In the book "C A reference Manual" (5th edition), the authors state
that a declaration like the following is invalid:

static struct s { int a; int b; }   ;

I can't find in the standard which constraint or semantics is
violated.
Can you help?

Luca Forlizzi wrote:
> In the book "C A reference Manual" (5th edition), the authors state
> that a declaration like the following is invalid:
>
> static struct s { int a; int b; }   ;
>
> I can't find in the standard which constraint or semantics is
> violated.
> Can you help?

'static' should be applied to declarations with declarators.
You are just declaring a 'tag', not a variable of type struct,
so it has not sense to apply a storage class specifier

-- 
Vincenzo Mercuri

Vincenzo Mercuri wrote:


> so it has not sense [...]

sorry, i meant 'it makes no sense'



-- 
Vincenzo Mercuri

On 1 Ago, 17:59, Vincenzo Mercuri <c...@lang.c> wrote:
> Vincenzo Mercuri wrote:
> > so it has not sense [...]
>
> sorry, i meant 'it makes no sense'
>
> --
> Vincenzo Mercuri

I agree that it makes no sense, but how one concludes that it is
invalid?

Luca Forlizzi <luca.forlizzi@gmail.com> writes:

> In the book "C A reference Manual" (5th edition), the authors state
> that a declaration like the following is invalid:
>
> static struct s { int a; int b; }   ;
>
> I can't find in the standard which constraint or semantics is
> violated.
> Can you help?

My best understanding is that it is legal and well-defined.
Certainly it's legal and well-defined without the 'static'
storage-class specifier, and a grep on 'storage.class' didn't
turn up anything wrong that I could see.

Luca Forlizzi wrote:
> On 1 Ago, 17:59, Vincenzo Mercuri<c...@lang.c>  wrote:
>> Vincenzo Mercuri wrote:
>>> so it has not sense [...]
>>
>> sorry, i meant 'it makes no sense'
>>
>> --
>> Vincenzo Mercuri
>
> I agree that it makes no sense, but how one concludes that it is
> invalid?

Beware that 'C: A Reference Manual' says that it is
"nonsensical and (in Standard C) illegal" where Standard C
means C89 (see page 4, 5th edition).
Now I can't say if it is illegal in C89, (i dont have the
Standard) but I can just say it is nonsensical in C99

-- 
Vincenzo Mercuri

On 1 Ago, 18:11, Tim Rentsch <t...@alumni.caltech.edu> wrote:
> Luca Forlizzi <luca.forli...@gmail.com> writes:
> > In the book "C A reference Manual" (5th edition), the authors state
> > that a declaration like the following is invalid:
>
> > static struct s { int a; int b; } =A0 ;
>
> > I can't find in the standard which constraint or semantics is
> > violated.
> > Can you help?
>
> My best understanding is that it is legal and well-defined.
> Certainly it's legal and well-defined without the 'static'
> storage-class specifier, and a grep on 'storage.class' didn't
> turn up anything wrong that I could see.

that's also my understanding. I also think that the presence of
'static' has no effect on the semantics of the declaration (as
suggested by the book authors). Do you agree?

On 1 Ago, 18:23, Vincenzo Mercuri <c...@lang.c> wrote:
> Luca Forlizzi wrote:
> > On 1 Ago, 17:59, Vincenzo Mercuri<c...@lang.c> =A0wrote:
> >> Vincenzo Mercuri wrote:
> >>> so it has not sense [...]
>
> >> sorry, i meant 'it makes no sense'
>
> >> --
> >> Vincenzo Mercuri
>
> > I agree that it makes no sense, but how one concludes that it is
> > invalid?
>
> Beware that 'C: A Reference Manual' says that it is
> "nonsensical and (in Standard C) illegal" where Standard C
> means C89 (see page 4, 5th edition).
> Now I can't say if it is illegal in C89, (i dont have the
> Standard) but I can just say it is nonsensical in C99
>
> --
> Vincenzo Mercuri

Hi Vincenzo,

 I checked C89 and I don't see in it anything that renders such a
declaration invalid. I think it is valid and a well-defined
declaration.
And that it has a meaning, because it declares the new struct type and
its tag.
Because of that, I wouldn't call it nonsensical. The only
"nonsensical" part, is that it uses a storage-class specifier that
could be omitted without changing the declaration's meaning. There are
cases like this. For instance in

extern int f(void);

the storage-class specifier could be omitted without changing
semantics, as well (so I think).
Would you call it nonsensical, too ?

