Some users have reported rare crashes when pressing the Enter key on the keyboard to confirm input in the normal software keyboard, particularly in Super Smash Bros. Ultimate while entering the name of a ruleset or controller layout.
It is suspected that the QLineEdit::returnPressed signal is causing a race condition as confirming input through other means does not produce the crash. Since Qt::QueuedConnection posts an event to the event queue of the callee's thread instead of executing it directly on the caller's thread, this eliminates any potential race conditions from occurring in this scenario.
Previously the text string for the inline software keyboard was being sent instead of the normal software keyboard, leading to empty text being sent all the time.
The Qt Software Keyboard frontend attempts to mimic the software keyboard rendered by the Nintendo Switch.
This frontend implements multiple keyboard types, such as the normal software keyboard, the numeric pad software keyboard and the inline software keyboard.
Keyboard and controller input is also supported in this frontend.
Keyboard input is handled as native keyboard input, and so the on-screen keyboard cannot be navigated with the keyboard arrow keys as the arrow keys are used to move the text cursor.
Controller input is translated into mouse hover movements on the onscreen keyboard or their respective button actions (B for backspace, A for entering the selected button, L/R for moving the text cursor, etc).
The text check dialogs can also be confirmed with controller input through the use of the OverlayDialog
Massive thanks to Rei for creating all the UI for the various keyboards and OverlayDialog. This would not have been possible without his excellent work.
Co-authored-by: Its-Rei <kupfel@gmail.com>
Since C++17, the introduction of deduction guides for locking facilities
means that we no longer need to hardcode the mutex type into the locks
themselves, making it easier to switch mutex types, should it ever be
necessary in the future.
Uses Qt's built-in interface instead of rolling our own separate one on
top of it. This also fixes a bug in reject() where we were calling
accept() instead of reject().
