On GCC 4.6.1, when I declare an instance of my own type that has a defaulted constructor, and if I instantiate an object of that type and initialize it with braces ( like Foo my_foo{}; ), the POD members in that class will only zero-initialize if there is no other constructor declared. If there is no other constructor except the defaulted one, they will zero-init like expected.
But, on GCC 4.7.3, the zero-initialize happens either way, which is the behavior I expected.
What's the difference here? Is this a compiler bug? Both of these GCC versions support defaulted constructors of the C++11 standard.
There's no real need to stick with old GCC versions, but I'd like to understand what's going on here.
note: I'm defaulting the main ctor, op=. and copy ctor simply to keep the type usable with variadic functions (clang demands this to classify the class as POD, although gcc let's me get away with using the type with variadic functions even with user-defined main ctor. bonus points if you can tell me why.)
Here is an example program to illustrate, including some output at the bottom (from binaries compiled with both GCC versions):
#include <cstdio>
// pod and pod_wctor are identical except that pod_wctor defines another ctor
struct pod {
pod( void ) = default;
pod( const pod& other ) = default;
pod& operator=( const pod& other ) = default;
int x,y,z;
};
struct pod_wctor {
pod_wctor( void ) = default;
pod_wctor( const int setx, const int sety, const int setz ) : x(setx), y(sety), z(setz) { }
pod_wctor( const pod_wctor& other ) = default;
pod_wctor& operator=( const pod_wctor& other ) = default;
int x,y,z;
};
int main ( void ) {
printf("the following shuold be uninitialized:\n");
pod pee;
printf( " %i,%i,%i\n", pee.x, pee.y, pee.z);
pod_wctor podtor;
printf( " %i,%i,%i\n", podtor.x, podtor.y, podtor.z);
printf("the following shuold be initialized to 0,0,0:\n");
pod peenit{};
printf( " %i,%i,%i\n", peenit.x, peenit.y, peenit.z );
pod_wctor podtornit{};
printf( " %i,%i,%i\n", podtornit.x, podtornit.y, podtornit.z );
return 0;
}
// compiled with: g++ m.cpp -std=gnu++0x
// g++ (Ubuntu/Linaro 4.6.1-9ubuntu3) 4.6.1 (i386)
/****************** output *******************
the following shuold be uninitialized:
10381592,134513249,134520820
134513969,134513504,0
the following shuold be initialized to 0,0,0:
0,0,0
7367877,134513945,8724468
*********************************************/
// compiled with: g++ m.cpp -std=gnu++0x
// gcc version 4.7.3 (Ubuntu/Linaro 4.7.3-2ubuntu4) (i386)
/****************** output *******************
the following shuold be uninitialized:
-1218358300,-1217268232,134520832
134514450,1,-1079827548
the following shuold be initialized to 0,0,0:
0,0,0
0,0,0
*********************************************/
By adding the constructor pod_wctor( const int setx, const int sety, const int setz ) : x(setx), y(sety), z(setz) { } to your class, it loses its status as an aggregate: [dcl.init.aggregate]/1
An aggregate is an array or a class (Clause 9) with no user-provided constructors
It still is a POD, because a trivial class only needs to have no non-trivial default ctors: [class]/6
A trivial class is a class that has a default constructor (12.1), has no non-trivial default constructors,
and is trivially copyable.
The interesting point here is that for an aggregate, the list-initialization pod peenit{}; performs aggregate-initialization:
List-initialization of an object or reference of type T is defined as follows:
If T is an aggregate, aggregate initialization is performed (8.5.1). [...]
Otherwise, if the initializer list has no elements and T is a class type with a default constructor, the object is value-initialized.
(Note: this is the revised order. AFAIK, in the Standard itself, the order of those two points is reversed, which must be a defect, as every aggregate has a default ctor -- the implicitly declared&defined one.)
