Re: [PATCH v2] Silence -Wpointer-to-int-cast warnings in newlib/libc/string/local.h.

Torbjorn SVENSSON <[email protected]> Thu, 23 Jul 2026 13:33:37 +0200
Newsgroups gmane.comp.lib.newlib
Message-ID <[email protected]>

On 2026-07-22 01:48, Jan Dubiec wrote:
> On 21.07.2026 23:35, Jeff Johnston wrote:
>> Hi Jan,
>>
>> The fallback for intptr_t is unnecessary.  According to libc/include/
>> sys/_intsup.h, gcc 3.2 or higher should support
>> the intptr info we need plus there is an #error statement if it isn't.
>> So, the code should just use intptr_t.
>>
>> -- Jeff J.
> 
> Thanks. Below is a simplified version of the patch.
> 
> /J.D.
> 
> 
> Signed-off-by: Jan Dubiec <[email protected]>
> ---
>   newlib/libc/string/local.h | 6 ++++--
>   1 file changed, 4 insertions(+), 2 deletions(-)
> 
> diff --git a/newlib/libc/string/local.h b/newlib/libc/string/local.h
> index 012a30d16..94c7c4a8f 100644
> --- a/newlib/libc/string/local.h
> +++ b/newlib/libc/string/local.h
> @@ -1,5 +1,6 @@
>   #include <_ansi.h>
>   #include <../ctype/local.h>
> +#include <stdint.h>
>   
>   /* internal function to compute width of wide char. */
>   int __wcwidth (wint_t);
> @@ -21,7 +22,7 @@ int __wcwidth (wint_t);
>    * This macro is used to skip a few bytes to find an aligned pointer.
>    * It's better to keep it as is even if _HAVE_HW_MISALIGNED_ACCESS is enabled,
>    * to avoid small performance penalties (if they are not zero).  */
> -#define UNALIGNED_X(X) ((long)(X) & (sizeof (long) - 1))
> +#define UNALIGNED_X(X) ((long)(intptr_t)(X) & (sizeof (long) - 1))
>   
>   #ifdef _HAVE_HW_MISALIGNED_ACCESS
>   /* Hardware performs unaligned operations with little
> @@ -30,7 +31,8 @@ int __wcwidth (wint_t);
>   #else /* _HAVE_HW_MISALIGNED_ACCESS */
>   /* Nonzero if either X or Y is not aligned on a "long" boundary.  */
>   #define UNALIGNED_X_Y(X, Y) \
> -  (((long)X & (sizeof (long) - 1)) | ((long)Y & (sizeof (long) - 1)))
> +  (((long)(intptr_t)X & (sizeof (long) - 1)) | \
> +   ((long)(intptr_t)Y & (sizeof (long) - 1)))

Not really what you change, but ideally, shouldn't this be:

+  (((long)(intptr_t)(X) & (sizeof (long) - 1)) | \
+   ((long)(intptr_t)(Y) & (sizeof (long) - 1)))

i.e. putting parenthesis around the macro arguments?


Kind regards,
Torbjörn