Re: [sigc] Renaming nil

[MHL Schulze t-online de (Martin Schulze)]

Régis Duchesne schrieb:

> Folks,
>
> I need to include libsigc++ (C++) header files from an Objective-C++ 
> file (yes, there is such a monstrosity). It fails to compile.
>
> The problem is that:
> o libsigc++ defines sigc::nil
> o 'nil' is a reserved keyword of Objective-C (just like 'return' is a 
> reserved keyword of C)
> o Objective-C++ is a superset of Objective-C
>
> For a few months, the workaround I used was to put this at the top of 
> my Objective-C++ files, _before_ including any sigc header, of course:
>
>    #define nil empty
>
> Now is the time for me to do things right, and submit a fix upstream 
> to libsigc++.
>
> There are a few questions though:
>
> 1) Which name should we use instead of 'nil'? Murray recommends 
> 'none'. I don't really care. I think the name is mostly seen during 
> compile failures and debugging sessions. So we want something short, 
> and we want something that is unlikely to conflict with C++-compatible 
> languages.
>
> 2) Should we do the rename unconditionally or only if the headers are 
> being read by an Objective-C++ compiler? I think it is fine to do it 
> conditionally, to avoid introducing any source incompatibility 
> (although I don't think people actually use 'nil' in their source code 
> in practice, it is theoretically possible they are).
>
> 3) Will this introduce any binary incompatibility? Murray and I think 
> it won't, because sigc::nil is only declared as a type, but it is 
> never instantiated. As such, we don't think the libsigc++ library 
> binary actually cares about it. And so far, my experiment seems to 
> confirm that (linking and running my Objective-C++ code against the 
> libsigc++ library binary seems to work fine). Is there a way to know 
> for sure?
>
> Your advise is welcome,


1) I think for the 2.0 branch the rename should be done conditionally.
2) 'none' is fine and probably closer to the actual meaning of the 
definition than 'empty'.
3) I don't know any reason why it should not be binary compatible.

Regards,

 Martin