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

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

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

revision 613 by ph10, Sat Jul 2 16:59:52 2011 UTC revision 691 by ph10, Sun Sep 11 14:31:21 2011 UTC
# Line 393  static const char error_texts[] = Line 393  static const char error_texts[] =
393    "internal error: previously-checked referenced subpattern not found\0"    "internal error: previously-checked referenced subpattern not found\0"
394    "DEFINE group contains more than one branch\0"    "DEFINE group contains more than one branch\0"
395    /* 55 */    /* 55 */
396    "repeating a DEFINE group is not allowed\0"    "repeating a DEFINE group is not allowed\0"  /** DEAD **/
397    "inconsistent NEWLINE options\0"    "inconsistent NEWLINE options\0"
398    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
399    "a numbered reference must not be zero\0"    "a numbered reference must not be zero\0"
# Line 409  static const char error_texts[] = Line 409  static const char error_texts[] =
409    "(*MARK) must have an argument\0"    "(*MARK) must have an argument\0"
410    "this version of PCRE is not compiled with PCRE_UCP support\0"    "this version of PCRE is not compiled with PCRE_UCP support\0"
411    "\\c must be followed by an ASCII character\0"    "\\c must be followed by an ASCII character\0"
412      "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
413    ;    ;
414    
415  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
# Line 545  static const unsigned char ebcdic_charta Line 546  static const unsigned char ebcdic_charta
546  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
547    
548  static BOOL  static BOOL
549    compile_regex(int, uschar **, const uschar **, int *, BOOL, BOOL, int, int *,    compile_regex(int, uschar **, const uschar **, int *, BOOL, BOOL, int, int,
550      int *, branch_chain *, compile_data *, int *);      int *, int *, branch_chain *, compile_data *, int *);
551    
552    
553    
# Line 577  return s; Line 578  return s;
578    
579    
580  /*************************************************  /*************************************************
581    *            Check for counted repeat            *
582    *************************************************/
583    
584    /* This function is called when a '{' is encountered in a place where it might
585    start a quantifier. It looks ahead to see if it really is a quantifier or not.
586    It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
587    where the ddds are digits.
588    
589    Arguments:
590      p         pointer to the first char after '{'
591    
592    Returns:    TRUE or FALSE
593    */
594    
595    static BOOL
596    is_counted_repeat(const uschar *p)
597    {
598    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
599    while ((digitab[*p] & ctype_digit) != 0) p++;
600    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
601    
602    if (*p++ != CHAR_COMMA) return FALSE;
603    if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
604    
605    if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
606    while ((digitab[*p] & ctype_digit) != 0) p++;
607    
608    return (*p == CHAR_RIGHT_CURLY_BRACKET);
609    }
610    
611    
612    
613    /*************************************************
614  *            Handle escapes                      *  *            Handle escapes                      *
615  *************************************************/  *************************************************/
616    
# Line 647  else Line 681  else
681      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
682      break;      break;
683    
684      /* \g must be followed by one of a number of specific things:      /* In a character class, \g is just a literal "g". Outside a character
685        class, \g must be followed by one of a number of specific things:
686    
687      (1) A number, either plain or braced. If positive, it is an absolute      (1) A number, either plain or braced. If positive, it is an absolute
688      backreference. If negative, it is a relative backreference. This is a Perl      backreference. If negative, it is a relative backreference. This is a Perl
# Line 664  else Line 699  else
699      the -ESC_g code (cf \k). */      the -ESC_g code (cf \k). */
700    
701      case CHAR_g:      case CHAR_g:
702        if (isclass) break;
703      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
704        {        {
705        c = -ESC_g;        c = -ESC_g;
# Line 885  else Line 921  else
921    }    }
922    
923  /* Perl supports \N{name} for character names, as well as plain \N for "not  /* Perl supports \N{name} for character names, as well as plain \N for "not
924  newline". PCRE does not support \N{name}. */  newline". PCRE does not support \N{name}. However, it does support
925    quantification such as \N{2,3}. */
926    
927  if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET)  if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
928         !is_counted_repeat(ptr+2))
929    *errorcodeptr = ERR37;    *errorcodeptr = ERR37;
930    
931  /* If PCRE_UCP is set, we change the values for \d etc. */  /* If PCRE_UCP is set, we change the values for \d etc. */
# Line 997  return -1; Line 1035  return -1;
1035    
1036    
1037  /*************************************************  /*************************************************
 *            Check for counted repeat            *  
 *************************************************/  
   
 /* This function is called when a '{' is encountered in a place where it might  
 start a quantifier. It looks ahead to see if it really is a quantifier or not.  
 It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}  
 where the ddds are digits.  
   
