/[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 643 by ph10, Fri Jul 29 15:56:39 2011 UTC revision 723 by ph10, Sat Oct 8 15:55:23 2011 UTC
# 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"    "\\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 681  else Line 681  else
681      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
682      break;      break;
683    
684      /* In a character class, \g is just a literal "g". Outside a character      /* 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:      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
# Line 921  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}. However, it does support  newline". PCRE does not support \N{name}. However, it does support
925  quantification such as \N{2,3}. */  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 &&
# Line 1506  for (;;) Line 1506  for (;;)
1506      case OP_CBRA:      case OP_CBRA:
1507      case OP_BRA:      case OP_BRA:
1508      case OP_ONCE:      case OP_ONCE:
1509        case OP_ONCE_NC:
1510      case OP_COND:      case OP_COND:
1511      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), utf8, atend, cd);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), utf8, atend, cd);
1512      if (d < 0) return d;      if (d < 0) return d;
# Line 1761  for (;;) Line 1762  for (;;)
1762        break;        break;
1763    
1764        case OP_THEN_ARG:        case OP_THEN_ARG:
1765        code += code[1+LINK_SIZE];        code += code[1];
1766        break;        break;
1767        }        }
1768    
# Line 1880  for (;;) Line 1881  for (;;)
1881        break;        break;
1882    
1883        case OP_THEN_ARG:        case OP_THEN_ARG:
1884        code += code[1+LINK_SIZE];        code += code[1];
1885        break;        break;
1886        }        }
1887    
# Line 1982  for (code = first_significant_code(code Line 1983  for (code = first_significant_code(code
1983    /* For a recursion/subroutine call, if its end has been reached, which    /* For a recursion/subroutine call, if its end has been reached, which
1984    implies a backward reference subroutine call, we can scan it. If it's a    implies a backward reference subroutine call, we can scan it. If it's a
1985    forward reference subroutine call, we can't. To detect forward reference    forward reference subroutine call, we can't. To detect forward reference
1986    we have to scan up the list that is kept in the workspace. This function is    we have to scan up the list that is kept in the workspace. This function is
1987    called only when doing the real compile, not during the pre-compile that    called only when doing the real compile, not during the pre-compile that
1988    measures the size of the compiled pattern. */    measures the size of the compiled pattern. */
1989    
1990    if (c == OP_RECURSE)    if (c == OP_RECURSE)
1991      {      {
1992      const uschar *scode;      const uschar *scode;
1993      BOOL empty_branch;      BOOL empty_branch;
1994    
1995      /* Test for forward reference */      /* Test for forward reference */
1996    
1997      for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)      for (scode = cd->start_workspace; scode < cd->hwm; scode += LINK_SIZE)
1998        if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;        if (GET(scode, 0) == code + 1 - cd->start_code) return TRUE;
1999    
2000      /* Not a forward reference, test for completed backward reference */      /* Not a forward reference, test for completed backward reference */
2001    
2002      empty_branch = FALSE;      empty_branch = FALSE;
2003      scode = cd->start_code + GET(code, 1);      scode = cd->start_code + GET(code, 1);
2004      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */      if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
2005    
2006      /* Completed backwards reference */      /* Completed backwards reference */
2007    
2008      do      do
2009        {        {
2010        if (could_be_empty_branch(scode, endcode, utf8, cd))        if (could_be_empty_branch(scode, endcode, utf8, cd))
# Line 2014  for (code = first_significant_code(code Line 2015  for (code = first_significant_code(code
2015        scode += GET(scode, 1);        scode += GET(scode, 1);
2016        }        }
2017      while (*scode == OP_ALT);      while (*scode == OP_ALT);
2018    
2019      if (!empty_branch) return FALSE;  /* All branches are non-empty */      if (!empty_branch) return FALSE;  /* All branches are non-empty */
2020      continue;      continue;
2021      }      }
# Line 2045  for (code = first_significant_code(code Line 2046  for (code = first_significant_code(code
2046    
2047    if (c == OP_BRA  || c == OP_BRAPOS ||    if (c == OP_BRA  || c == OP_BRAPOS ||
2048        c == OP_CBRA || c == OP_CBRAPOS ||        c == OP_CBRA || c == OP_CBRAPOS ||
2049        c == OP_ONCE || c == OP_COND)        c == OP_ONCE || c == OP_ONCE_NC ||
2050          c == OP_COND)
2051      {      {
2052      BOOL empty_branch;      BOOL empty_branch;
2053      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
# Line 2217  for (code = first_significant_code(code Line 2219  for (code = first_significant_code(code
2219      break;      break;
2220    
2221      case OP_THEN_ARG:      case OP_THEN_ARG:
2222      code += code[1+LINK_SIZE];      code += code[1];
2223      break;      break;
2224    
2225      /* None of the remaining opcodes are required to match a character. */      /* None of the remaining opcodes are required to match a character. */
# Line 2240  return TRUE; Line 2242  return TRUE;
2242  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
2243  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,
2244  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.
