Environment Variables

In general, there is one easy way to give Phonon devs all the information they need to help fix your problem.

Set these environment variables:

$ export PHONON_DEBUG=5
$ export PHONON_PULSEAUDIO_DEBUG=5
$ export PHONON_VLC_DEBUG=5
$ export PHONON_GST_DEBUG=5
$ export PHONON_XINE_DEBUG=5

Then run your program. The terminal will fill up with gobs of debugging output.

Phonon-GStreamer debugging

Even more verbose output

Phonon-gst has some fancier debug options available. In addition to PHONON_GST_DEBUG, there is PHONON_GST_GST_DEBUG. Setting it to 8 or so will cause the gstreamer libraries to produce copious quantities of debug output. We're talking thousands of lines of what seems to be useless noise. Roughly 1% of it is useful, but the phonon-gst devs can easily decipher it and drill down to the important bits.

Pulseaudio

Pulseaudio and phonon try their hardest to get along. Sometimes, it doesn't work. Luckily, there is a way to test gstreamer and pulseaudio to see who is at fault.

$ gst-launch filesrc location=/usr/share/sounds/KDE-Sys-Log-In.ogg ! decodebin2 ! audioresample ! audioconvert ! pulsesink
$ gst-launch filesrc location=/usr/share/sounds/KDE-Sys-Log-In.ogg ! decodebin2 ! audioresample ! audioconvert ! alsasink

If they both work flawlessly, blame Phonon-GStreamer.

If the first works but the second doesn't, you might have an exotic sound setup that involves tweaking the alsasink parameters to reflect what pulseaudio does to alsa.

If the second works but the first doesn't, you can blame pulseaudio.

Another way to confirm this is by setting the PHONON_GST_AUDIOSINK environment variable. Setting it to e.g. "pulsesink" uses the pulseaudio sink.

Recreating the phonon-gst pipeline

Phonon-GStreamer creates predicable pipelines. In general, they look like this:

filesrc -> decodebin2 -> queue -> audioresample -> audioconvert -> pulsesink
                   \
                    -> ffmpegcolorspace -> queue -> xvimagesink

If the input stream isn't coming from a file, it is likely coming in via KIO which pipes it into an abstractmediastream. If pulseaudio isn't used, then replace pulsesink with something of alsasink, osssink, or somesuch.

Recreating the playback stream of some video file can be done as such:

$ gst-launch filesrc location=/path/to/video ! decdebin2 name=dec ! audioresample ! audioconvert ! pulsesink \
    dec. ! ffmpegcolorspace ! xvimagesink

DVD playback is a bit different:

  (subpicture stream)
      /--->---\
rsndvdbin -> dvdspu -> ffmpegcolorspace -> queue -> xvimagesink
      \-> queue -> audioresample -> audioconvert -> pulsesink

This is built with the following:

$ gst-launch rsndvdbin ! dvdspu ! ffmpegcolorspace ! queue ! xvimagesink \
    rsndvdbin0 ! dvdspu0.subpicture \
    rsndvdbin0 ! queue ! audioresample ! audioconvert ! pulsesink