We already allow some xwayland-unmanaged surfaces to take focus on map,
if indicated by wlr_xwayland_or_surface_wants_focus(). But once these
surfaces lose focus, they never regain it again.
Add desktop_focus_view_or_surface() and call it in the appropriate
places to allow these views to regain focus in the usual ways (e.g.
clicking on them or focus-follows-mouse).
Commit 7e72bf975f changed behavior to not automatically focus xwayland
views using the "Globally Active" input model (WM_HINTS.inputs = false
but WM_TAKE_FOCUS listed in WM_PROTOCOLS).
One undesired side effect of this change is that when a dialog is
closed, the parent window is not re-focused if "Globally Active". This
issue is seen for example with JDownloader. It can be solved taking a
similar approach to what is done for unmanaged xwayland views: allow
automatic re-focus between views sharing the same PID.
Note that it's difficult to completely solve all of the focus issues
with Globally Active views without proper WM_TAKE_FOCUS support.
Implementing proper support is difficult since it requires wlroots
changes and would also mean waiting for a message round-trip in
desktop_focus_topmost_view().
Fixes (partially): 7e72bf975f
("view/xwayland: avoid focusing views that don't want focus")
We were checking for a locked session in desktop_focus_view(), but there
are several other call sites of seat_focus_surface() which were missing
such a check. Any one of those could cause the lock screen to lose focus
(making the session impossible to unlock) or another surface to gain it
(breaching the session lock).
To fix the issue, make any call to seat_focus_surface() no-op when the
session is locked. Add a specific seat_focus_lock_surface() function
which is the only way to bypass the check and is called only from
session-lock.c.
... especially regarding whether a (view *) parameter may be NULL. It's
confusing when some functions accept NULL and others don't, and could
trip someone up.
I'm partly to blame for the inconsistency, since (if memory serves) I
added view_is_tiled() and view_is_floating(), which do accept NULL.
In detail:
- Make view_is_tiled() and view_is_floating() no longer accept NULL.
- Rename view_isfocusable -> view_is_focusable for consistency with
other view_is_ functions.
- Eliminate view_inhibits_keybinds() as it only existed to safely accept
NULL and check a single flag, which can be checked directly.
- Add assert(view) to remaining public view_ functions to catch
accidentally passing NULL.
- Inline inhibit_keybinds() into view_toggle_keybinds(). It is closely
related and not called from anywhere else; inlining it allows
eliminating an extra assert() which is now impossible.
Some X11 applications (MATLAB is known to be one) apparently still use
the outdated "globally active" input focus model, in which they declare
they don't want the window manager to give them input focus, but expect
to be able to take it explicitly themselves via XSetInputFocus().
Such applications are not a good fit for the Wayland world, and may have
issues even with remotely modern X11 window managers that prevent such
"focus stealing". Labwc certainly doesn't (and won't) allow it. However,
to avoid breaking such applications entirely, let's still allow the user
to give focus by clicking in the window.
For the sake of applications that legitimately don't want to be given
input focus (such as taskbars or other "panels"), we still don't give
focus to them automatically when another view is closed, and they aren't
shown in Alt-Tab.
This reverts commit cae96b0cce.
XWayland views can self-declare that they don't want keyboard focus via
the ICCCM WM_HINTS property. Most of the logic is already in place to
avoid giving focus to such views (e.g. taskbars).
Add a couple of missing pieces to make this work:
- Hook up view_isfocusable() to look at WM_HINTS for XWayland views
- Adjust desktop_focus_topmost_mapped_view() to skip unfocusable views
- Connect to wlr_seat_keyboard_state's focus_change event.
- Add view_from_wlr_surface(), which does what the name says.
- Use focus_change event along with view_from_wlr_surface() to update
server->focused_view and set SSD states via view_set_activated().
- Eliminate desktop_focused_view() since server->focused_view should be
reliably up-to-date now.
- Eliminate view_focus/defocus() since we can now just call
seat_focus_surface() directly.
Make desktop_focus_view() always switch to the workspace containing the
view being focused. It doesn't make much sense for an invisible view to
have the keyboard focus.
Also add an optional "raise" parameter to desktop_focus_view(). This
allows the common pattern of desktop_focus_view() + view_move_to_front()
to be reduced to a single function call.
Our current approach to handling the focused/active view is a bit
confusing. In particular, it's hard to be sure when server->focused_view
is or isn't in sync with the real wlroots keyboard focus.
Try to clean things up a bit. In particular:
- Add comments to server->focused_view and desktop_focused_view() to
clarify that they should match, but it's not guaranteed.
- desktop_focused_view() now prints a warning if it detects that
server->focused_view is out of sync. We should keep an eye out for
this warning, and if we see it, try to figure out why it happened.
- For consistency, use only "focus/defocus" as the verbs in function
names rather than "activate". This is a bit arbitrary, but the idea is
that focus is the primary action while the active/inactive state is a
side effect.
- view_focus/defocus() replace view_set_activated() and now update both
focus and active/inactive state, to try to keep them in sync.
- Add comments at view_focus/defocus() to warn against calling them
directly (we should generally call the desktop.c functions).
- desktop_focus_view(NULL) is now forbidden and is no longer handled as
a special case to clear the focus. This was (at least to me) a
surprising behavior and caused trouble when working on another change.
- To maintain existing behavior, desktop_focus_topmost_mapped_view() now
explicitly clears the focus if there are no mapped views.
There should be no behavioral change here.
...to reduce code duplication.
