Re: PATCH to avoid bogus warning

[Jason Merrill <jason at cygnus dot com>]

>>>>> Mark Mitchell <mark@markmitchell.com> writes:

 > Well, it wasn't quite that simple.  I bit the bullet and cleaned up
 > the code to generate these warnings a bit, with the following patch,
 > which also allowed localizing the warning code.

Great, thanks.

 > !   if (warn_ctor_dtor_privacy
 > !       /* If the class has friends, those entities might create and
 > ! 	 access instances, so we should not warn.  */
 > !       && !(CLASSTYPE_FRIEND_CLASSES (t)
 > ! 	   || DECL_FRIENDLIST (TYPE_MAIN_DECL (t)))
 > !       /* We will have warned when the template was declared; there's
 > ! 	 no need to warn on every instantiation.  */
 > !       && !CLASSTYPE_TEMPLATE_INSTANTIATION (t))
 > !     {

I generally prefer the 

  if (! interesting)
    return;

style to

  if (interesting)
    {
      /* big ol' block */;
    }

and similarly for break in loops, to hold down the nesting level.

 > ! 	  for (i = 0; i < CLASSTYPE_N_BASECLASSES (t); i++)
 > ! 	    if (TREE_VIA_PUBLIC (TREE_VEC_ELT (binfos, i))
 > ! 		|| TREE_VIA_PROTECTED (TREE_VEC_ELT (binfos, i)))
 > ! 	      {
 > ! 		has_nonprivate_method = 1;
 > ! 		break;
 > ! 	      }

Seems odd that we would count a public base as suppressing this warning.  I
wonder why that was there.

 > ! 	  /* If a non-template class does not define a copy
 > ! 	     constructor, one is defined for it, enabling it to avoid
 > ! 	     this warning.  For a template class, this does not
 > ! 	     happen, and so we would normally get a warning on:
 > ! 
 > ! 	       template <class T> class C { private: C(); };  
 > ! 	  
 > ! 	     To avoid this asymmetry, we check TYPE_HAS_INIT_REF.  */ 
 > ! 	  if (!TYPE_HAS_INIT_REF (t))
 > ! 	    nonprivate_ctor = 1;

Perhaps the comment should clarify that TYPE_HAS_INIT_REF is always true
for a complete class these days.

Jason