On Tue, 12 Dec 2006 at 12:02 -0800, Linus Torvalds wrote:
> On Tue, 12 Dec 2006, Linus Torvalds wrote:
> >
> > Did you use "-m64" if you are doing this on an architecture with 64-bit
> > /usr/include?
>
> Never mind, that's not it.
>
> The problem is that the standard headers use _this_ for "int64_t":
>
> typedef int int64_t __attribute__ ((__mode__ (__DI__)));
>
> and sparse just says "ok, int64_t is an 'int'"
>
> The "__attribute__((__mode__()))" thing has always been a quick hack -
> sparse actually tries to parse it, but not very well.
>
> If you use
>
> typedef long long s64;
>
> sparse gets it right.
>
> In fact, sparse even gets it right if you do
>
> typedef int __attribute__((__mode__(__DI__))) int64_t;
>
> because then the attribute gets attached to the underlying type, before it
> gets bound to the typedef.
Great! This works for me. So I will probably use
#ifdef __CHECKER__
typedef int __attribute__((__mode__(__DI__))) int64_t;
#endif
until it will be fixed in sparse.
> I'll take a look if I can fix typedef to accept the crapola __attibute__
> syntax.
>
> Putting the attributes at the end really _does_ suck. It's as if you were
> to write
>
> typedef void * const_ptr_t const;
>
> and that obviously isn't supposed to work. Why gcc thinks attributes can
> go at the end will never be clear to me.
>
> Gcc attributes suck. Some gcc extensions really were thought out really
> really badly.
Thanks for quick reply and explanations!
--
Best regards,
Yura
-
To unsubscribe from this list: send the line "unsubscribe linux-sparse" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
