[PATCH v2] Implement C23 const-preserving standard library macros

[Collin Funk]

Joseph Myers <josmyers@redhat.com> writes:

> Another question would be whether some GCC extension should be added
> to support this feature better with macros that only expand each
> argument once.
>
> Tested for x86_64.
>
[...]
> +#if __HAVE_GENERIC_SELECTION
> +/* If PTR is a pointer to const, return CALL cast to type CTYPE,
> +   otherwise return CALL.  Pointers to types with non-const qualifiers
> +   are not valid.  This should not be defined for C++, as macros are
> +   not an appropriate way of implementing such qualifier-generic
> +   operations for C++.  */
> +# define __glibc_const_generic(PTR, CTYPE, CALL)	\
> +  _Generic (0 ? (PTR) : (void *) 1,			\
> +	    const void *: (CTYPE) (CALL),		\
> +	    default: CALL)
> +#endif
[...]
>  extern char *strstr (const char *__haystack, const char *__needle)
>       __THROW __attribute_pure__ __nonnull ((1, 2));
> +# if __GLIBC_USE (ISOC23) && defined __glibc_const_generic && !defined _LIBC
> +#  define strstr(HAYSTACK, NEEDLE)			\
> +  __glibc_const_generic (HAYSTACK, const char *,	\
> +			 strstr (HAYSTACK, NEEDLE))

Wouldn't this break a ton of existing code since HAYSTACK is evaluated
multiple times [1]? There are likely examples for the others.

I haven't thought about it for long, but I don't see a way to implement
them without doing that.

Collin

[1] https://codesearch.debian.net/search?q=strchr%28%2B%2B&literal=1