On Wed, Dec 13, 2017 at 12:54 PM, Martin Sebor <mse...@gmail.com> wrote:
> The attached update also fixes both instances of the ICE
> reported in bug 83322 and supersedes Jakub's patch for that
> bug (https://gcc.gnu.org/ml/gcc-patches/2017-12/msg00765.html).
> This passes bootstrap on x86_64 with no new regressions (there
> are an increasing number of failures on trunk at the moment
> but, AFAICS, none caused by this patch).
>
> Jason, I'm still trying to come up with a test case for templates
> that would illustrate the issue you're concerned about. If you
> have one that would be great (preferably one showing a regression).
I looked at the case I was concerned about, and found that it isn't an
issue because in that case we call duplicate_decls before applying
attributes.
But it looks like we'll still get this testcase wrong, because the
code assumes that if the old decl is a single _DECL, it must match.
[[gnu::noinline]] void f() { }
[[gnu::always_inline]] void f(int) { } // OK, not the same function
I think the answer is to use Nathan's new iterators unconditionally,
probably lkp_iterator.
Jason
