On 31 January 2013 07:37, Ian Malone <ibmalone at gmail.com> wrote: > On 31 January 2013 12:09, Tanu Kaskinen <tanuk at iki.fi> wrote: >> On Wed, 2013-01-30 at 09:30 +0100, Mikel Astiz wrote: >>> From: Mikel Astiz <mikel.astiz at bmw-carit.de> >>> >>> >>> Changes in v3: >>> - Patches should now apply cleanly on master. >>> - Comment modified considering David's proposal. >>> - name_owner memory leak fixed. >>> >>> Mikel Astiz (2): >>> reserve: Move get_name_owner() to the public rd_device API >>> reserve: Fix leaking NameLost signals after release+acquire >> >> Brendan and Ian, I believe these patches will finally solve the issues >> that you've been having with using pulseaudio and jack together. >> > > Thanks, and thanks to Mikel for these patches. I've been watching this > one with interest since it looks like it will handle the self-request > occurrence. Not sure when I'll have time to test it properly though. > Particularly good if it does fix it that these changes are contained > within the reserve code alone as I gather from earlier discussion > that's actually regarded as a self-contained library. > The 'move get_name_owner' and 'fix leaking namelost signals' commits can't be applied cleanly to 2.99.1 or 2.1 as they depend on other commits that have been made. Instead I lifted the complete reserve-monitor.c, reserve.c and reserve.h from the second commit http://cgit.freedesktop.org/pulseaudio/pulseaudio/commit/?id=58def1fd1de0de7d5732519539503a7ad83f24a2 and tried them in 2.99.1 and Fedora's 2.1-5 package (since it sounds like the fix should be self contained), and they do solve the problem (dbus interface org.freedesktop.ReserveDevice1.AudioN no longer destroyed while /org/freedesktop/ReserveDevice1/AudioN still held). Tested both on VM and hardware with multiple audio devices. -- imalone http://ibmalone.blogspot.co.uk
