Re: [PATCH 2/2] Simplify (and warn about) right shifts that result in zero

[Linus Torvalds <torvalds@linux-foundation.org>]

On Thu, 24 Apr 2008, Pavel Roskin wrote:
> 
> So I would suggest a similar warning is this case.  Maybe "right shift
> too big (%u) for source type %s" (if the source type is readily
> available) or "right shift count (%d) >= width of type (%d)"

That's fine, except we shouldn't talk about "type", since we're really 
doing some really stupid value analysis (the *type* will generally have 
been cast to a bigger one by the implicit C type evaluation rules).

> By the way, your patch has caught something interesting in
> net/mac80211/tkip.c:
> 
>         iv32 = data[hdr_len + 4] +
>                 (data[hdr_len + 5] >> 8) +
>                 (data[hdr_len + 6] >> 16) +
>                 (data[hdr_len + 7] >> 24);
>
> Wow!

Heh. That does look like somebody is shifting the wrong way, and 
apparently the new warning was worth something ;)

		Linus