Audio capturing for manual rendering mode

[hiroshihorie]

No description provided.

[hiroshihorie]

We need to test if I didn't break anything here 🙂

[pblazej]

Vibe-coded a quick prototype according to README livekit-examples/swift-example#77

It works well with one mode, but switching between manual/mic gives me either Engine is not running or a crash (on iOS). What am I doing wrong?

BTW maybe it's worth converting into a "proper panel" for future testing - feel free to destroy it 👍

[hiroshihorie]

We don't really support switching manual / device rendering mode at runtime, it's a complicated internal switch... although ideally it should work.

[pblazej]

We don't really support switching manual / device rendering mode at runtime, it's a complicated internal switch... although ideally it should work.

Maybe assertionFailure would be better in this case?

[pblazej]

BTW, are you gonna integrate https://github.com/livekit/webrtc-xcframework/releases/tag/137.7151.04?

[hiroshihorie]

Yes I think the options are:

1. Make it settable once only like ADM type.
2. Simply fix the AVAudioEngine ADM to be able to switch at run time. I think the errors you see are from the attached nodes in MixerEngineObserver.swift. I think the node attach/cleanup lifecycle is not working.

[hiroshihorie]

@pblazej Yes I think 0.4 is already being used.

[pblazej]

Make it settable once only like ADM type.

I think this is perfectly fine for now.

[pblazej]

Started getting Room.signalClient(_:didSubscribeTrack:) Could not find local track publication for subscribed event in manual mode, anything wrong with the example?