Qt state machines: more transition attempts than expected - c++

Question

I have created a simplified example of a state machine transitioning between 2 states on a button click (+ a checkbox showing the active state). I am seeing more calls to the transitions'eventTest method than I expect.
#include <QtStateMachine/QStateMachine>
#include <QtStateMachine/QSignalTransition>
#include <QtWidgets/QApplication>
#include <QtWidgets/QCheckBox>
#include <QtWidgets/QPushButton>
class MySignalTransition : public QSignalTransition
{
public:
MySignalTransition(QString name, QPushButton* button, QCheckBox* checkBox) :
QSignalTransition(button, &QAbstractButton::clicked),
cb(checkBox)
{
setObjectName(name);
}
protected:
bool eventTest(QEvent* e) override {
qDebug() << objectName() << "tested on" << e->type() << '|' << cb->isChecked();
return QSignalTransition::eventTest(e);
}
private:
QCheckBox* cb;
};
int main(int argc, char* argv[]) {
QApplication app(argc, argv);
app.setQuitOnLastWindowClosed(true);
QWidget w;
w.setFixedSize(200, 200);
QPushButton button(QStringLiteral("Click me"), &w);
QCheckBox checkbox(&w);
button.move(10, 80);
checkbox.move(10, 20);
QStateMachine m;
QState s1, s2;
m.addState(&s1);
m.addState(&s2);
m.setInitialState(&s1);
s1.assignProperty(&checkbox, "checked", false);
s2.assignProperty(&checkbox, "checked", true);
//button will be used to make the transition back and forth between states s1 and s2
MySignalTransition t1("Transition 1", &button, &checkbox), t2("Transition 2", &button, &checkbox);
t1.setTargetState(&s2);
t2.setTargetState(&s1);
s1.addTransition(&t1);
s2.addTransition(&t2);
w.show();
m.start();
return app.exec();
}
On top of the checkbox, I placed a qDebug() << ... to watch when Qt attempts to perform transitions.
The first attempt is right when the state machine starts, before I do any click on the button. I guess (but I am not sure) that this is expected.
The real problem happens when I actually do click on the button. I see:
eventTest is called from the starting state (event->type() == QEvent::None)
eventTest is called from the starting state (event->type() == QEvent::StateMachineSignal)
eventTest is called from the target state (event->type() == QEvent::None)
The one call I am most surprised about is point 1.
What causes the state machine to try the transition again when I click on the button with QEvent::None? This is especially surprising to me because it is always a transition the state machine has already tried (either on start or on the previous click).
If I were to create another transition out of s1 and/or s2, should I be worried that transition could have a higher priority than my signal transitions without any way for me to control it?
Edit
To address #Scheff's cat comment below, I changed the transition class to be:
class MySignalTransition : public QSignalTransition
{
public:
MySignalTransition(QString name, QPushButton* button, QCheckBox* checkBox) :
QSignalTransition(button, &QAbstractButton::clicked),
cb(checkBox)
{
setObjectName(name);
}
protected:
bool event(QEvent* e) override {
return false;
}
bool eventTest(QEvent* e) override {
qDebug() << objectName() << "tested on" << e->type() << '|' << cb->isChecked();
return transition;
}
private:
QCheckBox* cb;
public:
inline static bool transition = false;
};
and right after creating the push button, added:
QObject::connect(&button, &QPushButton::clicked, []() { MySignalTransition::transition = true; });
By doing that, the state machine lets me do 1 click before it starts an infinite loop.
When clicking, MySignalTransition::transition is set to true, then eventTest is called.
With a breakpoint, it is easy to see eventTest is never called with event->type() == QEvent::StateMachineSignal.
To me, this is a clear indication there is more to it than just probing.
First, it would need to do the probing with the actual event, not with a dummy event.
Second, it would not do the transition if the result of the probing was correct, I still should see the 2nd call to eventTest.
Last, I thought maybe the state machine tries to execute as many transitions as it can and will then pass the click signal to whatever other state it lands on.However, I am not very satisfied by that explanation because the attempt to execute an automatic transition is already done when entering the state (it is the third call to eventTest) and because I think my click should be applied on the state that is active at the very moment I do it, not on just "whatever other state it lands on".