Adding a virtual function causes strange result...

[jsulm]

@SPlatten said in Adding a virtual function causes strange result...:

clsXMLnode::strGetAttribute(QString)

You should check what is happening there

[SPlatten]

@jsulm , I didn't use the override keyword, should I?

[J.Hilk]

@SPlatten said in Adding a virtual function causes strange result...:

reinterpret_cast

when using virtual inheritance !?!?!?

use dynamic_cast or if QObject classes object_cast

reinterpret_cast takes the raw pointer and considers it as being of the derived type. However, because of the virtual inheritance, a slight adjustment must be done to the pointer to point to the correct method dispatch table, and that's precisely what dynamic_cast will do.

just as wrong, but probably wouldn't have resulted in a crash would be the use of static_cast.

[Christian Ehrlicher]

This post is deleted!

[SPlatten]

@J-Hilk said in Adding a virtual function causes strange result...:

dynamic_cast

I've replaced reinterpret_cast with dynamic_cast, same result.

[jsulm]

@SPlatten said in Adding a virtual function causes strange result...:

I didn't use the override keyword, should I?

Yes you should. To ask the compiler to check whether you're really overriding virtual method. Can help to detect bugs at compile time.

As I suggested: check what is happening in clsXMLnode::strGetAttribute(QString).
My guess is that the strGetData() method from base class is called which returns an empty string, so the code then fails later.

[Christian Ehrlicher]

@SPlatten said in Adding a virtual function causes strange result...:

I've replaced reinterpret_cast with dynamic_cast, same result.

For sure not.

btw: dynamic_cast was proposed to your exact same problem some weeks ago already: https://forum.qt.io/topic/132420/how-to-access/4

[SPlatten]

@jsulm, I have a breakpoint on the first line of the function getValue, execution doesn't get as far as the breakpoint before the issue occurs.

[SPlatten]

@Christian-Ehrlicher That issue was resolved, this is new and dynamic_cast wasn't used to resolve the last issue, it was resolved with the suggested qobject_cast.

[J.Hilk]

@SPlatten your stack trace actually points to your clsXMLnode class.
clsXMLnode::strGetAttribute

[jsulm]

@SPlatten I'm talking about clsXMLnode::strGetAttribute(QString).
If you check the stack trace you posted you will see that it is called shortly before the application stopps...

[Christian Ehrlicher]

@SPlatten said in Adding a virtual function causes strange result...:

it was resolved with the suggested qobject_cast.

Since your class is derived from QObject you can also use qobject_cast. Please learn the basics.

[SPlatten]

@Christian-Ehrlicher , I find your additions very insulting, I’ve been coding professionally since 1986.

[JonB]

@SPlatten
I'm not sure what part of your code you think is implicated. I see:

1  QListData::size() const
2  QList<std::pair<QString, QString>>::length() const 

and that is failing (according to your "the execution stops in qlist.h on line 115") on trying to get the length of the QList. Invalid QList at that point? That is what I see?

[Christian Ehrlicher]

@SPlatten said in Adding a virtual function causes strange result...:

I’ve been coding professionally since 1986.

But not with c++ (at least I hope so) - when I want to learn a new language I normally start to learn the basics. But you ignore those basics in nearly every 'problem' you have and then wonder why people complain about the lack of the basics.

[SPlatten]

@Christian-Ehrlicher that’s a huge assumption on your part.

[kshegunov]

@J-Hilk said in Adding a virtual function causes strange result...:

@SPlatten your stack trace actually points to your clsXMLnode class.
clsXMLnode::strGetAttribute

If he reinterpret_casts it's going to point to that class and method since that's what to "reinterpret" means.

@SPlatten said in Adding a virtual function causes strange result...:

I've replaced reinterpret_cast with dynamic_cast, same result.

You shouldn't use reinterpret_cast to begin with, as it's rather apparent you don't understand the implications of it; specifically to polymorphic calls and multiple inheritance. So while @Christian-Ehrlicher may seem brusque, you should do yourself a favor and listen to the advice.

As for the issue at hand, you have called strGetAttribute on an invalid object, I'm reasonably certain. The reasons may vary, and reinterpret_casting incompatible types is one. Dangling pointers and uninitialized memory are other candidates. You need to sift through the debugger to find it. Out of curiosity where does clsQtLayout::addButton call clsXMLnode::strGetAttribute and where does clsXMLnode::setWidget call clsQtLayout::addButton - provide the relevant code around the calls.

[SPlatten]

@kshegunov , using the debugger I have a break point on the last if condition below:

clsXMLinterface* pobjInterface(reinterpret_cast<clsXMLinterface*>(pobjButton));
clsXMLnode* pobjNode(nullptr);
if ( pobjInterface != nullptr ) {
    pobjNode = pobjInterface->pobjGetNode();
}
if ( pobjNode == nullptr ) {
    return;
}

pobjButton is a pointer to an instance of my derived class clsQtRadioButton:

    class clsQtRadioButton : public QRadioButton, public clsXMLinterface {
    Q_OBJECT

pobjNode is returned as not null, pobjGetNode:

clsXMLnode* pobjGetNode() { return mpobjNode; }

I can see in the debugger the pobjInterface looks ok and mpobjNode is correct as I have assigned a human readable description to it:

However pobjNode doesn't appear to be correct:

And then trying to call strGetAttribute fails and causes the problem, but its because pobjNode isn't correct, why ?
If I change the code to:

clsQtRadioButton* pobjRadioBtn(qobject_cast<clsQtRadioButton*>(pobjButton));
clsXMLnode* pobjNode(pobjRadioBtn->pobjGetNode());

This way pobjNode is valid and there is no problem, however I need a polymorphic way to access pobjGetNode because there can be lots of different types of object derived from the same base class and I need the functionality to work for all of them.

[jsulm]

@SPlatten said in Adding a virtual function causes strange result...:

why ?

We do not know whether and how you're initialising it...

"pobjNode is returned as not null" - doesn't mean it was initialised. You should initialise your pointers with nullptr before they are assigned proper pointers, else they can contain garbage.

[SPlatten]

@jsulm , here is the constructor for the class clsXMLinterface:

clsXMLinterface::clsXMLinterface(clsXMLnode* pobjNode) : mpobjNode(pobjNode) {

And clsQtRadioButton:

clsQtRadioButton::clsQtRadioButton(clsXMLnode* pobjNode, QString* pstrCSS
                                                       , QWidget* pParent)
                            : QRadioButton(pParent), clsXMLinterface(pobjNode) {

The formatting of this constructor is common to all instances. Widgets like instances of clsQtRadioButton are created by another call:

            pobjWidget = qobject_cast<QWidget*>(clsCNT::pCreate(this,
                                                                mstrName,
                                                                strCSS,
                                                                slstProperties,
                                                                pobjParWidget));

clsCNT::pCreate:

QWidget* clsCNT::pCreate(clsXMLnode* pobjNode, const QString& crstrType
                        ,QString& rstrCSS, QStringList& rslstProperties
                        ,QWidget* pobjParent) {
    QWidget* pobjWidget(nullptr);    
    if ( pobjNode != nullptr ) {
        if ( crstrType.compare(clsCNT::mscszButton) == 0 ) {
            pobjWidget = new clsQtPushBtn(pobjNode, &rstrCSS, pobjParent);

The above is just a sample, pobjNode is passed by pCreate as this.