a little question about polymorphism in constructor

Question

3.67/5 (3 votes)

See more:

class A{
public:
	A(){msg();}
	virtual void msg(){AfxMessageBox(_T("A"));}
};

class B: public A{
public:
	B():A(){}
private:
	virtual void msg(){AfxMessageBox(_T("B"));}
};

void test(){
	B b;	//will Message("A");  -------why? I think I can get Message("B")!-----

	A*pA=&b;
	pA->msg();//will Message("B");
}

Posted 21-Aug-13 23:38pm

jiazhiqiang

Add a Solution

Comments

Maciej Los 22-Aug-13 5:40am

And where is a question?

Thanks7872 22-Aug-13 5:41am

That's the puzzle. :(

Maciej Los 22-Aug-13 5:46am

;)

Stefan_Lang 22-Aug-13 7:39am

It is in the comments, first line of function test().

Richard MacCutchan 22-Aug-13 5:46am

This is easy enough to test for yourself.

pasztorpisti 22-Aug-13 6:03am

I don't think so... ;-)

Richard MacCutchan 22-Aug-13 6:09am

It took me less than 2 minutes.

pasztorpisti 22-Aug-13 6:16am

This will print "A" and then "B". The question of OP is why does it print "A" as the first message. Its not trivial at all and most C++ programmers unfortunately don't know the answer to this question. I have fixed a lot of bugs as a result of this.

Richard MacCutchan 22-Aug-13 6:23am

I didn't say it was trivial, I said it was easy to test. OP's message was not clear, so I had thought it was more of a "what will happen?".

pasztorpisti 22-Aug-13 7:10am

The question is indeed unusually embedded as a comment.

2 solutions

Add a Solution

Add your solution here

Treat my content as plain text, not as HTML

Preview 0

…

Existing Members

Sign in to your account

...or Join us

Download, Vote, Comment, Publish.

Your Email
Password
Forgot your password?

Your Email
This email is in use. Do you need your password?
Optional Password

I have read and agree to the Terms of Service and Privacy Policy
Please subscribe me to the CodeProject newsletters

When answering a question please:

Read the question carefully.
Understand that English isn't everyone's first language so be lenient of bad spelling and grammar.
If a question is poorly phrased then either ask for clarification, ignore it, or edit the question and fix the problem. Insults are not welcome.
Don't tell someone to read the manual. Chances are they have and don't get it. Provide an answer or move on to the next question.

Let's work to help developers, not make them feel stupid.

This content, along with any associated source code and files, is licensed under The Code Project Open License (CPOL)

CPallini · Accepted Answer · 2013-08-22T00:01:00

Solution 1

Quote:
B b; //will Message("A"); -------why? I think I can get Message("B")!-----

B constructor explicitely calls the A constructor, hence you get such a message.
Sorry, that was wrong, as pasztorpisti pointed out. You may find correct info here:
C++ FAQ: [23.5] When my base class's constructor calls a virtual function on its this object, why doesn't my derived class's override of that virtual function get invoked?[^]

Quote:
pA->msg();//will Message("B");

This is the usual polimorphic behaviour.

Posted 22-Aug-13 0:01am

CPallini

Updated 22-Aug-13 1:12am

v2

Comments

pasztorpisti 22-Aug-13 6:14am

No, the message "A" is a bug in the code and it has nothing to do with the explicit call. The explicit call does exactly the same that the auto-generated ctor upcall would do.

CPallini 22-Aug-13 7:15am

You are right I have updated my answer.

Stefan_Lang 22-Aug-13 7:36am

Great link. I thought to look that one up when I saw the question, but you obviously beat me to it :-)

P.S.: upvoted on the basis of version 2

CPallini 22-Aug-13 8:02am

Thank you.

pasztorpisti 22-Aug-13 7:53am

+5, much better now :-)

CPallini 22-Aug-13 8:02am

Thank you.

pasztorpisti · Accepted Answer · 2013-08-22T00:20:00

Solution 2

EDIT: Richard MacCutchan came up with a good idea of writing a tip/trick about this problem. I've just put together a small discussion with example code about this problem: Never call virtual methods from the constructor or destructor of your classes in C++/C#/Java![^]

I am lazy to write here the answer again as I have already answered this question a few times. The reason in a few words: Never call virtual methods directly or indirectly from (base class) constructors. Please read my posts regarding this:

http://www.codeproject.com/Messages/4619258/Re-Serial-Port-OK-in-class-constructor-but-not-out.aspx[^]
http://www.codeproject.com/Messages/4619708/Re-Serial-Port-OK-in-class-constructor-but-not-out.aspx[^]

Posted 22-Aug-13 0:20am

pasztorpisti

Updated 22-Aug-13 5:52am

v2

Comments

Stefan_Lang 22-Aug-13 7:35am

That first link is a bit ... compressed and may require checking previous parts of that thread. I'll give you a 5 for that second link however. Great posting.

pasztorpisti 22-Aug-13 7:51am

Thank you!

CPallini 22-Aug-13 8:02am

5.

pasztorpisti 22-Aug-13 8:06am

Thank you!

Richard MacCutchan 22-Aug-13 8:10am

5: The description in link 2 deserves publication as a Tip at the least.

pasztorpisti 22-Aug-13 8:12am

Good idea! A virtual +5. :-)

H.Brydon 25-Aug-13 22:29pm

[The only answer I give a +5, for the text "never call virtual functions from constructors..."]. IMHO a constructor should never call them, and any articles about doing it are really academic. My recommendation is to not call anything in the same class hierarchy except a private non-virtual method or (C++11) a delegating constructor.

pasztorpisti 26-Aug-13 5:54am

Thank you! Some people were uncomfortable about the "Never call..." stuff (so I adjusted the title of the final version) and I understand that to some degree because making it forbidden to use a core language feature is a bit strange. I rather made some language-specific recommendations. Still I think its a good practice to do what you have described, I've never been in the need of calling a virtual method in a constructor but I've fixed a lot of related bugs in C++ and debugged for a whole day in java to find my first related bug (because I've put a breakpoint in a derived virtual method, I've seen a lot of null member variables but I didn't expect the method call before the ctor so I was in a "WTF state" before finding out that the ctor hasn't yet been executed...).

In C++ calling a virtual method from ctor is basically a bug. It would be nice to be able to tell the compiler to generate the following ctor:
constructor()
{
auto generated: call the base class constructor (if we have a base class)
auto generated: initialize the table pointer to the fake-assert-vtable of this class
[your constructor code here]
auto generated: initialize the table pointer to the real-vtable of this class
}

This way your ctor code would be executed with a fake-assert-vtable that could help in catching most of these "bugs" at least at runtime when you create an instance... (because calling any virtual methods from the fake-assert-vtable would fire an assert.)