Question

I’m trying to use QDBusPendingCallWatcher to watch an async call. Some sample code like this:

{
    // interface = new QDBusInterface(...);
    QDBusPendingCall pcall = interface->asyncCall("query");
    QDBusPendingCallWatcher *watcher = new QDBusPendingCallWatcher(pcall, this);
    QObject::connect(watcher, SIGNAL(finished(QDBusPendingCallWatcher*)), this, SLOT(handler(QDBusPendingCallWatcher*)));
}

and handler function:

void Client::handler(QDBusPendingCallWatcher* call)
{       
    QDBusPendingReply<QString> reply = *call; 
    // do something
}

My questions are:

1. It looks like QDBusPendingCallWatcher uses shared data pointer inside, is it safe to not manually delete the watcher pointer? Just leave the scope and forget it?

2. If I can let the smart pointer of pendingcall to do all the tricks, can I use just one QDBusPendingCallWatcher pointer in my class to watch all the async calls? Like this:

   {
       QDBusPendingCall pcall = interface->asyncCall("query");
       watcher = new QDBusPendingCallWatcher(pcall, this);
       QObject::connect(watcher, SIGNAL(finished(QDBusPendingCallWatcher*)), this, SLOT(handleOne(QDBusPendingCallWatcher*)));
   
       pcall = interface->asyncCall("anotherQuery");
       watcher = new QDBusPendingCallWatcher(pcall, this);
       QObject::connect(watcher, SIGNAL(finished(QDBusPendingCallWatcher*)), this, SLOT(handleTwo(QDBusPendingCallWatcher*)));
   }
   

   Will this makes a disaster? Or should I use multiple pointers for each call?

Thanks!

Answer 1

Take a closer look at the QDBusPendingCallWatcher documentation:

The slot connected to by the above code could be something similar to the following:

void MyClass::callFinishedSlot(QDBusPendingCallWatcher *call)
{
    QDBusPendingReply<QString, QByteArray> reply = *call;
    if (reply.isError()) {
        showError();
    } else {
        QString text = reply.argumentAt<0>();
        QByteArray data = reply.argumentAt<1>();
        showReply(text, data);
    }
    call->deleteLater();
}

The call of QObject::deleteLater is the key: This means Qt will delete the Object as soon as execution returns to the event loop.

As long as you call deleteLater inside Client::handler(...), you don’t need to – more precisely you musn’t – call delete watcher; anywhere. The only thing you have to ensure is that noone uses the object behind call after the slot returns.