Comment 334 for bug 204996

Status update: I've gotten started on finding the regression via git-bisect. I was able to reproduce the bug on 2.6.27 vanilla, and unable to do so on 2.6.22 vanilla. I'm currently down to just under seven thousand candidate commits left. The whole thing is rather complicated by the fact that configuration options change between kernel versions, which has made me a bit leery of the possibility of actually isolating the code change that introduces the bug.

Also, older kernel versions seem to cause X to die every so often, and even apart from that, testing is slow, because finding out if a kernel is bad or not is like the Halting Problem--I can only get a "yes" answer, never a definitive "no".