There is a patch pending on this particular issue which I submitted last
week.
https://issues.apache.org/jira/browse/ZOOKEEPER-17
I only changed the docs though, not the semantics of the method. Could
you add your comments/suggestions there? Perhaps the patch should be
changed, I'm not sure, I really just tried to make the comments reflect
the current state of the api.
Patrick
Austin Shoemaker wrote:
The documentation in zookeeper.h states:
* clientid the id of a previously established session that this
* client will be reconnecting to. Pass 0 if not reconnecting to a
previous
* session. If the session timed out or the id is invalid, a new
* session will be automatically generated. Clients should check the
actual
* session id by calling \ref zoo_client_id.
The implementation of zookeeper_init appears to enter
EXPIRED_SESSION_STATE instead of automatically generating a new session id.
Would it be safe to remove this block in zookeeper.c check_events():
if (oldid != 0 && oldid != newid) {
zh->state = EXPIRED_SESSION_STATE;
errno = ESTALE;
return handle_socket_error_msg(zh,__LINE__,ZSESSIONEXPIRED,
"session %llx has expired.",oldid);
} else {
so that the new session continues to CONNECTED_STATE? It would be useful
to receive a session expired notification without aborting the connection.
Thanks,
Austin