 Arguments:  
   p         pointer to the first char after '{'  
   
 Returns:    TRUE or FALSE  
 */  
   
 static BOOL  
 is_counted_repeat(const uschar *p)  
 {  
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
 if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;  
   
 if (*p++ != CHAR_COMMA) return FALSE;  
 if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;  
   
 if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  
 while ((digitab[*p] & ctype_digit) != 0) p++;  
   
 return (*p == CHAR_RIGHT_CURLY_BRACKET);  
 }  
   
   
   
 /*************************************************  
1038  *         Read repeat counts                     *  *         Read repeat counts                     *
1039  *************************************************/  *************************************************/
1040    
# Line 1694  _pcre_find_bracket(const uschar *code, B Line 1699  _pcre_find_bracket(const uschar *code, B
1699  for (;;)  for (;;)
1700    {    {
1701    register int c = *code;    register int c = *code;
1702    
1703    if (c == OP_END) return NULL;    if (c == OP_END) return NULL;
1704    
1705    /* XCLASS is used for classes that cannot be represented just by a bit    /* XCLASS is used for classes that cannot be represented just by a bit
# Line 1974  for (code = first_significant_code(code Line 1980  for (code = first_significant_code(code
1980      }      }
1981    
1982    /* For a recursion/subroutine call, if its end has been reached, which    /* For a recursion/subroutine call, if its end has been reached, which
1983    implies a subroutine call, we can scan it. */    implies a backward reference subroutine call, we can scan it. If it's a
1984      forward reference subroutine call, we can't. To detect forward reference
1985      we have to scan up the list that is kept in the workspace. This function is
1986      called only when doing the real compile, not during the pre-compile that
1987      measures the size of the compiled pattern. */
1988    
1989    if (c == OP_RECURSE)    if (c == OP_RECURSE)
1990      {      {
1991      BOOL empty_branch = FALSE;      const uschar *scode;
1992      const uschar *scode = cd->start_code + GET(code, 1);      BOOL empty_branch;
1993    
1994        /* Test for forward reference */
1995    
1996        for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
1997          if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
1998    
1999        /* Not a forward reference, test for completed backward reference */
2000    
2001        empty_branch = FALSE;
2002        scode = cd->start_code + GET(code, 1);
2003      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2004    
2005        /* Completed backwards reference */
2006    
2007      do      do
2008        {        {
2009        if (could_be_empty_branch(scode, endcode, utf8, cd))        if (could_be_empty_branch(scode, endcode, utf8, cd))
# Line 1991  for (code = first_significant_code(code Line 2014  for (code = first_significant_code(code
2014        scode += GET(scode, 1);        scode += GET(scode, 1);
2015        }        }
2016      while (*scode == OP_ALT);      while (*scode == OP_ALT);
2017    
2018      if (!empty_branch) return FALSE;  /* All branches are non-empty */      if (!empty_branch) return FALSE;  /* All branches are non-empty */
2019      continue;      continue;
2020      }      }
# Line 2216  return TRUE; Line 2240  return TRUE;
2240  the current branch of the current pattern to see if it could match the empty  the current branch of the current pattern to see if it could match the empty
2241  string. If it could, we must look outwards for branches at other levels,  string. If it could, we must look outwards for branches at other levels,
2242  stopping when we pass beyond the bracket which is the subject of the recursion.  stopping when we pass beyond the bracket which is the subject of the recursion.
2243    This function is called only during the real compile, not during the
2244    pre-compile.
2245    
2246  Arguments:  Arguments:
2247    code        points to start of the recursion    code        points to start of the recursion
# Line 2266  where Perl recognizes it as the POSIX cl Line 2292  where Perl recognizes it as the POSIX cl
2292  "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,  "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2293  I think.  I think.
2294    
2295    A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2296    It seems that the appearance of a nested POSIX class supersedes an apparent
2297    external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2298    a digit.
2299    
2300    In Perl, unescaped square brackets may also appear as part of class names. For
2301    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2302    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2303    seem right at all. PCRE does not allow closing square brackets in POSIX class
2304    names.
2305    
2306  Arguments:  Arguments:
2307    ptr      pointer to the initial [    ptr      pointer to the initial [
2308    endptr   where to return the end pointer    endptr   where to return the end pointer
# Line 2280  int terminator;          /* Don't combin Line 2317  int terminator;          /* Don't combin
2317  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2318  for (++ptr; *ptr != 0; ptr++)  for (++ptr; *ptr != 0; ptr++)
2319    {    {
2320    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2321        ptr++;
2322      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2323      else
2324      {      {
     if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;  
2325      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2326        {        {
2327        *endptr = ptr;        *endptr = ptr;
2328        return TRUE;        return TRUE;
2329        }        }
2330        if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
2331             (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2332              ptr[1] == CHAR_EQUALS_SIGN) &&
2333            check_posix_syntax(ptr, endptr))
