On 01/05/2012 04:08 AM, MORITA Kazutaka wrote: > It is not good to make a conditional branch based on the drvier name > here. > > How about adding stat and jrnl_begin/end/recover to the store driver > interface? Then, we can remove get_store_dir(). > This is just a dirty hack that is supposed to removed later when we add more backend drivers. If we really need to address it right now, how about add a new hook for get_store_dir()? This is kind of dirty & strange too, asking underlying backend where is object stored and I think only simple store need to implement it, not a generic interface too me. > > IIUC, the farm driver updates epoch in farm_begin_recover(). Why do > the two drivers update it in the different places? > > Thanks, > > Kazutaka Ah, yeah. Thanks, Yuan |