[Bug debug/81936] ICE in dwarf2out_die_ref_for_decl, at dwarf2out.c:5543

[rguenther at suse dot de]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81936

--- Comment #14 from rguenther at suse dot de <rguenther at suse dot de> ---
On August 23, 2017 6:52:40 PM GMT+02:00, "vries at gcc dot gnu.org"
<gcc-bugzilla@gcc.gnu.org> wrote:
>https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81936
>
>--- Comment #13 from Tom de Vries <vries at gcc dot gnu.org> ---
>I've tried to see if I could workaround the problem by disabling the
>assert:
>...
>diff --git a/gcc/dwarf2out.c b/gcc/dwarf2out.c
>index 0c339bd46f0..b2df3b89c2e 100644
>--- a/gcc/dwarf2out.c
>+++ b/gcc/dwarf2out.c
>@@ -5538,9 +5538,12 @@ dwarf2out_die_ref_for_decl (tree decl, const
>char **sym,
>   while (die->die_parent)
>     die = die->die_parent;
>   /* For the containing CU DIE we compute a die_symbol in
>-     compute_comp_unit_symbol.  */
>+     compute_comp_unit_symbol.
>+     TODO: Remove flag_generate_offload from assert condition, and fix
>+     fallout (PR81936).
>+  */
>   gcc_assert (die->die_tag == DW_TAG_compile_unit
>-             && die->die_id.die_symbol != NULL);
>+             && (die->die_id.die_symbol != NULL ||
>flag_generate_offload));
>   *sym = die->die_id.die_symbol;
>   return true;
> }
>...
>
>That worked for pr66714.c, but for all other ICE-ing test-cases I run
>into
>another ICE, now in mkoffload:
>...
>      else
>        {
>          /* Pickle stray references.  */
>          t = lto_input_tree_1 (&ib_main, data_in, tag, 0);
>      gcc_assert (t && data_in->reader_cache->nodes.length () == from);
>        }
>...
That looks odd, I'll see if I can reproduce tomorrow.