GCC 10 introduced "access" attribute[1] to check against array
out-of-bounds.

I'd like to use it in my code, but I don't know how to write the code to
avoind warnings if the compiler doesn't support this attribute (GGC <10
or not GCC compiler).

Is a macro ATTRIBUTE_ACCESS defined in a compiler.h file a good solution?

#if GCC_GREATER_10
# define ATTRIBUTE_ACCESS(...) __attribute__((access(__VA_ARGS__)))
#else
# define ATTRIBUTE_ACCESS(...)
#endif

[1] https://gcc.gnu.org/onlinedocs/gcc/Common-Function-Attributes.html

pozz ha scritto:
> GCC 10 introduced "access" attribute[1] to check against array
> out-of-bounds.
>
> I'd like to use it in my code, but I don't know how to write the code to
> avoind warnings if the compiler doesn't support this attribute (GGC <10
> or not GCC compiler).
>
> Is a macro ATTRIBUTE_ACCESS defined in a compiler.h file a good solution?
>
> #if GCC_GREATER_10
> #  define ATTRIBUTE_ACCESS(...)   __attribute__((access(__VA_ARGS__)))
> #else
> #  define ATTRIBUTE_ACCESS(...)
> #endif
>
> [1] https://gcc.gnu.org/onlinedocs/gcc/Common-Function-Attributes.html

$ gcc --version | grep gcc
gcc (GCC) 11.4.0
^ ^ ^

$ cat /dev/null | gcc -E -dM - | grep "__GNUC"
#define __GNUC__ 11 <-----
#define __GNUC_WIDE_EXECUTION_CHARSET_NAME "UTF-16LE"
#define __GNUC_EXECUTION_CHARSET_NAME "UTF-8"
#define __GNUC_PATCHLEVEL__ 0 <-----
#define __GNUC_STDC_INLINE__ 1
#define __GNUC_MINOR__ 4 <-----

$ cat foo.c
#ifdef __GNUC__
#define INF "is gcc\n"
#if(__GNUC__ >= 10)
#define VER "minimum 10.0\n"
#if(__GNUC_MINOR__ >= 1)
#define SVER "minimum 10.1\n"
#define ATTRIBUTE_ACCESS(...) __attribute__((access(__VA_ARGS__)))
#else
#define SVER "bad subversion\n"
#define ATTRIBUTE_ACCESS(...)
#endif
#else
#define VER "bad version\n"
#define ATTRIBUTE_ACCESS(...)
#endif
#else
#define INF "no gcc\n"
#define ATTRIBUTE_ACCESS(...)
#endif
#include <stdio.h>

ATTRIBUTE_ACCESS(read_only, 1)
void foo(int arr[], size_t size)
{ for(int i=0;i<size;i++) printf("%d\n", arr[i]);
}

int main()
{ int iArr[]={1,2,3,4,5,6};
printf(INF VER SVER);
foo(iArr, 6);
}

$ ./foo
is gcc
minimum 10.0
minimum 10.1
1 2
3 4
5 6

....or I didn't understand the question, then sorry.

Il 04/12/2023 18:15, jak ha scritto:
> pozz ha scritto:
>> GCC 10 introduced "access" attribute[1] to check against array
>> out-of-bounds.
>>
>> I'd like to use it in my code, but I don't know how to write the code
>> to avoind warnings if the compiler doesn't support this attribute (GGC
>> <10 or not GCC compiler).
>>
>> Is a macro ATTRIBUTE_ACCESS defined in a compiler.h file a good solution?
>>
>> #if GCC_GREATER_10
>> #  define ATTRIBUTE_ACCESS(...)   __attribute__((access(__VA_ARGS__)))
>> #else
>> #  define ATTRIBUTE_ACCESS(...)
>> #endif
>>
>> [1] https://gcc.gnu.org/onlinedocs/gcc/Common-Function-Attributes.html
>
> $ gcc --version | grep gcc
> gcc (GCC) 11.4.0
>            ^ ^ ^
>
> $ cat /dev/null | gcc -E -dM - | grep "__GNUC"
> #define __GNUC__ 11                                    <-----
> #define __GNUC_WIDE_EXECUTION_CHARSET_NAME "UTF-16LE"
> #define __GNUC_EXECUTION_CHARSET_NAME "UTF-8"
> #define __GNUC_PATCHLEVEL__ 0                          <-----
> #define __GNUC_STDC_INLINE__ 1
> #define __GNUC_MINOR__ 4                               <-----
>
> $ cat foo.c
> #ifdef __GNUC__
>     #define INF "is gcc\n"
>     #if(__GNUC__ >= 10)
>         #define VER "minimum 10.0\n"
>         #if(__GNUC_MINOR__ >= 1)
>             #define SVER "minimum 10.1\n"
> #define ATTRIBUTE_ACCESS(...)   __attribute__((access(__VA_ARGS__)))
>         #else
>             #define SVER "bad subversion\n"
> #define ATTRIBUTE_ACCESS(...)
>         #endif
>     #else
>         #define VER "bad version\n"
> #define ATTRIBUTE_ACCESS(...)
>     #endif
> #else
>     #define INF "no gcc\n"
> #define ATTRIBUTE_ACCESS(...)
> #endif
> #include <stdio.h>
>
> ATTRIBUTE_ACCESS(read_only, 1)
> void  foo(int arr[], size_t size)
> {
>   for(int i=0;i<size;i++) printf("%d\n", arr[i]);
> }
>
> int main()
> {
>     int iArr[]={1,2,3,4,5,6};
>     printf(INF VER SVER);
>     foo(iArr, 6);
> }
>
> $ ./foo
> is gcc
> minimum 10.0
> minimum 10.1
> 1
> 2
> 3
> 4
> 5
> 6
>
> ...or I didn't understand the question, then sorry.
>
>

