Comment 248 for bug 445849

Ok, well, I'm mostly from LFS/Gentoo/Fedora side of the table, but this month I'm working off-site and has to use external HDD with Linux Mint 9 (Ubuntu 10.04 LTS based) Live CD copied into as the main working system. Speakers setup here is an ordinary 5.1 cheap set from Defender connected to the MB's build-in HDA codec. Having problems with PulseAudio had become such a typical thing so I hadn't been surprised to run into high-pith sound distortions I get as soon as set soundcard to work with Analog Out 5.1 profile. For a first few weeks I hadn't had enough time to investigate so the solution was to simply set "autospawn = no" into /etc/pulse/client.conf, 'echo "pcm.!default plug:dmix" > ~/.asoundrc"' and reconfigure all relevant stuff to use alsa instead of pulse (gstreamer-properties, mplayer, vlc, e.t.c.).

Today I finally had found some time to hunt down this bug. Quick google search lead me into this bug report, and reading a bunch of comments made me believe that the problem is related with incorrect PulseAudio SIMD optimizations and that it might be fixed in ubuntu-proposed and/or in the diwic/pulseaudio-testing ppa. Unfortunately both ubuntu-testing and diwic ppa hadn't been able to fix the problem. As for pusle* and libpulse* packages from lucid-testing - nothing seems to be changed after installing them. The problems with 5.1 distorted sound remain the same. Diwic's pulseaudio-testing ppa seems to be maverik-only, but the diwic/ppa seems to had fresh pulseaudio stuff. It is broken the same was as the packages in lucid-proposed though.

Trying to investigate a bit mode I had turned off the PA auto spawning again, killalled pulseaudio and run "PULSE_NO_SIMD=1 pulseaudio" in the terminal. Having SIMD optimizations disabled this way fixes the problem: no more sound distortions caused by "smart and cool daemon every desktop user should use" PulseAudio. For now I'm going to export PULSE_NO_SIMD=1 system wide and maybe wait another year till this thing would be finally fixed in so-called "LTS" release.

I'm quite disappointed in LTS support policy, really. This bug had been known for ages, and releasing such a simple thing as a workaround using aforementioned environment variable should had taken no longer than about a week. After making suffering people happy by applying workaround it would be reasonable to implement and test the real fix (for ex., rewrite SIMD assembly optimizations) and then finally release a patched version that "does it right". Looks like Ubuntu's (and thus derivate distros) support policy is totally the opposite, leaving people suffer from the bug and hunting around for workarounds themselves (like switching into using "Major Market Share OS" or following one of the numerous FAQ about "how to get the f***ing rid of this sh*tty PA on Ubuntu/Mint/Whatever"). *sadface smile*