Re: [PATCH] drop lchmod

[Rich Felker <dalias@libc.org>]

On Wed, Jan 04, 2017 at 10:23:55PM +0100, Julien Ramseier wrote:
> lchmod is not POSIX and not supported on Linux.
> Keeping it would probably do more harm than good: software checking 
> for its presence (i.e. libarchive) will use it and always fail.
> 
> 
> diff --git a/include/sys/stat.h b/include/sys/stat.h
> index 9d09662..0058338 100644
> --- a/include/sys/stat.h
> +++ b/include/sys/stat.h
> @@ -92,7 +92,6 @@ int futimens(int, const struct timespec [2]);
>  int utimensat(int, const char *, const struct timespec [2], int);
> 
>  #if defined(_GNU_SOURCE) || defined(_BSD_SOURCE)
> -int lchmod(const char *, mode_t);
>  #define S_IREAD S_IRUSR
>  #define S_IWRITE S_IWUSR
>  #define S_IEXEC S_IXUSR
> diff --git a/src/stat/lchmod.c b/src/stat/lchmod.c
> deleted file mode 100644
> index f324ba7..0000000
> --- a/src/stat/lchmod.c
> +++ /dev/null
> @@ -1,8 +0,0 @@
> -#define _GNU_SOURCE
> -#include <sys/stat.h>
> -#include <fcntl.h>
> -
> -int lchmod(const char *path, mode_t mode)
> -{
> -   return fchmodat(AT_FDCWD, path, mode, AT_SYMLINK_NOFOLLOW);
> -}

This can't be done because it's ABI. It will cause any program linked
with a reference to lchmod to failto execute with missing symbol
errors.

libarchive needs to be fixed to respect the errno value. fchmodat with
AT_FDCWD and the AT_SYMLINK_NOFOLLOW flag is semantically equivalent
to lchmod and is required by POSIX, so software should be prepared to
deal with systems where the error EOPNOTSUPP occurs.

Rich