More questions related to gimplification

[Nathanael Nerode]

Is it possible to have a MODIFY_EXPR going into gimplify_modify_expr,
such that
(1) The LHS is_gimple_reg_type (TREE_TYPE(lhs))
(2) The RHS, after gimplification, is *not*
or
(1) The RHS, after gimplification, is_gimple_reg_type (TREE_TYPE(lhs))
(2) The LHS is *not*

If this is indeed supposed to be impossible,
(a) could it be reasonable to add a sanity check for it into
gimplify_modify_expr?
(b) this would allow some logic simplification

If it is actually supposed to be possible, under what circumstances could it
happen?

(Note that is_gimple_reg_type means !AGGREGATE_TYPE_P && != COMPLEX_TYPE.)

-- 
This space intentionally left blank.