Re: [Qemu-arm] [PATCH for-4.1? 2/2] vmstate.h: Type check VMSTATE_STRUCT_VARRAY macros

[Peter Maydell]

On Thu, 25 Jul 2019 at 19:00, Dr. David Alan Gilbert
<address@hidden> wrote:
>
> * Peter Maydell (address@hidden) wrote:
> > On Thu, 25 Jul 2019 at 18:27, Dr. David Alan Gilbert
> > <address@hidden> wrote:
> > >
> > > * Peter Maydell (address@hidden) wrote:
> > > >  #define type_check_2darray(t1,t2,n,m) ((t1(*)[n][m])0 - (t2*)0)
> > > > +/* Check that t2 is an array of t1 of size n */
> > > >  #define type_check_array(t1,t2,n) ((t1(*)[n])0 - (t2*)0)
> > >
> > > I'd have to admit I don't understand why that does what you say;
> > > I'd expected something to index a t2 pointer with [n].
> >
> > Note that this is just a comment describing what the existing
> > macro does, as a way to distinguish its job from that of the
> > new macro I'm adding.
> >
> > What happens here is that t2 is a type like "foo [32]", ie
> > it is an array type already. t1 is the base 'foo' type; so the macro
> > is checking that t1[n] matches t2, where n is passed in to us
> > and must match the declared array size of the field (32 in
> > my example). (In C the size of the array is carried around as
> > part of its type, and must match on both sides of the expression;
> > so if you pass in the name of an array field that's the wrong size the
> > type check will fail, which is what we want.)
>
> Ah, OK that makes sense; what it really needs is that example to make
> me realise that t2 was already the array.

Would

/*
 * Check that type t2 is an array of type t1 of size n,
 * eg if t1 is 'foo' and n is 32 then t2 must be 'foo[32]'
 */

be clearer ?

thanks
-- PMM