[Bug lto/84213] [8 Regression] 521.wrf_r from SPEC 2017 fails to build (link) with LTO

["rguenth at gcc dot gnu.org" <gcc-bugzilla at gcc dot gnu dot org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84213

--- Comment #6 from Richard Biener <rguenth at gcc dot gnu.org> ---
In the Fortran case the DECL_VALUE_EXPR is
COMPONENT_REF<module_ra_gfdleta.eq.0, em1v>
and rtl_for_decl_location gets to a location via loc_list_from_tree
which eventually looks at DECL_VALUE_EXPR.

I'm quite sure the rev. is responsible for similar Ada -flto -g failures.  We
simply may _not_ add location attributes early.

Note adding a location refering to another entity makes this entity
non-removable.  Well - we kind of rely on late attribute pruning for
removed stuff here in the traditional model.  With LTO we'd need to
invent a new debug hook that tells the middle-end to perserve the
decl because it might have interesting debug info - but that _will_
cause refered stuff to be preserved.

So I think we need to revert this patch and instead make sure those
Ada variables do not get removed?  Like by marking them with
DECL_PRESERVE_P?

Trying that (figuring where those decls are created).  Pierre-Marie,
maybe you can help here as well.