  ptr = (const uschar *)pattern;  ptr = (const uschar *)pattern;
# Line 5906  to compile parts of the pattern into; th Line 6081  to compile parts of the pattern into; th
6081  no longer needed, so hopefully this workspace will never overflow, though there  no longer needed, so hopefully this workspace will never overflow, though there
6082  is a test for its doing so. */  is a test for its doing so. */
6083    
6084  cd->bracount = 0;  cd->bracount = cd->final_bracount = 0;
6085  cd->names_found = 0;  cd->names_found = 0;
6086  cd->name_entry_size = 0;  cd->name_entry_size = 0;
6087  cd->name_table = NULL;  cd->name_table = NULL;
# Line 5983  field. Reset the bracket count and the n Line 6158  field. Reset the bracket count and the n
6158  field; this time it's used for remembering forward references to subpatterns.  field; this time it's used for remembering forward references to subpatterns.
6159  */  */
6160    
6161    cd->final_bracount = cd->bracount;  /* Save for checking forward references */
6162  cd->bracount = 0;  cd->bracount = 0;
6163  cd->names_found = 0;  cd->names_found = 0;
6164  cd->name_table = (uschar *)re + re->name_table_offset;  cd->name_table = (uschar *)re + re->name_table_offset;
# Line 6033  while (errorcode == 0 && cd->hwm > cwork Line 6209  while (errorcode == 0 && cd->hwm > cwork
6209    if (groupptr == NULL) errorcode = ERR53;    if (groupptr == NULL) errorcode = ERR53;
6210      else PUT(((uschar *)codestart), offset, groupptr - codestart);      else PUT(((uschar *)codestart), offset, groupptr - codestart);
6211    }    }
6212    
6213  /* Give an error if there's back reference to a non-existent capturing  /* Give an error if there's back reference to a non-existent capturing
6214  subpattern. */  subpattern. */
6215    

Legend:
Removed from v.240  
changed lines
  Added in v.335

  ViewVC Help
Powered by ViewVC 1.1.5