On Wed, Jan 16, 2019 at 12:50:23PM +0100, Richard Biener wrote:
> I guess so. There's not much we can do about this other than making
> DECL_MODE dynamic the same way as TYPE_MODE. I still believe
> this is the wrong direction and instead RTL expansion should properly
> adjust DECL_RTL and generally on GIMPLE we shouldn't need to
> commit to modes early at all...
Are there cases where DECL_MODE (decl) should not be equal to
TYPE_MODE (TREE_TYPE (decl))? Is that for structs with flexible array
members where the initializer is filling up the flexible array member, or
something different?
E.g. in expand_debug* it shouldn't be hard to use TYPE_MODE if
VECTOR_TYPE_P (TREE_TYPE (decl)), but there are many other uses of
DECL_MODE.
Jakub
