[RFC weston] clients: Don't attach a buffer if mouse cursor surface is unchanged

[Daniel Stone]

Hi Jasper,

On Thu, 26 Jul 2018 at 03:53, Jasper St. Pierre <jstpierre at mecheye.net> wrote:
> From IRC conversations with krh a long time ago, this is indeed intentional and the cursor surface should "lose its role" in modern parlance.
>
> The original intention was to prevent glitching of the cursor surface. e.g. If the left side of the surface has a resize left cursor, you leave, and hover over the right, you don't want to see the resize left cursor for a split second before the resize right cursor appears.
>
> The original implementation of Weston respected this and would only change the cursor on set_cursor calls and would not even remember a per-client cursor surface. This behavior has probably been lost in numerous reactors by now.

Right, it's not about having the cursor surface stick to the normal
surface, but about having the buffer stick to the cursor surface. The
former definitely isn't controversial, and EFL is re-associating the
cursor surface with the normal surface. What's happening is that they
need to attach/damage/commit to that surface to get the content.

Cheers,
Daniel