Aggregate-initialization leads to value-initialization of the int members: [dcl.init.aggr]/7
If there are fewer initializer-clauses in the list than there are members in the aggregate, then each member not explicitly initialized shall be initialized from an empty initializer list
and [dcl.init.list]/3 "Otherwise, if the initializer list has no elements, the object is value-initialized"
However, for the non-aggregate pod_wctor, the list-initialization pod_wctor podtornit{} directly performs value-initialization, which calls the default ctor. [class.ctor]/6 specifies:
The implicitly-defined default constructor performs the set of initializations of the class that would be performed by a user-written default constructor for that class with no ctor-initializer (12.6.2) and an empty compound-statement.
and in [class.base.init]/8, we find:
In a non-delegating constructor, if a given non-static data member or base class is not designated by a mem-initializer-id (including the case where there is no mem-initializer-list because the constructor has no ctor-initializer) and the entity is not a virtual base class of an abstract class (10.4), then
[...]
otherwise, the entity is default-initialized (8.5).
The default ctor itself does not guarantee zeroing of the members because it only does default-initialization of the members.
The difference between default- and value-initialization: [dcl.init]
[7] To default-initialize an object of type T means:
if T is a (possibly cv-qualified) class type, the default constructor for T is called [...]
[...]
otherwise, no initialization is performed.
[...]
[8] To value-initialize an object of type T means:
if T is a (possibly cv-qualified) class type with either no default constructor or a default constructor that is user-provided or deleted, then the object is default-initialized;
if T is a (possibly cv-qualified) non-union class type without a user-provided or deleted default constructor, then the object is zero-initialized and, if T has a non-trivial default constructor, default-initialized;
[...]
otherwise, the object is zero-initialized.
(I admit this confused me, and I had to revise my answer.)
pod_wctor has a default-constructor that is not user-provided. Therefore, for the list-initialization pod_wctor podtornit{}, the second bullet of value-initialization applies. The object podtornit itself is zero-initialized, which leads to a zero-initialization of its members. Only then will it be default-initialized, and the default ctor will be called. The latter does nothing, but the former guarantees the members will be zeroed.
Related
Per cppreference, the syntax for value initialization is:
[..]
T object {}; (since C++11)
[..]
It's already known that value-initialization is performed when an object is constructed with an empty initializer.
Per, [dcl.init]/8 (emphasis mine)
To value-initialize an object of type T means:
(8.1) if T is a (possibly cv-qualified) class type ([class]), then
(8.1.1) if T has either no default constructor ([class.default.ctor]) or a default constructor that is user-provided
or deleted, then the object is default-initialized;
(8.1.2) otherwise, the object is zero-initialized and the semantic constraints for default-initialization are checked, and if T
has a non-trivial default constructor, the object is
default-initialized;
(8.2) [..]
(8.3) [..]
I interpret the term "no default constructor" as that there's no default constructor declared in the class. For example,
class S
{
long double d;
friend void f(const S&);
};
void f(const S& s) { std::cout << s.d; }
int main()
{
S s{ };
f(s); // 0
}
Since the class has "no default constructor", I'm expecting that the object s is default-initialized and the member s.d has indeterminate value. Why that's not the case?
I also have a confusion in understanding the point (8.1.1). How I can write this line of code T object {} without having a default constructor or with having a deleted default constructor? Notice the bold part, It's said that, "if T has either no default constructor or a default constructor that is deleted .."
Are there situations where objects of class types are value-initialized with deleted default constructor or without default constructor at all? Am I misreading 8.1.1?
Your class S does have a default constructor; it is just defined implicitly.
Per [class.default.ctor]/1:
A default constructor for a class X is a constructor of class X for which each parameter that is not a function parameter pack has a default argument (including the case of a constructor with no parameters). If there is no user-declared constructor for class X, a non-explicit constructor having no parameters is implicitly declared as defaulted ([dcl.fct.def]). An implicitly-declared default constructor is an inline public member of its class.
Therefore S{} will zero-initialize the object per [dcl.init]/8.1.2.
When [dcl.init]/8.1.1 refers to classes with no default constructor it means classes where the implicit default constructor doesn't get generated because user-defined constructors exist. That is, (8.1.1) would apply to the following classes:
struct NoDefaultCtor
{
NoDefaultCtor(int) {}
};
struct UserProvidedDefaultCtor
{
UserProvidedDefaultCtor() {}
};
struct DeletedDefaultCtor
{
DeletedDefaultCtor() = delete;
};
For all three of those classes, value-initialization will perform default-initialization. In the case of NoDefaultCtor and DeletedDefaultCtor that default-initialization will fail, of course, but it's important that (8.1.1) catches those types so they don't fall through and get zero-initialized by (8.1.2).
