[Ffmpeg-devel] [BUG] wrong fps in r_frame_rate ?

[Baptiste Coudurier]

Michael Niedermayer a ?crit :
> Hi
> 
> On Sat, Dec 09, 2006 at 12:51:27AM +0100, Baptiste Coudurier wrote:
>> Hi
>>
>> Michael Niedermayer a ?crit :
>>> Hi
>>>
>>> On Fri, Dec 08, 2006 at 04:08:14PM +0100, Baptiste Coudurier wrote:
>>>> Hi
>>>>
>>>> $subject
>>>>
>>>> latest svn. ffmpeg -i wrong_fps.mpg
>>>>
>>>> Seems that stream 0 comes from film source: 25.00 (25/1) -> 24.75 (99/4)
>>>> Input #0, mpeg, from 'wrong_fps.mpg':
>>>>  Duration: 00:00:04.3, start: 0.552822, bitrate: 15583 kb/s
>>>>  Stream #0.0[0x1e0]: Video: mpeg2video, yuv422p, 720x608, 15000 kb/s,
>>>> 24.75 fps(r)
>>>>  Stream #0.1[0x1c0]: Audio: mp2, 48000 Hz, stereo, 192 kb/s
>>>>  Stream #0.2[0x1c1]: Audio: mp2, 48000 Hz, stereo, 192 kb/s
>>>> Must supply at least one output file
>>>>
>>>> wrong_fps.mpg on mplayerhq as usual. File should be 25fps.
>>>>
>>>> in libavformat/utils.c if I remove || CODEC_ID_MPEG2VIDEO, or revert
>>>> commit r6345, it is correctly detected.
>>> neither is acceptable, the code should be executed
>>> find another solution :)
>>> also it might be interresting if you would dump the timestamps for the
>>> frames which are used in the r_frame_rate calculation ...
>>>
>> Ok that commit broke many more files than it solved. Can we revert it ?
> 
> no because this isnt true, the code before should have failed for pretty much 
> all telecined mpeg files
> 

I would prefer not to troll here, but I remember last time we were 
talking about policy, it was said that a commit should not break working 
scenario whatever it was, and that instead a clean a working for both 
solution should be commited.

If the sample I supplied is broken, then Im ok to leave this but if not, 
that file should be prefered over groundhog.vob, else this is a 
regression, and IMHO not acceptable.

-- 
Baptiste COUDURIER                              GnuPG Key Id: 0x5C1ABAAA
Key fingerprint                 8D77134D20CC9220201FC5DB0AC9325C5C1ABAAA
checking for life_signs in -lkenny... no