Re: G++ enums 'underlying type'

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

Paul Schlie <schlie@comcast.net> writes:

| > From: Paolo Carlini <pcarlini@suse.de>
| > 
| > Paul Schlie wrote:
| > 
| >> Agreed, to clarify: an enumeration (enumerated type, i.e. enum)
| >> representation need be no larger than the smallest compatible integer
| >> type, which I believe was the question.
| >> 
| > Actually, the question was rather different: which is the *actual*
| > representation
| > used by one specific implementation, that is the GNU C++ compiler? I know
| > what the standard says in general about this, but, having noticed that
| > even for an
| > empty enum the underlying type appear to be 4-bytes wide, I wondered whether
| > smaller underlying types are ever used in our implementation...
| > 
| > ... but, it's not a big deal: we have got more serious issues ;)
| 
| Understood, but regardless of whether or not GCC presently optimally
| sizes enums, it's likely a good idea to assume at it does, or risk
| complicating it's subsequent refinement to do so.

All that is moot; for to properly implement at the traits at issue
(is_enum), one needs hooks in the compiler.  Therefore, whatever is
choosed does not really matter.  

As far as Paolo's question is concerned, I think it would be hard for
him to have a proper implementation in 4.0.x -- unless we get a
seriously interesting interpretation of current rules for applying
a patch to mainline ;-)

-- Gaby