[pulseaudio-discuss] [PATCH v0 09/16] bluetooth: Move streams during shutdown

[Luiz Augusto von Dentz]

Hi Mikel,

On Wed, Oct 17, 2012 at 8:17 PM, Mikel Astiz <mikel.astiz.oss at gmail.com> wrote:
> Hi Tanu,
>
> On Wed, Oct 17, 2012 at 3:29 PM, Tanu Kaskinen <tanuk at iki.fi> wrote:
>> On Mon, 2012-10-15 at 04:28 +0200, Mikel Astiz wrote:
>>> Hi Tanu,
>>>
>>> On Sun, Oct 14, 2012 at 6:47 PM, Tanu Kaskinen <tanuk at iki.fi> wrote:
>>> > On Fri, 2012-09-28 at 17:45 +0200, Mikel Astiz wrote:
>>> >> From: Mikel Astiz <mikel.astiz at bmw-carit.de>
>>> >>
>>> >> If sink and sources exist during shutdown, the streams need to be moved
>>> >> exactly as when the profile is being set to off.
>>> >
>>> > Why? I thought that moving streams during profile change is done because
>>> > it's supposedly a good policy to keep the streams at the same device
>>> > before and after a profile switch. That doesn't apply during device
>>> > unload, so am I missing the real purpose for the moving, or is this
>>> > patch unnecessary?
>>>
>>> This might be a lack of knowledge from my side about stream moving.
>>>
>>> What I understood from the code (specially from the bluetooth module)
>>> is that the modules needed to report the core that something should be
>>> done with these streams, and then routing policies would apply
>>> (possibly implemented in some other policy module).
>>
>> No, there's no need to report that something needs to be done with the
>> streams. That information is already available through the sink unlink
>> hook - when a sink is about to get unlinked, policy modules can react to
>> that in some way. An example of that is module-rescue-streams, which
>> moves the streams somewhere.
>>
>> module-rescue-streams does not, however, implement the same policy as
>> what module-bluetooth-device does currently. The sink/source of the new
>> profile is not yet created when the sink/source of the old profile is
>> unlinked, and module-rescue-streams has no idea that the "right"
>> sink/source will appear soon, so it moves the streams away from the
>> bluetooth device.
>
> OK, I understand now. So the moving was probably there to handle
> HSP<->A2DP transitions for headset use-cases. I doubt however if this
> is really needed, assuming the routing policies would be able to route
> the audio independently from where it was routed before.
>
> Luiz, do you recall why this moving has been implemented?

Nope, in fact it might exist since the very beginning where things
like module-rescue-streams did not exist.

>> If it's deemed necessary, module-bluetooth-policy could react to the
>> profile change events and do what module-bluetooth-device is currently
>> doing internally.
>
> I wonder if PA has the necessary hooks to implement this, since
> SINK_HOOK_UNLINK wouldn't have enough information about the new card
> profile. Perhaps the core would need to be extended with something
> like HOOK_SET_CARD_PROFILE_START/_FINISH?
>
> For the moment, unless someone points out a good reason to have such a
> policy, I will submit another patchset including the removal of this
> stream-moving code.
>
> Cheers,
> Mikel



-- 
Luiz Augusto von Dentz