[fritsch]

Nice, this time it crashes in the renderer - the intel driver:

#0 0x00007f4d1affbaf9 in ?? () from /usr/lib/x86_64-linux-gnu/dri/i965_drv_video.so

Not nice - try to disable "Prefer VAAPI Output Method" ...

[fritsch]

If you don't already - please update your libva / libva-driver-intel to the ones wsnipex provides in his vaapi ppa. They should at least fix something, most likely not this issue but hunting false positives we could void by doing so.

[illiac4]

I did that already and since then the problems (crashes) seems more often on livetv.
Will remove latest patches and try with older.

[fritsch]

As you build yourself, add that one: https://github.com/FernetMenta/xbmc/issu...t-51034469

[bkuhls]

No, it does not - I just replied you on the bugtracker - you need the relevant ML patch in order to succesfully play that sample till the end. Wsnipex has packaged it within his vaapi ppa.

Sorry to bug you again, but I fail to see the difference between
the patch you mentioned
https://bugs.freedesktop.org/show_bug.cgi?id=81447#c18
and the patch I used
https://bugs.freedesktop.org/show_bug.cgi?id=81447#c17

Please note that I used the second version of libva-intel-driver-FD81447.patch
from commit "libva-intel-driver: better fix for FD81447":
https://github.com/OpenELEC/OpenELEC.tv/...4dc077bd41

Testing Joe_sample.mkv I see some distortion after 13s, the file finishes
fine and the screen returns back to xbmc after the playback ends.

[fritsch]

Yeah. All fine. I misread your reply. Happy movie watching :-)

[hal2100]

Hey all,

I just wanted to say that I "fixed" my vsync (or tearing) problem.
Although I don't know if it is a driver bug or not.

Like in the first post, I have a Xrandr line after xbmc starts. But mine included some "--tranmsformation (somenumbershere)". And this breaks the vsync for some reason.
I can also reproduce during the video, using xrandr from ssh. Seems like a driver bug to me.

[redstorm]

wsnipex has built new packages of the libva-driver-intel in his vaapi repository. That fixes the gpu hang with the Joe_Sample.mkv which resulted by non existing support for phantom frames. The new package handles those frames with less artifacts.

Code:

sudo apt-add-repository ppa:wsnipex/vaapi
sudo apt-get update
sudo apt-get dist-upgrade

Thanks very much, again :-)

I thought i would give these updated drivers a go, now after a period playback stutters and sound drops out completely.

xbmc.log

More info. when playback crashes the kernel log reports

Code:

Aug 14 22:57:33 xbmc kernel: [  155.039778] playback write error (DMA or IRQ trouble?)
Aug 14 22:57:43 xbmc kernel: [  165.038942] playback drain error (DMA or IRQ trouble?)

How do i remove them?

[fritsch]

Those drivers are not the problem - I see that your sound card is saying goodbye :-) and we cannot add another chunk of data onto the audio device.

Can you upgrade your kernel to 3.16.x or 3.15.9?

I have a little howto in the radeon oss vdpau thread (2nd post) it's a one liner.

[redstorm]

Those drivers are not the problem - I see that your sound card is saying goodbye :-) and we cannot add another chunk of data onto the audio device.

Can you upgrade your kernel to 3.16.x or 3.15.9?

I have a little howto in the radeon oss vdpau thread (2nd post) it's a one liner.

was working fine on the existing kernel till i tried the newer VA drivers.

I have backed them out but it still persists.
vainfo
xbmc.log

Last weekend i tried a newer kernel but had to back that out due to the DVB adapter drivers not being available for the newer kernels yet.

[fritsch]

If you reverted to the earlier drivers (which is the vainfo telling) and the problem persists then it's 100% sure not the reason of your issue.

[fritsch]

Please provide at least a Debug Log

[redstorm]

If you reverted to the earlier drivers (which is the vainfo telling) and the problem persists then it's 100% sure not the reason of your issue.

I agree with you that its not the drivers now, the reason i suspected them was when the audio sink crashed i had a Green Screen display during playback reminiscent of the issue early on in the Haswell days.

ill do some more trouble shooting tonight, as it only seemed to occur if i hammered changing channels on live TV, so ill see if i can reproduce it with debug enabled.

[fritsch]

wsnipex has updated the libva-driver-intel again and now incorporates a fix for that driver's master branch in combination with ffmpeg 2.3.x - if you are on 1.3.2 you are not affected by this bug - but you would have the joe_sample hang :-)

If you have nothing to loose give this new build a try.

Thanks to wsnipex again!

[ant_thomas]

Is it worth putting a list on the first post of the common CPU/GPU combinations around currently and what they're capable of?

Whether they can handle Yadif deinterlacing and/or Lanczos 3 scaling?

Is the Intel Celeron 2955U capable of both? With it having 10 EUs on the GPU.