Re: [PATCH net-next] tcp: tcp_sendmsg() wrong access to sk_route_caps

[Ben Hutchings <bhutchings@solarflare.com>]

On Mon, 2011-11-28 at 08:54 -0800, Joe Perches wrote:
> On Mon, 2011-11-28 at 17:04 +0100, Eric Dumazet wrote:
> > Le lundi 28 novembre 2011 à 02:46 -0800, Joe Perches a écrit :
> > > On Mon, 2011-11-28 at 11:27 +0100, Eric Dumazet wrote:
> > > > +	sg = !!(sk->sk_route_caps & NETIF_F_SG);
> > > As sg is now bool, using !! is unnecessary.
> > > A commit was done recently to remove one.
> > > 3ad9b358e03fd9dbf6705721490c811b666b0fe2
> > Hmm... I find it dangerous and error prone. Obviously not at the time we
> > commit such changes, but later, because a future reader might be fooled.
> > Using !!(expr) is pretty clear about the potential problem, and
> > generates no extra code if a bool is used for the target.
> 
> Though I don't think
> 	bool = expr;
> is particularly error prone,
> spatch shows 58 uses of
> 	bool = !!expr;
> in net-next drivers/net.
> 
> It might be useful to standardize on either the
> implicit or explicit !! style.
> 
> I think !! is an intentional and sensible style
> and should be the preferred use.

I see it as a workaround for the historical lack of a real bool type in
C.

> Perhaps a checkpatch warning should be issued
> when the implicit cast is used.

checkpatch can't do type-checking; maybe you mean sparse.

Ben.

-- 
Ben Hutchings, Staff Engineer, Solarflare
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.