Background music stops when playing 3D sounds #32096

New Issue

Alex Fraser · 2012-07-13T17:15:49+02:00

Alex Fraser commented

2012-07-13 17:15:49 +02:00

Category: Audio

%%%If a game plays 3D sounds, a looping 2D music track might stop playing.

Attached is a demo file (bg_music_test2.zip). It contains:

One 2D sound that plays in a loop (bird squawking), started with aud via Python
Three 3D sounds that play in a repeating sequence (squeaking noise), started from three different Sound actuators.

If you start the game, the bird should squawk endlessly. But instead, it stops after a few times. The number of times it squawks changes. On my computer it usually stops before it reaches 20. It doesn't wait until the end of a loop before stopping.

Because the test uses 3 sound actuators, I think this test could only ever try to play 4 sounds at once.

Workaround: change the 3D sounds to 2D.

Sorry about the slightly annoying choice of sound samples!

Versions:
Blender: 2.63a, 2.62, 2.61, 2.60a, 2.63.14 r48871
alsa-base 1.0.25+dfsg-0ubuntu1
pulseaudio 1:1.1-0ubuntu15.1
libopenal1: 1.13-2
libsdl1.2debian-pulseaudio: 1.2.14-6.1ubuntu4
libpulse0: 1.0-0ubuntu3.1
OS: Ubuntu 12.04 (64-bit)
Sound cards (happens with all):

USB: Bus 001 Device 028: ID 046d:ba05 Logitech, Inc. ("PureFi Mobile")
PCI: 00:1b.0 Audio device: Intel Corporation 5 Series/3400 Series Chipset High Definition Audio (rev 06)
PCI: 00:1b.0 Audio device: Intel Corporation 5 Series/3400 Series Chipset High Definition Audio (rev 05)
CPU: Intel Core i5 U 430%%%

**Category**: Audio %%%If a game plays 3D sounds, a looping 2D music track might stop playing. Attached is a demo file (bg_music_test2.zip). It contains: - One 2D sound that plays in a loop (bird squawking), started with aud via Python - Three 3D sounds that play in a repeating sequence (squeaking noise), started from three different Sound actuators. If you start the game, the bird should squawk endlessly. But instead, it stops after a few times. The number of times it squawks changes. On my computer it usually stops before it reaches 20. It doesn't wait until the end of a loop before stopping. Because the test uses 3 sound actuators, I think this test could only ever try to play 4 sounds at once. Workaround: change the 3D sounds to 2D. Sorry about the slightly annoying choice of sound samples! Versions: Blender: 2.63a, 2.62, 2.61, 2.60a, 2.63.14 r48871 alsa-base 1.0.25+dfsg-0ubuntu1 pulseaudio 1:1.1-0ubuntu15.1 libopenal1: 1.13-2 libsdl1.2debian-pulseaudio: 1.2.14-6.1ubuntu4 libpulse0: 1.0-0ubuntu3.1 OS: Ubuntu 12.04 (64-bit) Sound cards (happens with all): - USB: Bus 001 Device 028: ID 046d:ba05 Logitech, Inc. ("PureFi Mobile") - PCI: 00:1b.0 Audio device: Intel Corporation 5 Series/3400 Series Chipset High Definition Audio (rev 06) - PCI: 00:1b.0 Audio device: Intel Corporation 5 Series/3400 Series Chipset High Definition Audio (rev 05) CPU: Intel Core i5 U 430%%%

Alex Fraser commented

2012-07-13 17:15:49 +02:00

Changed status to: 'Open'

Joerg Mueller commented

2012-07-13 17:30:52 +02:00

%%%Not a bug. This is a limitation by OpenGL and is plattform depended. If it runs out of sounds it can play at once, it just stops the oldest ones, you have to limit your number sounds played at the same time!%%%

Dalai Felinto commented

2012-07-13 20:01:48 +02:00

%%%Joerg, is there a way to get how many 'available' sounds left the game can use? That would be a way to prevent that from the Python side%%%

Alex Fraser commented

2012-07-15 05:12:29 +02:00

%%%Exactly. I accept that there's a limit to the number of sounds, but how can I make sure my music is never killed? The only way I can see to do it now is to keep my own register of playing handles, and poll each for its status - and even then, I would have to guess at how many can be played at once. What would be really nice is to have a system of priorities, something like:

dev.play(factory, priority)

such that a new sound would never replace an already playing sound with a higher priority. What do you think?%%%

%%%Exactly. I accept that there's a limit to the number of sounds, but how can I make sure my music is never killed? The only way I can see to do it now is to keep my own register of playing handles, and poll each for its status - and even then, I would have to guess at how many can be played at once. What would be really nice is to have a system of priorities, something like: dev.play(factory, priority) such that a new sound would never replace an already playing sound with a higher priority. What do you think?%%%

