Summary: | [ilk] h.264 decoding broken / visual errors | ||
---|---|---|---|
Product: | libva | Reporter: | Tobias Jakobi <liquid.acid> |
Component: | intel | Assignee: | haihao <haihao.xiang> |
Status: | RESOLVED FIXED | QA Contact: | Sean V Kelley <seanvk> |
Severity: | major | ||
Priority: | medium | CC: | gb.devel |
Version: | unspecified | ||
Hardware: | x86-64 (AMD64) | ||
OS: | All | ||
Whiteboard: | |||
i915 platform: | i915 features: | ||
Attachments: |
picture showing an example of image corruption
naive fix for the issue |
Description
Tobias Jakobi
2012-12-30 01:14:34 UTC
"but I'm not" <- this should of course read "but I'm _NOW_" Ah, sry, another correction needed: The stream is H264, and _not_ MPEG2. Can a admin please change the title? Created attachment 72289 [details]
picture showing an example of image corruption
I also updated intel-driver to git, since there were some changes concerning H264. Not fixing anything though.
More corrections. I wasn't really awake then I entered the bug: The issue is introduced between 1.0.16 and 1.0.17. I also did a proper bisect this time, and the faulty commit is: h264: fix scan for bit offset to macroblock (4bbfe67d2098f4f2aaeb3c5ab2cd930d2acb1c26) Created attachment 72301 [details]
naive fix for the issue
So this is how I (naively) fixed the issue for now. No idea if this is correct at all, but it removes the image corruption completly for me.
Pushed. Thanks for your patch. Hi, what version of mplayer-vaapi (and FFmpeg) were you using? BTW, n == (i - j) :) The fix is needed for the following case: xx xx 00 00 03 aa bb cc 'aa' is the last byte of slice_header. (In reply to comment #7) > Hi, what version of mplayer-vaapi (and FFmpeg) were you using? It's built from the latest checkout of the gitorious repo. You remember that your build script forces a specific ffmpeg version? |
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.