Re: winedump: Cast-qual warnings fix

[Alexandre Julliard]

Andrew Talbot  writes:
> - const char *iter, *base_type, *catch_unsigned;
> - union
> - {
> -
>  const char    *constant;
> -
>  char          *nonconst;
> - } type_str;

That's not better than simply casting const away, it's just hiding the
problem from the compiler.

winedump: Cast-qual warnings fix

[David Laight]

On Fri, Jul 13, 2007 at 12:29:36PM +0200, Alexandre Julliard wrote:
> Andrew Talbot  writes:
> > - const char *iter, *base_type, *catch_unsigned;
> > - union
> > - {
> > -
> >  const char    *constant;
> > -
> >  char          *nonconst;
> > - } type_str;
>
> That's not better than simply casting const away, it's just hiding the
> problem from the compiler.

And I'm not sure that the compiler is required to treat the two
fields of the union as being the same data item.
Certainly if the fields of the union are 'void *' and 'intptr_t'
you can't assume that a value written to one field of the union 
can be immediately read from the other.

David

winedump: Cast-qual warnings fix

[Andrew Talbot]

David Laight wrote:
> And I'm not sure that the compiler is required to treat the two
> fields of the union as being the same data item.
> Certainly if the fields of the union are 'void *' and 'intptr_t'
> you can't assume that a value written to one field of the union
> can be immediately read from the other.
>
> David

My guess is that one cannot swap between different types (void * and
intptr_t, being a pointer type and an integer type, respectively), but I
would be inclined to think that char * and const char * do not differ in
type, only in qualifier, so perhaps one could, in that case.

However, I still have to consign this technique to the reject pile of those
that do not make the grade.