[Frank-NL]

So what's the current state of VPP with libva 1.3.1?
Is it still better to use software deinterlacing?

[fritsch]

Bob works with a patch not in 1.3.1 or is green otherwise. MADI / MACI are non functional. So - I only need to answer the first question I think :-)

[Frank-NL]

Yep. Will not touch my config

[alex.ba]

Hi all,

can someone confirm that XBMC won't start if no keyboard/mouse and network is connected? I want to use my system headless (with XBMC APP) and as soon as I plug out both keyboard and Network cable (I have a wireless network) XBMC won't start. After plugin in either keyboard or network it starts.

I installed according to the howto with Trusty.

Regards,

Alex

[fritsch]

cat /etc/network/interfaces | pastebinit

[alex.ba]

Hi Fritsch,

thanks a lot for the answer. I found out the problem. The behaviour is:

- If no network but keyboard is pluged in then XBMC starts normally
- Same for No keyboard but network
- If nothings pluged in then ther's a huge delay for starting which is set in /etc/init/failsafe.conf

As standard you have 20 + 40 + 60 + 1sec. So after about 2 minutes XBMC starts. You can of course reduce the delay.

Hope this also helps others.

Best Alex

[FernetMenta]

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?

Yes I can repro here though not using OE. Looks like it fails on all platforms. What role plays hdhomerun in this chain?

[puithove]

HDHomerun Prime - cablecard tuner is the actual capture device. It just pulls the stream off the cable line, decrypts it with the cablecard, and dumps it out the network port unmodified (other than the decryption of course). So the sample I gave you should be exactly what the cable company put on the line.

In my setup, MythTV controls the HDHomerun and dumps the stream to disk.

Just want to make sure you saw my note about forcing Deinterlace = ON instead of Auto clears up the issue. Seems to "De-Interlace" just fine with the 2x mode of Yadif and looks beautiful. But of course I don't want to set Deinterlace to always on as my default setting since I don't want to run progressive content through the deinterlacer.

[FernetMenta]

Yes, I saw this note. Looks much better this way but ffmpeg still complains about lots of errors. I guess there is a slightly different interpretation about the standard between your broadcaster and the guys at ffmpeg. I recommend that you open a bug report upstream https://trac.ffmpeg.org/ and link your sample.

[puithove]

OK, I can do that. Can you tell me what specific error you're seeing from ffmpeg so I can report accurately and a good way to tell them to reproduce - maybe a commandline example using those samples?

[FernetMenta]

I have those in xbmc.log http://xbmclogs.com/show.php?id=201408

Code:

19:38:03 T:139716592121600   DEBUG: Previous line repeats 4 times.
19:38:03 T:139716592121600   ERROR: ffmpeg[4DD7B700]: [NULL] non-existing SPS 0 referenced in buffering period
19:38:03 T:139716592121600   ERROR: ffmpeg[4DD7B700]: [NULL] non-existing PPS 0 referenced
19:38:03 T:139716592121600   ERROR: ffmpeg[4DD7B700]: [h264] non-existing SPS 0 referenced in buffering period
19:38:03 T:139716592121600   ERROR: ffmpeg[4DD7B700]: [h264] non-existing PPS 0 referenced
19:38:03 T:139716592121600   ERROR: ffmpeg[4DD7B700]: [h264] decode_slice_header error
19:38:03 T:139716592121600   ERROR: ffmpeg[4DD7B700]: [h264] no frame!

[FernetMenta]

disregard, this was another sample. there are no ffmpeg errors in the log. This is maybe related to an issue I have already opened: https://trac.ffmpeg.org/ticket/3452

There are some issues with video from ATSC countries.

[puithove]

Ok, because the samples I sent are MPEG2... I have seen those pullup errors mentioned in your ticket in my log even with the Deinterlace = ON setting. But at least with setting that, I don't seem to have the never-ending repeating of the Yadif messages - just the one time when the playback starts. With Auto, I really think it's constantly cycling the deinterlacer on and off rapidly for some reason. If I'm right about that, there must be something in the stream making it think it needs to change modes, or simply causing the detection to fail.

FYI - I downloaded that sample listed in your ticket to see how it would work on my system. It completely crashes XBMC within the first few seconds of playback With deinterlacing off or on, either way.

Edit 2: Ok, switching to Software Decode I'm able to play it... It definitely does act in a fairly similar manner. It's h.264 vs MPEG2 which may account for some of the difference (mine's just jerky - yours is jerky and blocky with Yadif). Plays fine with BOB, changing Auto to On makes "De-Interlace" better, but doesn't totally fix it.

[FernetMenta]

As I mentioned in the bug report, playing my sample through my vnsi parsers I get fine 59.98 fps. Maybe there is some missed access unit delimiter or some misinterpreted one. The payload unit start indicator of a PES packet doe not indicate the start of a frame. At this level h264 and mpeg2 are treated similar.

[puithove]

LOL - you went a little bit outside my knowledge there, but this part I understood:

At this level h264 and mpeg2 are treated similar.

So just out of curiosity - and to see if we really are playing with the same thing - does playing my sample through the vnsi parser work?

I guess this means I'd have to hope for an upstream ffmpeg fix though doesn't it?