On Sat, Oct 30, 2010 at 12:03 PM, Ralph Castain <r...@open-mpi.org> wrote: > Just wondering: would it make sense to provide a libevent API for this?
Yes, what kind of API did you have in mind? The easiest would be something to implement bufferevent_flush(BEV_FINSISHED) for sockets, and define it to a) write as much as possible, and b) if all data is written, shutdown() the socket. But it seems that what people usually want instead is something that will start writing, wait till all data is written, then shutdown() the socket. I'm a little less thrilled with writing one of these, since it's just a wrapper around other functionality, but it does seem to be the kind of thing that people. It should probably be called something like ... well, bufferevent_flush() is taken. bufferevent_flush_and_shutdown(), maybe ? It should call the event callback with a new event type, BEV_SENT_CLOSE, when it's finished. It should also respect write timeouts. Does anybody care strongly about this to try writing the code and tests? > I ask because proper shutdown is important to a number of us, and it seems a > shame that we all have to independently sprinkle the code throughout our > programs wherever sockets are used. On the surface, at least, it would appear > that a simple API could take care of this for us. Sure; it would be good to have something like this in 2.1. Somebody should write one. If it's _totally noninvasive_, well tested, and we can get it done inside the next week or so, it might even work out for 2.0. yrs, -- Nick *********************************************************************** To unsubscribe, send an e-mail to majord...@freehaven.net with unsubscribe libevent-users in the body.
