Replace PulseAudio with OSS v4?

Daniel Chen seven.steps at
Tue Jun 23 23:26:27 UTC 2009

On Sun, Jun 21, 2009 at 1:43 PM, Davyd McColl<davydm at> wrote:
> Oddly enough, pre-PA, I've never seen any kind of lockup on the SBLive. And

You're lucky. Some revisions of the EMU10k did awful, racy things.

> must point out that the latency issue, whilst more pressing for audio
> professionals, also steps into the user's realm when a game's audio doesn't
> align with the graphics on-screen. Someone playing a game, whilst not
> requiring sub 5ms latency, would probably appreciate sub-50ms latency.

Many of the sync issues are PulseAudio _and_ application bugs (e.g.,
the PulseAudio and xine-lib/MPlayer "pause" one from last dev cycle),
so it isn't that low latency is insignificant on the priority list for
PA but that reworking PA's mainloop and timer architectures have the
side effect of greatly improving both latency and resource use.

> able to contribute, if I can work the time in. So, point me at a good place
> to start, and perhaps I can be more help than just a lazy biscuit next to
> the hard-working tea.

Historically, Ubuntu has carried a shedload of backported (from PA
git) patches. I would like, and am working with Luke, to minimise
these patches for Karmic's PA. 0.9.16-test1 was tagged recently, and
it will be available for testing shortly.

Periodically, the question of how to contribute arises, so I'll address it here:

If you have C (and/or GTK) or C++ (and/or Qt) experience, then
consider working in upstream's Trac bug tracker. Some of the Launchpad
bugs affecting the pulseaudio source package are Ubuntu-specific; I'll
work on (and welcome assistance in) tagging them as (Ubuntu)
distro-specific. As Karmic's pulseaudio source sheds its
distro-specific bits, the benefits are apparent, since all Linux
distros face similar bugs.

If you don't feel comfortable contributing source code, then the Linux
audio realm is sorely lacking in test harness(es). There are no unit
tests in ALSA, PulseAudio, etc. There are no end-to-end tests defined
(e.g., "for this new HP Mini, attempt to stream this Ogg Vorbis file
to an identical HP Mini over an 802.11g network"). All distributions
will benefit by discussing and implementing them.


More information about the Ubuntu-devel-discuss mailing list