[lucid] SRU: OpenAL minor version bump

I should note I really want this version in Lucid as it will be impossible for me to cleanly ship my own separate (newer) openal in a PPA as openal needs to also be seeded into ia32-libs

Scott, we're talking about openal-soft, are we? (openal is the package released from creative, which to my knowledge was dropped in favor of openal-soft a few releases ago).

testing rdepends: can you provide a symbol diff for the library?

Reinhard: you're an uploader of openal-soft in Debian, mind to share your opinion?

Finally, based on my own observations from very long ago, I'm very sceptical to upgrade openal-soft due to its history. I think a full diff would at least be worthwhile to review. Can you attach one please?

Thanks,
   Stefan.

yes, this bug is obviously about openal-soft.

I have to admit that I didn't follow the openal package lately in debian, so I'm not sure I'm really qualified to comment on its inclusion. However, I do know that openal-soft upstream is fairly reasonable and cautious with his releases, so I think this is definitly worth investiaging further.

since we are now in final freeze, I think this needs to go via an SRU

I've tested a source package generated by uscan a fair bit. There are a heap of new symbols in the new version of the library, but they seem to have been added in an ABI-safe manner. I did not encounter any improvements in the pulse handling with the modified package (although I also didn't discover any regressions). I'd like a testcase to see a bug fixed by this change to have a strong opinion about it, but I'd rather not see it as planned-immediate-SRU, as ia32-libs is *painful* for users to update post-install.

Hi,

Just thought as the person who requested the openal-soft update I thought I'd chip in.

Since ubuntu 9.10 hangs indefinitely because of pulseaudio, as can be witnessed at

http://test.winehq.org/data/d4b853695276be6d1492a909a263bcc7ce4ca828/index_Wine.html#mmdevapi:render Because of this wine has decided to blacklist the pulseaudio device in openal-soft if a certain extension that's added in 1.12 is not found. Wine really needs this extension in order for robustness, and making sure that it will not interfere with applications that may happen to use openal, since wine also has a openal32->linux openal stub.

Furthermore openal 1.12 has several pulseaudio fixes not present in earlier releases. Not adding them will result in possible multi-second delays between an action and the sound played, because of the way pulseaudio works. so I really want that version, since wine will do a lot more with openal in the future, and supporting our own version would be a gigantic pain..

Cheers,
Maarten

Also to witness the hang in ubuntu 9.10, link this snipped with -lopenal:
extern void alcOpenDevice(const char *); int main() { alcOpenDevice("PulseAudio Software"); return 0; }

The program should immediately terminate, but instead hangs indefinitely.

I'll note that I want to SRU Wine 1.2 when it fully releases in a couple months to Lucid. So we either put OpenAL 1.12 in now or SRU it then. If we wait then in the meantime I will need to bundle openal 1.12 in the Wine PPA

bump? Kinda urgent. OpenAL-soft is horribly broken, and I added a testcase like you asked..

Hi,

I'd still like to see a symbol diff (Emmet, maybe you can attach yours to this bug?) and a full debdiff (updated source package in a ppa, so that I can debdiff myself will work as well).

Cheers,
   Stefan.

My PPA now has the OpenAL-soft version I tested:

https://launchpad.net/~scottritchie/+archive/ppa

There are a few changes in the headers that at least bear the potential for problems, so NACK from my side so close before the release.

Side note: Maybe carving out the diff of Alc/pulseaudio.c might improve the pulseaudio situation?

The version of OpenAL shipped in Lucid appears to be causing regressions in other apps. MPlayer, for instance, has encountered a very similar bug to what Wine developers were noting: https://bugs.launchpad.net/ubuntu/+source/openal-soft/+bug/575368

To fix this, I propose an SRU of the new minor version originally proposed here. Backporting portions of the code (for instance just the pulseaudio driver) is an option, however depending on how well we understand the underlying OpenAL code this may be more or less risky.

We have to be cautious here since libopenal1 has quite a few reverse dependencies. The changelog looks like it would retain full backwards compatibility with new releases, so I'm fine with getting this uploaded to -proposed now. But verification should not only include wine, but also some other rdepends like mplayer, flightgear, blender, and a couple of other games. So this will probably need a couple of weeks to mature in -proposed.

Note that the upstream release notes for the new minor version do not reflect the apparent gravity of the pulse fixes, namely this new version is critical for usable mplayer and wine among others. Without the pieces, both apps will hang, which is an extremely poor user experience.

This bug was fixed in the package openal-soft - 1:1.12.854-0ubuntu1

---------------
openal-soft (1:1.12.854-0ubuntu1) maverick; urgency=low

  * New upstream release (LP: #565071)
    - Fully backwards compatible
    - Fixed playback when the PulseAudio buffer is calculated to be more than 64KB.
    - Restored compatibility with some older PulseAudio libs.
    - Alternative buffer sizing for PulseAudio, specified using a new config option.
    - Improved buffer size calculations, to prevent drastic latency changes when certain properties (such as ALC_FREQUENCY) are modified.
    - "Not broken" unlike 1.11.753, according to upstream
  * Should fix some remaining crackling issues (LP: #351732, #516435)
  * Should be ok to sync over; no extra patches
 -- Scott Ritchie <email address hidden> Fri, 16 Apr 2010 15:20:20 -0700

Accepted openal-soft into lucid-proposed, the package will build now and be available in a few hours. Please test and give feedback here. See https://wiki.ubuntu.com/Testing/EnableProposed for documentation how to enable and use -proposed. Thank you in advance!

Sound seems to work fine for me in chromium-bsu and osmos (installed from .deb) which are listed under apt-cache rdepends libopenal1
Other games like World of Goo are not in the list, but I'm sure use it run fine but maybe they ship their own versions?
Some issues which I'm having which may or may not be related (sorry if not):
Osmos hangs completely including sound after about 30 mins of gameplay quite often - not sure this can be due to sound but sound certainly stops
0AD no longer starts though that's more probably related to upgrading to the 2nd alpha
Chromium-bsu - setting it to widescreen kills sound completely in spotify running in wine, although that is perhaps pulseaudio related?
Sorry if this is a bit nebulous :/

Wine doesn't yet use OpenAL so that is almost surely the same old pulseaudio Wine bug that's well known (and that OpenAL is an attempt to fix)

Osmos hanging completely could be due to any number of issues, but if it's not a regression we shouldn't be concerned in this report.

Thanks for testing, seems its working reasonably well then.

This bug was fixed in the package openal-soft - 1:1.12.854-0ubuntu1~lucid1

---------------
openal-soft (1:1.12.854-0ubuntu1~lucid1) lucid-proposed; urgency=low

  * New upstream release (LP: #565071)
    - Fully backwards compatible
    - Fixed playback when the PulseAudio buffer is calculated to be more than 64KB.
    - Restored compatibility with some older PulseAudio libs.
    - Alternative buffer sizing for PulseAudio, specified using a new config option.
    - Improved buffer size calculations, to prevent drastic latency changes when certain properties (such as ALC_FREQUENCY) are modified.
    - "Not broken" unlike 1.11.753, according to upstream
  * Should fix some remaining crackling issues (LP: #351732, #516435)
 -- Scott Ritchie <email address hidden> Thu, 13 May 2010 17:39:09 +0200