[RFC Wayland 1/2] protocol: double-buffered state for wl_surface

[Pekka Paalanen]

On Wed,  3 Oct 2012 16:14:15 +0300
Pekka Paalanen <ppaalanen at gmail.com> wrote:

> This change is breaks the protocol.

Hi all,

these two patches are my current plan, and after 2-3 days of careful
specification writing, I curse the typo quoted above. ;-)

I have implemented the wl_surface.attach double-buffering in weston and
demo clients so far.

All these protocol changes require some changes in all clients. I will
also fix Weston to live up to these specifications.

Now is the time to point out problems in the suggested protocol
changes, bikeshed the names and terminology, and make sure the language
is clear, understandable, and defines all relevant aspects.

These changes should fix all the races we might have for a single
wl_surface. I can see wl_surface.commit be used also for multi-buffer
surfaces and what not. There will be further work to specify and fix
shell protocols to be atomic, too, leveraging wl_surface.commit as much
as possible.

One open question I haven't really thought about yet is the frame
callback, should it fall under wl_surface.commit too?


Thanks,
pq