https://gcc.gnu.org/bugzilla/show_bug.cgi?id=90424
--- Comment #10 from Richard Biener <rguenth at gcc dot gnu.org> ---
(In reply to Andrew Pinski from comment #9)
> Hmm, we get:
> BIT_INSERT_EXPR <r_5(D), _1, 0>;
>
> Since r_5 is unintialized, can't we just do:
> {_1, 0}
>
> ?
Yes. Though I always get nervous when replacing a UNDEF with something else...
it will for sure end up confusing some users, possibly elide uninit diagnostics
and make those "two undef reads should end up the same value" folks unhappy
(with "two undef reads" often being two SSA uses of possibly the same source
mention).
Won't RTL do this anyway, because we have that initregs pass?
