Hello, About a year ago, there was a thread about deadlocking in bufferevent_openssl in multi-threaded mode ( http://archives.seul.org/libevent/users/Jan-2011/msg00019.html).
We've run into the same problem. Our minimal fix (as opposed to the patch mentioned, which adds atomic operations to evthread) for our specific use case was this : - expose event_base_lock() and event_base_unlock() to library users - in the user code make sure to lock the event_base before writing to an openssl bufferevent, so no callbacks can be fired during that write (which results in a deadlock). Does anybody have feedback on better ways to workaround this? Hi Nick, would you mind sharing the changes you had in mind? Thanks, Amarin
