On Mon, Jun 25, 2012 at 2:41 AM, Avi Bab <a...@trustwave.com> wrote: > > > > > I am using libevent-2.0.18-stable running on Liux (CentOs 6.2). > > Adding an event with a common timeout I get an assertion failure in > insert_common_timeout_inorder (event.c:2532): > “EVUTIL_ASSERT(is_same_common_timeout(&e->ev_timeout, &ev->ev_timeout));”
A nice bunch of data; that should help track it down! The assertion is triggering because it's trying to assert an event with a common timeout into one of the common timeout queues ... but one of the events in that queue had a different timeout, which is supposed to be impossible, since each queue only has events of a single common timeout! If your gdb results are right, it appears that the event in the queue has a timeout of 0,0: It looks like it was cleared without actually removing it from the queue! My guess here is that something re-set one of your pending events (e.g. by calling event_assign or event_set on it) without first calling event_del() to remove it from the event_base. There's actually a feature to try to detect this mistake for you: check out the documentation for event_enable_debug_mode(): it should detect that bug (if that's what's happening) at the point where the erroneous event_assign() is happening. If you're on Linux, you could also try running under valgrind; that would detect other kinds of errors that could be causing this behavior. yrs, -- Nick *********************************************************************** To unsubscribe, send an e-mail to majord...@freehaven.net with unsubscribe libevent-users in the body.
