Re: [Qemu-devel] Re: [PATCH 04/14] Zero initialize timespec struct explicitly

[Anthony Liguori]

On 08/30/2010 12:38 PM, Jes Sorensen wrote:
> On 08/30/10 18:56, malc wrote:
>    
> > On Mon, 30 Aug 2010, Anthony Liguori wrote:
> >
> >      
> > > On 08/30/2010 10:35 AM, address@hidden wrote:
> > >        
> > > > From: Jes Sorensen<address@hidden>
> > > > diff --git a/linux-aio.c b/linux-aio.c
> > > > index 68f4b3d..3240996 100644
> > > > --- a/linux-aio.c
> > > > +++ b/linux-aio.c
> > > > @@ -118,7 +118,7 @@ static void qemu_laio_completion_cb(void *opaque)
> > > >            struct io_event events[MAX_EVENTS];
> > > >            uint64_t val;
> > > >            ssize_t ret;
> > > > -        struct timespec ts = { 0 };
> > > > +        struct timespec ts = { 0, 0 };
> > > >
> > > >          
> > > I don't like these.  What's wrong with { } or { 0 }?  Implicit zeroing of
> > > members is a critical feature of structure initialization so if there is
> > > something wrong with this, it's important to know why because otherwise we've
> > > got a massive amount of broken code.
> > >
> > >        
> > Apart from gcc complaining about fields not being initialized explicitly
> > there's nothing wrong with it.
> >      
> Sure it compiles, it works, but it's not pretty. What does it mean if
> you write = { 1 } in the above case?
>    

Initialize first field to 1 and all remaining fields to 0.

That's precisely what it means and there's a lot of code written today 
that relies on this behavior.  No doubt, c99 initializers are an 
improvement but { 0 } still looks better to me than {}.

However, I wouldn't object to replacing {0} with {}.  Avoiding {} in 
favor of memset is crazy though.

Regards,

Anthony Liguori