tree 5eeb4cdc433eca111af0b45f477eb1451506fe67
parent 0b44b61a7a992874149b16b52e95801fce7c1eea
author Sébastien Blin <sebastien.blin@savoirfairelinux.com> 1659638684 -0400
committer Sébastien Blin <sebastien.blin@savoirfairelinux.com> 1659639390 -0400
gpgsig -----BEGIN PGP SIGNATURE-----
 
 iHUEABEIAB0WIQSF8dAXT8CbjGjWU3nIlLsB7rKpqQUCYuwWewAKCRDIlLsB7rKp
 qeYkAQDLEuAyDe8vvLUSsF7tESoM6Q54Ayl1afmWhz9mxz78kAEAlAQnLnJcAsYS
 sYUFLfwn36s7k/KsL5fIl4Euk3ccZK4=
 =9JtW
 -----END PGP SIGNATURE-----

ice_session: various fixes with TURN connections

This patch contains two fixes:
1. It moves a check for waiting passive connections to finish.
Indeed, we were waiting for passive connections (such as TURN) to
give a result only when we were a controlling agent (and nominating).
In fact, if the agent is not nominating (before getting a USE-CANDIDATE
or if it's controlled, the problem is the same, we need to wait that
the peer connects if we are the only one with TURN enabled).
Also, the RFC is only talking about waiting passive connections, not
if controlling or not.
2. It increases the number of connections attempt if we got a
connection reset by peer (if the TURN didn't authorize us). The previous
timeout was 4 seconds (12 now). It was not enough if we were sending
candidates over the DHT (the first put encrypted can take 10 seconds,
then it will be faster, causing some first call to fail).

For more details:
https://git.jami.net/savoirfairelinux/jami-daemon/-/issues/746

Note: squash with rfc6544 if rebased to pjproject 2.13

Change-Id: I3b8e807d9cf0722d59500eec9cb5ebdfe97133e2