On Mon, May 16, 2016 at 11:37 AM, Andy Lutomirski <l...@amacapital.net> wrote: > > All of those fixes were intended to fix incorrect times being > reported, not segfaults. Weird.
I'm assuming it's "time going backwards". I can easily see that causing segfaults. I've seen lots of code that timestamps events, and can easily imagine confusion if the end result is not ordered (ie walking off the beginning/end of a list or array or similar because the algorithm "knows" that the events are ordered). I agree that the original bisection result is a bit questionable, and it might just be about exposing a timing issue. Linus