* Call Broadcast in lock scope.
Seems like there is a possible window where things can hang forever
if a goroutine enters the Wait) after the lock is released but before
Broadcast gets called, it will never see that broadcast and will hang forever.
* RLock
This patch ensures that SIP headers provided in TransferSIPParticipantRequest requested by an agent are correctly forwarded to the internal SIP transfer request.
Previously, req.Headers was accepted but never passed along, causing headers from the initiating agent to be dropped.
This change adds a single line to include req.Headers in the internal request.
Tested with eyevinn client.
There are a few issues to figure out still
1. Simulcast - how?
2. For simulcast, how to know width, height so that adaptive stream can
work.
3. The layer added is dummy. It works, but connection quality scoring
would be incorrect (will always say excellent) without bitrate.
Will need some dynamic update of `TrackInfo` based on actual stream for
all of this to fit well into our system, but the simple video support
works for now.
Was setting the state to ACTIVE prematurely to enable the subscription
inter-lock in one shot signalling mode. But, that is incorrectly
changing state.
Use a callback to indicate subscriber ready and let the participant
ACTIVE happen when the connection actually establishes.
Had made the change to align `participant active` to after the ICE
connection is done and that log could list all candidates.
But, with one shot signalling, the state change has to be early to wait
on (auto) subscriptions of track of other participant. So, state has to
be changed early.
* Add support for WHIP ICE Trickle/Restart.
Tested a bit using the WHIP client at https://github.com/Eyevinn/whip,
but needs a lot more testing. ICERestart is not tested yet.
* comment
* clean up
* Exclude RED from enabled codecs for Flutter + 2.4.2 + Android.
That causes client to get confused and generate invalid SDP.
Also make SDK version rule check based on semantic versioning compare.
* deps
* Prevent migration race.
Comments in code. Briefly, due to race, the remote participant/track
could be closed early leading to missing subscription post-migration.
* fixes
* Unlabeled (pass through) data channels.
Support data channels than can pass through raw data without any LK
protocol marshaling/unmarshaling.
* statischeck
* test
* error -> warn
* reset data message callback
* Revert "Audio uses signal SignalCid and SdpCid. (#3564)"
This reverts commit cdfbb106d1.
* Revert "Keep simulcast information tied to receiver. (#3563)"
This reverts commit ed5e2f16b2.
* Revert "chore(logs): log VLS type for VP9/AV1 (#3561)"
This reverts commit ad010cfc43.
* Revert "fix(video): determine svc/simulcast from SDP for advanced codecs (#3549)"
This reverts commit 15f565510c.
* chore(deps): update protocol
This is a bit annoying and painful. Audio does not have the information
in the track info (because it does not use Codecs, it should probably at
some point to make it consistent).
* Keep simulcast information tied to receiver.
`simulcast` flag in `TrackInfo` is at track lavel. With codec simulcast,
the primary codec (in most cases) is SVC and the backup codec is
simulcast. Back up codec publish changing the track info setting to true
meant that the primary receiver was treated as simulcast if a subscriber
for primary codec joined after the backup codec was published.
Keep track of simulcast flag in receiver.
Also, TrackInfo Cids are from signal. So, keep track of SDP cids
separately. The `simulcastTrackIds` map uses SDP cid. Clean up by all
the SDP cids of a track
* clean up
* clean up
* clean up
* clean up
* test
* Store SdpCid and IsSimulcast in Trackinfo
* clean up
* mock
* Rework node stats a bit.
Related protocol PR - https://github.com/livekit/protocol/pull/1023
- Make a config for node stats measurements. Wanted to put the config in
`routing` package, but a circular dependency forced me to put in
config.go
- Make rate calculations explicit, i. e. requested via config.
Previously, it had some odd checks to decide when to calculate rate
and it would have been calculating over different windows.
- Report signal/data channel bytes every 5 seconds to stats collection
module. Previously, it was doing it every 30 seconds and that meant
some windows could have had a large spike
NOTE: Still need to think about this for load calculations as a large
number of participants leaving could flush in a small window and that
could report a large spike in bytes/packets. Maybe need to ignore
signal bytes for load calculation?
* deps
* use default node stats config if given config is nil
* split out node stats into a struct for re-use
* update config
* fix(video): determine svc/simulcast from SDP for advanced codecs
* fix(explicit-svc): cleanup
* fix(explicit-svc): remove from list on close/remove
* fix(explicit-svc): reorder VLS selection, cleanup
* fix(explicit-svc): todo comments for temporal layer selector
* fix(explicit-svc): remove from simulcastTrackIds even if client does not support unpublish
Seeing a lot of queuing delay based back offs. Trying a couple of things
1. Accept a bit more queuing.
2. An option to try a different pacer. Would like to try with pass
through. That will produce some out-of-order packets. Remains to be
seen if it will have a negative impact.