2334          return FALSE;
2335      }      }
2336    }    }
2337  return FALSE;  return FALSE;
# Line 2997  Arguments: Line 3041  Arguments:
3041    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
3042    reqbyteptr     set to the last literal character required, else < 0    reqbyteptr     set to the last literal character required, else < 0
3043    bcptr          points to current branch chain    bcptr          points to current branch chain
3044      cond_depth     conditional nesting depth
3045    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
3046    lengthptr      NULL during the real compile phase    lengthptr      NULL during the real compile phase
3047                   points to length accumulator during pre-compile phase                   points to length accumulator during pre-compile phase
# Line 3008  Returns:         TRUE on success Line 3053  Returns:         TRUE on success
3053  static BOOL  static BOOL
3054  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,  compile_branch(int *optionsptr, uschar **codeptr, const uschar **ptrptr,
3055    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,    int *errorcodeptr, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
3056    compile_data *cd, int *lengthptr)    int cond_depth, compile_data *cd, int *lengthptr)
3057  {  {
3058  int repeat_type, op_type;  int repeat_type, op_type;
3059  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */  int repeat_min = 0, repeat_max = 0;      /* To please picky compilers */
# Line 3017  int greedy_default, greedy_non_default; Line 3062  int greedy_default, greedy_non_default;
3062  int firstbyte, reqbyte;  int firstbyte, reqbyte;
3063  int zeroreqbyte, zerofirstbyte;  int zeroreqbyte, zerofirstbyte;
3064  int req_caseopt, reqvary, tempreqvary;  int req_caseopt, reqvary, tempreqvary;
3065  int options = *optionsptr;  int options = *optionsptr;               /* May change dynamically */
3066  int after_manual_callout = 0;  int after_manual_callout = 0;
3067  int length_prevgroup = 0;  int length_prevgroup = 0;
3068  register int c;  register int c;
# Line 3035  uschar *previous_callout = NULL; Line 3080  uschar *previous_callout = NULL;
3080  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
3081  uschar classbits[32];  uschar classbits[32];
3082    
3083    /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3084    must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3085    dynamically as we process the pattern. */
3086    
3087  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3088  BOOL class_utf8;  BOOL class_utf8;
3089  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
# Line 3043  uschar *class_utf8data_base; Line 3092  uschar *class_utf8data_base;
3092  uschar utf8_char[6];  uschar utf8_char[6];
3093  #else  #else
3094  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
 uschar *utf8_char = NULL;  
3095  #endif  #endif
3096    
3097  #ifdef PCRE_DEBUG  #ifdef PCRE_DEBUG
# Line 3215  for (;; ptr++) Line 3263  for (;; ptr++)
3263      previous_callout = NULL;      previous_callout = NULL;
3264      }      }
3265    
3266    /* In extended mode, skip white space and comments */    /* In extended mode, skip white space and comments. */
3267    
3268    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3269      {      {
# Line 4184  for (;; ptr++) Line 4232  for (;; ptr++)
4232      op_type = 0;                    /* Default single-char op codes */      op_type = 0;                    /* Default single-char op codes */
4233      possessive_quantifier = FALSE;  /* Default not possessive quantifier */      possessive_quantifier = FALSE;  /* Default not possessive quantifier */
4234    
4235      /* Save start of previous item, in case we have to move it up to make space      /* Save start of previous item, in case we have to move it up in order to
4236      for an inserted OP_ONCE for the additional '+' extension. */      insert something before it. */
4237    
4238      tempcode = previous;      tempcode = previous;
4239    
# Line 4208  for (;; ptr++) Line 4256  for (;; ptr++)
4256        }        }
4257      else repeat_type = greedy_default;      else repeat_type = greedy_default;
4258    
4259        /* If previous was a recursion call, wrap it in atomic brackets so that
4260        previous becomes the atomic group. All recursions were so wrapped in the
4261        past, but it no longer happens for non-repeated recursions. In fact, the
4262        repeated ones could be re-implemented independently so as not to need this,
4263        but for the moment we rely on the code for repeating groups. */
4264    
4265        if (*previous == OP_RECURSE)
4266          {
4267          memmove(previous + 1 + LINK_SIZE, previous, 1 + LINK_SIZE);
4268          *previous = OP_ONCE;
4269          PUT(previous, 1, 2 + 2*LINK_SIZE);
4270          previous[2 + 2*LINK_SIZE] = OP_KET;
4271          PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
4272          code += 2 + 2 * LINK_SIZE;
4273          length_prevgroup = 3 + 3*LINK_SIZE;
4274    
4275          /* When actually compiling, we need to check whether this was a forward
4276          reference, and if so, adjust the offset. */
4277    
4278          if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
4279            {
4280            int offset = GET(cd->hwm, -LINK_SIZE);
4281            if (offset == previous + 1 - cd->start_code)
4282              PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
4283            }
4284          }
4285    
4286        /* Now handle repetition for the different types of item. */
4287    
4288      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
4289      repeat item instead. If a char item has a minumum of more than one, ensure      repeat item instead. If a char item has a minumum of more than one, ensure
4290      that it is set in reqbyte - it might not be if a sequence such as x{3} is      that it is set in reqbyte - it might not be if a sequence such as x{3} is
# Line 4499  for (;; ptr++) Line 4576  for (;; ptr++)
4576        }        }
4577    
