I've tested this change with different modes from the monitor default
and validated that dpms still works, at least on my machine. If there's
a good reason why this exists, feel free to correct me, but this helps
get us closer to a flicker-free experience.
getMonitorFromVector can return nullptr on empty m_monitors, as such is
the case when the compositor is going down and a surface exist. return
early in vectorToWindowUnified if that happends.
* compositor: dont try to focus unmapped window
if lastwindow is unmapped it hits getWindowInDirection and nullptr
derefs window->m_workspace. and coredumps. triggered by dual monitor and
one client on each surface with a combination of animation and
killactive / movefocus keybind.
* keybindmgr: use newly added aliveAndVisible()
use newly added aliveAndVisible() over visible()
currently it damage the entire window if its floating and not x11 nor
fullscreen meaning damage isnt working at all for floating. im tracing
this back to a364df4 where the logic changed from damaging window only
if size was being forced to now unconditonally doing it.
change clampWindowSize to return as a bool if size changed and only
damage window if it got clamped.
ubsan reports under wonky situation a runtime error of uninitialized
value lookups because of m_capStatus isnt initialized. so default
initialize it.
OpenGL.cpp:3331:26: runtime error: load of value 190, which is not a valid value for type 'bool'
- Take out signal set up into a subroutine;
- Use `sigaction` instead of `signal` for consistent behavior across UNIX platforms;
- Enable a warning when a signal handler set up fails;
- Don't do anything to SIGKILL, since it cannot be handled.
* swipe: Fix previous workspace remembering in workspace gesture
Fixes a bug that previous workspace does not exist after swiping to a workspace
* tests: Test that `workspace previous` works after workspace gesture
* moveActiveToWorkspace: remember previous workspace unconditionally
Depedencies where not tracked in the pkgconfig leading to programs
who scan dependencies using it to fail/not track them.
I noticed this while building Hyprland on openSUSE where the -devel
package didn't include the dependencies it once had when Meson was
used previously.
