Hi Dirk,

On 2026-04-15T22:00:52+0200, Dirk Müller wrote:
> Hi Branden,
> 
> On Wed, Apr 15, 2026 at 8:24 PM G. Branden Robinson
> <g.branden.robinson@gmail.com> wrote:
> 
> > >       ssize_t readlink(size_t bufsiz;
> > >                        const char *restrict path,
> > >                        char buf[restrict bufsiz], size_t bufsiz);
> > >
> > > I believe this is a typo and it should be:
> > >
> > >       ssize_t readlink(const char *restrict path,
> > >                        char buf[restrict bufsiz], size_t bufsiz);
> > This is something of a FAQ on this mailing list.
> > https://lore.kernel.org/linux-man/adupQhfJQ7kws17U@debian/
> > It's a GCC syntax extension.
> >
> > https://gcc.gnu.org/onlinedocs/gcc/Variable-Length.html

Look at the very bottom of that page.

> 
> Oh, I understand that part.

I believe you didn't.

> let me try again. The Synpsis shows 4
> parameters to the readlink() call:
> 
>   ssize_t readlink(size_t, const char*, char buf[], size_t);

That's not what readlink(2) documents.  readlink(2) has a semicolon (;)
after the fisrt 'size_t bufsiz'.  That means it's a forward declaration
of a function parameter, and not a real parameter declaration.

> 
> I believe it should actually be 3 parameters instead:
> 
>   ssize_t readlink(const char*, char buf[], size_t);
> 
> with other words, the first argument is a pointer to the source link,
> not a repetition of bufsiz (which is at the end).
> 
> This page has the, what I consider correct synposis:
> 
> https://pubs.opengroup.org/onlinepubs/9699919799/functions/readlink.html
> 
> ssize_t readlink(const char *restrict path, char *restrict buf,
>        size_t bufsize);

This synopsis is compatible with the one from readlink(2).  See by
yourself:

	alx@devuan:~/tmp$ cat rl.c 
	#include <sys/types.h>

	ssize_t readlink(
	    const char *path, char *buf, size_t bufsiz);

	ssize_t readlink(size_t bufsiz;
	    const char *path, char buf[bufsiz], size_t bufsiz);
	alx@devuan:~/tmp$ gcc -Wall -Wextra -Wno-vla-parameter -S rl.c 
	alx@devuan:~/tmp$ 

> Again, I'm only commenting on the number of parameters, not the
> variadic length expression on parameter two (three in the original).

Both prototypes have 3 parameters.


Have a lovely night!
Alex

> 
> Thanks,
> Dirk
> 

-- 
<https://www.alejandro-colomar.es>