4578      /* If previous was a bracket group, we may have to replicate it in certain      /* If previous was a bracket group, we may have to replicate it in certain
4579      cases. Note that at this point we can encounter only the "basic" BRA and      cases. Note that at this point we can encounter only the "basic" bracket
4580      KET opcodes, as this is the place where they get converted into the more      opcodes such as BRA and CBRA, as this is the place where they get converted
4581      special varieties. */      into the more special varieties such as BRAPOS and SBRA. A test for >=
4582        OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
4583        ASSERTBACK_NOT, ONCE, BRA, CBRA, and COND. Originally, PCRE did not allow
4584        repetition of assertions, but now it does, for Perl compatibility. */
4585    
4586      else if (*previous == OP_BRA  || *previous == OP_CBRA ||      else if (*previous >= OP_ASSERT && *previous <= OP_COND)
              *previous == OP_ONCE || *previous == OP_COND)  
4587        {        {
4588        register int i;        register int i;
4589        int len = (int)(code - previous);        int len = (int)(code - previous);
4590        uschar *bralink = NULL;        uschar *bralink = NULL;
4591        uschar *brazeroptr = NULL;        uschar *brazeroptr = NULL;
4592    
4593        /* Repeating a DEFINE group is pointless */        /* Repeating a DEFINE group is pointless, but Perl allows the syntax, so
4594          we just ignore the repeat. */
4595    
4596        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
4597            goto END_REPEAT;
4598    
4599          /* There is no sense in actually repeating assertions. The only potential
4600          use of repetition is in cases when the assertion is optional. Therefore,
4601          if the minimum is greater than zero, just ignore the repeat. If the
4602          maximum is not not zero or one, set it to 1. */
4603    
4604          if (*previous < OP_ONCE)    /* Assertion */
4605          {          {
4606          *errorcodeptr = ERR55;          if (repeat_min > 0) goto END_REPEAT;
4607          goto FAILED;          if (repeat_max < 0 || repeat_max > 1) repeat_max = 1;
4608          }          }
4609    
4610        /* The case of a zero minimum is special because of the need to stick        /* The case of a zero minimum is special because of the need to stick
# Line 4537  for (;; ptr++) Line 4625  for (;; ptr++)
4625          **   goto END_REPEAT;          **   goto END_REPEAT;
4626          **   }          **   }
4627    
4628          However, that fails when a group is referenced as a subroutine from          However, that fails when a group or a subgroup within it is referenced
4629          elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it          as a subroutine from elsewhere in the pattern, so now we stick in
4630          so that it is skipped on execution. As we don't have a list of which          OP_SKIPZERO in front of it so that it is skipped on execution. As we
4631          groups are referenced, we cannot do this selectively.          don't have a list of which groups are referenced, we cannot do this
4632            selectively.
4633    
4634          If the maximum is 1 or unlimited, we just have to stick in the BRAZERO          If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4635          and do no more at this point. However, we do need to adjust any          and do no more at this point. However, we do need to adjust any
# Line 4726  for (;; ptr++) Line 4815  for (;; ptr++)
4815          }          }
4816    
4817        /* If the maximum is unlimited, set a repeater in the final copy. For        /* If the maximum is unlimited, set a repeater in the final copy. For
4818        ONCE brackets, that's all we need to do.        ONCE brackets, that's all we need to do. However, possessively repeated
4819          ONCE brackets can be converted into non-capturing brackets, as the
4820          behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to
4821          deal with possessive ONCEs specially.
4822    
4823        Otherwise, if the quantifier was possessive, we convert the BRA code to        Otherwise, if the quantifier was possessive, we convert the BRA code to
4824        the POS form, and the KET code to KETRPOS. (It turns out to be convenient        the POS form, and the KET code to KETRPOS. (It turns out to be convenient
# Line 4748  for (;; ptr++) Line 4840  for (;; ptr++)
4840          uschar *ketcode = code - 1 - LINK_SIZE;          uschar *ketcode = code - 1 - LINK_SIZE;
4841          uschar *bracode = ketcode - GET(ketcode, 1);          uschar *bracode = ketcode - GET(ketcode, 1);
4842    
4843            if (*bracode == OP_ONCE && possessive_quantifier) *bracode = OP_BRA;
4844          if (*bracode == OP_ONCE)          if (*bracode == OP_ONCE)
4845            *ketcode = OP_KETRMAX + repeat_type;            *ketcode = OP_KETRMAX + repeat_type;
4846          else          else
# Line 4799  for (;; ptr++) Line 4892  for (;; ptr++)
4892      there are special alternative opcodes for this case. For anything else, we      there are special alternative opcodes for this case. For anything else, we
4893      wrap the entire repeated item inside OP_ONCE brackets. Logically, the '+'      wrap the entire repeated item inside OP_ONCE brackets. Logically, the '+'
4894      notation is just syntactic sugar, taken from Sun's Java package, but the      notation is just syntactic sugar, taken from Sun's Java package, but the
4895      special opcodes can optimize it.      special opcodes can optimize it.
4896    
4897      Possessively repeated subpatterns have already been handled in the code      Possessively repeated subpatterns have already been handled in the code
4898      just above, so possessive_quantifier is always FALSE for them at this      just above, so possessive_quantifier is always FALSE for them at this
4899      stage.      stage.
4900    
4901      Note that the repeated item starts at tempcode, not at previous, which      Note that the repeated item starts at tempcode, not at previous, which
