":connect' error - "multiple definition..." ?

Anonymous_Banned275

@Chris-Kawa So I am reading it wrong , again

this (pointer) - source - is referring to a class which generates the SIGNAL

such class is called from the "parent" class which contains the desired destination SLOT

so I already have an instance of the destination and the source is in it.

If I create another class instance (pointer for the "destination class " - that is NOT where I want to send the message to.

I am actually looking n at this as a "parent / child" setting - I want to send a message generated by the child "up stream " to the parent..

Chris Kawa

this (pointer) - source - is referring to a class which generates the SIGNAL

No, not class. An instance of a class. An instance generates a signal. Class is just a type.

such class is called from the "parent" class which contains the desired destination SLOT

Classes are types. Types are not called. You can call o slot of an instance of a class.

If I create another class instance (pointer for the "destination class " - that is NOT where I want to send the message to.

Correct. Don't make a new instance. Get a pointer to the exact instance you want to send the signal to.

I am actually looking n at this as a "parent / child" setting

In general it doesn't have anything to do with parents/children. Connections are between instances of some classes. Those classes or instances don't have to be related in any way. The fact that in your case one of them is parent of the other is just anecdotal. It doesn't have special meaning for the connection.

I want to send a message generated by the child "up stream " to the parent

Although you technically can do that it's a bad design. Your operations should always be downstream, not up. Children are not supposed to know anything about their parents. This creates problems like yours - how do I get a pointer to parent. The answer is you shouldn't. It's the parent that should connect to its child's signal, because that's where you have the right context - pointers to both parent and child. When you're in the child's code you don't have easy access to the pointer to parent. Yes, you can call stuff like parentWidget() on this to get the parent and cast it to the right type, but that's ugly and hacky. The child gains knowledge about parent which it shouldn't have and you'll get into circular include problems fast.

I'd say move your connection to the parent class. this wil point to your parent class instance and you'll have a pointer to the child there, so you can do a connection like

connect(pointer_to_the_child, &BT_ChatServer::NewMessage, this, &BT_SPP_CA_MainWindow::PostMessage);

Anonymous_Banned275

@Chris-Kawa So I moved the "connect" to "parent" .... Left the emit and the old test "connect" within "child".
The old test stuff still "passes the test message within child", the new connect does not.

I am not sure if the "emit" which is in the child get "pickled up" by the "connect" in the "parent"...

I do agree with your comments about the software should be "from top down" , however the hardware runs just the opposite - from child to parent ....from down to up

I actually tried to use inheritance and got stuck with same issue - how to pass messages upstream.

However, after last few days I feel more confident I could try inheritance again.....

I am going to take few days off - saying final goodbye to old friend ... not sure if we are going to fly or drive west from Houston...

cheers

Chris Kawa

@AnneRanch said:

So I moved the "connect" to "parent" .... Left the emit and the old test "connect" within "child".

emit should be in the object that signals the change, so yes, it should stay in the child, as that is where the change takes place.
connect should be either where the slot is or somewhere outside entirely. For a parent/child connection it should be in the parent. For a connect of this to this it should be in that class, not in the parent, so what you did sounds right.

The old test stuff still "passes the test message within child", the new connect does not.

The type of connect doesn't matter. If it stopped working you just made something wrong.

however the hardware runs just the opposite - from child to parent

No, it doesn't. Hardware just does stuff. Child monitors the hardware and signals change in state. Parent picks up the signal and does something as a reaction. The information does flow upwards but the object structure is top to bottom.

I actually tried to use inheritance and got stuck with same issue - how to pass messages upstream.

It really doesn't sound like a job for inheritance. Inheritance represents "is" relation between classes. A dog is an animal. A car is a vehicle. Those are examples suited for inheritance. A dog has a collar. A car has a steering wheel. The "has" relation should be expressed as composition, not inheritance i.e. a parent has a child. In your case there is a MainWindow that has a ChatServer. That sounds like composition. MainWindow should not be a chat server.

Anonymous_Banned275

@Chris-Kawa I am stuck....

Here are my two test macros

    Q_ASSUME(connect(this, &BT_SPP_CA_MainWindow::NewMessage, this, &BT_SPP_CA_MainWindow::PostMessage));
    Q_ASSUME(connect(server,&BT_ChatServer::NewMessage,this,&BT_SPP_CA_MainWindow::PostMessage));

The top one works when I "emit" NewMessage in BT_SPP_CA_MainWindow instance.

