Important: Please read the Qt Code of Conduct -

[Solved] QGraphicsItem::mouseReleaseEvent not called

  • I have a QGraphicsItem subclass for which mousePressEvent is called, but mouseReleaseEvent isn't. Here's the situation:

    • A ComponentLayoutView class is derived from QGraphicsItem and represents, of course, a Component in the program.
    • Components contain Ports, so there's also a PortLayoutView, also derived from QGraphicsItem.
    • ComponentLayoutView objects create and own PortLayoutViews depending on the number and type of Ports in the Component object. @ComponentLayoutView::ComponentLayoutView(boost::shared_ptr<SimObjectBase> & bs) {
      // Set the input / output port views
      input1 = new PortLayoutView(this, bs->getConsumerPort(1), PortLayoutView::pdLeft);
      @ ComponentLayoutViews are owned by a QGraphicsScene object.
    • We set several flags in the ComponentLayoutView : @setFlags(ItemIsMovable | ItemIsSelectable | ItemSendsGeometryChanges);@ but not in the PortLayoutViews.
    • PortLayoutViews implement mousePressEvent, mouseReleaseEvent, and mouseMoveEvent.

    Most of this seems to work, including the mousePressEvent. However, the mouseReleaseEvent in the PortLayoutView is never called. Is this as expected? What must we do to detect a mouse button release?


  • I think you should use the events for the QGraphicsScene instead, which will propagate to the items.

  • I'm sorry, I'm fairly new to Qt. How do QGraphicsScene events propagate to the items?

  • Also in the QGraphicsItem "docs.":

  • A little more information: I put mouse event handlers in the scene like so: @void Layout2dScene::mousePressEvent (QGraphicsSceneMouseEvent * event)
    QGraphicsItem * item = mouseGrabberItem();

    void Layout2dScene::mouseReleaseEvent (QGraphicsSceneMouseEvent * event)
    QGraphicsItem * item = mouseGrabberItem();
    @ In both cases the item is the ComponentLayoutView object that contains the port view, not the PortLayoutView object. Nevertheless, the mousePressEvent is delivered to the PortLayoutView object that was clicked on. The mouseReleaseEvent goes to a generic handler in QGraphicsItem.

    Must we handle mouse events at the ComponentLayoutView level, then manually figure out which port, if any, the mouse is in? And if so why do mouse press, move, enter, leave, etc., events all seem to go to the PortLayoutView as we expected?


  • Oh, and just to be clear, PortLayoutView is declared as:
    @class PortLayoutView : public QGraphicsItem

    virtual void mousePressEvent (QGraphicsSceneMouseEvent * event);
    virtual void mouseReleaseEvent (QGraphicsSceneMouseEvent * event);
    virtual void mouseMoveEvent(QGraphicsSceneMouseEvent * event);

  • it's hard to see with all those code fragments.

    Can you make a minimal compilable example so we can try things out.

  • One way to determine what's wrong is good old debugging.

    Put a debug line in every mouseReleaseEvent / mousePressEvent of every QGraphicsView, scene, item you use.
    That way you will know where the action happens.

    Hope this helps.

  • Found a solution, but don't know if this is intentional behaviour on Qt's part. Note the commented-out line in the mousePressEvent handler:
    @void PortLayoutView::mousePressEvent ( QGraphicsSceneMouseEvent * event )
    boost::shared_ptr<PortBase> sp = port.lock();
    if (!sp) return;
    if (event->buttons() & Qt::LeftButton)
    startPos = event->pos();
    if (event->modifiers() & Qt::ControlModifier)
    // QGraphicsItem::mousePressEvent(event);
    Thanks again!

  • bq. Found a solution, but don’t know if this is intentional behaviour on Qt’s part

    You didn't use the mousepressEvent from the base class and you don't need to if you reimplement it.

    Good work! glad you made it.

  • What I meant by "intentional behaviour" is that the program behaves differently when the base class handler is called, even if the handler doesn't do anything. One would assume that the QGraphicsItem::mousePressEvent(event); statement would be harmless, if slightly inefficient. This seems to be the case for the other mouse events. Perhaps one day I'll have some time to trace through the Qt source and research this in more depth.


  • I'm sure that will be a good learning experience!

    For the main philosophy "this": is also a good read.

  • @deanc Thanks a lot for posting brief problem and solution. It solved my problem too.
    Thanks once again !

  • I have a similar problem. I have a QGraphicsScene which hold QGraphicsView as parent. then I got QGraphicsProxyWidget holding myWidget derived from QWidget which can not receive mouseReleaseEvent but no problem receiving mousePressEvent. I have a QLabel and a QPushButton in myWidget Class. I tried debugging but It confused me when tracing into Qt internal class. Then I just simply change myWidget parent from QWidget to QPushButton and it worked. Hope my solution helps.


  • @Eddy said:

    You didn't use the mousepressEvent from the base class and you don't need to if you reimplement it.

    That's not necessarily true. The base class events implement certain functionality -they may select or deselect items, move items, etc. If you do not call a base class event, you lose that functionality. Worse still, if you reimplement some events, but not other, you might stumble into undefined behavior, as e.g. the base class receives a mouseRelease event where it expected to receive a mousePress first.

    Before you decide not to call an event of the base class, you should

    • Check which functionality it implements, and whether you can afford to do without it
    • Check which events may be tied together, and whether you should reimplement all or none of them
    • Think about the alternative of either calling the base class event before or after your custom event code

Log in to reply