[pulseaudio-discuss] drain latency with echo-cancel

[David Henningsson]

On 2014-09-30 16:48, Alexander E. Patrakov wrote:
> 30.09.2014 19:35, Heero Yui wrote:
>> Hello
>>
>> I have write a simple echo suppressor inside the echo-cancel module (it
>> is based on "null" version).
>> When I activated it, I also open input and output channels, the module
>> work correctly.
>> I launch it with (same with null):
>> pacmd load-module module-echo-cancel aec_method=null
>> source_master=alsa_input.0.input-microphones adjust_time=1
>> pacmd set-default-source alsa_input.0.input-microphones.echo-cancel
>> pacmd set-default-sink alsa_output.0.output-speakers.echo-cancel
>>
>> The problem: When I open an other output, at the flush request, it take
>> like 1 or 2 second more time as when I have not activate the echo
>> canceller (work with a paplay)
>>
>> And I did not understand why!
>
> I also don't understand why. However, please note that this bug is not
> specific to module-echo-cancel. It can also be reproduced with
> module-virtual-sink that deliberately does absolutely no processing and
> just serves as a template for new virtual sinks.

The pa_sink_process_input_underruns stuff was something I added a while 
ago to speed up draining. I don't think I ever tested that with a chain 
of sinks, so maybe pa_sink_process_input_underruns needs to take filter 
sinks into account somehow?


-- 
David Henningsson, Canonical Ltd.
https://launchpad.net/~diwic