Suppose we have this code:
class A {
public:
A() = default;
A(const A&) = delete;
~A() = default;
};
class B {
public:
B() : a{} { }
A a[1];
};
int main()
{
B b;
}
This code compiles on latest GCC 9.2, Clang 9.2, and MSVC 19.22.
But when I change A default destructor to ~A() { } GCC returns error use of deleted function 'A::A(const A&)'. Clang and MSVC still compile.
When I write the copy constructor of A, GCC compiles, but at runtime this constructor was never called. What does GCC need the copy constructor for?
Is it GCC bug? (I've tried about all GCC versions on GodBolt.org, same error.)
This is a GCC bug.
The default constructor of B uses aggregate initialization to initialize a with no initializers. [dcl.init.aggr]/8:
If there are fewer initializer-clauses in the list than there are
elements in a non-union aggregate, then each element not explicitly
initialized is initialized as follows:
If the element has a default member initializer ([class.mem]), the element is initialized from that initializer.
Otherwise, if the element is not a reference, the element is copy-initialized from an empty initializer list ([dcl.init.list]).
Otherwise, the program is ill-formed.
[...]
So a[0] is copy-initialized from {}, which is copy-list-initialization. This is probably where GCC starts to get confused — copy-initialization does not necessarily involve the copy constructor.
[dcl.init.list]/3.4:
List-initialization of an object or reference of type T is defined as
follows:
[...]
Otherwise, if the initializer list has no elements and T is a class type with a default constructor, the object is
value-initialized.
[...]
Therefore, the default constructor of A is used directly. There's no copy constructor involved. Nor triviality.
In case you are worrying about the difference between C++11 and C++17, N3337 [dcl.init.aggr]/7:
If there are fewer initializer-clauses in the list than there are
members in the aggregate, then each member not explicitly initialized
shall be initialized from an empty initializer list ([dcl.init.list]).
[...]
Copy-initialization isn't even involved here. And N3337 [dcl.init.list]/3.1:
List-initialization of an object or reference of type T is defined
as follows:
If the initializer list has no elements and T is a class type with a default constructor, the object is value-initialized.
[...]
No change.
I don't understand how C++11 brace initialization rules work here.
Having this code:
struct Position_pod {
int x,y,z;
};
class Position {
public:
Position(int x=0, int y=0, int z=0):x(x),y(y),z(z){}
int x,y,z;
};
struct text_descriptor {
int id;
Position_pod pos;
const int &constNum;
};
struct text_descriptor td[3] = {
{0, {465,223}, 123},
{1, {465,262}, 123},
};
int main()
{
return 0;
}
Note, that the array is declared to have 3 elements, but only 2 initializers are provided.
However it compiles without error, which sounds strange, as the last array element's reference member will be uninitialized. Indeed, it has NULL value:
(gdb) p td[2].constNum
$2 = (const int &) #0x0: <error reading variable>
And now the "magic": I changed Position_pod to Position
struct text_descriptor {
int id;
Position_pod pos;
const int &constNum;
};
becomes this:
struct text_descriptor {
int id;
Position pos;
const int &constNum;
};
and now it gives the expected error:
error: uninitialized const member ‘text_descriptor::constNum'
My question: Why it compiles in the first case, when it should give an error (as in the second case).
The difference is, that Position_pod uses C - style brace initialization and Position uses C++11 - style initialization, which call Position's constructor. But how does this affect the possibility to leave a reference member uninitialized?
(Update)
Compiler:
gcc (Ubuntu 4.8.2-19ubuntu1) 4.8.2
It will be clear that
struct text_descriptor td[3] = {
{0, {465,223}, 123},
{1, {465,262}, 123},
};
is list-initialisation, and that the initialiser list is not empty.
C++11 says ([dcl.init.list]p3):
List-initialization of an object or reference of type T is defined as follows:
If the initializer list has no elements and T is a class type with a default constructor, the object is value-initialized.
Otherwise, if T is an aggregate, aggregate initialization is performed (8.5.1).
...
[dcl.init.aggr]p1:
An aggregate is an array or a class (Clause 9) with no user-provided constructors (12.1), no brace-or-equal-initializers for non-static data members (9.2), no private or protected non-static data members (Clause 11), no base classes (Clause 10), and no virtual functions (10.3).
td is an array, so it is an aggregate, so aggregate initialisation is performed.
