On 2019-08-26 00:15, Maxim Levitsky wrote: > On Sun, 2019-08-25 at 20:05 +0200, Zoltán Kővágó wrote: >> On 2019-08-25 11:44, Maxim Levitsky wrote: >>> On Wed, 2019-08-21 at 10:41 +0200, Gerd Hoffmann wrote: >>>> From: Kővágó, Zoltán <dirty.ice...@gmail.com> >>>> >>>> This means you should probably stop using -soundhw (as it doesn't allow >>>> you to specify any options) and add the device manually with -device. >>>> The exception is pcspk, it's currently not possible to manually add it. >>>> To use it with audiodev, use something like this: >>>> >>>> -audiodev id=foo,... -global isa-pcspk.audiodev=foo -soundhw pcspk >>> >>> Hi! >> >> Hi, >> >>> There is one corner case this breaks. >>> In qemu 4.1.0, there is no way to specify audiodev for a sound device, >>> specifying it >>> fails with error. >>> So some of my machines have audiodev (which is miles better that using old >>> env variables) >>> but also have sound devices without audiodev reference since this wasn't >>> supported. >>> >>> >>> In what will be qemu 4.2, you must specify it, thus this kind of breaks >>> backward compatibility. >>> Maybe we can have audiodev reference optional for a version or two? >>> >>> This is just a minor itch, as otherwise the sound improvements are really >>> good. The days >>> of installing that old realtek driver are finally gone :-) >> >> Hmm, this is what happens when you split a patch series. We could >> either revert this patch, or alternatively turn the error messages into >> warnings about using deprecated behavior. > Warning would be great in this case! >> >>> Another thing I noted, that there is no way for pulseaudio audiodev to >>> specify the 'client name', >>> it always shows up in pavucontrl as the socket path to the server. >>> Thus if I added two PA audiodevs, I can't really distinguish between them. >>> The in|out.name= seems to specify the pulseaudio source/sink to connect to, >>> which is not the same. >> >> We currently supply the constant "qemu" as a name to pa_stream_new. >> While it's still not ideal, shouldn't this end up as a client name in >> pulseaudio instead of a socket path? > > Actually it seems that pulseaudio has two names supplied for each stream > Maybe stream name and application name? > > This is how chromium playback looks versus qemu in pavucontrol and in gnome > volume control. > > https://imgur.com/a/I8HZhgx > > I do notice that 'qemu' now, in pavucontrol though.
I see. We currently pass the server socket to pa_context_new instead of the client name. I'll prepare a patch soon, thanks for the report! Regards, Zoltan