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

Diff of /code/trunk/pcre_compile.c

Parent Directory | Revision Log | View Patch Patch

-revision 333 by ph10,
Thu Apr 10 19:55:57 2008 UTC
+revision 335 by ph10,
Sat Apr 12 14:36:14 2008 UTC
 Line 1567 
 for (code = first_significant_code(code
    /* Groups with zero repeats can of course be empty; skip them. */
-   if (c == OP_BRAZERO || c == OP_BRAMINZERO)
+   if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)
      {
      code += _pcre_OP_lengths[c];
      do code += GET(code, 1); while (*code == OP_ALT);
 Line 1847 
 return -1;
  that is referenced. This means that groups can be replicated for fixed
  repetition simply by copying (because the recursion is allowed to refer to
  earlier groups that are outside the current group). However, when a group is
- 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
- 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
- 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
- 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
- the partially compiled regex must be temporarily terminated with OP_END.
+ is called, the partially compiled regex must be temporarily terminated with
+ OP_END.
  This function has been extended with the possibility of forward references for
  recursions and subroutine calls. It must also check the list of such references
-Line 1882 
 while ((ptr = (uschar *)find_recurse(ptr
+Line 1883 
 while ((ptr = (uschar *)find_recurse(ptr
    /* See if this recursion is on the forward reference list. If so, adjust the
    reference. */
    for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
      {
      offset = GET(hc, 0);
-Line 2458 
 for (;; ptr++)
+Line 2459 
 for (;; ptr++)
    /* Get next byte in the pattern */
    c = *ptr;
    /* If we are in the pre-compile phase, accumulate the length used for the
    previous cycle of this loop. */
-Line 3842 
 we set the flag only if there is a liter
+Line 3843 
 we set the flag only if there is a liter
        if (repeat_min == 0)
          {
-         /* 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
-         altogether. */
+         output altogether, like this:
-         if (repeat_max == 0)
-           {
-           code = previous;
-           goto END_REPEAT;
-           }
-         /* If the maximum is 1 or unlimited, we just have to stick in the
+         ** if (repeat_max == 0)
-         BRAZERO and do no more at this point. However, we do need to adjust
+         **   {
-         any OP_RECURSE calls inside the group that refer to the group itself or
+         **   code = previous;
-         any internal or forward referenced group, because the offset is from
+         **   goto END_REPEAT;
-         the start of the whole regex. Temporarily terminate the pattern while
+         **   }
-         doing this. */
+         However, that fails when a group is referenced as a subroutine from
+         elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it
+         so that it is skipped on execution. As we don't have a list of which
+         groups are referenced, we cannot do this selectively.
+         If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
+         and do no more at this point. However, we do need to adjust any
+         OP_RECURSE calls inside the group that refer to the group itself or any
+         internal or forward referenced group, because the offset is from the
+         start of the whole regex. Temporarily terminate the pattern while doing
+         this. */
-         if (repeat_max <= 1)
+         if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
            {
            *code = OP_END;
            adjust_recurse(previous, 1, utf8, cd, save_hwm);
            memmove(previous+1, previous, len);
            code++;
+           if (repeat_max == 0)
+             {
+             *previous++ = OP_SKIPZERO;
+             goto END_REPEAT;
+             }
            *previous++ = OP_BRAZERO + repeat_type;
            }
-Line 4138 
 we set the flag only if there is a liter
+Line 4149 
 we set the flag only if there is a liter
      bravalue = OP_CBRA;
      save_hwm = cd->hwm;
      reset_bracount = FALSE;
      /* First deal with various "verbs" that can be introduced by '*'. */
      if (*(++ptr) == '*' && (cd->ctypes[ptr[1]] & ctype_letter) != 0)
-Line 5127 
 we set the flag only if there is a liter
+Line 5138 
 we set the flag only if there is a liter
        -ESC_g is returned only for these cases. So we don't need to check for <
        or ' if the value is -ESC_g. For the Perl syntax \g{n} the value is
        -ESC_REF+n, and for the Perl syntax \g{name} the result is -ESC_k (as
-       that is a synonym). */
+       that is a synonym for a named back reference). */
        if (-c == ESC_g)
          {
          const uschar *p;
+         save_hwm = cd->hwm;   /* Normally this is set when '(' is read */
          terminator = (*(++ptr) == '<')? '>' : '\'';
          /* These two statements stop the compiler for warning about possibly
-Line 6197 
 while (errorcode == 0 && cd->hwm > cwork
+Line 6209 
 while (errorcode == 0 && cd->hwm > cwork
    if (groupptr == NULL) errorcode = ERR53;
      else PUT(((uschar *)codestart), offset, groupptr - codestart);
    }
  /* Give an error if there's back reference to a non-existent capturing
  subpattern. */

 Legend:



Removed from v.333
 


changed lines


 
Added in v.335
 Legend:



Removed from v.333
 


changed lines


 
Added in v.335
-Removed from v.333
+Added in v.335

webmaster@exim.org	ViewVC Help
Powered by ViewVC 1.1.12