[dcl.init.aggr]p7:
If there are fewer initializer-clauses in the list than there are members in the aggregate, then each member not explicitly initialized shall be initialized from an empty initializer list (8.5.4).
This is the case here, so td[2] is initialised from an empty initialiser list, which ([dcl.init.list]p3 again) means it is value-initialised.
Value-initialisation, in turn, means ([dcl.init]p7):
To value-initialize an object of type T means:
if T is a (possibly cv-qualified) class type (Clause 9) with a user-provided constructor (12.1), ...
if T is a (possibly cv-qualified) non-union class type without a user-provided constructor, then the object is zero-initialized and, if T's implicitly-declared default constructor is non-trivial, that constructor is
called.
...
Your class text_descriptor is a class with no user-provided constructor, so td[2] is first zero-initialised, and then its constructor is called.
Zero-initialisation means ([dcl.init]p5):
To zero-initialize an object or reference of type T means:
if T is a scalar type (3.9), ...
if T is a (possibly cv-qualified) non-union class type, each non-static data member and each base-class subobject is zero-initialized and padding is initialized to zero bits;
if T is a (possibly cv-qualified) union type, ...
if T is an array type, ...
if T is a reference type, no initialization is performed.
This is well-defined regardless of text_descriptor's default constructor: it just zero-initialises the non-reference members and sub-members.
Then the default constructor is called, if it is non-trivial. Here's how the default constructor is defined ([special]p5):
A default constructor for a class X is a constructor of class X that can be called without an argument. If there is no user-declared constructor for class X, a constructor having no parameters is implicitly declared
as defaulted (8.4). An implicitly-declared default constructor is an inline public member of its class. A defaulted default constructor for class X is defined as deleted if:
...
any non-static data member with no brace-or-equal-initializer is of reference type,
...
A default constructor is trivial if it is not user-provided and if:
its class has no virtual functions (10.3) and no virtual base classes (10.1), and
no non-static data member of its class has a brace-or-equal-initializer, and
all the direct base classes of its class have trivial default constructors, and
for all the non-static data members of its class that are of class type (or array thereof), each such class has a trivial default constructor.
Otherwise, the default constructor is non-trivial.
So, the implicitly defined constructor is deleted, as expected, but it is also trivial, if pos is a POD type (!). Because the constructor is trivial, it is not called. Because the constructor is not called, the fact that it is deleted is not a problem.
This is a gaping hole in C++11, which has since been fixed. It happened to have been fixed to deal with inaccessible trivial default constructors, but the fixed wording also covers deleted trivial default constructors. N4140 (roughly C++14) says in [dcl.init.aggr]p7 (emphasis mine):
if T is a (possibly cv-qualified) class type without a user-provided or deleted default constructor, then the object is zero-initialized and the semantic constraints for default-initialization are checked, and if
T has a non-trivial default constructor, the object is default-initialized;
As T.C. pointed out in the comments, another DR also changed so that td[2] is still initialised from an empty initialiser list, but that empty initialiser list now implies aggregate initialisation. That, in turn, implies each of td[2]'s members is initialised from an empty initialiser list as well ([dcl.init.aggr]p7 again), so would seem to initialise the reference member from {}.
[dcl.init.aggr]p9 then says (as remyabel had pointed out in a now-deleted answer):
If an incomplete or empty initializer-list leaves a member of reference type uninitialized, the program is ill-formed.
It is unclear to me that this applies to references initialised from the implicit {}, but compilers do interpret it as such, and there's not much else that could be meant by it.
The first version (the one with _pod suffix) still works but gives no error because for the z value, the default value of int is chosen (the 0). Idem for the const int reference.
In the second version, you cannot define a const reference without giving it a value. The compiler gives you an error because later you cannot assign any value to it.
In addition, the compiler you're using plays an important role here, maybe it's a bug, just because you are declaring a class member before an int member.
If I try to compile this code
struct A {
const int j;
};
A a;
I'll get an expected error:
error: uninitialized const member in ‘struct A’
but, if I try to compile this one:
struct A {
const int j;
};
A * a = new A();
I'll get a successful build.
The question is: why does new allocation allows creating a variable with const member without explicit constructor and stack allocation - doesn't ?
It's not because of the heap allocation, but because of the parenthesis you use when allocating. If you do e.g.
A* a = new A;
it would also fail.
The reason it works when you add the parenthesis is because then your structure is value initialized, and for a POD-type like A value-initialization value-initializes each member, and the default value-initialization for int is zero.
That means it would probably work creating the variable on the stack as well, if you just add the value-initialization parenthesis:
A a = A(); // watch out for the http://en.wikipedia.org/wiki/Most_vexing_parse
Though that brings other potential problems, better use uniform initialization if you can (requies C++11):
A a{};
This is ill-formed as of C++14. §12.1 [class.ctor] says that
4 A defaulted default constructor for class X is defined as deleted
if:
[...]
any non-variant non-static data member of const-qualified type (or array thereof) with no brace-or-equal-initializer does not have a
user-provided default constructor,
[...]
A default constructor is trivial if it is not user-provided and if:
its class has no virtual functions (10.3) and no virtual base classes (10.1), and
no non-static data member of its class has a brace-or-equal-initializer, and
all the direct base classes of its class have trivial default constructors, and
for all the non-static data members of its class that are of class type (or array thereof), each such class has a trivial default
constructor.
§8.5 [dcl.init] says in turn that
7 To default-initialize an object of type T means:
if T is a (possibly cv-qualified) class type (Clause 9), the default constructor (12.1) for T is called (and the initialization is
ill-formed if T has no default constructor or overload resolution
(13.3) results in an ambiguity or in a function that is deleted or
inaccessible from the context of the initialization);
[...]
8 To value-initialize an object of type T means:
if T is a (possibly cv-qualified) class type (Clause 9) with either no default constructor (12.1) or a default constructor that is
user-provided or deleted, then the object is default-initialized;
[...]
The empty pair of parentheses results in value-initialization. The default constructor for A is defined as deleted per [class.ctor]/p4. Thus, by [dcl.init]/p8, value-initialization means default-initialization, and by p7 the initialization is ill-formed because the constructor is deleted.
The C++11 version actually allowed value-initialization in this context; it says for value-initialization that, among other things,
if T is a (possibly cv-qualified) non-union class type without a
user-provided constructor, then the object is zero-initialized and, if
T’s implicitly-declared default constructor is non-trivial, that
constructor is called.
Since by the definition above the default constructor is trivial (even though it is deleted), it is not actually called. This was considered a defect in the standard and the relevant wording was changed by CWG issue 1301.
Compiler vendors usually do implement defect resolutions, so this could be considered a bug in GCC 4.8 that's been fixed in 4.9.
In the following example:
#include <iostream>
struct A {
int z;
A(std::string) {}
A() = default;
};
int main() {
char buf[1000];
std::fill(buf, buf + 1000, 'x');
auto a = new(buf) A{};
std::cerr << "A.z: " << a->z << std::endl;
}
Compiled with GCC 4.8 outputs zero (same behavior with Clang 3.4). This seems to indicate that a is being zero-initialized before the default constructor is called.
But according to value-initialization rules on cppreference.com, the object should not be initialized before the default constructor invocation. Class A qualifies for bullet point #1 under C++11:
1) If T is a class type with at least one user-provided constructor of any kind, the default constructor is called.
Another perhaps useful data point is that if we replace A() = default with A() {} in the above example, no zero-initialization happens as expected. This seems to indicate that bullet point #2 of value initialization is being followed in the initial example which would be wrong:
2) If T is an non-union class type without any user-provided constructors, then the object is zero-initialized and then the implicitly-declared default constructor is called (unless it's trivial)
The wording that you cite for value initialization in C++11 was considered defective, see Core Working Group DR 1368 and the resolution in Core Working Group DR 1301 that changed the wording (showing additions with bold and deletions with strike through):
To value-initialize an object of type T means:
if T is a (possibly cv-qualified) class type (Clause 9 [class]) with either no default constructor (12.1 [class.ctor]) or a default constructor that is user-provided or deleted constructor (12.1 [class.ctor]), then the object is default-initialized default constructor for T is called (and the initialization is ill-formed if T has no accessible default constructor);
if T is a (possibly cv-qualified) non-union class type without a user-provided or deleted default constructor, then the object is zero-initialized and, if T's implicitly-declared default constructor is T has a non-trivial default constructor, that constructor is called. default-initialized;
Some compilers - I believe GCC 4.8 and clang since 3.3ish - have already implemented the resolution of DR 1301.