Thank you very much for reporting. It turned out that -O3 optimization setting triggers compiler bug. Build .18 used -O2 optimization (as it was benchmarked to be faster at that time). Build .21 had optimization settings changed (back) to -O3 as motion compensation code was optimized for size and -O3 started providing positive performance effect. But if we take OSD bug into account, -O2 is clearly the winner now. The next build most likely will have working OSD. But changing optimization option still does not help with #763

I wonder how I missed this optimization option when I was analyzing bug #254, I'm almost sure I tried different optimization settings at that time. Well, nevermind, let's hope it will not show up again

With this additional details, it should be possible (though it can take a lot of time) to pinpoint the problem and find the problematic part of code by splitting code into pieces and compiling them with different optimization options. But I wonder if gcc developers or anyone else is interested in a bugreport against old version of gcc when more recent gcc seems to be ok?

This message is generated by gstreamer after it was not feeded with get audio data in time. This problem is a consequence of mplayer deadlock, but not its cause.