[ia64-abi] Issue reminder

[Matt Austern]

Martin von Loewis writes:
 > > But, I don't know whether we can do that, in the sense of whether we
 > > can treat __int64 as a typedef AND then use signed/unsigned prefixes.
 > > Does anyone know the rules?
 > 
 > This is an extension, so you (we) can do whatever we want. If you want
 > __int64 to be a full keyword, then specifying that formally is quite
 > involved. In [basic.fundamental]/2, you'd have a new signed integer
 > type. That gives you new standard conversion sequences, and so on...

That's my feeling too.  Language extension is a very complicated game
when the language is as complicated as C++.  I'd hate to see us forced
into an unplanned extension just because of unanticipated consequences
from some other decision.

The only thing that might force us to treat this is a language
extension is the business of "signed __int64" and "unsigned __int64".
How strongly committed are people to that syntax?  When I read that
the first time, I hadn't realized that it was supposed to be code.  I
thought it was just supposed to be an English-language description of
signed and unsigned versions of a 64-bit integer type.  We could
accomplish that just as easily with "__int64" and "__uint64", or with
the C99 integer type typedefs.

                        --Matt