[PATCH] shell: Enhance the basic random positioning algorithm

[Bill Spitzak]

Kristian Høgsberg wrote:

> That's a nice improvement indeed.  I've committed it since it's
> obviously a step forward, but I think we could use the input region
> extents for placement.  There's been talk of a outline region or such,
> something to define the "edge" of the window for snapping and
> placement, but I haven't yet been able to think of a case where that's
> different from the input region.  So we can just use the input region
> for now, and if a case for an outline region comes up, we can add that
> then.
> 
> Kristian

If this is what I was trying to descibe, this region I want is the 
"inside" of the edge. If you imagine a Windows95-style window but with a 
shadow, this region does not include the shadow and does not include the 
3 or 4 pixels that are the rounded-looking "edge". However it includes 
what X would call the "contents" and also the "flat" part of the "titlebar".

This would allow maximization to be done by the shell choosing a surface 
size that is larger than the output. The edges and shadows would be 
drawn in this extra area that is clipped by the compositor. The shell 
needs to know this region so it can choose this larger size.

The main reason for this is so that maximize-vertical can be supported 
and produces the same vertical size as full-screen maximize. It is also 
useful as a snap-to edge when moving windows, assuming they can be 
dragged off the screen at all.

Because it seems wasteful for clients to allocate and draw invisible 
window edges and shadows, I also propose that the shell tell the client 
both the desired surface size and the bounding box of this region. Most 
drawing api's can easily translate and clip off the unnecessary 
graphics, and more intelligent clients can actually skip the graphics or 
adjust the shading on still-visible ones.