Patch to alsa-plugins causes regression in java applet (webex)

[Tom Ellis]

# Issue

* Patch to alsa-plugins causes regression in java applet (webex)

## Environment

Ubuntu 10.04.2 LTS x86
sun-java6-plugin-6.24-1build0.10.04.1
linux-backports-modules-alsa-2.6.32-32-generic
libasound2-plugins-1.0.22-0ubuntu6
pulseaudio 1:0.9.22~0.9.21+stable-queue-32-g8478-0ubuntu14

Dell Latitude 2120

## Resolution

* (workaround) Disable pulseaudio and use alsa directly
or
* Revert patch debian/patches/lp464008lp503174.patch which ignores underruns

## Diagnostic Steps

1. Connect to a webex meeting
2. Speak to a user, confirm mic and audio is working correctly
3. Shout loudly/tap microphone and watch java app adjust mic volume
4. Observe that with headphones you can no longer hear the other user
5. Use package that reverts patch from http://people.canonical.com/~trellis/alsa/libasound2-plugins_1.0.22-0ubuntu6~tellis01_i386.deb
6. Repeat and notice all works well

## Other information

Ran pulseaudio debug logs for both good and bad packages and attaching to bug. Note you can see when the java plugin changes
the volume and after this point it stops putting audio through headphones/speakers.

[David Henningsson (diwic)]

We can't simply revert the patch as that would cause issues in other applications. I think we need to extend PulseAudio's protocol to include information about when the underrun happened in order to determine whether the underrun should be ignored or not.

[Tom Ellis (tellis)]

OK, will you work on a patch that helps this with upstream?

[Tom Ellis (tellis)]

Hi David, I see bug #503174 has a fix for underruns too, I'm going to test this with webex.

[Tom Ellis (tellis)]

Fix from #503174 had no affect. Reverting lp464008lp503174.patch still works.

[Tom Ellis (tellis)]

Thanks David, do you have any packages with these patches so I can test them on the Latitude 2120?
I can knock one together if the answer is no :)

[David Henningsson (diwic)]

Well, the PulseAudio patch is included in the next version of PulseAudio, which we are currently considering for Oneiric (see https://lists.launchpad.net/ubuntu-audio-dev/msg00212.html ), and the alsa-plugins patch is broken, but I'll attach a fixed version now. Note that the alsa-plugins patch would need the new PulseAudio to compile.

[Slava Goltser (slavikg)]

Since Oneiric is a good 2 months away, is there any chance of the "fixed" alsa-plugins being backported to Natty?

[David Henningsson (diwic)]

@Slava, given the changes to the PulseAudio protocol that are preconditions to this additional protocol update, backporting this into Natty seems highly unlikely. Sorry.

[Launchpad Janitor (janitor)]

This bug was fixed in the package alsa-plugins - 1.0.24-0ubuntu6

---------------
alsa-plugins (1.0.24-0ubuntu6) oneiric; urgency=low

  * Use the new PulseAudio API for better underrun handling (LP: #805940)
 -- David Henningsson <email address hidden> Wed, 24 Aug 2011 10:16:12 +0200

[Tom Ellis (tellis)]

is there anything that can be done to sort this for lucid? other than using a custom package?

[David Henningsson (diwic)]

@Tellis: Doing an SRU to change the behavior is not an option because that will regress other applications, and the better solution released in Oneiric requires the new PulseAudio API (which is not backportable to Lucid).

However, there might actually be ways to configure this behavior, but I don't know if such a configuration change would be easier to roll out than a custom package. I'm lacking knowledge of what is possible and not possible in your environment, so it's hard for me to know what to suggest.

So...maybe fixing webex not to abuse the API? :-)

[Tom Ellis (tellis)]

Thanks David, I don't think getting webex to change anything is possible.

It seems to me like the best approach is to run the slightly modified alsa-plugin package with the one patched removed and lock the package.

[Mantas Kriaučiūnas (mantas)]

It seems this bug also causes problems with WINE Windows compatibility layer, see http://wiki.winehq.org/Sound

[Chris J Arges (arges)]

This problem is still seen in 32-bit Precise. I've captured a log of pulseaudio of the failure. We are getting the same buffer Underrun.
I'm able to reproduce using webex using sun java 1.6.

[Sean DS (se4n-1)]

Unfortunately I still have this bug in 64 bit Precise.

[Billy_bonk621 (billy-bonk621)]

I confirm the instance of the bug under Ubuntu 12.04, wine 1.4.1
Using Civilization 4, Rise of Mankind - A New Dawn Mod

Sound will just drop out, or will play a 2x speed occasionally.

Most of play it remains fine, but will happen eventually.

[Billy_bonk621 (billy-bonk621)]

*Edit* Ubuntu 12.04 64 bit
Wine 1.4.1
Civilization IV, Beyond the Sword, Rise of Mankind - A New Dawn mod

[Chris J Arges (arges)]

I've done some additional testing recently using the above setup, and three versions of Ubuntu with and without Pulseaudio.

Java Version

Oracle JRE 7 (1.7.0_05) PulseAudio Enabled 12.04 i386 Fail [1]: 2
Oracle JRE 7 (1.7.0_05) PulseAudio Disabled 12.04 i386 Works: 6

Sun JRE (1.6.0_31) PulseAudio Enabled 11.10 i386 Works: 3
Sun JRE (1.6.0_31) PulseAudio Disabled 11.10 i386 Works: 2
Oracle JRE 7 (1.7.0_05) PulseAudio Enabled 11.10 i386 Works: 2
Oracle JRE 7 (1.7.0_05) PulseAudio Disabled 11.10 i386 Works:1

Sun JRE (1.6.0_31) PulseAudio Enabled 10.04 i386 Fail [1]: 1
Sun JRE (1.6.0_31) PulseAudio Disabled 10.04 i386 Works: 1

[Chris J Arges (arges)]

I hit the return key before completing my last comment accidentally.

The table shows the java version, if pulse audio is enabled, which version of Ubuntu, and if it failed or worked.
In summary, if pulse audio is disabled it works in all versions of Ubuntu (Lucid, Oneiric, Precise)
If Pulse audio is enabled, it only works in Oneiric and not in Precise nor Lucid.

[Launchpad Janitor (janitor)]

Status changed to 'Confirmed' because the bug affects multiple users.

[Roman Shipovskij (roman-shipovskij)]

After many tests, the problem was solved by editing /etc/pulse/default.pa and replacing string:
load-module module-udev-detect
by
load-module module-udev-detect tsched=0

seurce: https://wiki.archlinux.org/index.php/PulseAudio#Glitches.2C_skips_or_crackling

[Rolf Leggewie (r0lf)]

oneiric has seen the end of its life and is no longer receiving any updates. Marking the oneiric task for this ticket as "Won't Fix".

[Rolf Leggewie (r0lf)]

lucid has seen the end of its life and is no longer receiving any updates. Marking the lucid task for this ticket as "Won't Fix".

[Steve Langasek (vorlon)]

The Precise Pangolin has reached end of life, so this bug will not be fixed for that release