[FFmpeg-devel] [PATCH] audio conversion clipping/overflows

[Ronald S. Bultje]

Hi,

On Mon, Mar 1, 2010 at 11:47 AM, Vitor Sessak <vitor1001 at gmail.com> wrote:
> On Mon, Mar 1, 2010 at 3:18 PM, Ronald S. Bultje <rsbultje at gmail.com> wrote:
>> float 1.0 rounds to int INT_MIN (whatever int size). This leads to
>> considerable audio clipping artifacts.
>
> The way all the others codecs do is to clip in the range [-1;
> 32767./(1<<15)]

The convention of clipping based on the assumption that everything is
int16 is just asking for problems with proaudio zealots at some point.

Not that I care much, I just want the bug fixed... :-). Also, I think
doing this in decoders is bad practice, it should be done in common
code.

Ronald