widl: Fix the is_array_type function.

[Dan Hipschman]

On Tue, Jun 05, 2007 at 09:35:59PM +0100, Robert Shearman wrote:
> >This and your next patch are in the same area as the one I just sent in
> >yesterday.  Actually, my patch completely removes this function.  Can
> >you try doing whatever you're working on with my last patch applied and
> >see if that fixes your problems?
> 
> Ah, I was aware of that patch, but I didn't review it. The trouble is 
> that a big chunk of work in my tree will likely conflict with that patch.

It seems that it does.

> >  Generating the format string for
> >arrays of pointers is going to required a little more work because
> >FIXED_REPEAT, etc. isn't implemented yet.
> >  
> 
> And this is the chunk of work. I have a patch that implements writing of 
> typeformat strings for all embedded pointers.

OK.  The real point of my patch was to change the representation of
arrays, though.  Implementing the writing of the format string is almost
incidental.  It gets all arrays except bogus arrays working correctly,
or near-correctly, and conformant structures are just a few lines away.
I hope you've noticed that the structures in widl are very unnatural for
a compiler.  Types are represented as variables, and recursion is
difficult.  These are the real problems I was trying to fix.