On 02/29/2012 05:12 PM, Henrik NordstrÃ¶m wrote:
> ons 2012-02-29 klockan 16:43 -0700 skrev Steven Dake:
>> Possible but at the moment the error isn't captured by any API resulting
>> in a packet leak when this happens.
> It should already get picked up via _libssh2_packet_requirev by the one
> who sent the failing request.
>> I am not quite sure how to return an error code in this condition, so
>> not certain how to propose such a patch.
> Do you have any evidence of SSH_MSG_UNIMPLEMENTED piling up? If so, in
> response to what kind of requests?
Not piling up. There appears to be 1 or 2 extra on the session when I
close it. This was verified with the diagnostic patch I posted recently
which prints packets on the session during session close.
I was getting 1 or 2 3's and thousands of 82s in a 12 hour run. The 82s
were in response to the keep alive, not sure what triggered the 3s. the
3's remain after i turned off want_reply in my keepalive send operations.
> This is not the same case as keepalive with want_reply set. The
> keep-alive problem is libssh2_keepalive_send never trying to read the
> response to the request it sent. It's state machine needs to be extended
> to also read the reponse if want_reply is set.
Received on 2012-03-01