RE: [PATCH] incorrect use of sizeof() in ioctl definitions

[Tian, Kevin]

> Wrote from Maciej Zenczykowski [mailto:maze@xxxxxxx]
> 
> So as not to break userspace we must still support old values, at the
same
> time we want new programs to start using the new correct values -
hence
> the introduction of _backward compatibility_ values.

Thanks. :) Now I see... but are there any rules to decide which part
should be upgraded even breaking the backward compatibility? You know,
the latest 2.6 kernel will request many modules recompiled to run on it.
IMO, most ioctls defined in this bad manner seems to be not-widely used
ones, and... maybe it's worthy of some sacrifice on temporary
compatibility, thus to keep a clean and consistent environment. 

> the old was bad since it was sizeof(sizeof(...)) - it so happens that
by
> def sizeof(anything) is a size_t - thus replacing sizeof(sizeof(..))
with
> sizeof(size_t) changes nothing - just shortens the code...
> Of course what we probably should really have is the above (now) code
> defined as "BAD" and the previous (old) define without the sizeof as
the
> current (no BAD prefix).

Reasonable!

> Yap, both violate.  It is a mess and there is no easy fix due to the
need
> to retain the old invalid ioctl's as well... 

If without such need... :(

Thanks,
Kevin

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/