https://gcc.gnu.org/bugzilla/show_bug.cgi?id=124632
--- Comment #3 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Note, trying the same testcase with gcc 15, I do see ggc_freed stuff in
constexpr_call_table->m_entries[N]->bindings at fini_constexpr time:
<tree_vec 0x7fffe99f1b90 length:2 elt:0 <ggc_freed 0x7fffe99deab0>
elt:1 <integer_cst 0x7fffe9820a20 type <integer_type 0x7fffe98285e8 int>
constant 4>>
So to me this looks like a previously latent problem, it depends on how lucky
one is with the hashing, whether such entries are compared or not found when
searching for a hit.
