C++ PATCH: PR 20599 (1/3)

[Peter Dimov]

Gabriel Dos Reis wrote:
> Daniel Jacobowitz <drow@false.org> writes:
>
>> On Mon, Sep 25, 2006 at 04:32:03AM +0200, Gabriel Dos Reis wrote:
>>> Under the proposed plan, we're heading experimental route in the
>>> hope of luring more users into using GCC because it would be the
>>> first to implement some new features.  Such features come with
>>> command-line options, which triggers hard error (current GCC
>>> behaviour) when not supported.  Consequently, the additional macro
>>> feature-test is unnecessary bloat and distraction: if the feature
>>> is not supported, the compiler would not even get the chance to get
>>> to the source code that test it.
>>
>> Doesn't that miss the point?  He mentioned libraries; the library may
>> want to make some c++0x-only interfaces available, if the user
>> requests them, along the same lines as GCC.
>
> you can use that library code only if GCC gets to the point of
> compiling it -- which can't happen when the compiler does not
> understand the option. Concretely, the feature-test is illusory
> unless you also change how
> GCC handls unrecognized command-line options.

But the library could support both modes, with or without the option. 
Something like

struct X
{
    X( X const & x ); // copy

#ifdef __RVALUE_REF

    X( X && rx ); // move

#endif
};

So if I compile it with a plain C++98 compiler, I get no move support. If I 
compile it with a compiler that supports rvalue references 
(g++ --experimental-rvalue-ref), I get move support.