[FFmpeg-devel] [PATCH 2/2] libx265: Enable 12-bit encoding

[Stephen Hutchinson]

On 1/27/2016 10:48 AM, Derek Buitenhuis wrote:
> On 1/27/2016 12:10 AM, Stephen Hutchinson wrote:
>> The configure detection is bumped to X265_BUILD >= 68,
>> since API version 68 corresponds with the x265 1.8
>> release tarball.
>
> Was 68 the first API version it was in, or merely the tarball?
>
> I would prefer the actual version it was introduced in (or made useful).
>

The initial version of the patch bumped it only to version 60, since
that was the next API version after x265 added API support for querying
its 12-bit stuff (even though the actual 12-bit support was added
over a month after that). But it was suggested that it be bumped up
to where the 1.8 tarball was/would be so that the 12-bit support was
also considered ready on x265's side:

http://ffmpeg.org/pipermail/ffmpeg-devel/2015-August/177716.html
http://ffmpeg.org/pipermail/ffmpeg-devel/2015-August/177727.html

60 is the first API version after 12-bit stubs were added,
whether the actual encoding support was present or not.

68 is probably about where it became useful (or at least noteworthy),
and it's the same API version as the 1.8 tarball.

72 is the first API version after the remaining internal-to-x265
warning about 12-bit being experimental was removed, but as
noted in the previous discussion above, not much seemed to actually
change between 68 and its removal.  But if we wanted to make sure
no warning about it was emitted by libx265 at all, then 72 is
what we'd need.

I have no strong opinion on it.  68 seems like the compromise
position since it'd allow both 1.8 and 1.9 to be used, whereas
any of the earlier ones still fall within 1.8's dev cycle and would
only be used by those users making builds against Mercurial
(most of which are probably going to be using builds from
more recently than 5+ months ago anyway).