> Hi!
>
> On Fri, 6 Nov 2015 07:07:27 +0100, Jan Hubicka <hubi...@ucw.cz> wrote:
> > this patch fixes tripple thinko when in
> > ipa_polymorphic_call_context::restrict_to_inner_type when dealing with an
> > offset that is out of the range of the type considered. In this case
> > function
> > should return true only when type is dynamic (so there may be additional
> > type
> > after the known type) or derivations are allowed (so the type may get
> > bigger).
> > There is check that is supposed to make this happen, but it clears the flags
> > before checking htem that is not a good idea.
> >
> > There is also check in contains_type_p that is supposed to shortcut this
> > scenario,
> > but instead of checking TYPE_SIZE it checks type itself (ouch).
> >
> > Bootstrapped/regtested x86_64-linux, will commit it shortly.
>
> > --- testsuite/g++.dg/lto/pr68057_0.C (revision 0)
> > +++ testsuite/g++.dg/lto/pr68057_0.C (revision 0)
> > @@ -0,0 +1,23 @@
> > +// { dg-lto-do compile }
> > +/* { dg-extra-ld-options { -O2 -Wno-odr -r -nostdlib } } */
>
> I'm seeing "WARNING: lto.exp does not support dg-lto-do compile" as well
> as a few FAILs:
Ah, sorry, I must have messed up the testcase. Will fix it ASAP.
Honza
