Re: [boost] [atomic] Constraints on failure_order?