http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59322
--- Comment #4 from Jeffrey A. Law <law at redhat dot com> --- Thanks for the bug report. It hits a hunk of code which I've looked at a few times in the last couple months and haven't been able to convince myself it's correct or desirable. In create_edge_and_update_destination_phis, we clone a jump threading path onto the newly created edge. That path has a reference to an edge that later gets removed. Then we look at the path and fault.