2245  This function is called only during the real compile, not during the  This function is called only during the real compile, not during the
2246  pre-compile.  pre-compile.
2247    
2248  Arguments:  Arguments:
# Line 2295  I think. Line 2297  I think.
2297  A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.  A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
2298  It seems that the appearance of a nested POSIX class supersedes an apparent  It seems that the appearance of a nested POSIX class supersedes an apparent
2299  external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or  external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
2300  a digit. Also, unescaped square brackets may also appear as part of class  a digit.
2301  names. For example, [:a[:abc]b:] gives unknown class "[:abc]b:]"in Perl.  
2302    In Perl, unescaped square brackets may also appear as part of class names. For
2303    example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
2304    [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
2305    seem right at all. PCRE does not allow closing square brackets in POSIX class
2306    names.
2307    
2308  Arguments:  Arguments:
2309    ptr      pointer to the initial [    ptr      pointer to the initial [
# Line 2312  int terminator;          /* Don't combin Line 2319  int terminator;          /* Don't combin
2319  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2320  for (++ptr; *ptr != 0; ptr++)  for (++ptr; *ptr != 0; ptr++)
2321    {    {
2322    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2323      ptr++;      ptr++;
2324      else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2325    else    else
2326      {      {
2327      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)      if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
# Line 2325  for (++ptr; *ptr != 0; ptr++) Line 2333  for (++ptr; *ptr != 0; ptr++)
2333           (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||           (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2334            ptr[1] == CHAR_EQUALS_SIGN) &&            ptr[1] == CHAR_EQUALS_SIGN) &&
2335          check_posix_syntax(ptr, endptr))          check_posix_syntax(ptr, endptr))
2336        return FALSE;        return FALSE;
2337      }      }
2338    }    }
2339  return FALSE;  return FALSE;
# Line 3035  Arguments: Line 3043  Arguments:
3043    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)    firstbyteptr   set to initial literal character, or < 0 (REQ_UNSET, REQ_NONE)
3044    reqbyteptr     set to the last literal character required, else < 0    reqbyteptr     set to the last literal character required, else < 0
3045    bcptr          points to current branch chain    bcptr          points to current branch chain
3046    cond_depth     conditional nesting depth    cond_depth     conditional nesting depth
3047    cd             contains pointers to tables etc.    cd             contains pointers to tables etc.
