https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89149
--- Comment #3 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
That fold_build_pointer_plus_loc looks premature to me in any case given the
desire to delay folding. We have tons of those through the C as well as C++
FEs though.
E.g. pointer_int_sum calls:
fold_build2_loc (loc, MULT_EXPR
fold_build1_loc (loc, NEGATE_EXPR
fold_build_pointer_plus_loc
grep fold_build_pointer_plus c*/*.c | wc -l
24
I guess we should remove that all (though, we don't have e.g.
build_pointer_plus{,_loc} helpers at all) and verify that c_fully_fold_internal
and cp_fold handle those.
