Re: [PATCH] b43: Fix sparse warnings

[Pavel Roskin <proski@gnu.org>]

On Fri, 2009-08-14 at 17:29 -0400, Pavel Roskin wrote:
> On Fri, 2009-08-14 at 23:00 +0200, Michael Buesch wrote:
> > On Friday 14 August 2009 22:52:13 Pavel Roskin wrote:
> > > On Fri, 2009-08-14 at 22:15 +0200, Michael Buesch wrote:
> > > 
> > > > > -			b43_phy_mask(dev, 0x048A, (u16)~0x8000);
> > > > > +			b43_phy_mask(dev, 0x048A, (u16)(~0x8000 & 0xFFFF));
> > > > 
> > 
> > > I would just use 0x7fff here.
> > 
> > That does not work if 0x8000 is a #defined bit.
> 
> One approach would be to use a macro and tell sparse to ignore it
> 
> #define NEGATE(x) (__force typeof(x))(~x)

Scratch that.  It has no change to work for constants unless we hardcode
the size, e.g. by having NEGATE16, NEGATE32 etc.

The best I could do is:

#define NEGATE16(x) (0xFFFF & ~x)
b43_phy_mask(dev, 0x048A, NEGATE16(0x8000));

> Another approach would be to have b43_phy_mask() and similar functions
> accept int and do the bit cutting in one place.

I tend to think that it would be the best approach.

> It should also be possible to have separate functions e.g.
> b43_phy_unmask() that would do the negation, so that the caller won't
> need to do it.

This could make the code hard to read.

-- 
Regards,
Pavel Roskin