3048    lengthptr      NULL during the real compile phase    lengthptr      NULL during the real compile phase
3049                   points to length accumulator during pre-compile phase                   points to length accumulator during pre-compile phase
# Line 3075  uschar *save_hwm = NULL; Line 3083  uschar *save_hwm = NULL;
3083  uschar classbits[32];  uschar classbits[32];
3084    
3085  /* We can fish out the UTF-8 setting once and for all into a BOOL, but we  /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
3086  must not do this for other options (e.g. PCRE_EXTENDED) because they may change  must not do this for other options (e.g. PCRE_EXTENDED) because they may change
3087  dynamically as we process the pattern. */  dynamically as we process the pattern. */
3088    
3089  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 3086  uschar *class_utf8data_base; Line 3094  uschar *class_utf8data_base;
3094  uschar utf8_char[6];  uschar utf8_char[6];
3095  #else  #else
3096  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
 uschar *utf8_char = NULL;  
3097  #endif  #endif
3098    
3099  #ifdef PCRE_DEBUG  #ifdef PCRE_DEBUG
# Line 3137  for (;; ptr++) Line 3144  for (;; ptr++)
3144    int subfirstbyte;    int subfirstbyte;
3145    int terminator;    int terminator;
3146    int mclength;    int mclength;
3147      int tempbracount;
3148    uschar mcbuffer[8];    uschar mcbuffer[8];
3149    
3150    /* Get next byte in the pattern */    /* Get next byte in the pattern */
# Line 4250  for (;; ptr++) Line 4258  for (;; ptr++)
4258        ptr++;        ptr++;
4259        }        }
4260      else repeat_type = greedy_default;      else repeat_type = greedy_default;
4261    
4262      /* If previous was a recursion call, wrap it in atomic brackets so that      /* If previous was a recursion call, wrap it in atomic brackets so that
4263      previous becomes the atomic group. All recursions were so wrapped in the      previous becomes the atomic group. All recursions were so wrapped in the
4264      past, but it no longer happens for non-repeated recursions. In fact, the      past, but it no longer happens for non-repeated recursions. In fact, the
4265      repeated ones could be re-implemented independently so as not to need this,      repeated ones could be re-implemented independently so as not to need this,
4266      but for the moment we rely on the code for repeating groups. */      but for the moment we rely on the code for repeating groups. */
4267    
4268      if (*previous == OP_RECURSE)      if (*previous == OP_RECURSE)
4269        {        {
4270        memmove(previous + 1 + LINK_SIZE, previous, 1 + LINK_SIZE);        memmove(previous + 1 + LINK_SIZE, previous, 1 + LINK_SIZE);
# Line 4266  for (;; ptr++) Line 4274  for (;; ptr++)
4274        PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);        PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
4275        code += 2 + 2 * LINK_SIZE;        code += 2 + 2 * LINK_SIZE;
4276        length_prevgroup = 3 + 3*LINK_SIZE;        length_prevgroup = 3 + 3*LINK_SIZE;
4277    
4278        /* When actually compiling, we need to check whether this was a forward        /* When actually compiling, we need to check whether this was a forward
4279        reference, and if so, adjust the offset. */        reference, and if so, adjust the offset. */
4280    
4281        if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)        if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
4282          {          {
4283          int offset = GET(cd->hwm, -LINK_SIZE);          int offset = GET(cd->hwm, -LINK_SIZE);
4284          if (offset == previous + 1 - cd->start_code)          if (offset == previous + 1 - cd->start_code)
4285            PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);            PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
4286          }          }
4287        }        }
4288    
4289      /* Now handle repetition for the different types of item. */      /* Now handle repetition for the different types of item. */
4290    
4291      /* If previous was a character match, abolish the item and generate a      /* If previous was a character match, abolish the item and generate a
# Line 4575  for (;; ptr++) Line 4583  for (;; ptr++)
4583      opcodes such as BRA and CBRA, as this is the place where they get converted      opcodes such as BRA and CBRA, as this is the place where they get converted