The bottom one does not work - it seems to ignore the "emit New Meassage" in BT_ChatServer and I have several of them which works fine "locally" in BT_ChatServer .

My "server" is initialized and processes other calls...including local emit NewMessage...

What else can I do to identify / debug the issue ?

Chris Kawa

@AnneRanch Are you sure you're connecting the right instances? You can print out the pointer and compare the addresses i.e. qDebug() << server; where you make that connection and qDebug() << this; inside BT_ChatServer where you say it works. It should be the same address printed out. If not then you mixed your instances somewhere.

Anonymous_Banned275

@Chris-Kawa Here is a snippet of app output
The pointers match

void BT_SPP_CA_MainWindow::Connect_MainServer()
BT_ChatServer::BT_ChatServer(QObject *)
"\n\t TASK Run \nBT_ChatServer::BT_ChatServer(QObject *)\n"
!!!!!!!!!!!!!!!!!!! this BT_ChatServer(0x5653f9b74550)
conncet OK
TASK Construct BT_ChatServer
BT_ChatServer::BT_ChatServer(QObject *)
& line "84"
!!!!!!!!!!!!!!!!!!!!server BT_ChatServer(0x5653f9b74550)
!!!!!!!!!!!!!!!!!!! this BT_ChatServer(0x5653f9b74550)
"emit NewMessage(text)"
void BT_ChatServer::PostMessage(const QString)
"\n\t TASK Run \nvoid BT_ChatServer::PostMessage(const QString)\n !!!! REMOTE !!!! TEST MESSAGE "

Chris Kawa

@AnneRanch It's very difficult for me to decipher that stream of random text without the matching code, but you said earlier that you had this:

Q_ASSUME(connect(server,&BT_ChatServer::NewMessage,this,&BT_SPP_CA_MainWindow::PostMessage));

but the output says this:

!!!!!!!!!!!!!!!!!!!!server BT_ChatServer(0x5653f9b74550)
!!!!!!!!!!!!!!!!!!! this BT_ChatServer(0x5653f9b74550)

so which is it, because it's physically impossible that this output matches that connect. this can't be both BT_SPP_CA_MainWindow and BT_ChatServer at the same time, so what's the real code that produces this output?.

SimonSchroeder

@Axel-Spoerl said in ":connect' error - "multiple definition..." ?:

so wrapping it with Q_ASSUME(connect(...))will tell right away, if the connection was successful.

Will it, though? I had to quickly look up Q_ASSUME. From my understanding it just tells the compiler that you assert that the return value of the connect is always true. The compiler is allowed to use this assumption for optimization. If the assumption is wrong you have undefined behavior. Or does it actually do a check? I have seen people suggest Q_ASSERT, but this will only do the connect in the debug build, not in the release build. Is there a macro you could actually use here? Or am I mistaken about Q_ASSUME?

With the new connect syntax rarely if ever will the connect fail if it compiles. I have to admit that I never check the return value. Qt will write out warning to the console, though, when the connect fails. This is how you can figure this out at runtime as well.

JonB

@SimonSchroeder said in ":connect' error - "multiple definition..." ?:

If the assumption is wrong you have undefined behavior. Or does it actually do a check?

I believe you are quite correct here. It does not do a check, unlike Q_ASSERT, rather it tells the compiler what is going to happen.

The OP actually asked about this in Usage of Q_ASSUME in connect ?.

Axel Spoerl

@SimonSchroeder said in ":connect' error - "multiple definition..." ?:

Q_ASSUME?

Maybe I wasn't clear in my proposal to wrap QObject::connect() into Q_ASSUME().
This is for debugging and analyzing, if a connection was successful at all.
It's not recommended for production code, however.
Q_ASSUME() instructs the compiler to assume, that the wrapped condition is true. That's for optimizing purposes. The condition itself can be optimized out, even if that's not always the case. It remains the compiler's decision.
Q_ASSERT() will always kick in debug builds and is completely optimized out on production. In a debug build, Q_ASSUME becomes a Q_ASSERT, so they are treated identically.

Having said that, the safe way to assert a connection is to read it's return value into a variable, debug it and act on it. Connections should be designed such, that they always work. Once the plumbing is done, all the wrappers can go out. They create unnecessary overhead in stable code. That's why QObject::connect() is not marked nodiscard, like other functions that do not allow their return value to be ignored.

Anonymous_Banned275

I have added some"B@W" so we can see the compiler / linker actual output.
It still shows that the pointers match.

