https://gcc.gnu.org/bugzilla/show_bug.cgi?id=121593
--- Comment #13 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to Marc Mutz from comment #12)
> > Lemme throw ubsan/asan on the noinline code.
>
> Passes, no report :(
>
> > Does -fno-ipa-modref work?
> > Does -fno-strict-alias work?
>
> Both (independent of each other) fix the issue, yes.
That is good news and bad news.
The good news is that we now know there is an aliasing issue (could be still a
gcc internal aliasing bug) but now to pin point it. I will take a look tomorrow
some more. I might be able to figure out if there is some undefined code still
or some bug in gcc.
>
> Lemme try a C++20 build to exclude `enum class qchar8_t : uchar {};`.
