>    i don't like this change.  i think we can keep it as a small hack
>    #ifdef COMPAT_16 in sysctl_diskstats()... certainly it should only
>    be visible for COMPAT_16.

I also don't like the change, that was example to supoort source
compatiblity (but it just leads binary incompatible code).

> something like this appears to work just fine.  anyone mind if i
> commit this?

As I wrote in previous mail, I'd like to return error (EINVAL?) ifndef
COMPAT_16.  Otherwise, the new (and broken IMHO) binary which doesn't
pass the size of struct doesn't work with the kernel compiled with
COMPAT_16.

If we return an error, such broken binary just doesn't work.

enami.