[FFmpeg-cvslog] r16127 - trunk/libavformat/avidec.c

[Mike Melanson]

Michael Niedermayer wrote:
> On Sun, Dec 14, 2008 at 08:23:04PM -0800, Mike Melanson wrote:
>> Michael Niedermayer wrote:
>>> On Sun, Dec 14, 2008 at 10:33:49AM -0800, Mike Melanson wrote:
>>>> michael wrote:
>>>>> Author: michael
>>>>> Date: Sun Dec 14 16:48:48 2008
>>>>> New Revision: 16127
>>>>>
>>>>> Log:
>>>>> Fix non interleaved variable fps ODML avis.
>>>>> Fixed issue738
>>>> Problem: This causes an infinite loop when there is only 1 frame in an 
>>>> AVI file. This causes FATE to stall on this test:
>>>>
>>>> http://fate.multimedia.cx/index.php?test_spec=124
>>>>
>>>> (Problem file is linked from there.) The stall occurs on Linux 
>>>> x86_{32|64} but not on Linux/PPC or either Mac OS X configuration.
>>> should be fixed
>> BTW, this change also caused the following test spec to fail on all 
>> configurations (I bisected it to revision 16127):
>>
>> http://fate.multimedia.cx/index.php?test_spec=252
>>
>> $BUILD_PATH/ffmpeg -i $SAMPLES_PATH/VMnc/VS2k5DebugDemo-01-partial.avi 
>> -vsync 0 -pix_fmt rgb24 -f framecrc -
>>
>> Sample is linked from page. Is there a reason that the CRC output from 
>> the decoder would have changed? Note that this is the case with or 
>> without '-vsync 0'.
> 
> the demuxer output changed like:
> --- vold	2008-12-16 03:46:17.000000000 +0100
> +++ old	2008-12-16 03:47:42.000000000 +0100
> @@ -1,4 +1,3 @@
> -0, 0, 1655, 0x989d4766
>  0, 18000, 1655, 0x989d4766
>  0, 36000, 22, 0x2375044f
>  0, 90000, 1673, 0x2024483e
> 
> Thus it seems to me that te demuxer previously returned the first packet
> twice.
> Comment by the decoder maintainer is welcome, but without further
> investigation i suspect it was wrong and is more correct now.

That's all I needed to know. Thanks for looking into this. Barring 
further feedback, I will update the test spec.

-- 
     -Mike Melanson