Michael Niedermayer <[email protected]> added the comment:

On Mon, Mar 08, 2010 at 04:01:04PM +0000, Aleksey Nogin wrote:
> 
> Aleksey Nogin <[email protected]> added the comment:
> 
> > plese submit a reproduceabel bugreport, also make sure your version of
> > lame is stable & recent
> 
> I am using lame-3.98.2 on CentOS 5.4
> 
> > and you cannot just randomly change which bytes of a packet are 
> > returned to prevent a buffer from overflowing
> 
> You missed the point. It not random. When we *just* flushed with
> lame_encode_flush() then it is safe to return the complete contents of the
> buffer (they were just flushed).

you miss a point in your rant, your code is broken you return random data
completely independant of the bug or not bug issue.
what lame returns is not a mp3 frame at a time, but what we need (for
any container short of dumped raw .mp3) are proper mp3 frames, thats besides
that the libavcodec api mandates encoders to return properly framed data.
your patch disables the related code and you return a random chunk of
bytes that will lead to an invalid file that may or may not play correctly
depending on how tolerand the player is and what container is used.

[...]

________________________________________________
FFmpeg issue tracker <[email protected]>
<https://roundup.ffmpeg.org/issue803>
________________________________________________

Reply via email to