4902      might be the first part of a string whose (former) last char we repeated.      might be the first part of a string whose (former) last char we repeated.
4903    
# Line 4910  for (;; ptr++) Line 5003  for (;; ptr++)
5003        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
5004        namelen = (int)(ptr - name);        namelen = (int)(ptr - name);
5005    
5006          /* It appears that Perl allows any characters whatsoever, other than
5007          a closing parenthesis, to appear in arguments, so we no longer insist on
5008          letters, digits, and underscores. */
5009    
5010        if (*ptr == CHAR_COLON)        if (*ptr == CHAR_COLON)
5011          {          {
5012          arg = ++ptr;          arg = ++ptr;
5013          while ((cd->ctypes[*ptr] & (ctype_letter|ctype_digit)) != 0          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
           || *ptr == '_') ptr++;  
5014          arglen = (int)(ptr - arg);          arglen = (int)(ptr - arg);
5015          }          }
5016    
# Line 4931  for (;; ptr++) Line 5027  for (;; ptr++)
5027          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
5028              strncmp((char *)name, vn, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
5029            {            {
5030            /* Check for open captures before ACCEPT and convert it to            /* Check for open captures before ACCEPT and convert it to
5031            ASSERT_ACCEPT if in an assertion. */            ASSERT_ACCEPT if in an assertion. */
5032    
5033            if (verbs[i].op == OP_ACCEPT)            if (verbs[i].op == OP_ACCEPT)
# Line 4941  for (;; ptr++) Line 5037  for (;; ptr++)
5037                {                {
5038                *errorcodeptr = ERR59;                *errorcodeptr = ERR59;
5039                goto FAILED;                goto FAILED;
5040                }                }
5041              cd->had_accept = TRUE;              cd->had_accept = TRUE;
5042              for (oc = cd->open_caps; oc != NULL; oc = oc->next)              for (oc = cd->open_caps; oc != NULL; oc = oc->next)
5043                {                {
# Line 5242  for (;; ptr++) Line 5338  for (;; ptr++)
5338          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5339          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
5340          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
5341          cd->assert_depth += 1;          cd->assert_depth += 1;
5342          ptr++;          ptr++;
5343          break;          break;
5344    
# Line 5257  for (;; ptr++) Line 5353  for (;; ptr++)
5353            continue;            continue;
5354            }            }
5355          bravalue = OP_ASSERT_NOT;          bravalue = OP_ASSERT_NOT;
5356          cd->assert_depth += 1;          cd->assert_depth += 1;
5357          break;          break;
5358    
5359    
# Line 5267  for (;; ptr++) Line 5363  for (;; ptr++)
5363            {            {
5364            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
5365            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
5366            cd->assert_depth += 1;            cd->assert_depth += 1;
5367            ptr += 2;            ptr += 2;
5368            break;            break;
5369    
5370            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
5371            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
5372            cd->assert_depth += 1;            cd->assert_depth += 1;
5373            ptr += 2;            ptr += 2;
5374            break;            break;
5375    
# Line 5664  for (;; ptr++) Line 5760  for (;; ptr++)
5760    
5761                /* Fudge the value of "called" so that when it is inserted as an                /* Fudge the value of "called" so that when it is inserted as an
5762                offset below, what it actually inserted is the reference number                offset below, what it actually inserted is the reference number
5763                of the group. */                of the group. Then remember the forward reference. */
5764    
5765                called = cd->start_code + recno;                called = cd->start_code + recno;
5766                PUTINC(cd->hwm, 0, (int)(code + 2 + LINK_SIZE - cd->start_code));                PUTINC(cd->hwm, 0, (int)(code + 1 - cd->start_code));
5767                }                }
5768    
5769              /* If not a forward reference, and the subpattern is still open,              /* If not a forward reference, and the subpattern is still open,
5770              this is a recursive call. We check to see if this is a left              this is a recursive call. We check to see if this is a left
5771              recursion that could loop for ever, and diagnose that case. */              recursion that could loop for ever, and diagnose that case. We
5772                must not, however, do this check if we are in a conditional
5773                subpattern because the condition might be testing for recursion in
5774                a pattern such as /(?(R)a+|(?R)b)/, which is perfectly valid.
5775                Forever loops are also detected at runtime, so those that occur in
5776                conditional subpatterns will be picked up then. */
5777    
5778              else if (GET(called, 1) == 0 &&              else if (GET(called, 1) == 0 && cond_depth <= 0 &&
5779                       could_be_empty(called, code, bcptr, utf8, cd))                       could_be_empty(called, code, bcptr, utf8, cd))
5780                {                {
5781                *errorcodeptr = ERR40;                *errorcodeptr = ERR40;
# Line 5682  for (;; ptr++) Line 5783  for (;; ptr++)
5783                }                }
5784              }              }
5785    
5786            /* Insert the recursion/subroutine item, automatically wrapped inside            /* Insert the recursion/subroutine item. */
           "once" brackets. Set up a "previous group" length so that a  
           subsequent quantifier will work. */  
   
