Yes, I also agree with Lennart’s point. >>I agree with you that pausing may create timing havoc on the client; but we >>are missing a generic protocol to notify the clients and implement this >>legitimate use case. I am afraid there is no way to notify applications using ALSA due to the PA client is libasound_module_pcm_pulse.so at this time.
IMO, PA is for general OS but not for embedded system now. The requirement that internal pause an application is a natural to audio manager of embedded system. I guess we can only cover applications who will response the message "you should do pause now" from audio manager. ________________________________________ From: [email protected] [mailto:[email protected]] On Behalf Of pl bossart Sent: 2009年1月21日 0:13 To: General PulseAudio Discussion Subject: Re: [pulseaudio-discuss] why command_cork_playback_stream() will be invoked many times? Hi Lennart, Here is the use case Xing is referring to: you are listening to music, and a VoIP call starts. The user may not want to mix the music and the speech call. So the idea is to pause the music while the call takes place, and resume the music once the call finishes. PulseAudio receives both streams, and it would seem natural to configure said behavior in a PulseAudio module. So we either need the ability to pause a stream within PulseAudio, or we need a means to inform the client they need to pause. I agree with you that pausing may create timing havoc on the client; but we are missing a generic protocol to notify the clients and implement this legitimate use case. Regards, Pierre Bossart On Sun, Jan 18, 2009 at 11:16 AM, Lennart Poettering <[email protected]> wrote: On Fri, 09.01.09 21:10, Zhang, Xing Z ([email protected]) wrote: > Hi experts: > I worked on an audiomanager project based on PulseAudio. Now I am > blocked by a command_cork_playback_stream() issue. In our design, > an application may be corked when connects to pulseaudio if its > priority is low. I set a hook on PA_CORE_HOOK_SINK_INPUT_PUT and > invoke pa_sink_input_cork(..., TRUE). Unfortunately it doesn't > work. By GDB, I found application will call > command_cork_playback_stream() which invokes pa_sink_input_cork(..., > FALSE) several times, this make my hook is of no effect. I don't > look into PA for ALSA plugin, anyone know why > command_cork_playback_stream() need be called so frequently during > app initialization? Hmm, I think you are confusing a few things here. command_cork_playback_stream() is the code that dispatches client requests for corking/uncorking (when done via the native protocol). It is not used when corking something internally as for example by a hook function. Pausing a stream (i.e. corking) should be something that is controlled exclusively by the client. You should not intefere with it from inside the server. There is a state machine in the PA client code that follows the cork state. If you change the state underneath it might become invalid. Also it might confuse client applications due to the paused timing. Lennart -- Lennart Poettering Red Hat, Inc. lennart [at] poettering [dot] net ICQ# 11060553 http://0pointer.net/lennart/ GnuPG 0x1A015CC4 _______________________________________________ pulseaudio-discuss mailing list [email protected] https://tango.0pointer.de/mailman/listinfo/pulseaudio-discuss _______________________________________________ pulseaudio-discuss mailing list [email protected] https://tango.0pointer.de/mailman/listinfo/pulseaudio-discuss
