Re: [Flightgear-devel] Re: CVS: FlightGear/src/Sound voice.cxx,1.3,1.4

[Frederic B. <fre...@fr...>]

Melchior FRANZ a =E9crit :
> * Frederic Bouvier -- Saturday 18 February 2006 16:29:
>  =20
>> Modified Files:
>> 	voice.cxx=20
>> Log Message:
>> Not missing but at the wrong place
>>    =20
>
> Ah, OK. (Normally, one should keep #includes away from header
> files as much as possible, but as class definitions contain short
> code parts, this CONFIG_H is better put there. So I'd rather
> prefer to not have it in voice.cxx again, if it has no meaning
> anyway.)
>
>
> But frankly, I don't get this noise in all files:
>
>   #ifdef HAVE_CONFIG_H
>   #  include <config.h>
>   #endif
>
> Why the #ifdef? Is there a reasonable case where CONFIG_H is
> *not* available? Isn't it created by the build system? Doesn't
> its absence tell us that something went very wrong? So why do
> we even think about compiling files without it?
>
> And *iff* it can be missing in some relevant case. Why don't we
> have a much saner file fgfs.h that is to be included everywhere
> *instead*, which contains just:
>
>   /* fgfs.h */
>   #ifdef HAVE_CONFIG_H
>   #  include <config.h>
>   #endif
>
> m.
>
>  =20

It has always been the practice to enclose #include <config.h> by #ifdef
HAVE_CONFIG_H from the beginning of autoconf/automake.
It is already included in global.hxx, but people aren't including it eith=
er.
I don't mind which form it has, but this file that include system
peculiarities should be included at first place in every compiled c++
file one way or another.

-Fred