Yes, thank you. I was wondering if there was a "standard" way to manage
this compiler dependent things.

On Tue, 5 Dec 2023 08:53:15 +0100
pozz <pozzugno@gmail.com> wrote:
> Yes, thank you. I was wondering if there was a "standard" way to manage
> this compiler dependent things.

How standard is "standard" ? If it's compiler dependent then it cannot be
completely standard. I guess you could hope for a common convention for C
compilers to describe their version number but I don't know of such a
convention and I'm not even sure that such could exist in principle because
not all software projects use the same version format and version formats and
numbers won't necessarily mean the same thing between different projects
anyway.

<ukl1ft$3ggjc$1@dont-email.me> gave you what you need but you can also
have a look at gcc.gnu.org/onlinedocs/cpp/Common-Predefined-Macros.html .

--
The director of photography was probably a camcorder taped onto a skateboard
and pushed forward until it hits a wall.
www.imdb.com/review/rw1008884

On 12/5/2023 4:53 AM, Spiros Bousbouras wrote:
> On Tue, 5 Dec 2023 08:53:15 +0100
> pozz <pozzugno@gmail.com> wrote:
>> Yes, thank you. I was wondering if there was a "standard" way to manage
>> this compiler dependent things.
>
> How standard is "standard" ?

Indeed! :^)

> If it's compiler dependent then it cannot be
> completely standard. I guess you could hope for a common convention for C
> compilers to describe their version number but I don't know of such a
> convention and I'm not even sure that such could exist in principle because
> not all software projects use the same version format and version formats and
> numbers won't necessarily mean the same thing between different projects
> anyway.
>
> <ukl1ft$3ggjc$1@dont-email.me> gave you what you need but you can also
> have a look at gcc.gnu.org/onlinedocs/cpp/Common-Predefined-Macros.html .
>

Il 05/12/2023 13:53, Spiros Bousbouras ha scritto:
> On Tue, 5 Dec 2023 08:53:15 +0100
> pozz <pozzugno@gmail.com> wrote:
>> Yes, thank you. I was wondering if there was a "standard" way to manage
>> this compiler dependent things.
>
> How standard is "standard" ?

Indeed I wrote standard in quotes...

If it's compiler dependent then it cannot be
> completely standard. I guess you could hope for a common convention for C
> compilers to describe their version number but I don't know of such a
> convention and I'm not even sure that such could exist in principle because
> not all software projects use the same version format and version formats and
> numbers won't necessarily mean the same thing between different projects
> anyway.
>
> <ukl1ft$3ggjc$1@dont-email.me> gave you what you need but you can also
> have a look at gcc.gnu.org/onlinedocs/cpp/Common-Predefined-Macros.html .
>

On Tue, 5 Dec 2023 08:53:15 +0100, pozz wrote:
> Il 04/12/2023 18:15, jak ha scritto:
>> pozz ha scritto:
>>> GCC 10 introduced "access" attribute[1] to check against array
>>> out-of-bounds.
>>>
>>> I'd like to use it in my code, but I don't know how to write the code
>>> to avoind warnings if the compiler doesn't support this attribute (GGC
>>> <10 or not GCC compiler).
>>>
>>> Is a macro ATTRIBUTE_ACCESS defined in a compiler.h file a good
>>> solution?
>>>
>>> #if GCC_GREATER_10 #  define ATTRIBUTE_ACCESS(...)  
>>> __attribute__((access(__VA_ARGS__)))
>>> #else #  define ATTRIBUTE_ACCESS(...)
>>> #endif
[...]
> Yes, thank you. I was wondering if there was a "standard" way to manage
> this compiler dependent things.

Instead of checking for specific compiler versions (such checks are never
particularly accurate[1]), gcc provides a built-in way[2] to directly query
in the preprocessor whether or not a particular attribute is supported.
This is designed specifically for scenarios like yours:

#ifdef __has_attribute
# if __has_attribute(access)
# define use_newfangled_gcc_access_attribute 1
# endif
#endif

/* ... */

#if use_newfangled_gcc_access_attribute
/* use the attribute */
#else
/* don't use the attribute */
#endif

[1] for example, clang defines the gcc version macros pretending to be gcc 4.2,
but it does not implement every single feature of gcc 4.2, and it also
includes many features that are only available in newer versions of gcc.

[2] https://gcc.gnu.org/onlinedocs/gcc-13.2.0/cpp/_005f_005fhas_005fattribute.html