Re: block/bsg.c

[Andrew Morton <akpm@linux-foundation.org>]

On Mon, 16 Jul 2007 20:58:11 -0400 Jeff Garzik <jeff@garzik.org> wrote:

> Andrew Morton wrote:
> > On Mon, 16 Jul 2007 20:47:45 -0400 Jeff Garzik <jeff@garzik.org> wrote:
> > 
> >> Andrew Morton wrote:
> >>> The modern way of shutting up gcc is uninitialized_var().
> >>
> >> Should I convert my misc-2.6.git#gccbug repository over to this, and 
> >> push upstream?
> > 
> > Opinions differ (a bit) but personally I think the benefit of fixing the
> > warnings outweighs the risk that these suppressions will later hide a real
> > bug.
> 
> Tooting my own horn, but, anything in #gccbug I consider to be verified 
> to -not- be hiding a real bug.  Human-verified not machine-verified, of 
> course, so it's imperfect.  But at least it's been reviewed and 
> considered carefully.

Yup, but the concern (from Al, iirc) was that someone could change the code
later on, add a new bug and have that bug hidden by the unneeded
initialisation.

> I'll look into "tarting up" #gccbug for upstream...  I had missed the 
> introduction of uninitialized_var(), which was the genesis for this line 
> of questioning.

uninitialized_var() has the advantage that it generates no code, whereas "=
0" often adds instructions.  Plus of course it is self-documenting, greppable-for
and centrally alterable.