https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89900
--- Comment #2 from Paolo Carlini <paolo.carlini at oracle dot com> ---
I think that ultimately this boils down to this code in grokdeclarator:
if (type_was_error_mark_node && template_parm_flag)
/* FIXME we should be able to propagate the error_mark_node as is
for other contexts too. */
type = error_mark_node;
else
type = integer_type_node;
where we use integer_type_node as a catch-all type in case of error. Certainly
using error_mark_node avoids this ICE, and many other I suppose, but we are not
ready yet to resolve the FIXME (what about GCC-10?!?). We may want to fix this
specific ICE in a different way, for the time being.
