On Mon, Feb 18, 2013 at 8:58 PM, JiaQiang Xu <xjqkill...@gmail.com> wrote: > 2013/2/19 Nick Mathewson <ni...@freehaven.net>: >> On Mon, Feb 18, 2013 at 9:01 AM, JiaQiang Xu <xjqkill...@gmail.com> wrote: >>> Hi, >>> >>> I'm using libevent to write an http server for transferring files. I >>> use openssl for https. >> >> Have you tried looking at the BEV_OPT_DEFER_CALLBACKS option? Its >> main role it to prevent infinite chains of recursive callbacks like >> this. >> > > I'm going to try this. Would setting this option affect the > functionality of libevent openssl?
It shouldn't hurt anything. > And if this is need to make it work, why not set it in libevent > openssl by default? It is only needed to make it work in some cases, and when it's not needed, it adds an extra layer of indirection. At the time, I was worried about performance there. I doubt that layer would matter on openssl benchmarks, but I don't know that anybody's checked to make sure. I'm hoping that in 2.1 I (or someone else!) can use the new generic callback infrastructure so that the DEFER_CALLBACKS flags become generally unneeded, and can be replaced by an "avoid reentrant callback invocation" mechanism. -- Nick *********************************************************************** To unsubscribe, send an e-mail to majord...@freehaven.net with unsubscribe libevent-users in the body.
