Code review comment for lp:~kdub/mir/post-if-optimizable

On Wed, May 14, 2014 at 10:49 AM, Kevin DuBois
<email address hidden> wrote:
>>
>>
>> On Wed, May 14, 2014 at 3:54 AM, Kevin DuBois
>> <email address hidden> wrote:
>> >> So this makes it easier to bring up platforms in the short term,
>> >> and will
>> >> probably be a no-op in the long term?
>> >
>> > I'm confused as what will 'be a no-op in the long term'.
>> > post_renderables_if_optimizable() will be the hardware optimized
>> > interface that compositors should try if they know they're doing
>> > something simple, and post_update() will be the interface that the
>> > compositors use to do anything they want with GL
>>
>> What I mean is - as a platform (and, conceivably, Renderable) grows
>> in
>> sophistication, there will be fewer and fewer RenderableLists that
>> contain only unoptimisable Renderables. So what I mean by
>> “no-op”
>> is that post_renderables_if_optimizable() will end up always
>> returning
>> true.
>
> sure, I hope that it returns true more often as time goes on. I think
> though that we'll always need the post_update() interface for the
> compositors who know beforehand that they're doing something like
> wobbly windows that isn't representable by a RenderableList.

It kinda is - the results of the GL composition as a Renedeable, plus a
Cursor Renderable.

This does suggest that maybe the interaction with cursor drawing needs
thinking.

None of this is blocking, just thinking.