The function populates an array with views which meet any set of critera
from:
- current-workspace
- no-always-on-top
- no-skipWindowSwitcher (window-rule)
Make src/osd.c use this new interface. Note that always-on-top views are
still filtered out from the window-switcher and that desktop_cycle_view()
needs to be re-worked before always-on-top views can be opted in.
view_minimize() does not need to call desktop_move_to_back() because the
stacking order is not changed and the windowSwitcher uses the scene-tree
nodes anyway.
Note: Movement of xwayland sub-views still relies on keeping server->views
in sync with z-order
...to increase xwayland and xdg-shell encapsulation and to avoid passing a
function pointer as an argument in `xwayland_move_sub_views_to_front()`
which is inconsistent with labwc design patterns.
Rename view-impl.c to view-impl-common.c
Move function declarations that are common to view-implementations from
view.h into view-impl-common.h
- Process layer-shell keyboard interactivity in the map and commit
handlers only, rather than in layers_arrange(). This allows handling of
the layer-surface that emitted the event rather than iterating over all
surfaces in the output layer-tree, and therefore avoids having to guess
which surface should have keyboard preference (and it might not be the
last one added to the list which was the assumption previously).
- Prevent seat_focus_surface() from setting keyboard focus if a layer-shell
surface with exclusive keyboard-interactivity has the focus.
- Set cursor_context type for layer-surfaces without node-descriptors
in order to set keyboard focus correctly in cursor_button_press().
Tested satisfactorily with xfce4-panel and gtk-layer-demo.
Fixes#725 and #704
- Move xwayland-specific struct definitions to new xwayland.h header
- Move xwayland_move_sub_views_to_front() from desktop.c to xwayland.c
- Split out xwayland_server_init/finish() from server_init/finish()
- Rename new_xwayland_surface -> xwayland_new_surface and
xwayland_surface_new() -> handle_new_surface() for consistency
- Add "mapped" argument to xwayland_unmanaged_create() so that we can
make unmanaged_handle_map() private to xwayland-unmanaged.c
...to allow these surfaces to receive pointer button events
Test by running `gtk-layer-demo --keyboard exclusive`, then open the
'set margin' dialog and try setting the margin with the pointer.
...and thus simplify the usage of wlr_scene_node_at(). Specifically:
- desktop.c: in get_cursor_context() use node-description for
layer-surfaces and layer-popups. This lays the foundations for a
pointer-enter-event being sent when a new layer-surfaces appears under
the pointer (even if the pointer doesn not move).
- layers.c:
* Iterate over `struct wlr_scene_tree *layer_tree[]` rather than
`struct wl_list layers[]` when arranging layers to avoid surfaces
being out of sync with nodes
* Set signal handlers after scene node creation to avoid configure
race conditions
* Handle scene-node destroy event rather than event of
`struct wlr_layer_surface_v1`
* Arrange layers on map and unmap
* Handle client request for layer-change
Fixes issue #667
- Store a pointer to the `struct view` in `struct ssd`
- Pass `struct ssd *` instead of `struct view *` to ssd functions
- Add `ssd_get_margin()` convenience function
- Minimize includes in `ssd.h`
- Avoid repetitive `view->ssd.margin` pattern
- Use `struct ssd *` or `const struct ssd *` rather than `struct view *`
where convenient
Part of the motivation is to make it easier to separate `struct ssd`
from `struct view` in a future commit.
Add xdg_surface_from_view() + xwayland_surface_from_view() accessors
that assert() the view is of the expected type before returning.
Fix a real bug in xdg.c parent_of() that dereferenced
`view->xdg_surface->toplevel` without first checking `view->type`.
The goal of the new accessors is to catch similar bugs in future.
IMHO it encourages better design (by making dependencies more obvious)
to have source file/header file pairs like view.c/view.h, rather than a
monolithic header like labwc.h with everything in it.
I don't think we need to break up all of labwc.h at once, but maybe we
can start pulling it apart bit by bit as it's convenient.
Also:
- Move "struct border" to ssd.h so that view.h can use it without pulling
in all of labwc.h.
- Add a missing required #include within scaled_font_buffer.h (forward
declaration of "struct font" is not enough).
Fix a couple of glitches seen when exiting interactive move/resize:
- Cursor briefly set to left_ptr rather than the correct cursor image
- Cursor not updated if the view being moved/resized is destroyed
Also make sure to exit interactive mode if the view is going fullscreen
(labwc gets very confused otherwise).
Code changes in detail:
- Factor out set_server_cursor() which will set the correct cursor
image for non-client areas (either XCURSOR_DEFAULT or one of the
resize cursors).
- Unify the logic from cursor_rebase() and process_cursor_motion by
factoring out cursor_update_common(). This corrects some logic
discrepancies between the two, which should be a good thing(TM).
- Remove the extra cursor_set(XCURSOR_DEFAULT) from interactive_end()
and instead rely on cursor_update_focus() to do the right thing.
- Simplify cursor_button() by just calling interactive_end() when we
want to exit interactive mode.
- Call cursor_update_focus() from view_destroy() if the view had mouse
focus or was being interactively moved/resized.
v2: Eliminate force_reenter parameters and figure out automatically
when we need to re-enter the surface.
v3: Rename wlseat -> wlr_seat.
v4: Simplify client/server cursor logic.
The topmost view may not always be the focused view; for example,
when Audacious's main window is focused but the floating Search
Tool window remains on top of it. In that case the floating window
(not the main window) should be the first view selected in the
window switcher.