4584      into the more special varieties such as BRAPOS and SBRA. A test for >=      into the more special varieties such as BRAPOS and SBRA. A test for >=
4585      OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,      OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
4586      ASSERTBACK_NOT, ONCE, BRA, CBRA, and COND. Originally, PCRE did not allow      ASSERTBACK_NOT, ONCE, BRA, CBRA, and COND. Originally, PCRE did not allow
4587      repetition of assertions, but now it does, for Perl compatibility. */      repetition of assertions, but now it does, for Perl compatibility. */
4588    
4589      else if (*previous >= OP_ASSERT && *previous <= OP_COND)      else if (*previous >= OP_ASSERT && *previous <= OP_COND)
4590        {        {
4591        register int i;        register int i;
4592        int len = (int)(code - previous);        int len = (int)(code - previous);
4593        uschar *bralink = NULL;        uschar *bralink = NULL;
4594        uschar *brazeroptr = NULL;        uschar *brazeroptr = NULL;
4595    
4596        /* Repeating a DEFINE group is pointless, but Perl allows the syntax, so        /* Repeating a DEFINE group is pointless, but Perl allows the syntax, so
4597        we just ignore the repeat. */        we just ignore the repeat. */
4598    
4599        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)        if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
4600          goto END_REPEAT;          goto END_REPEAT;
4601    
4602        /* There is no sense in actually repeating assertions. The only potential        /* There is no sense in actually repeating assertions. The only potential
4603        use of repetition is in cases when the assertion is optional. Therefore,        use of repetition is in cases when the assertion is optional. Therefore,
4604        if the minimum is greater than zero, just ignore the repeat. If the        if the minimum is greater than zero, just ignore the repeat. If the
4605        maximum is not not zero or one, set it to 1. */        maximum is not not zero or one, set it to 1. */
4606    
4607        if (*previous < OP_ONCE)    /* Assertion */        if (*previous < OP_ONCE)    /* Assertion */
4608          {          {
4609          if (repeat_min > 0) goto END_REPEAT;          if (repeat_min > 0) goto END_REPEAT;
4610          if (repeat_max < 0 || repeat_max > 1) repeat_max = 1;          if (repeat_max < 0 || repeat_max > 1) repeat_max = 1;
4611          }          }
4612    
4613        /* 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
4614        OP_BRAZERO in front of it, and because the group appears once in the        OP_BRAZERO in front of it, and because the group appears once in the
# Line 4834  for (;; ptr++) Line 4842  for (;; ptr++)
4842          {          {
4843          uschar *ketcode = code - 1 - LINK_SIZE;          uschar *ketcode = code - 1 - LINK_SIZE;
4844          uschar *bracode = ketcode - GET(ketcode, 1);          uschar *bracode = ketcode - GET(ketcode, 1);
4845    
4846          if (*bracode == OP_ONCE && possessive_quantifier) *bracode = OP_BRA;          if ((*bracode == OP_ONCE || *bracode == OP_ONCE_NC) &&
4847          if (*bracode == OP_ONCE)              possessive_quantifier) *bracode = OP_BRA;
4848    
4849            if (*bracode == OP_ONCE || *bracode == OP_ONCE_NC)
4850            *ketcode = OP_KETRMAX + repeat_type;            *ketcode = OP_KETRMAX + repeat_type;
4851          else          else
4852            {            {
# Line 4887  for (;; ptr++) Line 4897  for (;; ptr++)
4897      there are special alternative opcodes for this case. For anything else, we      there are special alternative opcodes for this case. For anything else, we
4898      wrap the entire repeated item inside OP_ONCE brackets. Logically, the '+'      wrap the entire repeated item inside OP_ONCE brackets. Logically, the '+'
4899      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
4900      special opcodes can optimize it.      special opcodes can optimize it.
4901    
4902      Possessively repeated subpatterns have already been handled in the code      Possessively repeated subpatterns have already been handled in the code
4903      just above, so possessive_quantifier is always FALSE for them at this      just above, so possessive_quantifier is always FALSE for them at this
4904      stage.      stage.
4905    
4906      Note that the repeated item starts at tempcode, not at previous, which      Note that the repeated item starts at tempcode, not at previous, which
4907      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.
