On Tue, 2011-05-17 at 12:13 +0100, Colin Guthrie wrote: > Thanks, I'll test this patch and see if it helps with my test case. > > Assuming that it does, is this something we should really push to alsa > lib? When I first encountered the problem I did start hacking away at > alsa lib looking for the problem. > > Perhaps a new direction argument -2 and +2 for "nearest below bias" and > "nearest above bias" respectively? (the bias only coming into play when > there you are exactly inbetween two possible values). Nah, if we need rounding to the nearest step only for working around rounding errors that are caused by ourselves due to not-entirely-necessary format conversions, I don't think alsa-lib should be changed just for that. -- Tanu
