https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79463
--- Comment #4 from Jakub Jelinek <jakub at gcc dot gnu.org> --- Yeah. In this particular case it is only the check_abi_tags that assumes a VAR_DECL has always non-NULL TREE_TYPE, if I return in gdb from there early the result looks reasonable (likely because we now have early dwarf and the fact that a has type A has been emitted earlier.
