Re: __STRICT_ANSI__ checks in headers

From: Kyle Moffett
Date: Fri Sep 15 2006 - 09:28:04 EST


On Sep 15, 2006, at 02:42:58, David Woodhouse wrote:
On Fri, 2006-09-15 at 09:01 +0300, Ismail Donmez wrote:
Kernel headers currently uses __STRICT_ANSI__ check before defining a long
long variable because ANSI-C doesn't allow long long variables. But this
seems to harsh because any project including linux/videodev2.h ( and similar
ones ) and using -ansi flag will not compile because some types like __s64
will not be defined.

One possible fix is to let videodev2.h use int64_t, and in userspace
they can include <stdint.h>

Another is just to declare videodev2.h incompatible with -ansi, or maybe
just omit 'value64' from the union if __STRICT_ANSI__ is defined, and
replace it with an array of two __s32s.

A mildly better alternative is (on 32-bit architectures, 64-bit archs have no problem) change the typedef from this:

#if defined(__GNUC__) && !defined(__STRICT_ANSI__)
typedef unsigned long long __u64;
typedef signed long long __s64;
#endif

to this:

#if defined(__GNUC__)
__extension__ typedef unsigned long long __u64;
__extension__ typedef signed long long __s64;
#endif

GCC always supports __extension__ to indicate not to warn or error on GCC-only extensions. You only have to declare __extension__ on the typedef, any uses are considered OK. I think this also works for code-expressions like this:

int x = __extension__ ({ foo(); 1; })

but I don't remember exactly. In certain really complex expressions GCC can get confused and give bogus errors, but as long as you don't start sprinkling it in macros you should be fine.

Cheers,
Kyle Moffett

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/