Re: p = kmalloc(sizeof(*p), )

[Russell King <rmk+lkml@arm.linux.org.uk>]

On Tue, Sep 20, 2005 at 10:11:28AM -0700, Andrew Morton wrote:
> Russell King <rmk+lkml@arm.linux.org.uk> wrote:
> >
> >  Since some of the other major contributors to the kernel appear to
> >  also disagree with the statement, I think that the entry in
> >  CodingStyle must be removed.
> 
> Nobody has put forward a decent reason for doing so.  "I want to grep for
> initialisations" is pretty pointless because a) it won't catch everything
> anyway and b) most structures are allocated and initialised at a single
> place and many of those which aren't should probably be converted to do
> that anyway.
> 
> The broader point is that you're trying to optimise for the wrong thing. 
> We should optimise for those who read code, not for those who write it.
> 
> Every time I see such a type-unsafe allocation in a patch I have to go hunt
> down the definition of the lhs.  Which is sometimes in a header file, often
> one which hasn't been indexed yet.  Is a pita.

Well, as I've said, don't expect folk to change their style just
because something has been decided privately amongst a small select
group of folk (which is exactly what seems to have happened - maybe
not intentionally.)

And don't expect subsystem maintainers to accept the new "style"
guidelines without a fight.

However, if we really are concerned about type-unsafe allocation,
we should be using something like Alan's suggestion, where the
return type from the *alloc function is appropriately typed and
not void *.

-- 
Russell King
 Linux kernel    2.6 ARM Linux   - http://www.arm.linux.org.uk/
 maintainer of:  2.6 Serial core