In more than one of my Qt applications I've noticed that, whenever the menu bar is clicked, the last signal to have been sent from a widget within the GUI is re-sent before the menu action is invoked. Most of the time this doesn't matter; but on some occasions it matters very much.
In a few cases where the widget's signal is connected to one of its own slots, it's straightforward to begin the slot with a
if (hasFocus())
{
// ...
}
...block so that such spurious signals, not generated by the user actually clicking on the widget, can be ignored.
However, I've recently identified that this behaviour is responsible for several related bugs where the spurious signals are passed on through several layers of the program before being acted upon, so simply checking whether a particular widget has focus is not trivial to implement.
My question, therefore, is:
why on earth does clicking on a menu item cause a signal to be emitted from a widget elsewhere on the screen? I can't find this behaviour documented anywhere?
how do I stop it?
Many thanks,
Stephen.
As you deduced yourself, the QLineEdit::editingFinished signal can be emitted twice if the user press enter and the QLineEdit loses the focus later. An old bug was created but behaviour has not been changed : https://bugreports.qt.io/browse/QTBUG-40.
Your solution is fine : overload QLineEdit and check if the value has changed before triggering the signal. You can use the flag QLineEdit::isModified for the check : it has a default value of false and is changed to true whenever the user changes the line edit's contents. It has to be manually reset to false when you emit the signal.
You can also do that check in the receiver object, just checking if the value has a meaning (is different from the last recorded value for example) and process it or not.
Therefore the polluting signals would not be a problem. It is in my opinion the cleanest solution, then you have a robust model. Because the user could also hit enter a few times with the same text value and you have to handle this anyway.
You can also use QLineEdit::returnPressed but this requires the user to always press enter to validate a value. Not always intuitive but forces the user to explicitly validate the inputs.
Related
What's the difference between setting ContextMenuPolicy to DefaultContextMenu or CustomContextMenu? From what I can tell, they're the same things? Either way, you get to create a custom context menu if needed.
They are certainly not the same.
While they both could create a custom menu, their behavior is quite different. Consider the QWidget documentation:
The default value of this property is Qt::DefaultContextMenu, which means the contextMenuEvent() handler is called.
This means that you need to specifically override contextMenuEvent() in order to eventually show the context menu. This usually means that subclassing is required and only the widget will handle that event (which could also be ignored) and possibly its menu.
Note that overriding contextMenuEvent() could be done also for special cases of the event management, from the simplest one (accepting it, to avoid propagation to the parent) to more complex situations that need to set up some aspects before letting the default implementation to handle the event as usual.
With Qt::CustomContextMenu, the signal customContextMenuRequested() is emitted.
This means that any object can connect to that signal (and potentially more than once), even a completely unrelated one, and eventually do something from there. This also means that there is absolutely no access to the event, which will always be considered as accepted.
I am trying to fix a bug in a Qt app which I did not write. The window changes the background color of the entire window to red and puts up some buttons, dialog boxes, etc. When the escape key is pushed, the boxes and buttons go away, leaving an empty red screen. The Cancel button does the right thing in returning to the previous window. I think I need to somehow be notified of when the escape key is pushed, and then call the same function as the cancel pushbutton does. Hopefully, I can limit the scope of this special action to when the problem window is up. I am an experienced programmer but a complete Qt newbie. This app is purely C++. To my knowledge, it does not appear to use any QML. I am still searching through the Qt online documentation, but any suggestions / examples are appreciated.
This depends a lot on your specific Qt version and setup of your application. That said, I'll take a shot at helping. In the class where you're trying to intercept the escape key press, assuming it's inheriting QObject, simply override the virtual function eventFilter. Inside your overridden instance of eventFilter, check that the QEvent type is a QEvent::KeyPress, and then check whether the key of that KeyPress is the escape key, and handle as needed.
Be sure that you pass the event out of your function, else you'll see your overridden function eat all events. If you explicitly want the event to be "eaten", simply do not return it from your function. For specifics and examples, check out documentation of QObject::eventFilter().
My team is developing an UI for an apparatus with touch screen and we would like it to emit a sound (from a buzzer) each time the user correctly presses a button (so using the release event). Notice that I don't want to play the sound after each click on the interface, but only when the click is over a button.
We use many types of button, sometimes QPushButton and most of the times customized buttons derived from QAbstractButton. In most cases these buttons get an objectName.
So I supposed in order to do that, I would have to catch the MouseButtonRelease event and since I'm already working with a subclass of QApplication to handle excetions, I decided to do this in the notify function.
I tried, then, some methods to recognized when the MouseButtonRelease was related to a button but none of them were successfull. The best one, verifying the receiver's objectName was still not good enought not only because not all buttons had an objectName (which, of course, can be handled), but specially because not always the event was caught for buttons with names set. In other words, sometimes I would click in a button and it recognizes the event and sometimes I would click in the same button and the event is not recognized.
