Initializer list issue

[Christopher Benson-Manica]

I recently had an issue that I'll try to condense for your reviewing
pleasure:

class A
{
protected:
my_type Member;

public:
A() {/* initialize Member */}
};

class B : public A
{
private:
my_other_type MyMember;

public:
B() : A(), MyMember(Member) {}
};

This compiled, but crashed at run time. Was it because the order in
which the actions in the initializer list are taken is unspecified?

 

[John Harrison]

I recently had an issue that I'll try to condense for your reviewing
pleasure:

class A
{
protected:
my_type Member;

public:
A() {/* initialize Member */}
};

class B : public A
{
private:
my_other_type MyMember;

public:
B() : A(), MyMember(Member) {}
};

This compiled, but crashed at run time. Was it because the order in
which the actions in the initializer list are taken is unspecified?

No the order is defined but its a little different from what some people
expect.

Simplyfying a little; the order is that all bases are initialised first in
the order that they appear in the class declaration, not in the order that
they appear in the initialiser list. E.g.

class A : public B, public C
{
A() : C(), B() {} // B is initialised before C
};

Then all the members are initialised, again this is in the order that they
are in the class declaration, not in the initialiser list. E.g.

class A
{
A() : c(), b() {} // b is initialised before c
B b;
C c;
};

john

 

[Jeff Schwab]

I recently had an issue that I'll try to condense for your reviewing
pleasure:

class A
{
protected:
my_type Member;

public:
A() {/* initialize Member */}
};

class B : public A
{
private:
my_other_type MyMember;

public:
B() : A(), MyMember(Member) {}
};

This compiled, but crashed at run time. Was it because the order in
which the actions in the initializer list are taken is unspecified?

Nope. Sorry, no way to tell from this code what caused the crash.

 

[Christopher Benson-Manica]

Nope. Sorry, no way to tell from this code what caused the crash.

Well, I guess I'll tentatively blame my compiler (again), since I
sidestepped whatever problem there was by making the derived class
member a pointer and calling new in the constructor. I really didn't
see any reason why that should have made a difference, but it did...