davidliu

Seems like the `apply_alignment_to_all_simulcast_layers` solution seems to be more proper since it's fixed going forward from Android 12 (albeit slowly). Would simply setting `apply_alignment_to_all_simulcast_layers` to true on the bad devices...

Ah, I think I ran into an old issue with this now: Attempting to initEncode with odd dimensions will crash; this wasn't an issue before since alignment fixed this, but...

Here, the issue is that the base layer gets aligned, but the scaled layers are scaled without any further alignment (since the previous math dictated that no further alignment was...

Getting some weird outputs now: For a base layer of 600x800, with scaleResolutionDowns of [1, 1.66667, 3.33333], the alignment adjuster changes the scales like so: ``` 2022-11-18 01:39:08.331 30545-30680/io.livekit.android.composesample I/RTCModule:...

I think I understand what's going on here now; the issue is that no realignment seems to be done after the alignment is changed from the alignment adjuster. In the...

Ok, with a little more digging, I *think* I have a solution to this issue; in HardwareVideoEncoder, put back the width/height checks, but reduce `REQUIRED_RESOLUTION_ALIGNMENT` to 2 from 16. The...

Alternatively, changing the catch clause when configuring the codec in `HardwareVideoEncoder.initEncodeInternal()` from `IllegalStateException` to also handle `IllegalArgumentException` also works.

Seems good. Any other changes you think this will need? Or ready to merge?

Hmm, sounds like `room.release()` has already been called, resulting in the EglBase no longer being valid? If that's the case, I think this behavior should be expected (and will need...

Right now we don't have this capability. Can't really say when this will be implemented.