José Luis Millán
José Luis Millán
It's definitely odd. I wonder why we set that mask. Being a 3 byte field it should be `0x00FFFFFF` indeed. What's the reason behind `0x3FFFFFC0`, @bianxg ?
Indeed we need it to be a 23bit field, so the mask should be: `0x7FFFFF00`
There is indeed discussion/explanation about this here https://github.com/meetecho/janus-gateway/issues/1733#issuecomment-526553410 @ibc , but I don't see why we set that mask.
I'm checking it.
Got it: 0x1FFFFFC0 => 536870848 => 11111111111111111111111000000 Dividing such value by 64: 0x7FFFFF => 8388607 => 00000000011111111111111111111111 So by adding `0x1FFFFFC0` mask we make sure that after dividing the value...
Where do you then see the problem @bianxg ?
The difference is that mediasoup does NOT generate negative timestamp values. That's why we use 23 bits to represent it, so the most significant bit (sign) is always 0 and...
Anyway, have you seen this is an issue for libwebrtc processing this RTCP packets?
This was handled here https://github.com/versatica/mediasoup/pull/899
Thanks @fippo, we'll do some tests next week and see if there is any difference on the resulting BWE.