On Wed, 2015-10-28 at 17:03 +0100, Nicolas George wrote:
> Le septidi 7 brumaire, an CCXXIV, John Stebbins a écrit :
> > Sorry, I'm being dense.  I see what you mean.  The comment currently
> > implies that if you explicitly set frame_rate to 1/0, nobody in the
> > filter chain will modify it.  The code obviously does not do this
> > and I
> > see no attempt in the ffmpeg code to try to preserve 1/0 in any way.
> > In
> > addition, there are few filters or sources where you can explicitly
> > set
> > the frame_rate, and all of these with the exception of buffersrc
> > require
> > a valid framerate (i.e. they will not accept 1/0).
> > 
> > It seems easier to me to define 0/0 as the "unknown" value because
> > this
> > *is* preserved by the code as it currently stands.  There is no
> > filter
> > aside from filters that convert to a known constant frame_rate that
> > will
> > modify a value of 0/0.
> 
> Your proposal does not work because it does not distinguish "the frame
> rate
> is not known, maybe there is not even one" from "this filter did not
> bother
> setting the frame rate".
> 
> A lot of filters were written before the frame_rate field was added.
> Most of
> them are simple one-in-one-out filters where the output frame rate is
> the
> same as the input. Your proposal would require going all these adding
> the
> copy explicitly. This would be annoying, and you rightfully did not do
> it.
> 
> The cleared value 0/0 means that the filter did not bother setting it.
> If
> you assign it a different semantic, you create ambiguity and make the
> API
> more fragile.
> 
> The convention chosen in FFmpeg is: if 0/0, that means the filter did
> not
> bother to set it, then assume the simplest case and let the framework
> set
> the value as the same as the input.
> 
> Regards,
> 
> 

Ah, ok.  I misread some code.  Now it makes sense.  The code does
exactly what the comment says it does.  Sorry for all the spam.
_______________________________________________
libav-devel mailing list
[email protected]
https://lists.libav.org/mailman/listinfo/libav-devel

Reply via email to