[Ffmpeg-devel] cvs regression

Michael Niedermayer michaelni
Mon Jul 18 11:20:33 CEST 2005


Hi

On Monday 18 July 2005 10:05, Zoltan NAGY wrote:
> Hello!
>
> I'm playing with mp4's from disc, and streams, described by
> sdp files, generated by Apple's Darwin Streaming Server.
>
> here are my results:
>
> --> 0.4.8:
> ./configure --enable-shared --enable-faad
> make
> make install
>
> playing an mp4, with aac sound inside:
> plazma3:~/ffmpeg-0.4.8# ./ffplay -stats ../c770.mp4
> header damaged
> warning: first frame is no keyframe
> Input #0, mov, from '../c770.mp4':
>   Duration: 00:00:11.6, bitrate: 2246 kb/s
>   Stream #0.0: Video: mpeg4, 640x480, 29.97 fps
>   Stream #0.1: Audio: aac, 48000 Hz, stereo
> Fatal signal: Segmentation Fault (SDL Parachute Deployed)
> Xlib: unexpected async reply (sequence 0x67)!
>
> and then I have to kill the process.
>
> playing an sdp file:
> plazma3:~/ffmpeg-0.4.8# ./ffplay -an -stats ../darwin.sdp
> RTP: PT=60: bad cseq f20f expected=0001
> header damagedA-V:  0.000 aq=    0KB vq=    0KB
> warning: first frame is no keyframe
> Input #0, sdp, from '../darwin.sdp':
>   Duration: N/A, bitrate: N/A
>   Stream #0.0: Video: mpeg4, 380x480, 25.00 fps
> Title: C:\Program Files\Darwin Streaming Server\Playlists\teszt\teszt.s@
> warning: first frame is no keyframe
> 1121671779.95 A-V:  0.000 aq=    0KB vq=   13KB
>
> works fine.
>
> so, with 0.4.8, there is clearly a problem with aac.
> allright, that's an old one, moving on!
>
> --> 0.4.9-pre1:
> ./configure --enable-shared --enable-faad --enable-gpl
> make
> make install
> playing an mp4, with aac sound inside:
> plazma3:~/ffmpeg-0.4.9-pre1# ./ffplay -stats ../c770.mp4
> Input #0, mov,mp4,m4a,3gp, from '../c770.mp4':
>   Duration: 00:00:11.6, bitrate: 2246 kb/s
>   Stream #0.0: Video: mpeg4, 640x480, 29.97 fps
>   Stream #0.1: Audio: mpeg4aac, 48000 Hz, stereo
>    0.02 A-V: -6.956 aq=    0KB vq= 1083KB    KB
>
> great!
> small problem: the video is waaaaaaaaay toooo slow compared to the video
>
> let's see our lovely sdp:
> plazma3:~/ffmpeg-0.4.9-pre1# ./ffplay -an -stats ../darwin.sdp
> [NULL @ 0xb7e8f008]RTP: PT=60: bad cseq 38ee expected=0001
> Input #0, sdp, from '../darwin.sdp': vq=    0KB
>   Duration: N/A, bitrate: N/A
>   Stream #0.0: Video: mpeg4, 380x480, 25.00 fps
> Title: C:\Program Files\Darwin Streaming Server\Playlists\teszt\teszt.s@
>    4.07 A-V:  0.000 aq=    0KB vq=    0KB
>
> seems ok, but ffplay's window remains all black..
>
> so 0.4.9-pre1 is a regression compared to 0.4.8 in this way.
>
> let's see cvs as of friday:
> cvs -z3 -d:pserver:anonymous at mplayerhq.hu:/cvsroot/ffmpeg co -P ffmpeg
> cd ffmpeg
> ./configure --enable-shared --enable-faad --enable-gpl
> make
> make instlal
> plazma:~# ffplay -stats mp4\ samples/c770.mp4
> Input #0, mov,mp4,m4a,3gp,3g2, from 'mp4 samples/c770.mp4':
>   Duration: 00:00:11.6, start: 0.000000, bitrate: 2243 kb/s
>   Stream #0.0: Video: mpeg4, yuv420p, 640x480, 29.97 fps
>   Stream #0.1: Audio: aac, 48000 Hz, stereo
>    9.58 A-V: -0.016 aq=   31KB vq=  480KB
> plazma:~#
>
> it's perfectly ok!
> now, let's check out the sdp...
> plazma:~# ffplay darwin.sdp
> socket: Address family not supported by protocol
> darwin.sdp: Error while parsing header
> plazma:~#
>
> and that's it.
>
> so it's a regression regarding the handling of sdp's, and
> it's definitely an improvement regarding acc inside mp4.
>
> What could I do?
> How could I found out what change broke sdp?
> Where should I look regarding the handling of sdp's?
>
> give me some hints please, I'm willing to make a patch for
> cvs to fix sdp handling.

cvs binary search (split the time interval between working and non working in 
half and continue until only a single cvs commit is left) that has to be done 
several times if there are multiple bugs

and after sdp is fixed to prevent future breaks it is VERY important that 
someone writes a regression test for it so we will catch future breakage 
immedeatly

[...]
-- 
Michael





More information about the ffmpeg-devel mailing list