           *code = OP_ONCE;  
           PUT(code, 1, 2 + 2*LINK_SIZE);  
           code += 1 + LINK_SIZE;  
5787    
5788            *code = OP_RECURSE;            *code = OP_RECURSE;
5789            PUT(code, 1, (int)(called - cd->start_code));            PUT(code, 1, (int)(called - cd->start_code));
5790            code += 1 + LINK_SIZE;            code += 1 + LINK_SIZE;
   
           *code = OP_KET;  
           PUT(code, 1, 2 + 2*LINK_SIZE);  
           code += 1 + LINK_SIZE;  
   
           length_prevgroup = 3 + 3*LINK_SIZE;  
5791            }            }
5792    
5793          /* Can't determine a first byte now */          /* Can't determine a first byte now */
# Line 5813  for (;; ptr++) Line 5902  for (;; ptr++)
5902        skipbytes = 2;        skipbytes = 2;
5903        }        }
5904    
5905      /* Process nested bracketed regex. Assertions may not be repeated, but      /* Process nested bracketed regex. Assertions used not to be repeatable,
5906      other kinds can be. All their opcodes are >= OP_ONCE. We copy code into a      but this was changed for Perl compatibility, so all kinds can now be
5907      non-register variable (tempcode) in order to be able to pass its address      repeated. We copy code into a non-register variable (tempcode) in order to
5908      because some compilers complain otherwise. */      be able to pass its address because some compilers complain otherwise. */
5909    
5910      previous = (bravalue >= OP_ONCE)? code : NULL;      previous = code;                      /* For handling repetition */
5911      *code = bravalue;      *code = bravalue;
5912      tempcode = code;      tempcode = code;
5913      tempreqvary = cd->req_varyopt;     /* Save value before bracket */      tempreqvary = cd->req_varyopt;        /* Save value before bracket */
5914      length_prevgroup = 0;              /* Initialize for pre-compile phase */      length_prevgroup = 0;                 /* Initialize for pre-compile phase */
5915    
5916      if (!compile_regex(      if (!compile_regex(
5917           newoptions,                   /* The complete new option state */           newoptions,                      /* The complete new option state */
5918           &tempcode,                    /* Where to put code (updated) */           &tempcode,                       /* Where to put code (updated) */
5919           &ptr,                         /* Input pointer (updated) */           &ptr,                            /* Input pointer (updated) */
5920           errorcodeptr,                 /* Where to put an error message */           errorcodeptr,                    /* Where to put an error message */
5921           (bravalue == OP_ASSERTBACK ||           (bravalue == OP_ASSERTBACK ||
5922            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */
5923           reset_bracount,               /* True if (?| group */           reset_bracount,                  /* True if (?| group */
5924           skipbytes,                    /* Skip over bracket number */           skipbytes,                       /* Skip over bracket number */
5925           &subfirstbyte,                /* For possible first char */           cond_depth +
5926           &subreqbyte,                  /* For possible last char */             ((bravalue == OP_COND)?1:0),   /* Depth of condition subpatterns */
5927           bcptr,                        /* Current branch chain */           &subfirstbyte,                   /* For possible first char */
5928           cd,                           /* Tables block */           &subreqbyte,                     /* For possible last char */
5929           (lengthptr == NULL)? NULL :   /* Actual compile phase */           bcptr,                           /* Current branch chain */
5930             &length_prevgroup           /* Pre-compile phase */           cd,                              /* Tables block */
5931             (lengthptr == NULL)? NULL :      /* Actual compile phase */
5932               &length_prevgroup              /* Pre-compile phase */
5933           ))           ))
5934        goto FAILED;        goto FAILED;
5935    
