Copy constructors

K

keith

Hi, I've been through the FAQ-lite and can't see this mentioned, so
here goes...

I've got an abstract base class called Base which has no copy
constructor at all. In the derived class I have something like this:

// derived.h-----------------------------------------------------
class DerivedPrivate; // Not defined here
class Derived : public Base
{
private:
class DerivedPrivate* const p_;

public:
Derived();
Derived(const Derived& s);
// remainder snipped
}
// end--------------------------------------------------------------

// derived.cc----------------------------------------------------
// definition of DerivedPrivate skipped
Derived::Derived() : p_(new DerivedPrivate()) {}

Derived::Derived(const Derived& s) : p_(new DerivedPrivate())
{ *p_ = *(s.p_); }
// remainder skipped
// end--------------------------------------------------------------

Now this all compiles and works just fine, but when I turn on "-Wall -
W" in gcc, it tells me that:

derived.cc:134: warning: base class 'class Base' should be explicitly
initialized in the copy constructor

I'm afraid I'm being rather dense today, as I don't understand what
it's complaining about. Can someone explain for me please?

Thx!
 
B

Barry

Hi, I've been through the FAQ-lite and can't see this mentioned, so
here goes...

I've got an abstract base class called Base which has no copy
constructor at all. In the derived class I have something like this:

// derived.h-----------------------------------------------------
class DerivedPrivate; // Not defined here
class Derived : public Base
{
private:
class DerivedPrivate* const p_;

public:
Derived();
Derived(const Derived& s);
// remainder snipped
}
// end--------------------------------------------------------------

// derived.cc----------------------------------------------------
// definition of DerivedPrivate skipped
Derived::Derived() : p_(new DerivedPrivate()) {}

Derived::Derived(const Derived& s) : p_(new DerivedPrivate())
{ *p_ = *(s.p_); }
// remainder skipped
// end--------------------------------------------------------------

Now this all compiles and works just fine, but when I turn on "-Wall -
W" in gcc, it tells me that:

derived.cc:134: warning: base class 'class Base' should be explicitly
initialized in the copy constructor

I'm afraid I'm being rather dense today, as I don't understand what
it's complaining about. Can someone explain for me please?

It complains that you should write your Derived ctor in this way:

Derived::Derived()
: Base()
, p_(new DerivedPrivate())
{}
 
K

keith

It complains that you should write your Derived ctor in this way:

Derived::Derived()
: Base()
, p_(new DerivedPrivate())
{}

OK, this makes the warning go away - thanks. Now, what's happening?
I thought that by the time we reached Derived's constructor (in this
case, copy constructor), Base was guaranteed to have been fully
constructed. This result implies I am wrong...
 
J

Joe Greer

(e-mail address removed) wrote in @k35g2000prh.googlegroups.com:
OK, this makes the warning go away - thanks. Now, what's happening?
I thought that by the time we reached Derived's constructor (in this
case, copy constructor), Base was guaranteed to have been fully
constructed. This result implies I am wrong...

Don't read too much into the syntax. This basically just tells the
compiler which constructor to use for the base class and provides the
opportunity to provide parameters to the base class' constructor. By the
time your Derived member initialization occurs and your constructor body is
executed, the base class will have been initialized.

joe
 
R

robin

OK, this makes the warning go away - thanks. Now, what's happening?
I thought that by the time we reached Derived's constructor (in this
case, copy constructor), Base was guaranteed to have been fully
constructed. This result implies I am wrong...- Hide quoted text -

- Show quoted text -

Hmmm... but we write such code:

class base
{
public:
base(int bn){}
~base(){}
};

class derived : public base
{
public:
derived(int dn) : base(dn){} // call base class constructor
~derived(){}
};

When it comes to the execution of derived(int dn){}, according to your
understanding, it is meaningless to call base() since base's been
constructed. But this really is what our code usually looks like.

I think by the time we reach Derived's constructor, it is just the
start to construct a Derived object, which is to say "we are going to
construct a Derived object now, and we should start from the
construction of its base".

Regards,
-robin
 
K

keith

(e-mail address removed) wrote in @k35g2000prh.googlegroups.com:


Don't read too much into the syntax. This basically just tells the
compiler which constructor to use for the base class and provides the
opportunity to provide parameters to the base class' constructor. By the
time your Derived member initialization occurs and your constructor body is
executed, the base class will have been initialized.

Ahhh... My understanding was _nearly_ correct; By the time we reach
Derived's constructor's BODY, Base is guaranteed to have been fully
constructed. But while we are still in the initialisation list, we
can (and apparently should) explicitly call Base's constructor.

Have I got it right now?!
 
G

Guest

Hi, I've been through the FAQ-lite and can't see this mentioned, so
here goes...

