Re: gtkmm and C++0x

[Reece Dunn <msclrhd googlemail com>]

On 30 March 2010 20:46, Murray Cumming <murrayc murrayc com> wrote:
> On Tue, 2010-03-30 at 21:18 +0200, Mark Roberts wrote:
>> I wouldn't want gtkmm to drop compatibilty with my old gcc just
>> because
>> there is a newer gcc available that I know I'm not going to install.
>
> We don't know of any reason that we would need to drop support.
>
> I'm mostly just asking for people to
> a) Check that gtkmm builds with C++0x.

C++0x should be backward compatible with C++98/03.

> b) Check that example code works when it uses C++0x features, and check
> if that suggests any need for improvements in the gtkmm API.

Likewise, the examples should work unmodified.

> c) Explore what C++0x features, if any, we might want to use in the API,
> which might require dropping support.

C++0x language features that should be useful:
  1/ auto, esp. in for loops over iterator ranges -- this should just
be internal implementation/external usage, no API/ABI breakage [1];
  2/ rvalue references for speeding up returning containers/complex
types by value -- this should work without modification [1];
  3/ lambda for signal/slot handlers -- this should work without modification;
  4/ variadic template parameters for tuple and function/bind types
(signal/slot functors) -- better support for N-ary template arguments;
  5/ scoped and/or sized enumeration types -- better named/scoped
enumerations -- gtk::justification::left;
  6/ delagating constructors -- simplify handling of default
arguments/special types, like in Java and C#;
  7/ ???

[1] There will be the usual ABI breakage when using a newer version of
standard types (std::string, std::vector, ...).

- Reece