There's an edge case in finnicky networks where ALG-like firewalls
tamper with USE-CANDIDATE STUN packets and, consequently, bork ICE-lite
connectivity establishment. The odd part is that client-side gathering
seems to complete if intermediate STUN bindings work (before the final
USE-CANDIDATE), which may cause the peer not to generate relay
candidates == connectivity fails.
This adds the `public.kurento.gatheringTimeout` option to forcefully extend
the candidate gathering window in peers that act as offerers. The
behavior is as follows: if the flag is set (ms), the peer will wait
either the gathering completed stage or, _at most_,
public.kurento.gatheringTimeout ms before proceeding with calls chained
to setLocalDescription.
This option is disabled by default and intentionally ommited from the
base settings.yml file as to not encourage its use. Don't use it unless
you know what you're doing :).
Reconnection timers are far too long for abrupt failures because we
are waiting the original timeouts to elapse (30-60s) before trying it
again - even if a connection worked N-sessions back in that session's
history. The ideal thing to have is another intermediate, smaller and
fixed reconnection timer for sessions that had a working screen share
at least once.
The UI is also not being updated to the reconnecting state on negotiation
failures.
* Add an intermediate reconnection timer for abrupt failures set to 8s.
This should improve reconnection times.
* Lower default connection timers values (base 20s down from 30s, max
25s down from 60s)
* Set screen share UI to reconnecting on abrupt failures as well - we
were only tracking ICE states prior to this, not negotiation errors
The reconnect routine is stopping for viewers if a broker cannot
re-connect in the first try. That is wrong: viewers should try to
reconnect as long as there'sigaling data that mandates so.
The reconnect trigger is changed from broker's started attribute to the
presence of a scheduled reconnection timeout - if there isn't one (not
schedule), always re-schedule it.
RTCRTPSender exposes DSCP marking via `networkPriority` in the encodings
configuration dictionaries. That should allow us to control
QoS priorities for different media streams, eg audio with higher network
priority than video. The only browser that implements that right
now is Chromium.
To use this, the public.app.media.networkPriorities configuration in
settings.yml. Audio, camera and screenshare priorities can be controlled
separately. For further info on the possible values, see:
- https://www.w3.org/TR/webrtc-priority/
- https://datatracker.ietf.org/doc/html/rfc8837#section-5
If a viewer session failed mid-call, it was being scheduled for a reconnect via
the min-max connection timers (30s-60s), which is terrible UX.
This commit makes screen sharing viewers try to reconnect immediately when
appropriate.
Outbound/presenter screen sharing reconnect was broken from inception, so it's
being removed until it´s properly re-implemented.
This also fixes an issue where presenter disconnections would be silent for the
end user - now an error toast is shown and the error properly logged.
Fixes an issue where subsequent failures might lead to wrong error codes being
reported;
Splits the screen sharing bridge stop method into a reconnect-safe version and
a public one - should also address some quirks with inbound stream reconnection.
There could be a race condition where the local getDisplayMedia stream ends
(eg via Chrome`s stop sharing toast) while the broker hasn't finished starting.
That would lead to a scenario where the broker wouldn't emit an end event,
causing screen sharing to be flagged as started with a blank/invalid stream.
There could be a scenario where the local gDM stream wasnt cleaned up;
eg.: SFU is offline.
This commit guarantees all tracks from the local stream are stopped.
I have growing concerns about gain node`s effect on audio quality the way it
was implemented, so I opted to fall back to HTMLMediaElement`s volume control
for the time being until we can gauge quality impacts properly later on
Add a new configuration flag enableVolumeControl, false by default while the
feature undergoes a field trial
- forceRelayOnFirefox: whether TURN/relay usage should be forced to work
around Firefox's lack of support for regular nomination when dealing with
ICE-litee peers (e.g.: mediasoup).
* See: https://bugzilla.mozilla.org/show_bug.cgi?id=1034964
- iOS endpoints are ignored from the trigger because _all_ iOS browsers
are either native WebKit or WKWebView based (so they shouldn't be affected)
Splits screenshare stream into video and audio and adds gain node to audio
stream in order to permit volume control by the user. Volume is normalized
between [0, 2](muted and 2x boost).
ICE lite servers (eg mediasoup) dont need candidates signaled out-of-band; neither does KMS in certain scenarios
Disable their signaling saves us some ticks in bbb-webrtc-sfu and some bandwidth all around
Added support for getStats in screenshare's service. This works similar
to the getStats for video provider, and the information retrieved from
screenshare is added to the video information for cameras.
Scenario: presenter`s client could crash when the presenter changed while they were sharing their screen
That is due to a race condition on the stop procedure in the bridge: two stops can be triggered (one from the server-side websocket tear off and another from the client itself detecting the presenter change)
That could create a scenario where the broker was cleaned in one stop procedure after the second had checked its availability, causing an attribute access of a null member
Added new SFU broker for screen sharing
Removed kurento-extension entirely
Added inbound and outbound reconnection procedures
Improve UI responsiveness when sharing
Add reconnection UI states
Redo error handling
Refactor actions-bar screen share components. Make it smarter with less prop drilling and less re-rendering. Also more readable. Still work to do in that I think
Add a connection retry procedure for screen presenters when they are sharing; try a configurable amount of times when failure is triggered, with configurable min and max reconn timeouts and timeout increase factor
Make local preview attachment smarter
ADD PARTIAL SUPPORT FOR AUDIO SHARING VIA SCREEN SHARING WITH GET DISPLAY MEDIA, RECORDING STILL NOT SUPPORTED!!!
Refactored STUN/TURN fetch to be done only once, when successful, per session and cache it in mem to avoid too many reqs. Current way is a bit dumb, this should increase reliability a bit more. The caching is configurable so folks who want to use very short lived TURN credentials can disable it
Add a fallback STUN config option to be used when the default STUN/TURN fetch fails
Clean the safari/no candidate generation pre flight check from 3rd party STUNs
Fix deadlock in audio join when STUN/TURN fetch failed
This adds support for screensharing on Safari 12.1+ or whichever has getDisplayMedia enabled
To be conservative and backwards compatible, I`ve maintained the current gUM-based code available as a fallback for those still using older versions of FF or Chrome
Made screenshare constraints configurable. The constraints config will be piped directly to gDM, so it`s just a regular gDM constraint dictionary