Joerg Mueller commented

2012-07-23 19:21:13 +02:00

%%%Well, priorities is an API layer above what audaspace currently supports. It's possible with OpenAL to get a hint about how many sources might be supported, but not a definite number/guarantee. For now you should best limit your sounds to 32 at the same time maximum.%%%

Alex Fraser commented

2012-07-28 17:09:04 +02:00

%%%Hmm, I'm sure this is happening in my game with very few sounds playing - about 2-4 sounds, I think, including the music. But I'm having trouble putting together a repeatable test case. Can you recommend a strategy for debugging? Maybe a couple of places to add printf statements to the code?%%%

Alex Fraser commented

2012-07-29 04:26:44 +02:00

%%%OK, I think I have a good test now; see bg_music_test2.zip. It has:

One 2D sound that plays in a loop (bird squawking), started with aud via Python
Three 3D sounds that play in a repeating sequence (squeaking noise), started from three different Sound actuators.

If you start the game, the bird should squawk endlessly. But instead, it stops after a few times. The number of times it squawks changes. On my computer it usually stops before it reaches 20. It doesn't wait until the end of a loop before stopping.

Because the test uses 3 sound actuators, I'm pretty sure this test could only ever try to play 4 sounds at once. Is that right?

Sorry about the slightly annoying choice of sound samples!

Versions:
Blender 2.63.14 r48921
Ubuntu 12.04 (64-bit)
alsa-base 1.0.25+dfsg-0ubuntu1
pulseaudio 1:1.1-0ubuntu15.1

Sound cards (happens with both):

USB: Bus 001 Device 028: ID 046d:ba05 Logitech, Inc. ("PureFi Mobile")
PCI: 00:1b.0 Audio device: Intel Corporation 5 Series/3400 Series Chipset High Definition Audio (rev 06)

Probably not important: sometimes I get the following message printed to the console:

  AL lib: pulseaudio.c:331: PulseAudio returned minreq > tlength/2; expect break up

If I restart pulseaudio that message goes away, but the test still fails (bird stops squawking).%%%

%%%OK, I think I have a good test now; see bg_music_test2.zip. It has: - One 2D sound that plays in a loop (bird squawking), started with aud via Python - Three 3D sounds that play in a repeating sequence (squeaking noise), started from three different Sound actuators. If you start the game, the bird should squawk endlessly. But instead, it stops after a few times. The number of times it squawks changes. On my computer it usually stops before it reaches 20. It doesn't wait until the end of a loop before stopping. Because the test uses 3 sound actuators, I'm pretty sure this test could only ever try to play 4 sounds at once. Is that right? Sorry about the slightly annoying choice of sound samples! Versions: Blender 2.63.14 r48921 Ubuntu 12.04 (64-bit) alsa-base 1.0.25+dfsg-0ubuntu1 pulseaudio 1:1.1-0ubuntu15.1 Sound cards (happens with both): - USB: Bus 001 Device 028: ID 046d:ba05 Logitech, Inc. ("PureFi Mobile") - PCI: 00:1b.0 Audio device: Intel Corporation 5 Series/3400 Series Chipset High Definition Audio (rev 06) Probably not important: sometimes I get the following message printed to the console: ``` AL lib: pulseaudio.c:331: PulseAudio returned minreq > tlength/2; expect break up ``` If I restart pulseaudio that message goes away, but the test still fails (bird stops squawking).%%%

Alex Fraser commented

2012-07-29 06:39:34 +02:00

%%%I've done some testing with more versions. The official releases are harder to trigger the bug in. They tend to take longer, and sometimes it's necessary to restart Blender to get the bug to trigger. Doing other things like switching between windows seemed to make it more likely to trigger.
Official releases tested: 2.63a, 2.62, 2.61, 2.60a
Also 2.63 r49322 from http://builder.blender.org/download/blender-2.63-r49322-linux-glibc27-x86_64.tar.bz2

I couldn't get the bug to trigger in 2.59.%%%

%%%I've done some testing with more versions. The official releases are harder to trigger the bug in. They tend to take longer, and sometimes it's necessary to restart Blender to get the bug to trigger. Doing other things like switching between windows seemed to make it more likely to trigger. Official releases tested: 2.63a, 2.62, 2.61, 2.60a Also 2.63 r49322 from http://builder.blender.org/download/blender-2.63-r49322-linux-glibc27-x86_64.tar.bz2 I couldn't get the bug to trigger in 2.59.%%%

Alex Fraser commented

2012-08-04 08:50:43 +02:00

%%%I just confirmed this bug on another computer running Ubuntu 12.04 with Blender 2.63a.

