Login at Kodi Home

**Prof Yaffle** · 2011-05-01, 11:45

Yup, that was my initial guess and remains my most-likely view. I just need to eliminate everything (while still trying to get to a working system that does what I want it to). I'll keep this thread updated, as a diary as much as anything.

Did you find anything of note in the log files, btw? I didn't see anything relevant, so I still suspect that XBMC thinks that everything's fine.

Cheeers....

opdenkamp · 2011-05-01, 13:05

nope, nothing in there.

**Prof Yaffle** · 2011-05-01, 23:40

Well, I've just downloaded and compiled the current GIT code - man, did that take a while on a 1.6GHz Atom processor! - and, lo and behold, the corruption is there. So it's clearly an issue with the mainline XBMC code, which points the finger all the way back to the original ffmpeg (multi-thread?) query.

Thanks for the thoughts, guys, I'll go and raise a TRAC ticket.

**Prof Yaffle** · 2011-05-02, 00:21

http://trac.xbmc.org/ticket/11499

MindGuerrillas · 2011-06-15, 20:19

I'm getting this issue too. My setup is the same as that of Prof Yaffle

Has anyone had any luck in sorting it out?
Thanks

loggio · (This post was last modified: 2011-06-16, 04:55 by loggio.)

I've been getting this for a while on my windows Machine... I've started several posts and created trac tickets but gotten nowhere with it.

This is a Multi-platform issue.... As I'm running windows 7, and i get corruption on one channel... Bottom quarter of screen when DXVA is enabled.

I'm using 10.1 builds from Margro, but i've tried recent pre builds and the issue is still present. It's xbmc's Hardware accelerated ffmpeg that's doing it...

I'm also running ION.
The main PC in my living room was running ATI HD graphics and when i played back the same file the corruption wasn't there.... But then i realsed that the ati graphics card i had did not support DXVA for Mpeg2... so it was just the fact that XBMC disabled dxva while playing it back on that machine.

That graphics card stopped working not long ago, so i bought an NVIDIA and... what do you know, the file plays back corrupt because it's using DXVA.

Here's "one of" my original Threads. http://forum.xbmc.org/showthread.php?tid=100220

XBMC GUYS really need to fix this!

opdenkamp · 2011-06-16, 10:45

we need to do a lot of things. and you're using a pre-release/development version, so issues can be expected.

loggio · 2011-06-16, 10:51

One thing i do know is that the version of xbmc being used has no effect on this... Stable, pre-release, development, experimental... I've tried them all. the recorded .ts file plays back corrupt on all.

Actually, there is one thing i did that solved my issue... That was when i tested the DSPlayer branch a while back.
I wonder what it is that's causing this!

Strange.

Peace,
Loggio.

**Prof Yaffle** · 2011-06-17, 00:33

Well, I can't say I'm disappointed to hear that other folks are getting similar issues, although that's entirely selfish and no comfort at all to the other folks in question!

I think Dushmaniac's comment is fair enough - this is pre-release, you can expect some things to not work. I suppose the question, then, is what we can collectively do to help nail the problem down and make sure it's fixed in the 11.0 release. If nothing else, we don't want a whole world of people suffering the same issue if this goes out as a mainstream release...

Has anyone had any success trying other accelerated programs (Mplayer, VLC, ...)? ffplay?

What about compiling with external ffmpeg, does that help anyone (I seem to recall that it's easier said than done with the git code)? Any alternatives?

I always meant to try FernetMenta's code to see if his VDPAU method made any difference, but never had the chance. But that's another alley to stumble down.

Any other thoughts on how we isolate this so the devs know what to look at?

**Prof Yaffle** · 2011-06-21, 19:39

okay, I thought I'd try again...

Git pull of the current XBMC - compiles fine but corruption still shows (XBMC reports pre11.0 git:20110501-bebae84)

Current FernetMenta code, the same issue (pre11.0 git:20110616-79c3394)

I've just tried compiling the latest SVN ffmpeg (26402), which seemed to work, but I'm getting too many errors trying to compile FernetMenta's code with --enable-external-ffmpeg. According to this, there may be some trunk changes yet to ripple down to the PVR codebase, so I'm going to revert to the standalone XBMC and see if I can get that to use the external libraries.

A job for another time, though, the family needs the HTPC back so it's time to reboot to the production image :-)

FernetMenta · 2011-06-21, 20:50

There's been some changes in mainline and pvr in order to adapt to newer (external) versions of ffmpeg. My repo was not up to date.

TheSwissKnife did lots of tests with new ffmpeg flags for frame dropping. This has shown some weird behavior for interlaced material which led to complete corruption of playback. He has found some work arounds to cope with this problem.

Not sure if this problem is related to the one you are observing. I still have your samples and will run some tests.

**Prof Yaffle** · 2011-06-22, 15:12

Thanks. I did try to compile the current mainline git code for external ffmpeg last night, but with no success. I suspect that I either need to do some more reading around the errors I received, or else need to wait a little for some new code to trickle into the branches before trying again.

Anyone else more makefile-and-dependency literate want to have a go to see if we can spread the effort?

FernetMenta · 2011-06-22, 15:52

Update: We have tried your samples on two systems using the same xbmc code.

- GT220 no corruption
- ION corruption

TheSwissKnife · 2011-06-22, 17:08

ION2 nvidia drivers 195.36.24 was corrupt for the flv file to be specific.

FernetMenta · 2011-06-22, 17:20

I have compiled against vdpau 0.4.1