[FFmpeg-cvslog] h264: check the pixel format directly and force a reinit on mismatches.

Guillaume POIRIER poirierg at gmail.com
Sun Feb 17 20:03:39 CET 2013


Moin-moin,

On Tue, Jan 29, 2013 at 4:58 AM, Michael Niedermayer <git at videolan.org> wrote:
> ffmpeg | branch: master | Michael Niedermayer <michaelni at gmx.at> | Tue Jan 29 04:17:48 2013 +0100| [11c99c78bafa77f679a1a3ba06ad00984b9a4cae] | committer: Michael Niedermayer
>
> h264: check the pixel format directly and force a reinit on mismatches.
>
> The existing checks are insufficient to detect a pixel format
> changes in case of some damaged streams.
> Fixes inconsistency and later out of array accesses

This fix breaks Hardware-accelerated decoding on OSX through VDA.

I get these kind of errors:
A:   1.0 V:   0.0 A-V:  1.025 ct:  0.000   0/  0 ??% ??% ??,?% 218 0
[h264_vda @ 0x10a0b91a0]Reinit context to 1920x1088, pix_fmt: 81
[h264_vda @ 0x10a0b91a0]Missing reference picture, default is 0
[h264_vda @ 0x10a0b91a0]decode_slice_header error
[h264_vda @ 0x10a0b91a0]Reinit context to 1920x1088, pix_fmt: 81
[h264_vda @ 0x10a0b91a0]reference picture missing during reorder
[h264_vda @ 0x10a0b91a0]reference picture missing during reorder
[h264_vda @ 0x10a0b91a0]reference picture missing during reorder
[h264_vda @ 0x10a0b91a0]reference picture missing during reorder
[h264_vda @ 0x10a0b91a0]Missing reference picture, default is 0
[h264_vda @ 0x10a0b91a0]decode_slice_header error

Reverting just that commit restores H.264 decoding via VDA.

So what can I do to have the best of both worlds, that is, VDA
decoding and the pixel format changes issue fixed by this commit?

mit freundlichen Grüßen,

Guillaume
-- 
If a cluttered desk is the sign of a cluttered mind, what is the
significance of a clean desk?
Laurence J. Peter


More information about the ffmpeg-cvslog mailing list