[FFmpeg-devel] [PATCH] lavfi: add pp filter.

[Michael Niedermayer]

On Wed, Dec 12, 2012 at 08:25:48PM +0100, Clément Bœsch wrote:
> On Wed, Dec 12, 2012 at 10:18:02AM +0100, Michael Niedermayer wrote:
> > On Wed, Dec 12, 2012 at 04:49:49AM +0100, Clément Bœsch wrote:
> > > Ported from MPlayer. Original author is A'rpi, with various
> > > contributions from Michael Niedermayer. The original documentation was
> > > mostly written by Diego Biurrun. See the MPlayer history for full
> > > credits.
> > > 
> > > The filter is under GPL like the original filter, even if it differs
> > > quite a lot. There is not much point in making it LGPL since pp is under
> > > GPL.
> > > 
> > > TODO: fix FATE
> > > TODO: bump minor
> > > TODO: Changelog
> > > ---
> > > Some insight about current FATE failure: the 3 first pp tests fail because of a
> > > mismatch between linesize and width (for the chroma planes), which is not
> > > present with mp=pp, where linesize==width all the time for some reason. It
> > > might be a bug in libpostproc, but I need someone familiar with it to confirm
> > > this.
> > 
> > which individual filter(s) exactly have this mismatch problem ?
> > 
> 
> Seems to be the dering one, I didn't find another mismatching one yet.

I suspect the dering filter reads over the left and right image edges
Its not overreading the array, a buffer is used for these cases but
not where such overread seemed harmless, sadly that breaks bitexact
regression tests.

If above hypothesis is correct, the solution is to write special
cases for the first and last columns or use a buffer for them.
or just drop dering from the tests as this doesnt affect anything but
such test

[...]

-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

DNS cache poisoning attacks, popular search engine, Google internet authority
dont be evil, please
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: Digital signature
URL: <http://ffmpeg.org/pipermail/ffmpeg-devel/attachments/20121212/e59bd92a/attachment.asc>