[Bug lto/59723] [4.9 Regression] ICE: in lto_output_tree, at lto-streamer-out.c:1390 when compiling some Fortran tests with -flto

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

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59723

--- Comment #11 from Aldy Hernandez <aldyh at gcc dot gnu.org> ---
Created attachment 31807
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=31807&action=edit
preliminary patch

I wonder why we can't do what we do for IMPORTED_DECL and avoid special casing
NAMELIST_DECL.  It looks like the constructor elements even get streamed
correctly this way.

This untested patch gets rid of the majority of the failures, although I still
see 2-3 errors in the dwarf code in tests like namelist_70.cf90:

FAIL: gfortran.dg/namelist_70.f90  -O3 -g  (internal compiler error)
FAIL: gfortran.dg/namelist_70.f90  -O3 -g  (test for excess errors)
Excess errors:
lto1: internal compiler error: in force_decl_die, at dwarf2out.c:20105
0x6b4319 force_decl_die
        /source/gcc/gcc/dwarf2out.c:20105
0x6b5b59 gen_namelist_decl
        /source/gcc/gcc/dwarf2out.c:20632
0x6b5379 gen_decl_die
        /source/gcc/gcc/dwarf2out.c:20435
0x6b391e process_scope_var
        /source/gcc/gcc/dwarf2out.c:19944
0x6b399a decls_for_scope
        /source/gcc/gcc/dwarf2out.c:19969
0x6ae9ac gen_subprogram_die
        /source/gcc/gcc/dwarf2out.c:18354
etc

I'll continue looking at this later, but comments welcome :).