5936      if (bravalue >= OP_ASSERT && bravalue <= OP_ASSERTBACK_NOT)      if (bravalue >= OP_ASSERT && bravalue <= OP_ASSERTBACK_NOT)
5937        cd->assert_depth -= 1;        cd->assert_depth -= 1;
5938    
5939      /* At the end of compiling, code is still pointing to the start of the      /* At the end of compiling, code is still pointing to the start of the
5940      group, while tempcode has been updated to point past the end of the group      group, while tempcode has been updated to point past the end of the group
# Line 6083  for (;; ptr++) Line 6174  for (;; ptr++)
6174          }          }
6175    
6176        /* \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).
6177        We also support \k{name} (.NET syntax) */        We also support \k{name} (.NET syntax).  */
6178    
6179        if (-c == ESC_k && (ptr[1] == CHAR_LESS_THAN_SIGN ||        if (-c == ESC_k)
           ptr[1] == CHAR_APOSTROPHE || ptr[1] == CHAR_LEFT_CURLY_BRACKET))  
6180          {          {
6181            if ((ptr[1] != CHAR_LESS_THAN_SIGN &&
6182              ptr[1] != CHAR_APOSTROPHE && ptr[1] != CHAR_LEFT_CURLY_BRACKET))
6183              {
6184              *errorcodeptr = ERR69;
6185              break;
6186              }
6187          is_recurse = FALSE;          is_recurse = FALSE;
6188          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
6189            CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?            CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?
# Line 6291  Arguments: Line 6387  Arguments:
6387    lookbehind     TRUE if this is a lookbehind assertion    lookbehind     TRUE if this is a lookbehind assertion
6388    reset_bracount TRUE to reset the count for each branch    reset_bracount TRUE to reset the count for each branch
6389    skipbytes      skip this many bytes at start (for brackets and OP_COND)    skipbytes      skip this many bytes at start (for brackets and OP_COND)
6390      cond_depth     depth of nesting for conditional subpatterns
6391    firstbyteptr   place to put the first required character, or a negative number    firstbyteptr   place to put the first required character, or a negative number
6392    reqbyteptr     place to put the last required character, or a negative number    reqbyteptr     place to put the last required character, or a negative number
6393    bcptr          pointer to the chain of currently open branches    bcptr          pointer to the chain of currently open branches
# Line 6304  Returns:         TRUE on success Line 6401  Returns:         TRUE on success
6401  static BOOL  static BOOL
6402  compile_regex(int options, uschar **codeptr, const uschar **ptrptr,  compile_regex(int options, uschar **codeptr, const uschar **ptrptr,
6403    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,
6404    int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr, compile_data *cd,    int cond_depth, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
6405    int *lengthptr)    compile_data *cd, int *lengthptr)
6406  {  {
6407  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
6408  uschar *code = *codeptr;  uschar *code = *codeptr;
# Line 6342  pre-compile phase to find out whether an Line 6439  pre-compile phase to find out whether an
6439    
6440  /* If this is a capturing subpattern, add to the chain of open capturing items  /* If this is a capturing subpattern, add to the chain of open capturing items
6441  so that we can detect them if (*ACCEPT) is encountered. This is also used to  so that we can detect them if (*ACCEPT) is encountered. This is also used to
6442  detect groups that contain recursive back references to themselves. Note that  detect groups that contain recursive back references to themselves. Note that
6443  only OP_CBRA need be tested here; changing this opcode to one of its variants,  only OP_CBRA need be tested here; changing this opcode to one of its variants,
6444  e.g. OP_SCBRAPOS, happens later, after the group has been compiled. */  e.g. OP_SCBRAPOS, happens later, after the group has been compiled. */
6445    
6446  if (*code == OP_CBRA)  if (*code == OP_CBRA)
