[FernetMenta]

could you please post a log running this version?

Sure, would you like it playing MythTV content, that sample, or both?

can you reproduce this playing some file you can share with us?

[puithove]

Never trust myth, but go ahead.

Can it be convinced to have 50hz, by putting that xorg.conf to /storage/.config ?

Edit: Btw. please try with one of those nightlies. OE has patched out ffmpeg 2.2 out of the FM patches for 4.0 and "mixed" something up.

No, it appears my TV doesn't like 50p (believe I had verified that at some point in the past with the same results.

Updated to: OpenELEC-Generic.x86_64-devel-20140511141602-r18367-g0627396
So here's xbmc-xrandr with it playing the sample above using that xorg.conf minus the 50p mode (this does make the new 60p mode the preferred so I may keep this in place)
http://sprunge.us/SFNc

And the debug log:
https://spideroak.com/storage/OB2WS5DIN5...e6b03bd785

And for the MythTV content with this OE version (still seeing jerky issue):
xrandr: http://sprunge.us/jZcA
Log: https://spideroak.com/storage/OB2WS5DIN5...1db5b88b7c

Confirmed that both were using the "De-interlace" method

[fritsch]

Same error :-)

If you find lots of time - can you follow this howto here and recheck?

Actuall you have something in your logs which has long been patched and is non existent anymore. So we currently search what OE is missing.

[fritsch]

Here, please try that one: http://snapshots.openelec.tv/OpenELEC_He...5520e3.tar Thx much

[puithove]

can you reproduce this playing some file you can share with us?

Sure, Myth records live tv and retains the files for a day. Here's the exact .MPG that was played in that last log I posted. I just copied it to the box and played it again via "Files" and saw the same results - so we can at least rule out it being an issue with Myth handing off the data. I'd have replied sooner but it took forever to upload this on my slow-ass uplink. 329MB https://spideroak.com/storage/OB2WS5DIN5...ab518ade85

It'll be really interesting to see if you get the same results.

Same error :-)

If you find lots of time - can you follow this howto here and recheck?

Actuall you have something in your logs which has long been patched and is non existent anymore. So we currently search what OE is missing.

I might, but it's unlikely to happen before this weekend if at all. I'd like to help but time is a factor. I'll see what I can do. I'll have to do it during a time when wifey isn't home.

That's interesting... I might actually help identify something useful

Here, please try that one: http://snapshots.openelec.tv/OpenELEC_He...5520e3.tar Thx much

Ok, updated to that Helix build. Still same issue using that same sample file (played from "Files"). Here's the log - playback of the entire sample:
https://spideroak.com/storage/OB2WS5DIN5...220cf13ba4

[puithove]

Idea: Any chance you have an OE build with the new BBC deinterlacer enabled? Could you let me try it?

[fritsch]

You don't want that. Performance is so bad - that I get barely 30 fps on my core i5 3400 Mhz running and visible difference is zero.

[puithove]

You don't want that. Performance is so bad - that I get barely 30 fps on my core i5 3400 Mhz running and visible difference is zero.

Ah, no thanks then.

[fritsch]

But, here is the deal:

Use the above sample I linked and deinterlace it via ffmpeg command line, once with weston and one time with yadif and compare :-) Use some libx264 encoding with a crf of 18 so that you don't introduce other artifacts and compare those.

Like: ffmpeg -i deinterlacing-test.mp4 -c:v libx264 -crf 19 -preset slow -vf "yadif=1:-1:1" -acodec copy "deinterlacing-test-yadif.mp4"
vs: ffmpeg -i deinterlacing-test.mp4 -c:v libx264 -crf 19 -preset slow -vf "w3fdif=complex:interlaced" -acodec copy "deinterlacing-test-weston.mp4"

You need ffmpeg 2.x of course.

[puithove]

Oh, I believe you. Was just wondering if it'd be useful to see if W3 handled the content better. If not, no big deal.

[fritsch]

You can build that yourself, the change is a one liner, only change needed is to replace the above string. xbmc uses that filter directly.

[fritsch]

Here is a comparison:

Transformers 576i: https://dl.dropboxusercontent.com/u/5572...formers.ts
Transformers 576p (w3fdif aka bbc deinterlacer): https://dl.dropboxusercontent.com/u/5572...weston.mp4
Transformers 576p (Yadif): https://dl.dropboxusercontent.com/u/5572...-yadif.mp4

[puithove]

Here is something interesting. Using my same sample file from MythTV... If I change the Deinterlace Video setting from "Auto" to "On" and Deinterlace Method set to "De-Interlace" I get buttery smooth playback. The FPS still varies up and down around 48 for most of the file, jumping up to 59.94 for a short segment. Seems normal, right?

Is the Auto Detect perhaps seeing some parts of the video as interlaced and some not - causing it to constantly switch the deinterlacer on and off?

I did this after rolling back to the OpenELEC-Generic.x86_64-devel-20140511141602-r18367-g0627396 build

Here's a debug log from this run:
https://spideroak.com/storage/OB2WS5DIN5...77fcd295de

[FernetMenta]

I have tried your sample here. ffmpeg complains a lot. Either the recorded video stream got corrupted or ffmpeg has an issue. Do you have a chance to record something without using myth?

[puithove]

Sure. I use HDHomerun Prime as the capture device. This sample is captured straight from the hdhomerun_config utility. Playback shows the same issue.

https://spideroak.com/storage/OB2WS5DIN5...13996fc08e

Are you able to reproduce what I'm seeing on your end in OE? What about other XBMC builds?