On Thu, Jul 31, 2025 at 8:11 AM Alexander Monakov <amona...@ispras.ru> wrote:
>
>
> On Thu, 31 Jul 2025, H.J. Lu wrote:
>
> > > We have a few pre-existing tests that verify that the attribute is
> > > honored,
> > > such as those added by commit 1d561e1851c4. Do they actually not fail with
> > > this patch?
> >
> > There are no regressions.
>
> Sorry, I don't see how that would be possible, that's literally the purpose of
> the half of those tests, named vis-*-gd.c. Is there some issue with the tests
> then?
decl_default_tls_model has
/* Local dynamic is inefficient when we're not combining the
parts of the address. */
else if (is_local && optimize_dyn_tls_for_decl_p (decl))
kind = TLS_MODEL_LOCAL_DYNAMIC;
else
kind = TLS_MODEL_GLOBAL_DYNAMIC;
(gdb) call optimize_dyn_tls_for_decl_p (decl)
$3 = false
(gdb)
There are no issues for those tests.
> Alexander
--
H.J.
