[Freeciv-Dev] Re: (PR#11213) Making client headers compatible with C++

["Jason Short" <jdorje@xxxxxxxxxxxxxxxxxxxxx>]

<URL: http://rt.freeciv.org/Ticket/Display.html?id=11213 >

Frédéric Brière wrote:

>>ever came of it except broken code.  Renaming elements to avoid keywords
>>is okay, but adding casts to void pointers is just ugly.  Can't you use
> 
> 
> I did nothing of the sort.  If you look at the patch, you'll see that I
> cast a void pointer *to* a SPECLIST_TYPE (or SPECVEC_TYPE) pointer,
> which is the return type of the function.  There was already an implicit
> cast going on, and I merely made it explicit.  (C++ doesn't allow
> implicit conversion from void pointers, to prevent bad surprises.)

You cast a void pointer as a SPECLIST_TYPE one.  C allows implicit 
conversion from void pointers, to...prevent bad surprises.  Forcing a 
cast is annoying but not fatal.  But it's best to avoid where possible. 
  Having it only in the headers is okay since this should be a rare 
situation (inline functions and a few macros).

>>C bindings for Qt?  Or is it possible some of these errors are just
>>warnings, and can be ignored?
> 
> Given how much time I spent Googling to make sure I knew which part of
> the standard was involved for each item, I'm pretty confident.  (The
> story behind disallowing forward declaration of enums is quite
> interesting, BTW.)  I didn't want to bother you with too many details,
> but if you'd like a more thorough explanation of some or all of the
> problems I raised, I'd be glad to oblige you.

The enum is a minor issue.

Before we start changing things I think we should agree on a standard. 
The last time someone started writing a C++ client I believe all sorts 
of code got changed for the worse.  But from what you say it seems only 
the headers need changing.  Can we make only the headers (.h files) C++ 
compatible, and allow C++ clients to be compiled?  If so that should be 
the standard.

You should provide patches for each element one at a time so we can 
discuss them.  I believe there's already a patch for stdbool in RT 
somewhere, which looks okay.

jason