← Back to team overview

ubuntu-audio-dev team mailing list archive

Re: Karmic Caveats: "Another process locking the sound card"

 

On Tue, Apr 13, 2010 at 12:37 AM, David Henningsson
<launchpad.web@xxxxxxxxxxxxxx> wrote:
> Brad Figg wrote:
>> I'm curious to know how specific to karmic this particular issue
>> really is and how frequently we are seeing it.
>
> Perhaps Daniel has a better overview, but I would say not that
> frequently. I believe the main problem with sl-modem-daemon has been
> solved (thanks to Daniel), and timidity is not used *that* much. But
> once in a while somebody is using an unusual app where the developer of
> that app prefers to use "plughw:0" instead of "default", which then
> turns out to cause trouble.

The above description is correct. Also, in dist-upgrades to 10.04 LTS,
I've seen a number of stale ~/.asoundrc* generated by (explicitly
user-invoked) asoundconf (which was removed from our and Debian's
alsa-utils package) that cause issues. The fix in that context is to
remove ~/.asoundrc*. As more onboard HDA ships without support for
multiopen in the driver, it becomes more important that we catch such
cases (apport hook?). However, it isn't any more a pressing issue now
than back in 5.10.

When, in 10.10, we finally begin the migration to a unified audio
platform (everything routed through PA - Kubuntu will finally be on
board thanks to Qt changes by Colin G), things will be better (and
worse - more bug reports!).

> There is also another difficulty - if the sound card is busy when
> Pulseaudio is starting up, PA skips the sound card (and adds a null sink
> if no other sound cards are available), but if the sound card becomes
> busy e g if an app uses plughw directly, that will just cause PA to be
> unable to output to that sink for the time that the app uses it. This is
> somewhat inconsistent, and perhaps something to discuss upstream with
> PA. I would prefer PA to add sinks (and sources) even if they are busy
> and try to grab them later, should they become available.

Last week, I raised this issue with Colin and Lennart in the pulse irc
channel. While there isn't movement now, we'll revisit when 10.10
opens.

Another sticking point is the race in PA between
module-default-device-restore and module-udev-detect (cf. bug 557421).

-Dan



References