Description of problem: libpulsedsp.so is needed to run OSS applications in compatibility mode. As such, to run 32-bit OSS apps the 32-bit libpulsedsp.so is needed. Currently libpulsedsp.so is packaged with pulseaudio-utils, which contains binary executables, and thus 32-bit and 64-bit versions cannot be installed in parallel. Moving this to pulseaudio-lib (which is installed multilib by default) would allow apps such as Real Player to work, which I've tested by extracting the .so file and putting it in /usr/lib Version-Release number of selected component (if applicable): pulseaudio-0.9.7-0.17.svn20071017.fc8 How reproducible: Always Steps to Reproduce: 1. On an x86_64 system, try and run RealPlayer (unstable version needed) when the sound device is in use by PulseAudio 2. Now set Real Player's output device to OSS, and start RealPlayer using 'padsp realplay' 3. Extract libpulsedsp.so from the 32-bit pulseaudio-utils and drop it in /usr/lib, and repeat (2) Actual results: (1) would fail, (2) would also fail as the 32-bit lib is not present. (3) works. Expected results: (2) should work out-of-the-box if libpulsedsp.so is moved to pulseaudio-lib Additional info:
Created attachment 254851 [details] Patch spec file to move libpulsedsp.so
+1 for additionally providing a 32bit version of libpulsedsp.so on x86_64 systems.
But it doesn't belong in pulseaudio-lib. It is not a client library anyone should link to. It belongs to the padsp script and nothing else. And padsp is part of pulseaudio-utils. multilib is our downfall. I guess this whole multilib mess leaves me with no other option than to add yet another package "pulseaudio-utils-lib" to our already large package farm.
*** Bug 426882 has been marked as a duplicate of this bug. ***
(In reply to comment #3) > But it doesn't belong in pulseaudio-lib. It is not a client library anyone > should link to. It belongs to the padsp script and nothing else. And padsp is > part of pulseaudio-utils. > > multilib is our downfall. > > I guess this whole multilib mess leaves me with no other option than to add yet > another package "pulseaudio-utils-lib" to our already large package farm. well its ugly but its worse than non working sound for i386 oss aps.
Actually also to run i386 alsa programs over a pulseaudio alsa device the i386 version of libasound_module_pcm_pulse.so is needed in multilib which it currently is not (alsa-plugins-pulseaudio package).
#6 addresses a separate bug -- Andreas, could you file that bug?
(In reply to comment #3) > But it doesn't belong in pulseaudio-lib. It is not a client library anyone > should link to. It belongs to the padsp script and nothing else. And padsp is > part of pulseaudio-utils. > > multilib is our downfall. > > I guess this whole multilib mess leaves me with no other option than to add yet > another package "pulseaudio-utils-lib" to our already large package farm. I could also be fixed by adding pulseaudio-utils to the multilib whitelist.
No, because pulseaudio-utils contains platform-specific binaries. The solution would be to split the libraries out of pulseaudio-utils
(In reply to comment #9) > No, because pulseaudio-utils contains platform-specific binaries. The solution > would be to split the libraries out of pulseaudio-utils that should not be a problem rpm should take care of this (on x86_64 the x86_64 binaries would be preferred).
What's the status on this? The following two packages need to be added to the whitelist, at least: pulseaudio-esound-compat-0.9.10-1.fc9.i386 pulseaudio-utils-0.9.10-1.fc9.i386
pulseaudio-utils has been whitelisted