Document deprecated g++ extensions [was Re: C++ PATCH: Disallow floating-point literals in integral-constant expressions]

[Jonathan Wakely <cow at compsoc dot man dot ac dot uk>]

On Thu, Feb 03, 2005 at 10:51:09PM -0800, Mark Mitchell wrote:

> Jonathan Wakely wrote:
> >On Tue, Feb 01, 2005 at 11:55:58AM -0800, Mark Mitchell wrote:
> >Is this OK?  Have I correctly described the extensions?  
> 
> Thank you for taking a stab at the documentation.

Not a very good stab, but I'll get there in the end  :-)

Thanks for the help getting it right.

> Not quite.  There are two completely orthogonal extensions: (1) we allow 
> floating-point literals in integral constant expressions, and (2) we 
> allow in-class initialization of static data members of floating-point type.
> 
> An example of (1) without (2) is something like:
> 
>   enum E { e = int(2.2 + 3.7) };
> 
> which is invalid ISO C++, but is accepted by G++, without -pedantic.

Of course - doesn't have to be in an initializer. Duh.

> >Should I try to
> >fix the alignment of the menu at the top of the file?
> 
> Joseph Myers is probably the right person to ask about that.

Then I might do that separately some other time.

> I think that you should also move these descriptions under "Deprecated 
> Features", as nobody objected to deprecating this extension.

That makes the patch much smaller. New version attached.

jon


-- 
No sig today

Attachment: gcc-doc_non_integral_init.2.patch
Description: Text document