------- Comment #1 from jakub at gcc dot gnu dot org 2007-11-13 12:12 ------- The set_packs_to_error stuff is very problematic in this case, because trees are shared and the C++ FE certainly doesn't expect error_mark_nodes to appear at random places. In this case T* type is shared (as the pointer to T has been cached). check_for_bare_parameter_packs should have errored about this twice, but will do only once because of the sharing and in the second case just won't fail, so doesn't let the caller to deal with the errorneous type.
-- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34056
