Re: [RFA/dwarf-2] Fix for the null record problem

[Daniel Jacobowitz <drow at false dot org>]

On Wed, Feb 25, 2004 at 06:31:08PM -0800, Joel Brobecker wrote:
> >  > This is a followup on the thread that started with:
> >  > 
> >  >    http://sources.redhat.com/ml/gdb-patches/2004-02/msg00058.html
> >  > 
> >  > The test proposed under that thread was dropped because an empty
> >  > struct is not legal C. However, it is legal in Ada, and I've seen
> >  > a message saying that it is also legal in C++.
> > 
> > Seriously, I'd like to see a testcase that FAIL->PASS with this patch.
> > Can somebody get a C++ testcase, at least?
> > 
> > the patch looks sensible, but I would like to see the testcase go in
> > at the same time, or we'll forget.
> 
> I tried to update one of the C++ testcase to include an empty struct,
> but my C++ is completely rusty. 
> 
> In class2.cc, I tried adding
> 
>         struct empty_struct {};
> 
> Is that a struct type definition or a class definition. In any case,
> GDB has no trouble at all printing the description of this type:
> 
>         (gdb) ptype empty_struct 
>         type = class empty_struct {
>         }
> 
> Does anybody know of a way to produce the "<incomplete type>" bug
> with C++? Otherwise, there is my upcoming null_record.exp test,
> but I need a bit more time before the first Ada testcase can be
> committed.

It occurs to me that C++ probably will not show the problem.  The empty
struct is legal, but has size 1; the language requires:
 struct empty_struct { } array[10];
 assert (&array[0] != &array[1]);

This is one of the many quirks of the GNU C extension in question.

I'd like to have the empty struct test anyway.  Remember to add a
variable of that type if you add a type to class2.cc; or newer GCCs
will just elide the type.

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer