Switching a stream from a remote tunnel to a local sink takes some time

[sjoerd]

To reproduce.. Switch a stream to a remote tunnel, switch it back to a local sink.. Audio output stops for multiple seconds and then continues..

Below is what my local pulse server outputs when moving the stream back to a local sink..

I: resampler.c: Forcing resampler 'copy', because of fixed, identical sample rates.
I: resampler.c: Using resampler 'copy'
I: resampler.c: Using float32le as working format.
D: module-suspend-on-idle.c: Sink tunnel.spring.local.alsa_output.platform_snd_powermac_alsa_playback_0 becomes idle.
D: module-suspend-on-idle.c: Sink tunnel.spring.local.alsa_output.platform_snd_powermac_alsa_playback_0 becomes idle.
I: module-alsa-sink.c: Trying resume...
I: module-alsa-sink.c: Resumed successfully...
D: module-suspend-on-idle.c: Sink alsa_output.pci_8086_284b_alsa_playback_0 becomes busy.
D: sink-input.c: Successfully moved sink input 54 from tunnel.spring.local.alsa_output.platform_snd_powermac_alsa_playback_0 to alsa_output.pci_8086_284b_alsa_playback_0.
I: module-volume-restore.c: Saving sink for <pulsecore/protocol-native.c$Rhythmbox>
D: memblock.c: Memory block too large for pool: 49152 > 16368
I: module-alsa-sink.c: Starting playback.
W: module-tunnel.c: Server signalled buffer underrun.
I: module-suspend-on-idle.c: Sink tunnel.spring.local.alsa_output.platform_snd_powermac_alsa_playback_0 idle for too long, suspending ...

[Esmil]

..and in latest (> 0.9.21) git master branch switching a stream from remote tunnel to a local sink is completely broken. Playback just stops until you switch the stream back to the tunnel sink.

Oddly switching from local sink to tunnel sink works great though.

[accumulator]

I had exactly this problem, with both ubuntu provided 0.9.19 and 0.9.22 from PPA/ricotz/unstable.

The problem manifested only after the target system was running for a while (asrock nettop), mapping sound from the desktop to the asrock always works (but with audio that tries to catch up, i.e. much faster speed), but mapping it back to a local device the player is paused for a while, then continues. The delay before playing continues is a bit random, with a few sequential tests I observed 20s-120s.

restarting pulseaudio there did not make any difference.

But.. I then synced both system clocks using NTP and voila, perfect handovers!

So somehow absolute clock is used in PA for relative stream positioning?

[accumulator]

of course the NTP thing is a workaround, so a) hoping/assuming this is easy to fix and b) since this bug is so old already.. I mark this as milestone