# Line 6384  for (;;) Line 6481  for (;;)
6481    into the length. */    into the length. */
6482    
6483    if (!compile_branch(&options, &code, &ptr, errorcodeptr, &branchfirstbyte,    if (!compile_branch(&options, &code, &ptr, errorcodeptr, &branchfirstbyte,
6484          &branchreqbyte, &bc, cd, (lengthptr == NULL)? NULL : &length))          &branchreqbyte, &bc, cond_depth, cd,
6485            (lengthptr == NULL)? NULL : &length))
6486      {      {
6487      *ptrptr = ptr;      *ptrptr = ptr;
6488      return FALSE;      return FALSE;
# Line 6817  do { Line 6915  do {
6915       case OP_EXACT:       case OP_EXACT:
6916       scode += 2;       scode += 2;
6917       /* Fall through */       /* Fall through */
6918    
6919       case OP_CHAR:       case OP_CHAR:
6920       case OP_PLUS:       case OP_PLUS:
6921       case OP_MINPLUS:       case OP_MINPLUS:
# Line 6830  do { Line 6928  do {
6928       case OP_EXACTI:       case OP_EXACTI:
6929       scode += 2;       scode += 2;
6930       /* Fall through */       /* Fall through */
6931    
6932       case OP_CHARI:       case OP_CHARI:
6933       case OP_PLUSI:       case OP_PLUSI:
6934       case OP_MINPLUSI:       case OP_MINPLUSI:
# Line 6990  utf8 = (options & PCRE_UTF8) != 0; Line 7088  utf8 = (options & PCRE_UTF8) != 0;
7088    
7089  /* Can't support UTF8 unless PCRE has been compiled to include the code. The  /* Can't support UTF8 unless PCRE has been compiled to include the code. The
7090  return of an error code from _pcre_valid_utf8() is a new feature, introduced in  return of an error code from _pcre_valid_utf8() is a new feature, introduced in
7091  release 8.13. It is passed back from pcre_[dfa_]exec(), but at the moment is  release 8.13. It is passed back from pcre_[dfa_]exec(), but at the moment is
7092  not used here. */  not used here. */
7093    
7094  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 7020  if ((options & PCRE_UCP) != 0) Line 7118  if ((options & PCRE_UCP) != 0)
7118    
7119  /* Check validity of \R options. */  /* Check validity of \R options. */
7120    
7121  switch (options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))  if ((options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) ==
7122         (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))
7123    {    {
7124    case 0:    errorcode = ERR56;
7125    case PCRE_BSR_ANYCRLF:    goto PCRE_EARLY_ERROR_RETURN;
   case PCRE_BSR_UNICODE:  
   break;  
   default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;  
7126    }    }
7127    
7128  /* Handle different types of newline. The three bits give seven cases. The  /* Handle different types of newline. The three bits give seven cases. The
# Line 7111  outside can help speed up starting point Line 7207  outside can help speed up starting point
7207  ptr += skipatstart;  ptr += skipatstart;
7208  code = cworkspace;  code = cworkspace;
7209  *code = OP_BRA;  *code = OP_BRA;
7210  (void)compile_regex(cd->external_options, &code, &ptr, &errorcode, FALSE,  (void)compile_regex(cd->external_options, &code, &ptr, &errorcode, FALSE,
7211    FALSE, 0, &firstbyte, &reqbyte, NULL, cd, &length);    FALSE, 0, 0, &firstbyte, &reqbyte, NULL, cd, &length);
7212  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;
7213    
7214  DPRINTF(("end pre-compile: length=%d workspace=%d\n", length,  DPRINTF(("end pre-compile: length=%d workspace=%d\n", length,
# Line 7185  of the function here. */ Line 7281  of the function here. */
7281  ptr = (const uschar *)pattern + skipatstart;  ptr = (const uschar *)pattern + skipatstart;
7282  code = (uschar *)codestart;  code = (uschar *)codestart;
7283  *code = OP_BRA;  *code = OP_BRA;
7284  (void)compile_regex(re->options, &code, &ptr, &errorcode, FALSE, FALSE, 0,  (void)compile_regex(re->options, &code, &ptr, &errorcode, FALSE, FALSE, 0, 0,
7285    &firstbyte, &reqbyte, NULL, cd, NULL);    &firstbyte, &reqbyte, NULL, cd, NULL);
7286  re->top_bracket = cd->bracount;  re->top_bracket = cd->bracount;
7287  re->top_backref = cd->top_backref;  re->top_backref = cd->top_backref;

Legend:
Removed from v.613  
changed lines
  Added in v.691

  ViewVC Help
Powered by ViewVC 1.1.5