[vlc-devel] [PATCH 01/34] picture: add a video context to the picture

[Steve Lhomme]

On 2019-11-12 11:26, Rémi Denis-Courmont wrote:
> I agree that picture_t.context and picture_t.sys should be unified when 
> push buffers are ready. But that's not the problem.
> 
> There're 3 places that a picture can refer to its video context: in 
> picture_t, in the common part of the picture context, or in the 
> type-specific part of the picture context (where the proto-video_context 
> currently are). Since there should not be pictures with video context 
> and without picture context, I don't see the benefit of the first 
> option. It just adds a potentially useless member.

OK, I'll give this a try.

> Le 12 novembre 2019 09:24:18 GMT+02:00, Steve Lhomme <robux4 at ycbcr.xyz> 
> a écrit :
> 
>     On 2019-11-09 3:40, Rémi Denis-Courmont wrote:
> 
>         Le vendredi 8 novembre 2019, 16:40:05 EET Steve Lhomme a écrit :
> 
>             The video context is only held if the picture is created
>             successfully. It is
>             released after the picture is destroyed.
> 
>         Seems redundant with existing picture context, which typically
>         refers to the
>         video context in a way or other - unless the goal is to remove
>         the picture
>         context - which seems unrealistic to me.
> 
>         Maybe we should actually "standardize" the reference from the
>         picture context
>         to the video context, but I don't see the need for yet another
>         member in
>         crowded picture_t.
> 
> 
>     Adding the video context to the picture is exactly what was decided with
>     push. The fact it's also used the first time the vout is initialized is
>     a necessary mean to get there. In some of the proposed patches the video
>     context from the picture is already used as such instead of getting it
>     from the decoder/filter.
> 
>     I agree there's probably one too many between the picture_sys_t, the
>     picture_context_t and the video_context. At least the video context
>     exists on its own without a picture (as already used in current master).
>     Adding the destroy/clone callbacks from the picture_context_t would make
>     no sense when used without a picture.
> 
>     The picture_sys is probably a better candidate to remove from the
>     picture. In D3D the picture_context_t already has a picture_sys_t
>     because they may exist in some cases and not others (depending if they
>     came from the decoder or the display). All this will be gradually
>     removed and ultimately there won't be any picture_sys_t there. I think
>     it can be the same for other formats.
> 
>     This cleaning should happen at the end though. We can only guarantee the
>     picture_sys_t is not needed anymore once eveything is safely moved there
>     and the picture_sys_t are never used anymore. For now I'm concentrating
>     on architectural changes.
>     ------------------------------------------------------------------------
>     vlc-devel mailing list
>     To unsubscribe or modify your subscription options:
>     https://mailman.videolan.org/listinfo/vlc-devel
> 
> 
> -- 
> Envoyé de mon appareil Android avec Courriel K-9 Mail. Veuillez excuser 
> ma brièveté.
> 
> _______________________________________________
> vlc-devel mailing list
> To unsubscribe or modify your subscription options:
> https://mailman.videolan.org/listinfo/vlc-devel
>