Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > comp.lang.c > #151249 > unrolled thread

direct-declarator or direct-abstract-declarator?

Back to article view | Back to comp.lang.c

Contents

  direct-declarator or direct-abstract-declarator? Johann Klammer <klammerj@NOSPAM.a1.net> - 2020-04-09 13:58 +0200
    Re: direct-declarator or direct-abstract-declarator? Tim Rentsch <tr.17687@z991.linuxsc.com> - 2020-04-09 06:52 -0700
    Re: direct-declarator or direct-abstract-declarator? James Kuyper <jameskuyper@alumni.caltech.edu> - 2020-04-09 11:23 -0400
      Re: direct-declarator or direct-abstract-declarator? Johann Klammer <klammerj@NOSPAM.a1.net> - 2020-04-09 19:20 +0200

#151249 — direct-declarator or direct-abstract-declarator?

the n2176 draft
has on page 100 the following exmples:

> typedef signed int t;

..and later..

> t f(t (t));
> long t;

About the first line they say: a function f is declared with type “function returning signed int with one unnamed parameter with type pointer
to function returning signed int with one unnamed parameter with type signed int ”

so this segments into:
declaration
declaration-specifiers init-declarator-list[opt]
type-specifier declaration-specifiers[opt] init-declarator-list[opt]
typedef-name init-declarator
typedef-name declarator
typedef-name direct-declarator
typedef-name identifier (parameter-type-list)
typedef-name identifier (parameter-list)
typedef-name identifier (parameter-declaration)                *1
typedef-name identifier (declaration-specifiers declarator)  
typedef-name identifier (typedef-name direct-declarator)
typedef-name identifier (typedef-name (declarator))
typedef-name identifier (typedef-name (identifier))            *2


or alternatively starting from *1
typedef-name identifier (declaration-specifiers abstract-declarator[opt])
typedef-name identifier (typedef-name pointer[opt] direct-abstract-declarator)
typedef-name identifier (typedef-name (abstract-declarator)(parameter-type-list)) *3


*2 is wrong as the last one is a declarator which won't ever resolve to a type-specifer/typedef-name.
*3 won't match because the there's no (*) before the parameter list.
It might be made to match if the direct abstract declarator were allowed to consist of solely 
a parameter-type list.

So how do I get to the typedef name 't' there towards the end with the grammar specified?

[toc] | [next] | [standalone]

#151257

Johann Klammer <klammerj@NOSPAM.a1.net> writes:

> [considering this code, excerpted from n2176, page 100:]
>
>    typedef signed int t;
>    t f(t (t));
>
> [About the declaration of f it is said:]

> a function f is declared with type
> "function returning signed int
>     with one unnamed parameter
>     with type pointer to function returning signed int
>         with one unnamed parameter with type signed int"
>
> [...]
>
> So how do I get to the typedef name 't' there towards the end
> with the grammar specified?


  t  f( t                                    ( t                      )  );
          ^                                 ^  ^                     ^ ^
          + direct-abstract-declarator{opt} +  + parameter-type-list + |
          |                                                            |
          |                                                            |
          +-------------  direct-abstract-declarator  -----------------+


Does this help clarify things?

[toc] | [prev] | [next] | [standalone]

#151264

On 4/9/20 7:58 AM, Johann Klammer wrote:
> the n2176 draft
> has on page 100 the following exmples:
> 
>> typedef signed int t;
> 
> ..and later..
> 
>> t f(t (t));
>> long t;
> 
> About the first line they say: a function f is declared with type �function returning signed int with one unnamed parameter with type pointer
> to function returning signed int with one unnamed parameter with type signed int �
> 
> so this segments into:
> declaration
> declaration-specifiers init-declarator-list[opt]
> type-specifier declaration-specifiers[opt] init-declarator-list[opt]
> typedef-name init-declarator
> typedef-name declarator
> typedef-name direct-declarator
> typedef-name identifier (parameter-type-list)
> typedef-name identifier (parameter-list)
> typedef-name identifier (parameter-declaration)                *1
...
> typedef-name identifier (declaration-specifiers abstract-declarator[opt])
> typedef-name identifier (typedef-name pointer[opt] direct-abstract-declarator)

The next steps are
typedef-name identifier (typedef-name direct-abstract-declarator[opt] (
parameter-typelist ) )
typedef-name identifier (typedef-name ( parameter-list ) )
typedef-name identifier (typedef-name ( parameter-declaration ) )
typedef-name identifier (typedef-name ( declaration-specifiers ) )
typedef-name identifier (typedef-name ( type-specifier ) )
typedef-name identifier (typedef-name ( typedef-name ) )

...
> It might be made to match if the direct abstract declarator were allowed to consist of solely 
> a parameter-type list.

A direct-abstract-declarator IS allowed to consist solely of a
parenthesized parameter type list. See the last grammar production under
6.7.7p1. The above declaration is equivalent to

    t f(t g(t h));


"A declaration of a parameter as ‘‘function returning type’’ shall be
adjusted to ‘‘pointer to function returning type’’, as in 6.3.2.1"
(6.7.6.3p8). Therefore, that declaration would be adjusted to

    t f(t (*g)(t h));

This is directly analogous to the more familiar "A declaration of a
parameter as ‘‘array of type’’ shall be adjusted to ‘‘qualified pointer
to type’’...". Thus, for example:

    func( int callback(double), char array[20]);

is adjusted to

    func( int (*callback)(double), char *array);

[toc] | [prev] | [next] | [standalone]

#151271

On 04/09/2020 05:23 PM, James Kuyper wrote:
> A direct-abstract-declarator IS allowed to consist solely of a
> parenthesized parameter type list. See the last grammar production under
> 6.7.7p1. The above declaration is equivalent to
> 
>     t f(t g(t h));

Ah, thank you. I missed an opt there.
My real problem seems to be that the other interpretation as
declarator is valid if I can not use a rule wot says identifier-not-typedef-name.

But due to one of the preceding examples(on p99):

typedef signed int t;
typedef int plain;
struct tag {
unsigned t:4;
const t:5;
plain r:5;
};

namely the "const t:5;"
where the t has to be a typedef-name
I can not do this. They use the same declarator
production in there.. tho maybe it's not the same exactly.
Will try to write separate ones for struct/union and otherwise.

[toc] | [prev] | [standalone]

Back to top | Article view | comp.lang.c

csiph-web