Static types

[cpp_weenie]

In the class below, what does the keyword "static" do when applied to the
struct? It seems it does *not* make any member variables declared to be of
type "bar" static...

class foo
{
public:
static struct bar
{
int data_1;
int data_2;
};

bar data;
};

 

[cpp_weenie]

In the class below, what does the keyword "static" do when applied to the

struct? It seems it does *not* make any member variables declared to be of
type "bar" static...

class foo
{
public:
static struct bar
{
int data_1;
int data_2;
};

bar data;
};

Let me update my previous post. I think this is a compiler problem (VC++
6.0) - it should not be let through. It doesn't compile under g++ and I
can't find any reference to this type of construct in the Standard.

If it is indeed a compiler problem, sorry to have asked a non-question.
Hope nobody burned too much time on this one...

 

[Ron Natalie]

In the class below, what does the keyword "static" do when applied to the
struct? It seems it does *not* make any member variables declared to be of
type "bar" static...

class foo
{
public:
static struct bar
{
int data_1;
int data_2;
};

bar data;
};

In this case, absolutey nothing.
If it said:
static struct bar {
int data_1;
int data_2;
} data;

the static applies to data (making it a static member).

 

[Rob Williscroft]

cpp_weenie wrote in

Let me update my previous post. I think this is a compiler problem
(VC++ 6.0) - it should not be let through. It doesn't compile under
g++ and I can't find any reference to this type of construct in the
Standard.

I don't know which is right g++ tells you it doesn't make any sense
and msvc 6.0 just ignores the static. It may well be that this is
implementation defined behaviour, so both could be conforming.

If it is indeed a compiler problem, sorry to have asked a
non-question. Hope nobody burned too much time on this one...

Perhapse this will help it all make sense note the 'variable'
identifiyer:

#include <iostream>

class foo
{
public:
static struct bar
{
int data_1;
int data_2;
} variable;

bar data;
};

foo::bar foo::variable = { 10, 12 };

int main()
{
std::cerr << foo::variable.data_1 << "\n";
}

Rob.

 

[Ron Natalie]

I don't know which is right g++ tells you it doesn't make any sense
and msvc 6.0 just ignores the static. It may well be that this is
implementation defined behaviour, so both could be conforming.

It's not conforming. If you place a storage class specifier (such
as static), then you have to go ahead and declare a variable.
Standard section 7.1.1. Says so in the third sentence of that section.

 

[cpp_weenie]

It's not conforming. If you place a storage class specifier (such

as static), then you have to go ahead and declare a variable.
Standard section 7.1.1. Says so in the third sentence of that section.

As an additional note, I just tried it under VC++ 7.1 as well. I found that
it let this non-conforming construct through without so much as a warning!

Bill did pretty good this time around, but I have found several little
things like this...

 

[Josephine Schafer]

As an additional note, I just tried it under VC++ 7.1 as well. I found that
it let this non-conforming construct through without so much as a warning!

Yep..VC++ 7 doesn't crib even with /Za (disable language extensions).
Many times compilers differ in their behavior on the same code.
IMO, one should compile code (standard C++) on atleast two compilers.
Referring to the standard definitely tells which compiler is correct but may
times the interpretation is left to the reader.
In such cases I personally trust Comeau's implementation. Nice to see a compiler
which can claim to be almost standard compliant.
( I read somewhere that Comeau has some problems with Koenig lookup)

HTH,
J.Schafer