On 9/3/26 4:58 PM, Patrick Palka wrote:
On Fri, 31 Jul 2026, Marek Polacek wrote:
On Fri, Jul 31, 2026 at 01:58:15PM -0400, Jason Merrill wrote:
On 7/31/26 1:49 PM, Patrick Palka wrote:
On Fri, 24 Jul 2026, Marek Polacek wrote:
Bootstrapped/regtested on x86_64-pc-linux-gnu, ok for trunk?
-- >8 --
This is a wrong-code problem starting with the recent check_initializer
simplification (r17-1661). I thought the fix would be to bring some of
those dropped conditions back, but now I think the change just uncovered
a latent bug.
FWIW it seems the aforementioned latent bug causes PR126472 and your
patch fixes it on trunk! I wonder if backporting this patch to the 16
branch would make sense without r17-1661?
Nice.
I think so, but let's wait a bit to see if the assert breaks anything.
Sounds good, I'll wait 1-2 weeks. For trunk, I think we might want to
add the https://gcc.gnu.org/bugzilla/show_bug.cgi?id=126472#c2 test.
Marek
Maybe we could backport this fix to older release branches as well, since the
original non-preprocessed testcase from PR126472 works with 10.5 and 11.2 but
not with 11.3 or 12+ (after an std::optional refactoring that triggers the front
end bug), so I guess it's a regression from the libstdc++ side necessitating
the front end fix?
Makes sense.
Jason
