> Wait a minute, post() is a no-op as is flip() for Android. You must be
> planning to change this in future if DisplayGroup is to have any function at
> all...?
>
> Previously Android correctly had an empty flip() (now an empty post()),
> because it's a high-level platform abstraction that just needs to
> gl_swap_buffers.
>
> I suggest part of the problem then is that you've misappropriated flip() on a
> platform that has no flip().
Right, a no-op flip() seemed inappropriate for the android platform, as it did a "flip" operation from within gl_swap_buffers(). So, a user of the interface might be calling gl_swap_buffers(), which drives the display on android, and on mesa, it would just be rendering without posting to screen. It seems more appropriate to have gl_swap_buffers handle updating the context, and have flip/post() drive the content making it to the screen.
> Wait a minute, post() is a no-op as is flip() for Android. You must be
> planning to change this in future if DisplayGroup is to have any function at
> all...?
>
Yes, the subsequent branch uses the interface changes made here: bazaar. launchpad. net/~kdub/ mir/android- multimonitor/ view/head: /src/platforms/ android/ server/ display_ group.cpp# L68
http://
> Previously Android correctly had an empty flip() (now an empty post()),
> because it's a high-level platform abstraction that just needs to
> gl_swap_buffers.
>
> I suggest part of the problem then is that you've misappropriated flip() on a
> platform that has no flip().
Right, a no-op flip() seemed inappropriate for the android platform, as it did a "flip" operation from within gl_swap_buffers(). So, a user of the interface might be calling gl_swap_buffers(), which drives the display on android, and on mesa, it would just be rendering without posting to screen. It seems more appropriate to have gl_swap_buffers handle updating the context, and have flip/post() drive the content making it to the screen.