This uses the new `view->outputs` bitset to calculate the intersection
with *all* outputs the view is currently visible on.
This ensures that the invisible resize handle works across outputs
while still making sure that it won't leak into neighboring ones just
because it is positioned closely to an output edge (either manually,
maximized or snapped via SnapToEdge or SnapToRegion).
Co-Authored-By: @johanmalm
Fixes: #1486
Reported-By: @lurch
...to edges actually encountered by motion during interactive moves and
resizes.
In addition, ignore edge resistance and attraction for minimized views.
...in view_move_to_edge. If a view is fully within the usable area of
its original output, it should be fully within the usable area (if
possible) on its new output.
This ensures that client-side decorations (and maybe other client
behaviors) behave intuitively during interative moves, but means that
drag to snapped position -> drag to maximize -> un-maximize
will put the window back to its natural geometry (saved before the first
snap) rather than back to the snapped position. This is a change in
behavior, but a) simplifies the logic a bit, because the tiled state is
always in sync with the window geometry and b) in some sense is "more
correct", because the window has been "visually" un-tiled the minute the
window starts dragging.
Note that maximizing using an action (including the button) a window
that has been snapped and then un-maximizing the window will put the
window back to the snapped position, as it always has.
This also fixes a bug wherein dragging a window and pressing a hot-key
to maximize or fullscreen a window could leave a snap-region highlight
visible after the interactive move was canceled.
...on view destruction because focus_change_notify() in seat.c changes
server->active_view so the logic introduced by 6c6e406 (which checks if
view =! active_view) is no longer right.
The glitches described in the commit below have not come back.
6c6e406507Fixes#1393
If keyboard-layout-per-toplevel-window is used, reset the group (index)
for each window on --reconfigure whenever the keymap has changed.
Refactor to use a common configure function for reconfigure and
keyboard-group creation.
Co-authored-by: @johanmalm
Fixes#1407
Add the -m|--merge-config command line option to iterate backwards over
XDG Base Dir paths and read config/theme files multiple times.
For example if both ~/.config/labwc/rc.xml and /etc/xdg/labwc/rc.xml
exist, the latter will be read first and then the former (if
--merge-config is enabled).
When $XDG_CONFIG_HOME is defined, make it replace (not augment)
$HOME/.config. Similarly, make $XDG_CONFIG_DIRS replace /etc/xdg when
defined.
XDG Base Dir Spec does not specify whether or not an application (or a
compositor!) should (a) define that only the file under the most important
base directory should be used, or (b) define rules for merging the
information from the different files.
ref: https://specifications.freedesktop.org/basedir-spec/basedir-spec-latest.html
In the case of labwc there is a use-case for both positions, just to be
clear, the default behaviour, described by position (a) above, does NOT
change.
This change affects the following config/theme files:
- rc.xml
- menu.xml
- autostart
- environment
- themerc
- themerc-override
- Theme buttons, for example max.xbm
Instead of caching global config/theme directories, create lists of paths
(e.g. '/home/foo/.config/labwc/rc.xml', '/etc/xdg/labwc/rc.xml', etc).
This creates more common parsing logic and just reversing the direction
of iteration and breaks early if config-merge is not wanted.
Enable better fallback for themes. For example if a particular theme does
not exist in $HOME/.local/share/themes, it will be searched for in
~/.themes/ and so on. This also applies to theme buttons which now
fallback on an individual basis.
Avoid using stat() in most situations and just go straight to fopen().
Fixes#1406
...and notify the client of the preferred output scale when doing so.
This should allow clients to better determine an optimal size if they
are initially configured (unmapped) with zero size.
In particular, this fixes an issue with foot:
https://codeberg.org/dnkl/foot/issues/1579
...(calling `wlr_session_change_vt()`) because when the session is
switched, the access to the keyboard is lost and therefore the RELEASE
event will not be passed to the compositor.
Fixes bug whereby compositor crashes on VT change on FreeBSD.
Fixes#1424
This builds on the work of @Consolatis in #1018.
Co-authored-by: Consolatis <35009135+Consolatis@users.noreply.github.com>
Co-authored-by: Andrew J. Hesford <ajh@sideband.org>
