On Tue, Mar 17, 2026 at 01:06:11PM +0100, Richard Biener wrote:
> > I'd put it on the trunk, we still have at least a month to find issues in it
> > and it will be widely deployed till then to get real-world coverage.
>
> OK, I'll be prepared to eventually revert though when there is larger
> fallout that cannot be quickly addressed.
Sure. Note, my most important reason to want this in is that we've switched
to -std=gnu++20 and it affects the C++20 std::variant and I think it is
pretty normal to use multiple variants of the same variable in the same
function in different areas of the function. So it is more dangerous in GCC
16 to leave it unfixed compared to GCC 15.
Jakub
