Re: [PATCH] cgraph: Don't verify semantic_interposition flag for aliases [PR105399]

[Jan Hubicka <hubicka@kam.mff.cuni.cz>]

> On Thu, Apr 28, 2022 at 01:54:51PM +0200, Jan Hubicka wrote:
> > > --- gcc/cgraph.cc.jj	2022-04-20 09:24:12.194579146 +0200
> > > +++ gcc/cgraph.cc	2022-04-27 11:53:52.102173154 +0200
> > > @@ -3488,7 +3488,9 @@ cgraph_node::verify_node (void)
> > >  	     "returns a pointer");
> > >        error_found = true;
> > >      }
> > > -  if (definition && externally_visible
> > > +  if (definition
> > > +      && externally_visible
> > > +      && (!alias || thunk || !in_lto_p)
> > 
> > The alias check seems OK to me.
> > Why we need thunk?  Here thunk may eventually gain a gimple body and we
> > will be interested in its opt_for_fn flags.  So if we do not set
> > DECL_FUNCTION_SPECIFIC_OPTIMIZATION we may get unexpected surprises
> > (such as blocked inlining).
> 
> I've added || thunk because free_lang_data sets DECL_FUNCTION_SPECIFIC_OPTIMIZATION
> for thunks, the guarding condition is:
> if (gimple_has_body_p (decl) || (node && node->thunk))
> I wasn't sure if node->alias and node->thunk can be both set or not.
> If they can't, I can take it out.

Aha, I misread is as !alias || !thunk.
But node can not be both alias and thunk, so I would leave it out.

Honza

> 
> 	Jakub
>