Re: [PATCH] PR c++/80473 allow suppressing notes about over-aligned new

[Jonathan Wakely <jwakely at redhat dot com>]

On 20/04/17 17:33 +0200, Marek Polacek wrote:
> On Thu, Apr 20, 2017 at 05:22:00PM +0200, Jakub Jelinek wrote:
> > On Thu, Apr 20, 2017 at 04:09:20PM +0100, Jonathan Wakely wrote:
> > > --- a/gcc/cp/init.c
> > > +++ b/gcc/cp/init.c
> > > @@ -3128,11 +3128,14 @@ build_new_1 (vec<tree, va_gc> **placement, tree type, tree nelts,
> > >      {
> > >        warning (OPT_Waligned_new_, "%<new%> of type %qT with extended "
> > >  	       "alignment %d", elt_type, TYPE_ALIGN_UNIT (elt_type));
> > > -      inform (input_location, "uses %qD, which does not have an alignment "
> > > -	      "parameter", alloc_fn);
> > > -      if (!aligned_new_threshold)
> > > -	inform (input_location, "use %<-faligned-new%> to enable C++17 "
> > > -				"over-aligned new support");
> > > +      if (diagnostic_report_warnings_p (global_dc, input_location))
> > > +	{
> > > +	  inform (input_location, "uses %qD, which does not have an alignment "
> > > +		  "parameter", alloc_fn);
> > > +	  if (!aligned_new_threshold)
> > > +	    inform (input_location, "use %<-faligned-new%> to enable C++17 "
> > > +				    "over-aligned new support");
> > > +	}
> >
> > This looks weird.  I'd expect instead:
> >       if (warning (OPT_Waligned_new_, "%<new%> of type %qT with extended "
> > 		   "alignment %d", elt_type, TYPE_ALIGN_UNIT (elt_type)))
> > 	{
> > 	  inform (input_location, "uses %qD, which does not have an alignment "
> > 		  "parameter", alloc_fn);
> > 	  if (!aligned_new_threshold)
> > 	    inform (input_location, "use %<-faligned-new%> to enable C++17 "
> > 				    "over-aligned new support");
> > 	}
> > That is a standard idiom used if some inform or later warning/error depends
> > on whether earlier warning/error has been diagnosed.
>
> Yes.
>
> > If that works, this is ok for trunk and 7.1 (we don't have a rc1 yet, it is
> > ok now).
>
> One more thing, the test passes even without the patch.  Did you mean to add
> // { dg-options "-Wall -w" }
> ?

Good catch. My original testcase used <type_traits> but I changed it
to not depend on the library, and didn't fix the dg-options.