replace snd_pcm_hw_params_set_* with snd_pcm_set_params.

It seems that snd_pcm_hw_params_set_* API does not work well any more in the latest ubuntu like 12.04. My tests show that it will freeze the pulseaudio for a while and get into underrun state. And it seems snd_pcm_set_params works better for the same purpose.

This patch also fixes the choppy audio issue for webrtc in ubuntu 12.04.

BUG=152232, 153505
TEST=use pepper flash (./out/Debug/chrome -ppapi-flash-path=/your/path/libpepflashplayer.so)
open link http://www.youtube.com/watch?v=F7pYHN9iC9I&feature=youtu.be in two tabs, audio should be good.
uses apprtc.appspot.com to make a loopback call and verify the audio is good on ubuntu 12.04

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=159680

[DaleCurtis, 2012-10-01 18:59:15]

lgtm, but this is mostly a rubber stamp. Local testing shows this fixes the
fast-forward issue for me.

http://codereview.chromium.org/11019010/diff/1/media/audio/linux/alsa_wrapper.cc
File media/audio/linux/alsa_wrapper.cc (right):

http://codereview.chromium.org/11019010/diff/1/media/audio/linux/alsa_wrapper...
media/audio/linux/alsa_wrapper.cc:79: SND_PCM_ACCESS_RW_INTERLEAVED,
What is the difference between SND_PCM_ACCESS_MMAP_INTERLEAVED and this?

[no longer working on chromium, 2012-10-01 19:51:13]

I will update the code and description before land it.
thanks.

http://codereview.chromium.org/11019010/diff/1/media/audio/linux/alsa_wrapper.cc
File media/audio/linux/alsa_wrapper.cc (right):

http://codereview.chromium.org/11019010/diff/1/media/audio/linux/alsa_wrapper...
media/audio/linux/alsa_wrapper.cc:79: SND_PCM_ACCESS_RW_INTERLEAVED,
On 2012/10/01 18:59:16, DaleCurtis wrote:
> What is the difference between SND_PCM_ACCESS_MMAP_INTERLEAVED and this?

Sorry, it was one of some quick fixes I tried, I should use access instead of
hard coding SND_PCM_ACCESS_RW_INTERLEAVED, though it is the same since we hard
code to SND_PCM_ACCESS_RW_INTERLEAVED in alsa_util:: OpenDevice() anyway.

The difference between them is on the way how to get access to the data.
From Alsa API:
SND_PCM_ACCESS_MMAP_INTERLEAVED:
   mmap access with simple interleaved channels SND_PCM_ACCESS_RW_INTERLEAVED:
  snd_pcm_readi/snd_pcm_writei access

[scherkus (not reviewing), 2012-10-01 19:52:59]

OOC have you reached you to pulseaudio community to see if there's a possible
regression in pulseaudio?

[no longer working on chromium, 2012-10-02 10:09:47]

On 2012/10/01 19:52:59, scherkus wrote:
> OOC have you reached you to pulseaudio community to see if there's a possible
> regression in pulseaudio?

Not yet, I just got my gprecise a few days ago, and had lots of issues with
audio in the new machine. But it is still not clear to me if the problems are
mainly on the ubuntu settings or pulse server.

Regarding to this change, snd_pcm_set_params is a wrapper built upon the
snd_pcm_hw_params_set_* APIs, but it is much easier to use and robust. If
snd_pcm_set_params works but not snd_pcm_hw_params_set_*, then part of the
problem might be on our side.

[no longer working on chromium, 2012-10-02 10:18:02]

Given we constantly get audio issue with pulseaudio, it makes me believe that it
is very hard to make a ALSA implementation fully compatible with all versions of
pulse. For the long term, I think we should use native pulse implementation
instead, this will gives us optimal performance regarding native sample rate and
device selection, also robust sound on all platforms running pulse server.

[commit-bot: I haz the power, 2012-10-02 10:18:16]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/xians@chromium.org/11019010/2002

[commit-bot: I haz the power, 2012-10-02 12:12:41]

Change committed as 159680

[scherkus (not reviewing), 2012-10-02 13:18:22]

I agree :)

Know anyone who is available?
On Oct 2, 2012 3:18 AM, <xians@chromium.org> wrote:

> Given we constantly get audio issue with pulseaudio, it makes me believe
> that it
> is very hard to make a ALSA implementation fully compatible with all
> versions of
> pulse. For the long term, I think we should use native pulse implementation
> instead, this will gives us optimal performance regarding native sample
> rate and
> device selection, also robust sound on all platforms running pulse server.
>
>
http://codereview.chromium.**org/11019010/<http://codereview.chromium.org/110...
>

[no longer working on chromium, 2012-10-03 09:10:24]

On Tue, Oct 2, 2012 at 3:18 PM, Andrew Scherkus <scherkus@chromium.org>wrote:

> I agree :)
>
> Know anyone who is available?
>
>
I have already spent some of my spare time on putting the code together,
and have CLs which work with basic functionalities like recording and
playing sound.
But they only work on my lucid, but not gprecise (recording is choppy
there).

My plan is to start with the input side, since I have more control here. If
it works well, then we can extend it to the output.

But honestly this won't be ready for review within a few weeks, since I
need to finish some UI and media stream work first.