On 14/10/15 14:45, wm4 wrote:
> On Wed, 14 Oct 2015 14:39:31 +0200
> Luca Barbato <[email protected]> wrote:
> 
>> On 14/10/15 11:33, Vittorio Giovara wrote:
>>> This structure served as a bridge between data pointers and frames, but
>>> it suffers from several limitations:
>>> - it is not refcounted and data must be copied to every time
>>> - it cannot be expanded without ABI break due to being used on the stack
>>> - its functions are just wrappers to imgutils which add a layer of
>>>   unneeded indirection, and maintenance burden
>>> - it allows hacks like embedding uncompressed data in packets
>>> - its use is often confusing to our users
>>>
>>> AVFrame provides a much better API, and, if a full blown frame is not
>>> needed, it is just as simple and more straightfoward to use data and
>>> linesize arrays directly.
>>>
>>> Signed-off-by: Vittorio Giovara <[email protected]>
>>> ---
>>>  libavcodec/avcodec.h    | 50 
>>> +++++++++++++++++++++----------------------------
>>>  libavcodec/avpicture.c  |  3 ++-
>>>  libavcodec/imgconvert.c |  2 ++
>>>  3 files changed, 25 insertions(+), 30 deletions(-)
>>
>> Ok, I'll add few helpers to avframe later.
> 
> Which helpers?

Probably wrapper around av_image_copy_to_buffer to make Derek happier
even if `av_frame_image_to_buffer` sounds horrible and an
`av_frame_to_buffer` would require to make av_samples_copy too smart.

lu
_______________________________________________
libav-devel mailing list
[email protected]
https://lists.libav.org/mailman/listinfo/libav-devel

Reply via email to