On Fri, Aug 11, 2017 at 6:05 AM, Pierre-Marie de Rodat <dero...@adacore.com> wrote: > On 08/07/2017 08:42 PM, Jason Merrill wrote: >> >> Perhaps the DECL_ABSTRACT_ORIGIN check in my patch should be >> decl_ultimate_origin instead, which should return null in that case? > > > It looks like it worked, thanks! Note that I had to tweak a bit your change > in process_scope_vars to avoid a crash in a couple of Fortran tests (at -O3 > -g). Also, as the gcc_unreachable appears in an IF block, I turned it into a > gcc_assert and remove the corresponding IF. > > Here is the final patch version, bootstrapped and regtested on x86_64-linux.
OK. Jason