What steps will reproduce the problem?
1. Take two browsers and two users logged into gmail.
2. Close browser of user 1.
3. Have user 2 call user 1.
4. Times out after a large delay, maybe 50 seconds based on unwritable stream
delay in sessionmanager.cc.
What is the expected output? What do you see instead?
If a user goes offline via a loss of connectivity, and you try to place a call
to this user, the library has a default delay for writable stream. Since all
requests are acked in the jingle protocol, I think it's an reasonable
enhancement to actually track the receivedinitiate_ack as a new session state
in session.cc, and also add a new error for an ack timeout, which could be much
smaller, say 3 seconds. In the case of phones for example, waiting for this ack
would signify that it's ringing on the other side, and before you receive the
ack, you would be in a call setup or connecting state.
Diff of proposed changes attached.
Original issue reported on code.google.com by l...@tuenti.com on 17 Oct 2012 at 11:24
Original issue reported on code.google.com by
l...@tuenti.com
on 17 Oct 2012 at 11:24Attachments: