[Bug c++/126483] [13/14/15/16/17 Regression] Wrong active union member in constant evaluation of variable template initializers since r13-6422

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=126483

--- Comment #3 from GCC Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Patrick Palka <[email protected]>:

https://gcc.gnu.org/g:9b0341a7a6a38912e667982003bcbfebdfb86335

commit r17-3303-g9b0341a7a6a38912e667982003bcbfebdfb86335
Author: Patrick Palka <[email protected]>
Date:   Fri Aug 14 15:53:10 2026 -0400

    c++: testsuite: Add some opt/is_constant_evaluated3.C variants

    r13-6422 added opt/is_constant_evaluated3.C to verify that we constant
    fold copy-init such as a1.  (Later, r14-6506 enabled constant folding of
    the a2-a5 inits.)  r13-6422 turns out to trigger a latent bug and causes
    PR c++/126483, premature folding of is_constant_evaluated in copy-init
    inside a lambda scope.

    This patch adds a couple extra variants of this test, one where the
    containing function is constexpr (3a) and one where it's a lambda (3b).

    The new 3a.C demonstrates however that we still don't constant fold a1's
    copy-init when inside a constexpr function scope.  3b.C demonstrates that
    we do fold it when inside a (maybe-constexpr) lambda scope.

    gcc/testsuite/ChangeLog:

            * g++.dg/opt/is_constant_evaluated3a.C: New test.
            * g++.dg/opt/is_constant_evaluated3b.C: New test.

    Reviewed-by: Jason Merrill <[email protected]>