Re: Does "complex xxx" work?

[Gabriel Dos Reis <gdr at integrable-solutions dot net>]

"H. J. Lu" <hjl@lucon.org> writes:

| On Sat, Jun 21, 2003 at 10:02:58PM +0100, Joseph S. Myers wrote:
| > On Sat, 21 Jun 2003, H. J. Lu wrote:
| > 
| > > If I added
| > > 
| > >   record_builtin_type (RID_LONG, "__int64",
| > >                        long_integer_type_node);
| > >  
| > > should the parser accept "unsigned __int64"? If somehow I make it
| > 
| > No, at least not properly in all cases, because the parser knows which
| > keywords are type specifiers; __int64 seems to be designed as a keyword,
| > and implementing it as something else isn't a robust solution.  I don't
| > believe we want to add this extra keyword to GCC.  If the Intel compiler
| 
| Can I add __int64 as a keyword to ia64?

I do not believe that is the right thing to do.

| > makes __int64 a type compatible with long long and unsigned __int64
| > compatible with unsigned long long, or if it makes them compatible with
| > long / unsigned long (with 64-bit long), then just use the compatible
| > types.  If it is a type incompatible with all standard types of that size,
| > more consideration may be needed.
| 
| For LP64, __int64 is long and unsigned __int64 is unsigned long. The
| problem is __int64 and unsigned __int64 are used in the ia64 intrinsics.

In that case, please make __int64 a typedef, not a core keyword.

-- Gaby