[pulseaudio-discuss] [PATCH v5 0/7] New API for Bluetooth A2DP codecs

[Pali Rohár]

On Saturday 26 January 2019 00:09:17 Pali Rohár wrote:
> On Friday 25 January 2019 11:04:22 Tanu Kaskinen wrote:
> > On Wed, 2019-01-23 at 18:52 +0100, Pali Rohár wrote:
> > > On Wednesday 23 January 2019 19:16:16 Luiz Augusto von Dentz wrote:
> > > > Hi Pali,
> > > > On Sat, Jan 19, 2019 at 7:19 PM Pali Rohár <pali.rohar at gmail.com> wrote:
> > > > > I think that this patch series is now ready for review and testing. All
> > > > > needed parts are finished.
> > > > > 
> > > > > The only problem is with asynchronous profile switching as function
> > > > > pa_card_set_profile() does not support it. All details are in patch
> > > > > "[PATCH v5 5/7] bluetooth: Modular API for A2DP codecs".
> > > > > 
> > > > > So can you look at this patch series and review it?
> > > > 
> > > > I tested this today, it seems to work fine with sbc but with
> > > > faststream there seems to be some problem somewhere as it does seems
> > > > to consume too much cpu time. Regarding set_profile that is something
> > > > I think we will need to address otherwise the UI is pretty much
> > > > useless to switch as it doesn't seems to react well when -EAGAIN is
> > > > returned, we could perhaps block on SetConfiguration since we do block
> > > > on Acquire anyway, or you have already tried that?
> > > 
> > > Yes, I tried it. Pulseaudio blocks communication and therefore it also
> > > does not receive following DBus requests from bluez and so
> > > SetConfiguration timeout.
> > > 
> > > In fact -EAGAIN is just dropped in pulseaudio bluetooth module and
> > > replaced by -1. I just wanted to "highlight" that part of code, so I
> > > used -EAGAIN, but it has same meaning as any other non-zero value.
> > > 
> > > I was thinking about it and we could block this on another layer.
> > > pa_card_set_profile() is called when application request to change
> > > profile via PA socket. So extend pa_card_set_profile() API to pass
> > > callback function which would be called after profile is changed
> > > asynchronously. And just block user request until callback is called.
> > > 
> > > Maybe some other people who understand pulseaudio better can help? How
> > > to implement asynchronous profile switching?
> > 
> > To me it seems that either pa_card_set_profile() should be made
> > asynchronous, which would be nice, because it would allow errors to be
> > reported properly, or the bluetooth code should just initiate the
> > profile change operation and pretend that the switching always
> > succeeds. In the latter option, in case of failure the bluetooth code
> > would switch to the off profile.
> > 
> > I'm fine with either option.
> 
> Seems that immediate switch to profile is easier as implementing whole
> asynchronous operations.

Implemented in v6 and it is working fine!

If asynchronous operations are needed, it should be implemented in
future and bluetooth code can be then switched to that API.

-- 
Pali Rohár
pali.rohar at gmail.com
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 195 bytes
Desc: not available
URL: <https://lists.freedesktop.org/archives/pulseaudio-discuss/attachments/20190203/61338e85/attachment.sig>