[FFmpeg-devel] [PATCH] avcodec/libvpx: set CODEC_CAP_EXPERIMENTAL correctly instead of testing at codec open time

[Timothy Gu]

On Monday, December 9, 2013, Reimar Döffinger wrote:

> On Mon, Dec 09, 2013 at 06:35:17PM -0300, James Almer wrote:
> > On 09/12/13 6:08 PM, James Zern wrote:
> > > On Sun, Dec 8, 2013 at 2:14 PM, James Almer <jamrial at gmail.com<javascript:;>>
> wrote:
> > >> On 08/12/13 10:19 AM, Michael Niedermayer wrote:
> > >>> Signed-off-by: Michael Niedermayer <michaelni at gmx.at <javascript:;>>
> > >>> ---
> > >>>  libavcodec/libvpx.c    |   13 ++++---------
> > >>>  libavcodec/libvpx.h    |    2 +-
> > >>>  libavcodec/libvpxdec.c |    4 +---
> > >>>  libavcodec/libvpxenc.c |    4 +---
> > >>>  4 files changed, 7 insertions(+), 16 deletions(-)
> > >>>
> > >>
> > >> Maybe we should leave a log message about libvpx < 1.3.0 as an
> > >> AV_LOG_WARNING line.
> > >> The idea is that the warning will show up after running ffmpeg with
> > >> -strict -2, so the user knows that what's experimental is VP9 encoding
> > >> with their libvpx version and not ffmpeg's wrapper, and that they
> > >> should consider updating it.
> > >
> > > This now happens at registration time, so the log would be issued even
> > > if you weren't using vp9.
> >
> > The log can for example be in vp9_init() (Both encoder and decoder). That
> > way it will only trigger for VP9 and libvpx < 1.3.0.
> > It will however need an extra version check there.
>
> Stupid question: If libvpx < 1.3.0 is only experimental, why would we
> want to support it at all now that something non-experimental is
> available?

It seems a lot of effort for something that to me looks like a
> completely useless (if not worse) feature...
>

We shouldn't. But some people might need pre1.3 libvpx for testing or
something,
or they might are too lazy to upgrade, or maybe the person's distro is too
outdated, or ... I don't know. But it's still better not to break something
that worked.

Timothy