https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81897
--- Comment #7 from Jeffrey A. Law <law at redhat dot com> --- So just a note. We *could* pick this up without waiting on Aldy's work. After the second DOM pass we're failing to merge a pair of blocks because there are still SSA_NAMEs queued for renaming. If we were to manage to get those blocks merged, then the CFG would be cleaned up enough that the second VRP pass would see the final jump thread and ultimately eliminate the unexecutable path. I'm not actually working on this.