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

[Al Viro <viro@ftp.linux.org.uk>]

On Tue, Sep 20, 2005 at 12:47:32PM +0300, Pekka J Enberg wrote:
> On Tue, 20 Sep 2005, Al Viro wrote:
> 
> > On Tue, Sep 20, 2005 at 11:53:42AM +0300, Pekka Enberg wrote:
> > > On 9/18/05, Robert Love <rml@novell.com> wrote:
> > > > 5.  Contrary to the above statement, such coding style does not help,
> > > >     but in fact hurts, readability.  How on Earth is sizeof(*p) more
> > > >     readable and information-rich than sizeof(struct foo)?  It looks
> > > >     like the remains of a 5,000 year old wolverine's spleen and
> > > >     conveys no information about the type of the object that is being
> > > >     created.
> > > 
> > > Yes it does. The semantics are clearly "I want enough memory to hold
> > > the type this pointer points to." While sizeof(struct foo) might seem
> > > more readable, it is in fact not as you have no way of knowing whether
> > > the allocation is correct or not by looking at the line. So for
> > > spotting allocation errors with grep, the shorter form is better (and
> > > arguably less error-prone).
> > 
> > Huh???   How do you use grep to find something of that sort?
> 
> To find candidates, something like:
> 
> grep "kmalloc(sizeof([^*]" -r drivers/ | grep -v "sizeof(struct"
> 
> And then use my eyes to find real bugs.

"grep for kmallocs that do not have _either_ form and look for bugs among
them" is hardly usable as an argument in favour of one of them...