Enable pulse as the default IO handling in Chrome

This patch enables pulse as the default IO handling in Chrome.

Right now we have the dynamic linking for pulse library, for the machine without pulse, it will automatically fall back to ALSA.

In order to enable pulse as the default IO, libpulse-dev package is required to build Chrome on linux. Previously libpulse-dev is only required if building chromeOS, right now it is also required by building chrome.


Note, we still have some small issues for the pulse implementations, so this patch needs to be landed after https://codereview.chromium.org/12328097/,
https://codereview.chromium.org/12316131/, https://codereview.chromium.org/12310102/ and
https://codereview.chromium.org/12613005/


BUG=178101
TEST=compile on linux

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=187677

[no longer working on chromium, 2013-02-25 14:45:45]

Hi Dale and Andrew,

I am going to enable pulse as the default IO in Chrome Linux, for the machine
without pulse, it will automatically fall back to ALSA.
 
Could you please review this patch?

Thanks
SX

[DaleCurtis, 2013-02-25 19:11:23]

This seems reasonable, but I thought we were going to add a
GetNativeSampleRate() method for Pulse and use that for output configuration to
avoid glitching?

[no longer working on chromium, 2013-02-25 20:05:26]

On 2013/02/25 19:11:23, DaleCurtis wrote:
> This seems reasonable, but I thought we were going to add a
> GetNativeSampleRate() method for Pulse and use that for output configuration
to
> avoid glitching?

If we land https://codereview.chromium.org/12328084/, the audio is good.
But I agree that we should use the native sample rate to further optimize the
audio.

[no longer working on chromium, 2013-02-26 18:31:50]

I updated the descriptions, and this CL is depending on 
Dale's CL https://codereview.chromium.org/12328097/, which fixes the glitching;
https://codereview.chromium.org/12316131, which makes the Pulse to use the
native sample rate.

[scherkus (not reviewing), 2013-02-26 21:39:41]

Deferring to dalecurtis@ to review.

Please send a PSA to chromium-dev@  a few days before you plan on landing this.

Example:
https://groups.google.com/a/chromium.org/forum/#!msg/chromium-dev/Nv5xigbqWls...

[no longer working on chromium, 2013-03-06 19:28:55]

Dale, I tested the pulse a bit today, and it sounds OK. Do you want to me test
more before landing this?

Please send me the links if you want me to do more tests.

Thanks
SX

[DaleCurtis, 2013-03-06 20:18:05]

Let me give it a once over tomorrow and then we can land the day after that.
Please go ahead and write up a PSA to send out to chromium-dev as scherkus
suggested.

[DaleCurtis, 2013-03-06 20:18:24]

(the PSA should be sent out before we land this)

[cmp, 2013-03-07 17:55:22]

Please file a ticket at crbug.com using the build infrastructure template and
request that all of the bots be updated to re-run install-build-deps.sh to pick
up this change.

You will need to change the order of operations here.  You should land the
change that modifies install-build-deps.sh first, wait for the bots to be
updated, then land the change that uses the package.  LMK if you have any
questions.

[DaleCurtis, 2013-03-11 22:06:46]

fischman also mentioned that if we statically link Pulse into Chrome, it will
work properly on systems that don't have PulseAudio installed.  Do you know much
about this xians?

Assuming the Pulse drivers work out, we should consider deleting the ALSA path
in favor of statically linking Pulse in.  According to the web Pulse has built
in support for: ALSA, Jack, ESD, and OSS.

I spent a couple hours listening to music, watching videos, playing games, and
trying out WebAudio demos today.  I haven't found any issues with the playback
quality, so once the bot infrastructure is in place and you add a UMA stat:
lgtm.

https://codereview.chromium.org/12310101/diff/13001/media/audio/linux/audio_m...
File media/audio/linux/audio_manager_linux.cc (right):

https://codereview.chromium.org/12310101/diff/13001/media/audio/linux/audio_m...
media/audio/linux/audio_manager_linux.cc:326: AudioManager* manager =
AudioManagerPulse::Create();
fischman had a great suggestion: You should add a UMA stat for how often this
route is taken so we know our relative percentage of pulse users.

[no longer working on chromium, 2013-03-11 23:24:05]

Thanks Dale.