Luca Forlizzi wrote:
> On 1 Ago, 18:23, Vincenzo Mercuri<c...@lang.c>  wrote:
>> Luca Forlizzi wrote:
>>> On 1 Ago, 17:59, Vincenzo Mercuri<c...@lang.c>    wrote:
>>>> Vincenzo Mercuri wrote:
>>>>> so it has not sense [...]
>>
>>>> sorry, i meant 'it makes no sense'
>>
>>>> --
>>>> Vincenzo Mercuri
>>
>>> I agree that it makes no sense, but how one concludes that it is
>>> invalid?
>>
>> Beware that 'C: A Reference Manual' says that it is
>> "nonsensical and (in Standard C) illegal" where Standard C
>> means C89 (see page 4, 5th edition).
>> Now I can't say if it is illegal in C89, (i dont have the
>> Standard) but I can just say it is nonsensical in C99
>>
>> --
>> Vincenzo Mercuri
>
> Hi Vincenzo,
>
>   I checked C89 and I don't see in it anything that renders such a
> declaration invalid. I think it is valid and a well-defined
> declaration.
> And that it has a meaning, because it declares the new struct type and
> its tag.
> Because of that, I wouldn't call it nonsensical. The only
> "nonsensical" part, is that it uses a storage-class specifier that
> could be omitted without changing the declaration's meaning. There are
> cases like this. For instance in
>
> extern int f(void);
>
> the storage-class specifier could be omitted without changing
> semantics, as well (so I think).
> Would you call it nonsensical, too ?

Hi Luca,

	...I think I got your point. My answer is: no, I wouldn't call it
'nonsensical' and I'm going to tell you why. The 'extern' storage-class
specifier is unnecessary in function prototype declarations, but
unnecessary doesn't mean nonsensical. For some programmers it's
just good style. Something different happens with:

	static struct s { int a; int b; };
	
here the problem is that the storage-class specifier will be ignored
by the compiler (see H&S pag 89) -- gcc warns you: 'useless storage
class specifier in empty declaration'.
Again, as H&S says:
"you might think that a later declaration of the form":

	struct s x, y;
	
"will cause x and y to have the storage class static. It will not."


-- 
Vincenzo Mercuri

Luca Forlizzi <luca.forlizzi@gmail.com> writes:

> On 1 Ago, 18:11, Tim Rentsch <t...@alumni.caltech.edu> wrote:
>> Luca Forlizzi <luca.forli...@gmail.com> writes:
>> > In the book "C A reference Manual" (5th edition), the authors state
>> > that a declaration like the following is invalid:
>>
>> > static struct s { int a; int b; }   ;
>>
>> > I can't find in the standard which constraint or semantics is
>> > violated.
>> > Can you help?
>>
>> My best understanding is that it is legal and well-defined.
>> Certainly it's legal and well-defined without the 'static'
>> storage-class specifier, and a grep on 'storage.class' didn't
>> turn up anything wrong that I could see.
>
> that's also my understanding. I also think that the presence of
> 'static' has no effect on the semantics of the declaration (as
> suggested by the book authors). Do you agree?

Yes, reading the relevant sections (6.2.2 and 6.2.4, IIRC), makes
it clear that the result is not changed by adding 'static'.

On Aug 4, 2:27=A0pm, Tim Rentsch <t...@alumni.caltech.edu> wrote:
> Luca Forlizzi <luca.forli...@gmail.com> writes:
> > On 1 Ago, 18:11, Tim Rentsch <t...@alumni.caltech.edu> wrote:
> >> Luca Forlizzi <luca.forli...@gmail.com> writes:
> >> > In the book "C A reference Manual" (5th edition), the authors state
> >> > that a declaration like the following is invalid:
>
> >> > static struct s { int a; int b; } =A0 ;
>
> >> > I can't find in the standard which constraint or semantics is
> >> > violated.
> >> > Can you help?
>
> >> My best understanding is that it is legal and well-defined.
> >> Certainly it's legal and well-defined without the 'static'
> >> storage-class specifier, and a grep on 'storage.class' didn't
> >> turn up anything wrong that I could see.
>
> > that's also my understanding. I also think that the presence of
> > 'static' has no effect on the semantics of the declaration (as
> > suggested by the book authors). Do you agree?
>
> Yes, reading the relevant sections (6.2.2 and 6.2.4, IIRC), makes
> it clear that the result is not changed by adding 'static'.
Agreed.  Functions and objects have linkage.  Objects have storage
duration.  A 'struct' tag has neither of these, but "storage class
specifier" in the syntax is (ab)used to include 'typedef'.  One might
wish to use 'typedef' along with their 'struct' declaration.