On Sat, 2013-01-26 at 12:47 +0200, Tanu Kaskinen wrote: > On Fri, 2013-01-25 at 08:05 +0200, Tanu Kaskinen wrote: > > On Fri, 2013-01-25 at 07:48 +0200, Tanu Kaskinen wrote: > > > On Fri, 2013-01-25 at 11:40 +1100, Luke Yelavich wrote: > > > > On Mon, Jan 21, 2013 at 05:16:44PM EST, Tanu Kaskinen wrote: > > > > > Are you sure you are testing a properly patched alsa-lib? If you are, > > > > > then this is some other alsa-lib bug. ucm_get_modifiers() in pulseaudio > > > > > definitely doesn't do any double-freeing. > > > > > > > > Yes, I have verified that I was testing with your referenced patch, > > > > and alsa 1.0.26 twice, and I still get the crash. > > > > > > > > For reference, I am also attaching the UCM files that are used in > > > > Ubuntu for the PandaBoardES, which is what I am using. I haven't > > > > verified their syntax beyond making sure they are packaged. > > > > > > With that configuration (with hw:PandaBoardES replaced with hw:Intel) I > > > get a UCM crash too, but it's a different backtrace. I'll investigate > > > this. > > > > That was with Debian's alsa-lib 1.0.25. After installing the current > > development version, I don't get the crash anymore, nor do I get a crash > > with 1.0.26. I could try editing the UCM configuration a bit more to get > > rid of the errors that I'm currently getting (naturally PandaBoard > > configuration doesn't work very well with Intel HDA). > > Sorry, the reason why I didn't get crashes with newer alsa-lib versions > was that the location of the UCM configuration files changed from /usr > to /usr/local. I moved the configuration files to the right place, and > I'm getting a crash again (this time in a non-UCM related malloc() call, > which would be rather nasty to debug, but luckily we already know the > likely place where the invalid memory freeing is happening). I sent a fix to alsa-devel: http://thread.gmane.org/gmane.linux.alsa.devel/104990 -- Tanu
