On 09/01/13 17:25, Claudio Carbone wrote:
> I traced my program through all the function
> zmq::socket_base_t::connect
>
> and it goes through it flawlessly until line 556
> add_endpoint (addr_, (own_t *) session);
>
> which translates to
> launch_child (endpoint_);
> endpoints.insert (endpoints_t::value_type (std::string (addr_),
> endpoint_));
>
> the second is the one that triggers the assert, although the code is
> alien to me
> _GLIBCXX_CONSTEXPR pair(const _T1& __a, const _T2& __b)
> : first(__a), second(__b) { }
>
> I don't know enough about this to understand what is happening.
> The address that is being passed though is
> epgm://192.168.2.113;239.192.1.1:5678
>
>
Ok maybe I'm going somewhere: it may be failing on taking ownership of
the thread
void zmq::own_t::launch_child (own_t *object_)
{
// Specify the owner of the object.
object_->set_owner (this);
// Plug the object into the I/O thread.
send_plug (object_);
// Take ownership of the object.
send_own (this, object_);
}
But if that's the case then there are two questions:
- how do I use multi threading correctly to avoid this?
- why it doesn't fail when run normally? ft the problem is that the
debugger somehow manipulates thread ownerships, then how do you debug
such an application?
Claudio
_______________________________________________
zeromq-dev mailing list
[email protected]
http://lists.zeromq.org/mailman/listinfo/zeromq-dev
