Re: ioctl-patch for Darwin and FreeBSD

[Eric Blake]

On 02/22/2011 12:34 PM, Markus Gothe wrote:
> Please see attached patch for ioctl() on FreeBSD and Darwin. Their definition 
> differ from (int, int, ...) and the gnulib variant doesn't work well on 
> 64-bit Darwin with this proto.

Can you please convince your mailer to send patches with MIME type
text/plain, rather than encoded as application/octet-stream?

> --- lib/ioctl.c.old   2011-02-22 20:21:11.000000000 +0100
> +++ lib/ioctl.c       2011-02-22 20:24:38.000000000 +0100
> @@ -28,7 +28,11 @@
>  /* Provide a wrapper with the POSIX prototype.  */
>  # undef ioctl
>  int
> +#if __FreeBSD__ || __Darwin__
> +rpl_ioctl (int fd, unsigned long request, ... /* {void *,char *} arg */)
> +#else
>  rpl_ioctl (int fd, int request, ... /* {void *,char *} arg */)
> +#endif
This part is wrong - the replacement should ALWAYS match the POSIX
signature, and the type munging take place within the replacement,
rather than declaring the replacement with the wrong type.

> +++ m4/ioctl.m4       2011-02-22 20:26:10.000000000 +0100
> @@ -24,7 +24,13 @@
>        [AC_COMPILE_IFELSE(
>           [AC_LANG_PROGRAM(
>              [[#include <sys/ioctl.h>]],
> -            [[extern int ioctl (int, int, ...);]])
> +            [[
> +             #if __FreeBSD__ || __Darwin__
> +             extern int ioctl (int, unsigned long, ...);
> +             #else
> +             extern int ioctl (int, int, ...);
> +             #endif
> +     ]])
>           ],
>           [gl_cv_func_ioctl_posix_signature=yes],
>           [gl_cv_func_ioctl_posix_signature=no])

This is wrong as well - the whole point of this test is to reject the
FreeBSD/Darwin ioctl signature as non-compliant, so that the rest of the
code will provide a correct signature wrapper in the form of rpl_ioctl.

What is the exact failure you are seeing, and on which project?

-- 
Eric Blake   address@hidden    +1-801-349-2682
Libvirt virtualization library http://libvirt.org

signature.asc
Description: OpenPGP digital signature