https://gcc.gnu.org/bugzilla/show_bug.cgi?id=119534
--- Comment #8 from Jakub Jelinek <jakub at gcc dot gnu.org> --- vector(8) <signed-boolean:1> mask__2.12; <signed-boolean:1> _72; <signed-boolean:1> _78; _72 = BIT_FIELD_REF <mask__2.12_63, 8, 0>; pr119534.c:1:6: error: integral result type precision does not match field size of ‘bit_field_ref’ _78 = BIT_FIELD_REF <mask__2.12_63, 8, 8>;