Sound card (similar to mine, but different revision)

PCI: 00:1b.0 Audio device: Intel Corporation 5 Series/3400 Series Chipset High Definition Audio (rev 05)

Versions:
Blender: 2.63a
libopenal1: 1.13-2
libsdl1.2debian-pulseaudio: 1.2.14-6.1ubuntu4
libpulse0: 1.0-0ubuntu3.1

I also compiled a fresh checkout of OpenAL Soft 1.14. It fails too.

Reopening this ticket; I hope you don't mind. If you think this is a problem in one of the libraries, I could file a bug on Launchpad too.%%%

%%%I just confirmed this bug on another computer running Ubuntu 12.04 with Blender 2.63a. Sound card (similar to mine, but different revision) - PCI: 00:1b.0 Audio device: Intel Corporation 5 Series/3400 Series Chipset High Definition Audio (rev 05) Versions: Blender: 2.63a libopenal1: 1.13-2 libsdl1.2debian-pulseaudio: 1.2.14-6.1ubuntu4 libpulse0: 1.0-0ubuntu3.1 I also compiled a fresh checkout of OpenAL Soft 1.14. It fails too. Reopening this ticket; I hope you don't mind. If you think this is a problem in one of the libraries, I could file a bug on Launchpad too.%%%

Alex Fraser commented

2012-08-04 09:05:18 +02:00

%%%Cleaned up description, and removed rejected test case.%%%

Alex Fraser commented

2012-08-06 14:11:44 +02:00

%%%Adding a new test case that uses only aud via Python; no Sound actuators: see bg_music_test3.zip

In this test, aud.Handle.location is used to set the location of the 3D sounds. The locations are updated repeatedly after the sounds start. The script as uploaded updates the locations 1000 times per logic tick, which is very likely to trigger the bug - but on my machine, it happens even when only updating the locations once per tick. You can change the frequency by editing the "range" expression on line 80.

What seems to happen is this: if the location of a sound is changed just as it finishes playing, some other sound will be killed. In this test, the string "W: Tried to set position of stopped sound" is printed when the bug is triggered. It is not triggered if the sound has already been stopped for a while; to test this, comment out the "continue" statement on line 71.

Workaround: This seems to be a race condition. It seems to be fixed by locking:

Lock the device.
Test Handle.status. If the sound has stopped, skip the next step.
Set Handle.location.
Unlock the device.

The test includes the workaround; to see it in action, attach the "Always" sensor to the "SAFE" controller instead of the "BUG" controller.%%%

%%%Adding a new test case that uses only aud via Python; no Sound actuators: see bg_music_test3.zip In this test, aud.Handle.location is used to set the location of the 3D sounds. The locations are updated repeatedly after the sounds start. The script as uploaded updates the locations 1000 times per logic tick, which is very likely to trigger the bug - but on my machine, it happens even when only updating the locations once per tick. You can change the frequency by editing the "range" expression on line 80. What seems to happen is this: if the location of a sound is changed just as it finishes playing, some other sound will be killed. In this test, the string "W: Tried to set position of stopped sound" is printed when the bug is triggered. It is not triggered if the sound has already been stopped for a while; to test this, comment out the "continue" statement on line 71. Workaround: This seems to be a race condition. It seems to be fixed by locking: 1. Lock the device. 2. Test Handle.status. If the sound has stopped, skip the next step. 3. Set Handle.location. 4. Unlock the device. The test includes the workaround; to see it in action, attach the "Always" sensor to the "SAFE" controller instead of the "BUG" controller.%%%

Joerg Mueller commented

2012-10-24 23:34:50 +02:00

%%%Hey!

I tried your file and could trigger the bug finally. I've implemented a fix in revision 51599, could you check if it works for you? If not reopen the bug please.

Regards%%%

%%%Hey! I tried your file and could trigger the bug finally. I've implemented a fix in revision 51599, could you check if it works for you? If not reopen the bug please. Regards%%%

Joerg Mueller commented

2012-10-24 23:34:50 +02:00

Changed status from 'Open' to: 'Resolved'

Joerg Mueller closed this issue

2012-10-24 23:34:50 +02:00

Alex Fraser commented

2012-11-04 23:13:35 +01:00

%%%Hi Joerg! Thanks for the fix. I just tested it again in r51726, and it works fine on my computer.

Cheers :-)%%%

%%%Hi Joerg! Thanks for the fix. I just tested it again in r51726, and it works fine on my computer. Cheers :-)%%%

Sign in to join this conversation.

No Label

Download

What's New

Blender Studio

Manual

Developers Blog

Documentation

Benchmark

Blender Conference

Development Fund

One-time Donations

Background music stops when playing 3D sounds #32096