4908    
# Line 4997  for (;; ptr++) Line 5007  for (;; ptr++)
5007        previous = NULL;        previous = NULL;
5008        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
5009        namelen = (int)(ptr - name);        namelen = (int)(ptr - name);
5010    
5011        /* It appears that Perl allows any characters whatsoever, other than        /* It appears that Perl allows any characters whatsoever, other than
5012        a closing parenthesis, to appear in arguments, so we no longer insist on        a closing parenthesis, to appear in arguments, so we no longer insist on
5013        letters, digits, and underscores. */        letters, digits, and underscores. */
5014    
5015        if (*ptr == CHAR_COLON)        if (*ptr == CHAR_COLON)
5016          {          {
# Line 5008  for (;; ptr++) Line 5018  for (;; ptr++)
5018          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
5019          arglen = (int)(ptr - arg);          arglen = (int)(ptr - arg);
5020          }          }
5021    
5022        if (*ptr != CHAR_RIGHT_PARENTHESIS)        if (*ptr != CHAR_RIGHT_PARENTHESIS)
5023          {          {
5024          *errorcodeptr = ERR60;          *errorcodeptr = ERR60;
# Line 5022  for (;; ptr++) Line 5032  for (;; ptr++)
5032          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
5033              strncmp((char *)name, vn, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
5034            {            {
5035            /* Check for open captures before ACCEPT and convert it to            /* Check for open captures before ACCEPT and convert it to
5036            ASSERT_ACCEPT if in an assertion. */            ASSERT_ACCEPT if in an assertion. */
5037    
5038            if (verbs[i].op == OP_ACCEPT)            if (verbs[i].op == OP_ACCEPT)
# Line 5032  for (;; ptr++) Line 5042  for (;; ptr++)
5042                {                {
5043                *errorcodeptr = ERR59;                *errorcodeptr = ERR59;
5044                goto FAILED;                goto FAILED;
5045                }                }
5046              cd->had_accept = TRUE;              cd->had_accept = TRUE;
5047              for (oc = cd->open_caps; oc != NULL; oc = oc->next)              for (oc = cd->open_caps; oc != NULL; oc = oc->next)
5048                {                {
# Line 5040  for (;; ptr++) Line 5050  for (;; ptr++)
5050                PUT2INC(code, 0, oc->number);                PUT2INC(code, 0, oc->number);
5051                }                }
5052              *code++ = (cd->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT;              *code++ = (cd->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT;
5053    
5054                /* Do not set firstbyte after *ACCEPT */
5055                if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
5056              }              }
5057    
5058            /* Handle other cases with/without an argument */            /* Handle other cases with/without an argument */
# Line 5052  for (;; ptr++) Line 5065  for (;; ptr++)
5065                goto FAILED;                goto FAILED;
5066                }                }
5067              *code = verbs[i].op;              *code = verbs[i].op;
5068              if (*code++ == OP_THEN)              if (*code++ == OP_THEN) cd->external_flags |= PCRE_HASTHEN;
               {  
               PUT(code, 0, code - bcptr->current_branch - 1);  
               code += LINK_SIZE;  
               }  
5069              }              }
5070    
5071            else            else
# Line 5067  for (;; ptr++) Line 5076  for (;; ptr++)
5076                goto FAILED;                goto FAILED;
5077                }                }
5078              *code = verbs[i].op_arg;              *code = verbs[i].op_arg;
5079              if (*code++ == OP_THEN_ARG)              if (*code++ == OP_THEN_ARG) cd->external_flags |= PCRE_HASTHEN;
               {  
               PUT(code, 0, code - bcptr->current_branch - 1);  
               code += LINK_SIZE;  
               }  
5080              *code++ = arglen;              *code++ = arglen;
5081              memcpy(code, arg, arglen);              memcpy(code, arg, arglen);
5082              code += arglen;              code += arglen;
# Line 5333  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 5348  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 5358  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 5608  for (;; ptr++) Line 5613  for (;; ptr++)
5613    
5614            temp = cd->end_pattern;            temp = cd->end_pattern;
5615            cd->end_pattern = ptr;            cd->end_pattern = ptr;
5616            recno = find_parens(cd, name, namelen,            recno = find_parens(cd, name, namelen,
5617              (options & PCRE_EXTENDED) != 0, utf8);              (options & PCRE_EXTENDED) != 0, utf8);
5618            cd->end_pattern = temp;            cd->end_pattern = temp;
5619            if (recno < 0) recno = 0;    /* Forward ref; set dummy number */            if (recno < 0) recno = 0;    /* Forward ref; set dummy number */
# Line 5764  for (;; ptr++) Line 5769  for (;; ptr++)
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. We              recursion that could loop for ever, and diagnose that case. We
