mplayer sticky pause

[nevion]

Not sure if this is a PA problem or an mplayer problem but... I've got an intel8x0 on my desktop and an hda intel on my laptop. I currently use opensuse 11.1 and MPlayer-1.0rc2_r27637-3.pm.3 and pulseaudio-0.9.12-9.5 . Whenever I unpause mplayer, it likely gets stuck there and while mplayer still reacts to user input, it unfreezes if you seek around. Other times it resumes for an instant, the audio pauses and the video keeps going for IDK, 1.5 seconds - then the audio skips and everything catches back up. And maybe 10% of the time it actually resumes with no problems.

I set ao=pulse or ao=alsa and the result is the same. It does not happen if I prevent alsa from using PA and output directly to it (=killing PA). Didn't happen with older versions way back when so this is a new development. And one that drives me nuts.

[yar1984]

I also vote for this, because I have same problem

[coling]

Works fine for me on -ao alsa with MPlayer 1.0-1.rc2 and pulse 0.9.10 and git master and alsa-plugins 1.0.18.

With -ao pulse does sometimes exhibit this behaviour but this is likely a problem in the mplayer pulse output layer itself and I'm not sure who maintains that.

[nevion]

coling,

I imagine it happens with alsa because I have alsa output by default to pulseaudio - it definitely works if I change that.

[thingwath]

I'm having same problem, if there is a video, mplayer will sometimes continue to play, but never if there is only an audio.

  2 Thread 0xb3727b90 (LWP 13663)  0xb8035416 in __kernel_vsyscall ()
* 1 Thread 0xb792d700 (LWP 13662)  0xb8035416 in __kernel_vsyscall ()

#0  0xb8035416 in __kernel_vsyscall ()
#1  0x004f4996 in nanosleep () from /lib/libpthread.so.0
#2  0x081a9784 in usec_sleep (usec_delay=11000) at osdep/timer-linux.c:25
#3  0x08085b2e in fill_audio_out_buffers () at mplayer.c:2032
#4  main (argc=2, argv=0xbfa34684) at mplayer.c:3705

#0  0xb8035416 in __kernel_vsyscall ()
#1  0xb7b85777 in __poll (fds=0xa295ca0, nfds=2, timeout=332) at ../sysdeps/unix/sysv/linux/poll.c:87
#2  0x04488937 in poll_func (ufds=0xa295ca0, nfds=2, timeout=332, userdata=0xa28ead0)
    at pulse/thread-mainloop.c:71
#3  0x04478320 in pa_mainloop_poll (m=0xa28ea30) at pulse/mainloop.c:857
#4  0x04478707 in pa_mainloop_iterate (m=0xa28ea30, block=1, retval=0x0) at pulse/mainloop.c:923
#5  0x0447878c in pa_mainloop_run (m=0xa28ea30, retval=0x0) at pulse/mainloop.c:941
#6  0x044889b8 in thread (userdata=0xa28ea10) at pulse/thread-mainloop.c:90
#7  0x044bc4f8 in internal_thread_func (userdata=0xa295c38) at pulsecore/thread-posix.c:72
#8  0x004ed51f in start_thread (arg=0xb3727b90) at pthread_create.c:297
#9  0xb7b9004e in clone () at ../sysdeps/unix/sysv/linux/i386/clone.S:130

It just sleep for a while in fill_audio_out_buffers, then asks for pa_stream_writable_size for the stream, result is zero so there is nothing to write, and repeats it forever.

I have mplayer SVN-r28234 and Fedora 10 pulseaudio (pulseaudio-0.9.13-6.fc10.i386).

[nevion]

Got a report over at mplayer for this bug http://bugzilla.mplayerhq.hu/show_bug.cgi?id=1408

Seems there's a reliable method to trigger it now (comments 6 and 8). It would seem that PA is at fault but this is not necessarily the case.

[lennart]

This is fixed now in http://git.0pointer.de/?p=pulseaudio.git;a=commit;h=e61728e67a656c0bbd12b13bb4c1777d0e3163f4

Sorry for the long delay.

[quitr]

I am experiencing nearly identical symptoms using MPlayer SVN-r30325-4.4.2 and pulseaudio 0.9.21 on arch linux. However, in my case it seems to be limited to sticking/freezing after unpause (the audio doesn't go out before the video etc).

Additionally, I get skipping audio in vlc media player (somewhat like OP says he has in mplayer). VLC never freezes on unpause, but the audio will periodically stop, then resume fast to catch up, then play at normal speed, then stop again in a cycle. If I watch pulseaudio's Volume Control's Playback tab, I can see the VLC alsa plugin disappear coinciding with the lapses in audio, and reappearing when the audio returns. VLC version is 1.0.4.

Both of these apps are using alsa (which in turn is using pulseaudio's alsa plugin). Alsa version is 1.0.22.

[orbisvicis]

quitr: Sounds similar to Arch bug#19322 @ http://bugs.archlinux.org/task/19322 which is now closed.

Until pulseaudio moves to extra I build mplayer myself to get the mplayer pulseaudio output driver/module thingy.

I tend to agree that the root problem is in pulseadio. If the mplayer alsa output works fine with alsa, then it shouldn't cause stuttering with pulseaudio, right? The alsa -> pulseaudio -> alsa -> soundcard process should be a transparent drop-in?

On the flip side, with the mplayer audio output set to pulse, I get some audio glitches now and then. Posit that mplayer alsa output can't handle glitching sound, but the cause of the glitching sound is pulse?

mplayer-svn 31147 pulseaudio 0.9.21

[coling]

The mplayer sticky pause issue is definitely fixed. It wasn't to do with lagging or anything like that, it properly hung the mplayer stream.

It sounds like for some reason, the alsa connection is being closed and reopened, but it's certainly not this bug.