On Tue, Jan 26, 2016 at 03:46:50PM -0500, Jason Merrill wrote:
> On 01/26/2016 03:32 PM, Jakub Jelinek wrote:
> >>>+if (CHECKING_P)
> >>>+ verify_constructor_flags (t);
> >>>+else
> >>>+ recompute_constructor_flags (t);
>
> >But I don't understand this. Either the flags are supp
On 01/26/2016 03:32 PM, Jakub Jelinek wrote:
>+ if (CHECKING_P)
>+ verify_constructor_flags (t);
>+ else
>+ recompute_constructor_flags (t);
But I don't understand this. Either the flags are supposed to be already
correct here, then I'd expect to see
if (CHECKING_P)
On Tue, Jan 26, 2016 at 03:20:04PM -0500, Jason Merrill wrote:
> Tested x86_64-pc-linux-gnu. Are the tree.c changes OK for trunk?
The tree.c changes are ok. But I have nits and one bigger issue in
constexpr.c:
> --- a/gcc/cp/constexpr.c
> +++ b/gcc/cp/constexpr.c
> @@ -2214,6 +2214,9 @@ cxx_eval
The problem in this bug was that the constexpr code builds a lot of
CONSTRUCTORs and then fills in the elements later without ever going
back and updating TREE_CONSTANT and TREE_SIDE_EFFECTS.
This patch adds middle end functions recompute_constructor_flags and
verify_constructor_flags, and fix