5772              must not, however, do this check if we are in a conditional              must not, however, do this check if we are in a conditional
5773              subpattern because the condition might be testing for recursion in              subpattern because the condition might be testing for recursion in
5774              a pattern such as /(?(R)a+|(?R)b)/, which is perfectly valid.              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              Forever loops are also detected at runtime, so those that occur in
5776              conditional subpatterns will be picked up then. */              conditional subpatterns will be picked up then. */
5777    
5778              else if (GET(called, 1) == 0 && cond_depth <= 0 &&              else if (GET(called, 1) == 0 && cond_depth <= 0 &&
# Line 5779  for (;; ptr++) Line 5784  for (;; ptr++)
5784              }              }
5785    
5786            /* Insert the recursion/subroutine item. */            /* Insert the recursion/subroutine item. */
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;
# Line 5903  for (;; ptr++) Line 5908  for (;; ptr++)
5908      be able to pass its address because some compilers complain otherwise. */      be able to pass its address because some compilers complain otherwise. */
5909    
5910      previous = code;                      /* For handling repetition */      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        tempbracount = cd->bracount;          /* Save value before bracket */
5915      length_prevgroup = 0;                 /* Initialize for pre-compile phase */      length_prevgroup = 0;                 /* Initialize for pre-compile phase */
5916    
5917      if (!compile_regex(      if (!compile_regex(
5918           newoptions,                      /* The complete new option state */           newoptions,                      /* The complete new option state */
5919           &tempcode,                       /* Where to put code (updated) */           &tempcode,                       /* Where to put code (updated) */
5920           &ptr,                            /* Input pointer (updated) */           &ptr,                            /* Input pointer (updated) */
# Line 5917  for (;; ptr++) Line 5923  for (;; ptr++)
5923            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */
5924           reset_bracount,                  /* True if (?| group */           reset_bracount,                  /* True if (?| group */
5925           skipbytes,                       /* Skip over bracket number */           skipbytes,                       /* Skip over bracket number */
5926           cond_depth +           cond_depth +
5927             ((bravalue == OP_COND)?1:0),   /* Depth of condition subpatterns */             ((bravalue == OP_COND)?1:0),   /* Depth of condition subpatterns */
5928           &subfirstbyte,                   /* For possible first char */           &subfirstbyte,                   /* For possible first char */
5929           &subreqbyte,                     /* For possible last char */           &subreqbyte,                     /* For possible last char */
# Line 5928  for (;; ptr++) Line 5934  for (;; ptr++)
5934           ))           ))
5935        goto FAILED;        goto FAILED;
5936    
5937        /* If this was an atomic group and there are no capturing groups within it,
5938        generate OP_ONCE_NC instead of OP_ONCE. */
5939    
5940        if (bravalue == OP_ONCE && cd->bracount <= tempbracount)
5941          *code = OP_ONCE_NC;
5942    
5943      if (bravalue >= OP_ASSERT && bravalue <= OP_ASSERTBACK_NOT)      if (bravalue >= OP_ASSERT && bravalue <= OP_ASSERTBACK_NOT)
5944        cd->assert_depth -= 1;        cd->assert_depth -= 1;
5945    
5946      /* 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
5947      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.
5948      and any option resetting that may follow it. The pattern pointer (ptr)      The pattern pointer (ptr) is on the bracket.
     is on the bracket. */  
5949    
5950      /* If this is a conditional bracket, check that there are no more than      If this is a conditional bracket, check that there are no more than
5951      two branches in the group, or just one if it's a DEFINE group. We do this      two branches in the group, or just one if it's a DEFINE group. We do this
5952      in the real compile phase, not in the pre-pass, where the whole group may      in the real compile phase, not in the pre-pass, where the whole group may
5953      not be available. */      not be available. */
# Line 6177  for (;; ptr++) Line 6188  for (;; ptr++)
6188            ptr[1] != CHAR_APOSTROPHE && ptr[1] != CHAR_LEFT_CURLY_BRACKET))            ptr[1] != CHAR_APOSTROPHE && ptr[1] != CHAR_LEFT_CURLY_BRACKET))
6189            {            {
6190            *errorcodeptr = ERR69;            *errorcodeptr = ERR69;
6191            break;            break;
6192            }            }
6193          is_recurse = FALSE;          is_recurse = FALSE;
6194          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
6195            CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?            CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?
