[FFmpeg-devel] [PATCH] matroskaenc: Add support for writing video projection.

Vittorio Giovara vittorio.giovara at gmail.com
Tue Jan 31 12:05:58 EET 2017


On Sat, Jan 28, 2017 at 4:13 AM, James Almer <jamrial at gmail.com> wrote:
> On 1/27/2017 11:21 PM, Aaron Colwell wrote:
>> On Fri, Jan 27, 2017 at 5:46 PM James Almer <jamrial at gmail.com> wrote:
>>
>> yeah. I'm a little confused why the demuxing code didn't implement this to
>> begin with.
>
> Vittorio (The one that implemented AVSphericalMapping) tried to add this at
> first, but then removed it because he wasn't sure if he was doing it right.

Hi,
yes this was included initially but then we found out what those
fields were really for, and I didn't want to make the users get as
confused as we were. As a matter of fact Aaron I mentioned this to you
when I proposed that we probably should have separated the classic
equi projection from the tiled one in the specification, in order to
simplify the implementation.

>>> I know you're the one behind the spec in question, but wouldn't it be a
>>> better idea to wait until AVSphericalMapping gets a way to propagate this
>>> kind of information before adding support for muxing video projection
>>> elements? Or maybe try to implement it right now...
>>>
>>
>> I'm happy to implement support for the projection specific info. What would
>> be the best way to proceed. It seems like I could just place a union with
>> projection specific structs in AVSphericalMapping. I'd also like some
>
> I'm CCing Vittorio so he can chim in. I personally have no real preference.

The best way in my opinion is to add a third type, such as
AV_SPHERICAL_TILED_EQUI, and add the relevant fields in
AVSphericalMapping, with a clear description about the actual use case
for them, mentioning that they are used only in format. By the way,
why do you mention adding a union? I think four plain fields should
do.

Please keep me in CC.
-- 
Vittorio


More information about the ffmpeg-devel mailing list