emit_group_store vs complex

[Geoff Keating]

Richard Henderson <rth@cygnus.com> writes:

> The problem is that emit_group_store forces /s set, no matter
> what the incoming memory says it's supposed to be.  This is
> clearly a recipe for disaster.  And there's a comment that
> says that the reason we make that change is because store_bit_field
> will abort otherwise.  And as far as I can tell, there's no
> reason for store_bit_field to do that except to try to catch
> presumed bugs elsewhere in the compiler.

Hmmm.  Can anyone say what MEM_IN_STRUCT_P is _for_?

I can't think of anything, other than alias analysis which is surely
much better handled by the frontend.  What impact could
MEM_IN_STRUCT_P possibly have on the generated code?
-- 
- Geoffrey Keating <geoffk@cygnus.com>