I have been trying to track down an issue which began after the installation of gstreamer-1.4.5 and associated plugin. I run the linux version of thinkorswim which is a java program. I stream audio from it which ends up in pulse via alsa-plugins. All was working fine until December 26th. I have been running the gnome-overlay version of gnome-3.14 for a few weeks with no issues. Now when sound is coming from thinkorswim and any other sound source plays the volume on the java alsa stream goes up to clipping level and stays there until the other sound ends. Then it returns to normal. I suspect the 1.4.5 gstreamer, but I was running 1.4.3 prior without problems. Cannot easily go back to 1.4.3 because it is out to the tree and overlay. There is some sort of strange interaction and I suspect a possible buffer overrun. Any one else seeing anything like this?
Maybe this is caused by some gst plugins not yet bumped to 1.4.x in the tree... I saw some of them being bumped in last days, maybe they fix this issues (anyway, I think some are still pending)
This is probably from orc bump regressing PA, upstream bug here: https://bugzilla.gnome.org/show_bug.cgi?id=742271 I expect an analysis and solution will happen upstream soon
orc was indeed the problem orc-0.4.23 installed on Dec 24th and problems started on 25th. I downgraded orc to 0.4.19 and rebuilt pulse and all the issues are resolved. I would suggest you mask orc-0.4.23 until pulse is fixed upstream.
note. This bug effects more than just java sound. On my system it also effected mangler and in general causes alsa based sounds to be boosted to clipping while any direct sound source playing. I could even trigger it by hitting a backspace on an empty line in gnome-terminal.
Going back to 0.4.19 sounds too much of a downgrade, as we didn't have other previous versions in. If nothing appears from upstream by the time I get to my development machine and have time, I'll look into adding in a 0.4.22 version, which doesn't have this issue as known by arch linux people
Should be fixed in media-sound/pulseaudio-5.0-r7, but please test - I never was able to find a way to actually reproduce this in an audible way. However without the applied patch (which arch people are saying fixes it), a relevant svolume fails upon PA rebuild with tests, and it starts to pass after applying that patch. So pretty sure we should be good now.
The new pulseaudio resolves the issue for me with orc-0.4.23