6196            CHAR_APOSTROPHE : CHAR_RIGHT_CURLY_BRACKET;            CHAR_APOSTROPHE : CHAR_RIGHT_CURLY_BRACKET;
6197          goto NAMED_REF_OR_RECURSE;          goto NAMED_REF_OR_RECURSE;
6198          }          }
6199    
6200        /* Back references are handled specially; must disable firstbyte if        /* Back references are handled specially; must disable firstbyte if
6201        not set to cope with cases like (?=(\w+))\1: which would otherwise set        not set to cope with cases like (?=(\w+))\1: which would otherwise set
# Line 6335  for (;; ptr++) Line 6346  for (;; ptr++)
6346        else firstbyte = reqbyte = REQ_NONE;        else firstbyte = reqbyte = REQ_NONE;
6347        }        }
6348    
6349      /* firstbyte was previously set; we can set reqbyte only the length is      /* firstbyte was previously set; we can set reqbyte only if the length is
6350      1 or the matching is caseful. */      1 or the matching is caseful. */
6351    
6352      else      else
# Line 6382  Arguments: Line 6393  Arguments:
6393    lookbehind     TRUE if this is a lookbehind assertion    lookbehind     TRUE if this is a lookbehind assertion
6394    reset_bracount TRUE to reset the count for each branch    reset_bracount TRUE to reset the count for each branch
6395    skipbytes      skip this many bytes at start (for brackets and OP_COND)    skipbytes      skip this many bytes at start (for brackets and OP_COND)
6396    cond_depth     depth of nesting for conditional subpatterns    cond_depth     depth of nesting for conditional subpatterns
6397    firstbyteptr   place to put the first required character, or a negative number    firstbyteptr   place to put the first required character, or a negative number
6398    reqbyteptr     place to put the last required character, or a negative number    reqbyteptr     place to put the last required character, or a negative number
6399    bcptr          pointer to the chain of currently open branches    bcptr          pointer to the chain of currently open branches
# Line 6396  Returns:         TRUE on success Line 6407  Returns:         TRUE on success
6407  static BOOL  static BOOL
6408  compile_regex(int options, uschar **codeptr, const uschar **ptrptr,  compile_regex(int options, uschar **codeptr, const uschar **ptrptr,
6409    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,
6410    int cond_depth, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,    int cond_depth, int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr,
6411    compile_data *cd, int *lengthptr)    compile_data *cd, int *lengthptr)
6412  {  {
6413  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
# Line 6434  pre-compile phase to find out whether an Line 6445  pre-compile phase to find out whether an
6445    
6446  /* 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
6447  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
6448  detect groups that contain recursive back references to themselves. Note that  detect groups that contain recursive back references to themselves. Note that
6449  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,
6450  e.g. OP_SCBRAPOS, happens later, after the group has been compiled. */  e.g. OP_SCBRAPOS, happens later, after the group has been compiled. */
6451    
6452  if (*code == OP_CBRA)  if (*code == OP_CBRA)
# Line 6476  for (;;) Line 6487  for (;;)
6487    into the length. */    into the length. */
6488    
6489    if (!compile_branch(&options, &code, &ptr, errorcodeptr, &branchfirstbyte,    if (!compile_branch(&options, &code, &ptr, errorcodeptr, &branchfirstbyte,
6490          &branchreqbyte, &bc, cond_depth, cd,          &branchreqbyte, &bc, cond_depth, cd,
6491          (lengthptr == NULL)? NULL : &length))          (lengthptr == NULL)? NULL : &length))
6492      {      {
6493      *ptrptr = ptr;      *ptrptr = ptr;
# Line 6727  do { Line 6738  do {
6738    
6739     /* Other brackets */     /* Other brackets */
6740    
