[alsa-devel] [PATCH v2 1/2] ALSA: jack: create jack kcontrols for every jack input device

[Takashi Iwai]

At Fri, 20 Mar 2015 12:50:33 +0000,
Jie, Yang wrote:
> 
> > -----Original Message-----
> > From: Takashi Iwai [mailto:tiwai at suse.de]
> > Sent: Friday, March 20, 2015 8:27 PM
> > To: Jie, Yang
> > Cc: perex at perex.cz; broonie at kernel.org; alsa-devel at alsa-project.org;
> > Girdwood, Liam R; Liam Girdwood
> > Subject: Re: [PATCH v2 1/2] ALSA: jack: create jack kcontrols for every jack
> > input device
> > 
> > At Fri, 20 Mar 2015 12:22:10 +0000,
> > Jie, Yang wrote:
> > >
> > > > > +}
> > > > > +
> > > > > +static int snd_jack_new_kctl(struct snd_card *card, struct
> > > > > +snd_jack *jack, int type) {
> > > > > +	struct snd_kcontrol *kctl;
> > > > > +	struct snd_jack_kctl *jack_kctl;
> > > > > +	int i, err, index, state = 0 /* use 0 for default state ?? */;
> > > > > +
> > > > > +	INIT_LIST_HEAD(&jack->kctl_list);
> > > > > +	for (i = 0; i < fls(SND_JACK_BTN_0); i++) {
> > > > > +		int testbit = 1 << i;
> > > > > +		if (type & testbit) {
> > > >
> > > > With this implementation, you'll get multiple boolean kctls for a
> > > > headset.  I thought we agreed with creating a single boolean for headset?
> > > [Keyon] We agreed with creating multiple kctls for combo jack, e.g.
> > headset.
> > > Furthermore, e.g., imagine that type = SND_JACK_HEADSET  |
> > > SND_JACK_BTN_0, we will create 3 kctls for the jack, when BTN_0 is
> > > pressed, we will report to the 3rd kctl.
> > 
> > Hm, I don't remember that I agreed with multiple kctls...
> > 
> > The multiple kctls have a significant drawback (multiple event calls for a single
> > headset) and its behavior is incompatible with the current code (both the
> > name change and the behavior change).  That is, your patch will very likely
> > break the existing applications.
> [Keyon] I am not very clear with the existing applications that using these 
> kctl events(seems Android use input subsystem event? Which we didn't
> Change here. If I understand correctly, Pulseaudio uses jack switch controls, 
> via the name, then we can use different name for headphone and mic here.)

PA uses jack kctls.

If you rename, how would you guarantee that the existing application
will work as expected?  PA doesn't have the definition of "Headset
Speaker Jack" or such.

And, no, we have no option "fix PA".  Other way round: we are not
allowed to break the current PA (or any user-space) behavior in
general.

> we will generate 2 event calls(one headphone, one mic) when Headset 
> plug in/out, applications will receive these 2 events,  and they can do anything,
> e.g. decide to switch on/off speaker/headphone.

Won't this break any user-space stuff?

> BTW, I haven't implemented the generating of combo jack kctls' name yet,
> currently, they looked like below:
> numid=12,iface=CARD,name='Headset Jack'
> numid=13,iface=CARD,name='Headset Jack',index=1
> numid=14,iface=CARD,name='Headset Jack',index=2
> 
> once we have come to agreement, we can modify it in snd_jack_new_kctl(),
> e.g., "Headset Jack Mic" and "Headset Jack Speakers".

.... and how the existing user-space works without changing its code?


Keyon, the most important point at this moment is to keep the
compatibility.  HD-audio is no new driver.  It's a driver that has
been present over a decade with (literally) thousands of variants.
Please keep this in mind, and reconsider whether your patch will
retain the compatibility, especially with PulseAudio.


thanks,

Takashi