On 2013/03/11 22:06:46, DaleCurtis wrote:
> fischman also mentioned that if we statically link Pulse into Chrome, it will
> work properly on systems that don't have PulseAudio installed.  Do you know
much
> about this xians?
> 

+ Ami for more details.

That is not what I know. Doing pulse static linking will fail launching chrome
on machines without pulse lib, that is why we use dynamic linking here.


> Assuming the Pulse drivers work out, we should consider deleting the ALSA path
> in favor of statically linking Pulse in.  According to the web Pulse has built
> in support for: ALSA, Jack, ESD, and OSS.
>

If pulse can run on machines without pulse, I will be happy to do so. But it is
a bit hard for me to believe, our pulse IO is playing sound via pulse server,
for machine without pulse, I can't imagine a way to inject pulse server to
clients' machine.

 
> I spent a couple hours listening to music, watching videos, playing games, and
> trying out WebAudio demos today.  I haven't found any issues with the playback
> quality, so once the bot infrastructure is in place and you add a UMA stat:
> lgtm.
> 
>
https://codereview.chromium.org/12310101/diff/13001/media/audio/linux/audio_m...
> File media/audio/linux/audio_manager_linux.cc (right):
> 
>
https://codereview.chromium.org/12310101/diff/13001/media/audio/linux/audio_m...
> media/audio/linux/audio_manager_linux.cc:326: AudioManager* manager =
> AudioManagerPulse::Create();
> fischman had a great suggestion: You should add a UMA stat for how often this
> route is taken so we know our relative percentage of pulse users.

Great idea, I will address it tomorrow.

[Ami GONE FROM CHROMIUM, 2013-03-11 23:59:23]

My comment to Dale was that pulse can obviously drive ALSA internally
(that's how it plays audio on glucid/gprecise desktops, for example), and
that I _believe_ one can link that functionality directly into one's
binary, so that instead of dynamically linking a client (for a server that
may or may not be present), one effectively links in the pulse server.
But I've not done this, so take it with a grain of salt.  It would be good
to verify whether this is possible or not.

The reason I care is that now we're up to _6_ audio paths (in/out *
cras/alsa/pulse), for a total userbase that's overall pretty small; IWBN to
cut out at least 2 of those, and if pulse could subsume alsa, we could drop
the alsa code.

[no longer working on chromium, 2013-03-12 16:08:29]

On 2013/03/11 23:59:23, Ami Fischman wrote:
> My comment to Dale was that pulse can obviously drive ALSA internally
> (that's how it plays audio on glucid/gprecise desktops, for example), and
> that I _believe_ one can link that functionality directly into one's
> binary, so that instead of dynamically linking a client (for a server that
> may or may not be present), one effectively links in the pulse server.
> But I've not done this, so take it with a grain of salt.  It would be good
> to verify whether this is possible or not.
> 
> The reason I care is that now we're up to _6_ audio paths (in/out *
> cras/alsa/pulse), for a total userbase that's overall pretty small; IWBN to
> cut out at least 2 of those, and if pulse could subsume alsa, we could drop
> the alsa code.

OK, I see. I know we can use ALSA to drive Pulse, but not sure if we can do it
the other way around. It is definitely interesting if it is possible.

I agree that no one likes so many IO implementations for Linux. That is also one
of the reasons why we use ALSA for such a long time even thought having lots of
audio issues for years.
I have already added the UMA histogram, lets get some numbers before we decide
if we should take one more step here.

https://codereview.chromium.org/12310101/diff/13001/media/audio/linux/audio_m...
File media/audio/linux/audio_manager_linux.cc (right):

https://codereview.chromium.org/12310101/diff/13001/media/audio/linux/audio_m...
media/audio/linux/audio_manager_linux.cc:326: AudioManager* manager =
AudioManagerPulse::Create();
On 2013/03/11 22:06:46, DaleCurtis wrote:
> fischman had a great suggestion: You should add a UMA stat for how often this
> route is taken so we know our relative percentage of pulse users.

Done.

[commit-bot: I haz the power, 2013-03-12 18:00:35]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/xians@chromium.org/12310101/24001

[commit-bot: I haz the power, 2013-03-12 22:57:00]

Change committed as 187677