6741     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_COND)     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_ONCE_NC ||
6742                op == OP_COND)
6743       {       {
6744       if (!is_anchored(scode, bracket_map, backref_map)) return FALSE;       if (!is_anchored(scode, bracket_map, backref_map)) return FALSE;
6745       }       }
# Line 6831  do { Line 6843  do {
6843    
6844     /* Other brackets */     /* Other brackets */
6845    
6846     else if (op == OP_ASSERT || op == OP_ONCE)     else if (op == OP_ASSERT || op == OP_ONCE || op == OP_ONCE_NC)
6847       {       {
6848       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;       if (!is_startline(scode, bracket_map, backref_map)) return FALSE;
6849       }       }
# Line 6901  do { Line 6913  do {
6913       case OP_SCBRAPOS:       case OP_SCBRAPOS:
6914       case OP_ASSERT:       case OP_ASSERT:
6915       case OP_ONCE:       case OP_ONCE:
6916         case OP_ONCE_NC:
6917       case OP_COND:       case OP_COND:
6918       if ((d = find_firstassertedchar(scode, op == OP_ASSERT)) < 0)       if ((d = find_firstassertedchar(scode, op == OP_ASSERT)) < 0)
6919         return -1;         return -1;
# Line 6910  do { Line 6923  do {
6923       case OP_EXACT:       case OP_EXACT:
6924       scode += 2;       scode += 2;
6925       /* Fall through */       /* Fall through */
6926    
6927       case OP_CHAR:       case OP_CHAR:
6928       case OP_PLUS:       case OP_PLUS:
6929       case OP_MINPLUS:       case OP_MINPLUS:
# Line 6923  do { Line 6936  do {
6936       case OP_EXACTI:       case OP_EXACTI:
6937       scode += 2;       scode += 2;
6938       /* Fall through */       /* Fall through */
6939    
6940       case OP_CHARI:       case OP_CHARI:
6941       case OP_PLUSI:       case OP_PLUSI:
6942       case OP_MINPLUSI:       case OP_MINPLUSI:
# Line 7083  utf8 = (options & PCRE_UTF8) != 0; Line 7096  utf8 = (options & PCRE_UTF8) != 0;
7096    
7097  /* 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
7098  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
7099  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
7100  not used here. */  not used here. */
7101    
7102  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 7116  if ((options & PCRE_UCP) != 0) Line 7129  if ((options & PCRE_UCP) != 0)
7129  if ((options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) ==  if ((options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) ==
7130       (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))       (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))
7131    {    {
7132    errorcode = ERR56;    errorcode = ERR56;
7133    goto PCRE_EARLY_ERROR_RETURN;    goto PCRE_EARLY_ERROR_RETURN;
7134    }    }
7135    
# Line 7202  outside can help speed up starting point Line 7215  outside can help speed up starting point
7215  ptr += skipatstart;  ptr += skipatstart;
7216  code = cworkspace;  code = cworkspace;
7217  *code = OP_BRA;  *code = OP_BRA;
7218  (void)compile_regex(cd->external_options, &code, &ptr, &errorcode, FALSE,  (void)compile_regex(cd->external_options, &code, &ptr, &errorcode, FALSE,
7219    FALSE, 0, 0, &firstbyte, &reqbyte, NULL, cd, &length);    FALSE, 0, 0, &firstbyte, &reqbyte, NULL, cd, &length);
7220  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;  if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;
7221    
# Line 7282  re->top_bracket = cd->bracount; Line 7295  re->top_bracket = cd->bracount;
7295  re->top_backref = cd->top_backref;  re->top_backref = cd->top_backref;
7296  re->flags = cd->external_flags;  re->flags = cd->external_flags;
7297    
7298  if (cd->had_accept) reqbyte = -1;   /* Must disable after (*ACCEPT) */  if (cd->had_accept) reqbyte = REQ_NONE;   /* Must disable after (*ACCEPT) */
7299    
7300  /* If not reached end of pattern on success, there's an excess bracket. */  /* If not reached end of pattern on success, there's an excess bracket. */
7301    

Legend:
Removed from v.643  
changed lines
  Added in v.723

  ViewVC Help
Powered by ViewVC 1.1.5