On Tue, 2017-02-14 at 10:51 +0100, Georg Chini wrote: > On 14.02.2017 10:25, Tanu Kaskinen wrote: > > On Sat, 2017-02-04 at 19:34 +0100, Georg Chini wrote: > > > The fixed latency for the HSP sink was set to 125 ms plus one write block size. > > > pa_sink_get_latency() however returns 28 ms like it does for A2DP and also for > > > the HSP source. > > > Therefore this patch corrects the fixed latency of the HSP sink to 25 ms plus > > > one write block size, assuming that the original value of 125 ms was a typo. > > > > I believe latency reporting in module-bluez5-device is broken, so it > > can't be used as a justification for changing the constants. module- > > bluez4-device includes the fixed latency in its latency reports. > > module-bluez5-device does that too, but it only uses the A2DP constants > > (even when HSP is active). That's probably because originally module- > > bluez5-device only supported A2DP, and when HSP support was added > > later, updating the latency reporting code was just forgotten. > > > > Thanks for the review. You are right, the correct justification would be, > that it can be expected that HSP recording produces the same delay as > playback. > > Do you think it is OK if I send a fix for the other bug (which looks easy > to fix) and include that change? Or would you prefer to leave the constants > as they are? I'd prefer leaving the constants as they are if there's no hard evidence of them being wrong. I think this should be quite easy to test: just play a video using HSP with old and new constant values, and if the new value results in better A/V sync, that's good enough evidence for me. -- Tanu https://www.patreon.com/tanuk
