discarding buffered packets at the start of the stream #570
Conversation
| p.stats.Streams.Add(1) | ||
| p.mediaReceived.Break() | ||
| log := p.log.WithValues("ssrc", ssrc) | ||
| log.Infow("accepting RTP stream") |
There was a problem hiding this comment.
Moved to include discarded packets number, and first RTP seq
| // such as in outbound calls that take a sec to answer, we have a ton of junk in the stream. | ||
| // We need to discard all of it. | ||
| now := time.Now() | ||
| if lastPacketTime.IsZero() || now.Sub(lastPacketTime) < time.Millisecond { |
There was a problem hiding this comment.
Note, for inbound calls, we will discard the first packet as well, even if there isn't anything buffered in there.
We can, of course introduce code to correct for that, but the added complexity was deemed not justified due to non-existing impact.
| } | ||
| if catchup { | ||
| // When a remote writer has been sending data but the local stream didn't start the RTP loop, | ||
| // such as in outbound calls that take a sec to answer, we have a ton of junk in the stream. |
There was a problem hiding this comment.
I'm not sure I'm following here.
There's presumably some garbage sent before the call is established. But the code in the PR doesn't even check the RTP headers or the payload. Why are we dropping it unconditionally?
Assuming there's some audio there (silence? ringing?) or even just audio junk, it won't propagate through the media pipeline until it's actually enabled (we have multiple SwitchWriters in the pipeline that can cut the stream).
So why is this necessary?
No description provided.