I've got an abstract base class called Base which has no copy
constructor at all. In the derived class I have something like this:

// derived.h-----------------------------------------------------
class DerivedPrivate; // Not defined here
class Derived : public Base
{
private:
class DerivedPrivate* const p_;

public:
Derived();
Derived(const Derived& s);
// remainder snipped
}
// end--------------------------------------------------------------

// derived.cc----------------------------------------------------
// definition of DerivedPrivate skipped
Derived::Derived() : p_(new DerivedPrivate()) {}

Derived::Derived(const Derived& s) : p_(new DerivedPrivate())
{ *p_ = *(s.p_); }

Just a question, does not DerivedPrivate have a copy-constructor? You
should be able to use something like

Derived::Derived(const Derived& s)
: Base(),
p_(new DerivedPrivate(*(s.p))) // Instead of assignment in the body
{}
 
B

Bo Persson

(e-mail address removed) wrote:
:: Hi, I've been through the FAQ-lite and can't see this mentioned,
:: so here goes...
::
:: I've got an abstract base class called Base which has no copy
:: constructor at all. In the derived class I have something like
:: this:
::
:: // derived.h-----------------------------------------------------
:: class DerivedPrivate; // Not defined here
:: class Derived : public Base
:: {
:: private:
:: class DerivedPrivate* const p_;
::
:: public:
:: Derived();
:: Derived(const Derived& s);
:: // remainder snipped
:: }
:: //
:: end--------------------------------------------------------------
::
:: // derived.cc----------------------------------------------------
:: // definition of DerivedPrivate skipped
:: Derived::Derived() : p_(new DerivedPrivate()) {}
::
:: Derived::Derived(const Derived& s) : p_(new DerivedPrivate())
:: { *p_ = *(s.p_); }
:: // remainder skipped
:: //
:: end--------------------------------------------------------------
::
:: Now this all compiles and works just fine, but when I turn on
:: "-Wall - W" in gcc, it tells me that:
::
:: derived.cc:134: warning: base class 'class Base' should be
:: explicitly initialized in the copy constructor
::
:: I'm afraid I'm being rather dense today, as I don't understand what
:: it's complaining about. Can someone explain for me please?
::

It is just a warning that it is very unusual to have a copy
constructor that does not call the copy constructor of the base class.
You default construct the base and then copy the derived class.

More idiomatic would be:

Derived::Derived(const Derived& s) : Base(s), p_(new
DerivedPrivate(*(s.p_)))
{ }


Bo Persson
 
J

Joe Greer

(e-mail address removed) wrote in

Ahhh... My understanding was _nearly_ correct; By the time we reach
Derived's constructor's BODY, Base is guaranteed to have been fully
constructed. But while we are still in the initialisation list, we
can (and apparently should) explicitly call Base's constructor.

Have I got it right now?!

You can picture it like that as long as you realize that the base class
(es) constructors will get called in the order they are declared in the
class and not in any arbitrary order, then the member initialization(s)
for your derived class will occur, and then the body of the constructor
will be invoked. Remember that you can have multiple inheritance, so
you may have more than one base class constructor to specify.

So, for example:

#include <iostream>
#include <string>

class Base {
int m_i;
public:
Base(int i) : m_i(i) {std::cout << "Base" << std::endl;}
};


class Base1 {
int m_i;
public:
Base1(int i) : m_i(i) {std::cout << "Base1" << std::endl;}
};

class Item1 {
public:
Item1() { std::cout << "Item1" << std::endl;}
};

class Item2 {
public:
Item2() { std::cout << "Item2" << std::endl;}
};

class Derived : public Base1, public Base {
Item1 m_Item1;
Item2 m_Item2;
public:
Derived(int i) : Base(i), Base1(i), m_Item2(), m_Item1() {std::cout
<< "Derived" << std::endl;}
};


int main()
{
Derived d(5);
}

The output of this program is:

Base1
Base
Item1
Item2
Derived

As you can see, the bases are initialized in the order declared in the
class definition, not in the order they appear in the constructor list
and the items are initialized in the order they appear in the class, not
in the order of the list, but all of the items in the list are done
before the body of the constructor is executed.



Hope that helps,
joe
 
P

Pete Becker

Ahhh... My understanding was _nearly_ correct; By the time we reach
Derived's constructor's BODY, Base is guaranteed to have been fully
constructed. But while we are still in the initialisation list, we
can (and apparently should) explicitly call Base's constructor.

Have I got it right now?!

Yes, except that you don't need to explicitly mention Base's
constructor if you want the default constructor. That's the one that
the compiler will use if you don't mention Base's constructor at all.
Busybody compilers might give you a warning for doing that, but the
behavior is well defined.
 
J

James Kanze

Hi, I've been through the FAQ-lite and can't see this mentioned, so
here goes...
I've got an abstract base class called Base which has no copy
constructor at all.

No you don't. Every class has a copy constructor declared.
Always. The most you can do is to make it private, and not
provide an implementation.
In the derived class I have something like this:
// derived.h-----------------------------------------------------
class DerivedPrivate; // Not defined here
class Derived : public Base
{
private:
class DerivedPrivate* const p_;
public:
Derived();
Derived(const Derived& s);
// remainder snipped}
// end--------------------------------------------------------------

// derived.cc----------------------------------------------------
// definition of DerivedPrivate skipped
Derived::Derived() : p_(new DerivedPrivate()) {}
Derived::Derived(const Derived& s) : p_(new DerivedPrivate())
{ *p_ = *(s.p_); }
// remainder skipped
// end--------------------------------------------------------------
Now this all compiles and works just fine, but when I turn on "-Wall -
W" in gcc, it tells me that:
derived.cc:134: warning: base class 'class Base' should be explicitly
initialized in the copy constructor
I'm afraid I'm being rather dense today, as I don't understand what
it's complaining about. Can someone explain for me please?

The only explination I can see is that the author of this
warning message doesn't understand C++. You're Derived copy
constructor calls (and should call) the default constructor of
the Base class, which is probably what it should do (if e.g. the
Base class has no data members). At the very least, the warning
is only relevant if the Base class has data members.
 

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments. After that, you can post your question and our members will help you out.

Ask a Question

Members online

Forum statistics

Threads
473,755
Messages
2,569,536
Members
45,020
Latest member
GenesisGai

Latest Threads

Top