BT_ChatServer::BT_ChatServer(QObject *)
& line "84"
DEBUG connect void BT_SPP_CA_MainWindow::Connect_MainServer() @ line "1399" this BT_SPP_CA_MainWindow(0x558ce48df8f0, name="BT_SPP_CA_MainWindow")
DEBUG connect void BT_SPP_CA_MainWindow::Connect_MainServer() @ line "1400" server BT_ChatServer(0x558ce4df7170)
!!!!!!!!!!!!!!!!!!!!server BT_ChatServer(0x558ce4df7170)
DEBUG connect void BT_ChatServer::startServer(const QBluetoothAddress &) @ line "109" this BT_ChatServer(0x558ce4df7170)
"emit NewMessage(text)"
void BT_ChatServer::PostMessage(const QString)

As far as Q_ASSUME goes - it does check the syntax , not the actual function of the "connect" .
This is helpful.

I do understated there may be alternatives to accomplish passing messages between objects, but my goal is to find out WHY this does not work.
( The BT_ChatServer is a member of BT_SPP_CA_MainWindow )

If I have a multiple and WRONG "connect" would it still do something , even if incorrect ?
It is (remotely) possible that the "new" syntax does not works BETWEEN objects ?
( I just cannot believe this "passing messages", was never tried by some other coder ....)

Let's keep in focus, it does not matter what / where the error is, I just want to resolve it,
putting the blame on something / somebody IS NOT my goal.

.

Anonymous_Banned275

Maybe Q_ASSUME will do more...

I went back to the "old style" too,
The old style complains about missing parameter.

Then my usage of Q_ASSUME is also wrong - I need to fix that and see if it will help to locate the problem...

..and yes,, I hand two instances of "server" , but either one should pass the test message....

QObject::connect: No such signal BT_ChatServer::&BT_ChatServer().NewMessage() in bt_spp_ca_mainwindow.cpp:1466
QObject::connect:  (receiver name: 'BT_SPP_CA_MainWindow')
ASSERT failure in Q_ASSUME(): "Assumption in Q_ASSUME("connect (server,SIGNAL(&BT_ChatServer().NewMessage()), this,SLOT(&BT_SPP_CA_MainWindow().PostMessage()) );") was not correct", file bt_spp_ca_mainwindow.cpp, line 1470
11:16:58: /mnt/07b7c3f8-0efb-45ab-8df8-2a468771de1f/PROJECTS/BT_JAN13_/Qt-5.15.2/widgets/mainwindows/mdi/mdi crashed.


            Q_ASSUME(
                    connect
                            (server,SIGNAL(&BT_ChatServer().NewMessage()),
                             this,SLOT(&BT_SPP_CA_MainWindow().PostMessage())
                             );

                    );


           //Q_ASSUME(connect(this, &BT_SPP_CA_MainWindow::NewMessage, this, &BT_SPP_CA_MainWindow::PostMessage));

            Q_ASSUME(connect(server,&BT_ChatServer::NewMessage,this,&BT_SPP_CA_MainWindow::PostMessage));

            qDebug() <<"DEBUG connect " << Q_FUNC_INFO << "@ line " << QString::number(__LINE__) << " server  " << server;

Anonymous_Banned275

more debugging....
Added another class member , "standard" Qt Designer Form class with single TEST button emitting test message.

SUCCESS The message gets passed to the "parent" using "connect"....

What is puzzling - the debug outputs are in reverse sequence...

So - I must have a wrong "connect " somewhere.... in my original code....

Anonymous_Banned275

SOLVED

Eureka !!
The original QT example code has multiple instances of ChatServer and no "delete".

Now a minor curiosity question remains:
if the test code emits TWO messages , why do I get only ONE "postmessage " ?

void BT_ChatServer::PostMessage(const QString)
"\n\t TASK Run \nvoid BT_ChatServer::PostMessage(const QString)\n !!!! REMOTE !!!! TEST MESSAGE "
void BT_ChatServer::PostMessage(const QString)
"\n\t TASK Run \nvoid BT_ChatServer::PostMessage(const QString)\n !!!! REMOTE !!!! TEST MESSAGE "
void BT_SPP_CA_MainWindow::PostMessage(const QString)
"\n\t TASK Run \nvoid BT_SPP_CA_MainWindow::PostMessage(const QString)\n !!!! REMOTE !!!! TEST MESSAGE "