On Tue, Mar 12, 2013 at 09:08:21AM +0100, Holger Hans Peter Freyther wrote: > from a quick look it appears to be that a very long running connection > could overflow lchan->s to a negative number. Could you either make this > code robust or explain why it is not needed? E.g. if we assume that this > happens once per multiframe the counter will overflow within 1.3 > hours?
clang lately got a nice feature for integer over/under-flow detection[1] during runtime, but of course it makes the resulting code extremely slow. I am not sure if it is usable in this case. Kind regards, -Alexander Huemer [1] http://blog.regehr.org/archives/905
