Rui He

> @ruihe774, I've just noted: `--x11-bypass-compositor=never` ( that is disabling bypass of "XRandr scaling" by meaning of the words) gets rid of delay (which is when as you wrote "XRandr...

I've added `ao_stop_streaming` to handle data after EOF. Now consecutive playback works fine.

@sfan5 I added support for `set_pause` and EOF handling in `ao_audiotrack`. I'm sorry that I have no idea how to build mpv for Android. Would you plz test my implementation?...

@sfan5 I wonder if there are any updates. 😃

> This breaks `--loop` on ao_pipewire, I can't test ao_audiotrack or ao_avfoundation. Weird. It works sometimes and doesn't work in other times. I'm looking into this.

> > This breaks `--loop` on ao_pipewire, I can't test ao_audiotrack or ao_avfoundation. > > Weird. It works sometimes and doesn't work in other times. I'm looking into this. @llyyr...

@sfan5 BTW I wonder why `ao_audiotrack` is a pull-based AO. It uses a thread to feed data to `AudioTrack_write`. IMO a push-based AO is more suitable.

> What I'm not sure about is how avfoundation and audiotrack (which both use a thread) normally block. `AudioTrack_write` might block, maybe?? I didn't get you. What's the problem if...

> @ruihe774 can you split the ao_audiotrack change into a separate PR? We can then consider this ready for merging. Ok. > It isn't. Inversely it would be a problem...

> each write call returns immediately I believe it is just the case for `AudioTrack_write`. It returns immediately. > Then, when resuming, you can almost effortlessly continue writing the buffer...