The window module re-entered the same shared_mutex while refreshing IPC state:
update() took the lock and then called queryActiveWorkspace(), which tried to
lock it again. That is undefined behavior for std::shared_mutex and could
manifest as a deadlock.
Remove the recursive lock path and reset the derived window state before each
IPC refresh. That keeps solo/floating/swallowing/fullscreen classes from
sticking around when the client lookup fails or a workspace becomes empty.
Signed-off-by: Austin Horstman <khaneliman12@gmail.com>
- Replaced pass-by-value std::string parameters with const std::string&
or std::string_view to prevent SSO overallocations.
- Refactored static mapping functions in UPower to return
std::string_view instead of constructing std::string literals, enabling
perfect cache locality.
- Optimized string concatenation in hot loops (network IPs, inhibitor
lists, sway window marks) by using std::string::append() and
pre-reserving capacity instead of overloaded operator+ which produces
temporary heap instances.
These optimizations reduce high-frequency memory churn and overall heap
fragmentation within the main rendering loops.
Signed-off-by: Austin Horstman <khaneliman12@gmail.com>
Child exec failure paths were returning success, which masked command launch
errors from callers.
I switched the child-side failure exits to _exit(127) and added errno-specific
logging so failures propagate with actionable diagnostics.
Signed-off-by: Austin Horstman <khaneliman12@gmail.com>
Hyprland IPC had fd lifecycle risks on failure/shutdown paths and used a
spin-sleep listener model.
I initialized fd state defensively, tightened connect/close/shutdown handling,
moved to blocking read with newline framing, and added RAII-style fd cleanup in
socket1 reply paths.
Signed-off-by: Austin Horstman <khaneliman12@gmail.com>
SafeSignal could queue events forever when worker threads emitted faster than
the main loop could consume, which risks memory growth and stale updates.
I added a queue cap with a drop-oldest policy so growth stays bounded under
burst load, plus a regression test that validates bounded delivery.
Signed-off-by: Austin Horstman <khaneliman12@gmail.com>
SleeperThread control flags were shared across threads without consistent
synchronization.
I converted the run/signal flags to atomics and updated wait predicates and
lifecycle transitions to use explicit atomic loads/stores.
Signed-off-by: Austin Horstman <khaneliman12@gmail.com>
Reassigning SleeperThread could replace a joinable std::thread and trigger
std::terminate.
I now stop and join any existing worker before reassignment, then reset control
state before starting the replacement worker.
Signed-off-by: Austin Horstman <khaneliman12@gmail.com>
The keyboard-state module crashes with SIGSEGV in libinput_device_ref
when a new input device appears in /dev/input/.
Three bugs fixed:
1. Missing NULL check: tryAddDevice() calls libinput_path_add_device()
which returns NULL on failure, then immediately passes the result to
libinput_device_ref() without checking. On laptops, virtual input
devices (power buttons, lid switch, etc.) appear and disappear in
/dev/input/ triggering the hotplug handler; if libinput can't open
one of these, the NULL return causes the segfault.
2. Missing cleanup on device removal: The IN_DELETE handler erased
devices from the map without calling libinput_path_remove_device(),
leaving dangling pointers in the libinput context.
3. Thread safety: libinput_devices_ was accessed from 3 threads
(main/GTK, libinput_thread_, hotplug_thread_) without any mutex.
Fixes#4851
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Move GTK operations from IPC thread to GTK main thread in Window module
- Move GTK operations from IPC thread to GTK main thread in WindowCount module
- Move GTK style class operations from IPC thread to GTK main thread in Submap module
- Language and Workspaces modules already safe (only update internal state)
Co-authored-by: Alexays <13947260+Alexays@users.noreply.github.com>
The monitor signal handlers (signal_monitor_added, signal_monitor_removed)
were never disconnected during SIGUSR2 reload. Each reload accumulated
additional handlers, causing multiple layer surfaces to be created when
monitors were hotplugged.
This fix:
- Stores signal connections as class members
- Disconnects them before reconnecting in bindInterfaces()
- Clears stale outputs_ on reload
