This is an implementation of named, pre-declared workspaces. With this
implementation, workspaces can be declared in the configuration file by
name:
```
workspace "name" {
open-on-output "winit"
}
```
The `open-on-output` property is optional, and can be skipped, in which
case the workspace will open on the primary output.
All actions that were able to target a workspace by index can now target
them by either an index, or a name. In case of the command line, where
we do not have types available, this means that workspace names that
also pass as `u8` cannot be switched to by name, only by index.
Unlike dynamic workspaces, named workspaces do not close when they are
empty, they remain static. Like dynamic workspaces, named workspaces are
bound to a particular output. Switching to a named workspace, or moving
a window or column to one will also switch to, or move the thing in
question to the output of the workspace.
When reloading the configuration, newly added named workspaces will be
created, and removed ones will lose their name. If any such orphaned
workspace was empty, they will be removed. If they weren't, they'll
remain as a dynamic workspace, without a name. Re-declaring a workspace
with the same name later will create a new one.
Additionally, this also implements a `open-on-workspace "<name>"` window
rule. Matching windows will open on the given workspace (or the current
one, if the named workspace does not exist).
Signed-off-by: Gergely Nagy <niri@gergo.csillger.hu>
- Keep a root surface cache to be accessible in surface destroyed()
- Only snapshot during / right before closing, rather than every frame
- Store textures rather than elements to handle scale and alpha properly
Most visibly, fixes screen not immediately redrawing upon layer-shell
popup commits.
There's still a number of places with questionable handling left, mostly
to do with subsurfaces (like, find_popup_root_surface() doesn't go up to
subsurfaces), and session-lock. I don't have good clients to test these.
Ivan Molodetskikh