I did some research and another method I found was to set an event filter in the MainWindow, but not all widgets have the MainWindow as their parent which means I would have to Ctrl+c / Ctrl+V the same code time after time when I obviously want something more localized (i.e. in only one spot).
So why it happens that the notify not always handles the events? And how could I do this? Any suggestions are appreciated specially one that is less heavier then handling the events globally.
As info, the other two ways I tried to catch the events with similar or even worst results inside notify were with receiver->inherits("...") and qobject_cast< QAbstractButton* >(receiver).
I have an issue with associating a keypress to a QAction. I'm mapping Enter to an action like this:
myAction->setShortcut(Qt::Key_Return);
The problem is that the QAction is also triggered when I type a value in a QSpinBox and then press Enter. I was expecting the spinbox to consume the event (so it's not caught by the mainwindow) but it's not working like that.
I've seen that I can add a context to my shortcut, I tried all values but this does not solve my issue.
#Matthew It is indeed rejecting the event:
case Qt::Key_Enter:
case Qt::Key_Return:
d->edit->d_func()->control->clearUndo();
d->interpret(d->keyboardTracking ? AlwaysEmit : EmitIfChanged);
selectAll();
event->ignore();
emit editingFinished();
return;
I find this behavior quite strange.
From my knowledge the QSpinBox default behavior does wait until you press return to accept the value. Technically the value is set/accepted as soon as you change it.
If you're looking to change that behaviour, i.e. when you hit return on your QSpinBox you the value is actually set/accepted, then you will most likely have to handle the associated key press event yourself on the QSpinBox and mark it as accepted (i.e. event->accepted() ).
Otherwise, this sounds like it is working as originally designed although not as you want.
I have exactly the same problem, only in PyQt and with QLineEdit
It may be that QActions are always considered menu actions, or global, and have a higher priority than Widget Default keys.
My "solution" was to create a global list of my actions with important keys (Return, Tab, Backspace etc.) as shortcuts.
Then I hook into QLineEdit focusInEvent and focusOutEvent. Each time the event happens I loop over the actions-list and deactivate them all on focusIn and activate them again on focusOut. Then I give the event to the original focus event handlers.
I really hope there is a better solution.
I need help to understand to use QEvents in QT, this is driving me crazy.
I am writting an application using custom events, but as in QApplication::postEvent function, it's necesary to specify the target object.
As I understand, it's possible to post events to Qt's event loop with
QApplication::postEvent(obj_target, QEvent myevent);
This means that I'm trying to catch "myevent" event in obj_target an do some stuff.
But I need to post events without specify a target object, as QMouseEvent or QKeyEvent do
I mean, when clicking in a QMainWindow with a lot of buttons, how is that I can click
any button and that button is pressed?
What is the target object when the click event is posted?
It's possible to register objects to "listen" for a specific event?
I'm really confused, it's possible to post an event without specifying a target object?
Thank you very much in advance
There is no trivial way to post events "globally", as Dan has said. All of the event dispatching of native events is done by private Qt implementation code.
The important distinction is:
There are native messages/events, delivered by the operating system, usually received by a window-specific event loop.
There are QEvents.
Internally, Qt keeps track of the top-level Widgets (windows, really), so when it receives an event from the OS, it knows which window it should go to - it can match it using the platform window id, for example.
QEvent delivery makes no sense without a receiving object, since sending an event to an object really only means that QObject::event(QEvent*) method is called on that object. It's impossible to call this method without having an object instance!
If you want to synthesize a global key press or mouse click event, then you have to figure out what object the event goes to. Namely:
Identify what top-level window (widget) the event should go to. You can enumerate top level widgets via qApp->topLevelWidgets().
Identify the child widget the event should go to. If it's a keyboard event, then sending the event to currently focused widget via qApp->focusWidget() is sufficient. You need to enumerate the child widgets to find the deepest one in the tree that overlaps the mouse coordinates.
Send the correct QEvent subclass to the widget you've just identified. Events delivered to top-level widgets will be routed to the correct child widget.
When sending mouse events, you also need to synthesize relevant enter and leave events, or you risk leaving the widgets in an invalid state. The application.cpp source file should give you some ideas there.
This doesn't give you access to native graphical items, such as menus on OS X.
Please tell us exactly what you're trying to do. Why do you want to post a broadcast event? Who receives it? Since your own QObject-derived classes will receive those broadcasts, I presume, it's easy enough to use signal-slot mechanism. You'd simply connect(...) those receiver classes to some global broadcaster QObject's signal(s).
For this purpose, I have a specific singleton class which I call GuiSignalHub. It regroups all the application-wide signals.
Objects that want to trigger an application-level action (such as opening context help) just connect their signal to the GuiSignalHub signal. Receivers just connect the GuiSignalHub to their slot.