https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93102
--- Comment #3 from Richard Biener <rguenth at gcc dot gnu.org> --- Implementation-wise the issue is we're committing too early to piecewise init of C00002029 during gimplification where IMHO we should always emit an aggregate copy C00002029 = *.LC0; which would allow optimizers to elide C00002029 in favor of *.LC0.