On Mon, Apr 18, 2016 at 02:58:51PM -0400, Jason Merrill wrote:
> My change to type_has_constexpr_default_constructor broke build_vec_init,
> which asserts that it implies that default-initialization is the same as
> value-initialization, which is not the case for the "maybe constexpr" case.
> So this patch reverts the change and adds a new "maybe constexpr" entry
> point when we don't need a firm answer.
>
> Tested x86_64-pc-linux-gnu, applying to trunk. OK for 6.1?
Ok, thanks.
> commit 695043a6eb769f6510c4d495dacd5d70514c89ab
> Author: Jason Merrill <ja...@redhat.com>
> Date: Mon Apr 18 14:29:19 2016 -0400
>
> PR c++/70690
>
> PR c++/70528
> * class.c (type_maybe_constexpr_default_constructor): New.
> (type_has_constexpr_default_constructor): Revert.
Jakub
