RE: [CMake] issue with CHECK FUNCTION EXISTS() and friends

[Alexander Neundorf]

> Von: "Thompson, David C" <dcthomp at sandia.gov> 
> An: "Alexander Neundorf" <a.neundorf-work at gmx.net>, cmake at www.cmake.org 
> Betreff: RE: [CMake] issue with CHECK FUNCTION EXISTS() and friends 
> Datum: Wed, 29 Jun 2005 04:30:21 -0600 
>  
> > there is the family of CHECK_XXX_EXISTS() macros in the cmake 
Modules.   
> > They work quite well but I have a problem with them.   
> > If the item which was searched for has been found HAVE_FOO is set to 
> "1".   
> > This is ok and works in all cases:   
> > in config.h.cmake:   
> >  #define HAVE_FOO ${HAVE_FOO}   
>  
> Doesn't 
>  
>    #cmakedefine HAVE_FOO 
>  
> in config.h.cmake do the right thing? 
 
I didn't know this exists. Is it documented somewhere ? 
 
So this produces either 
 
#define HAVE_FOO 
or  
#undef HAVE_FOO 
 
So it does the right thing for this usage. But it doesn't work for 
defines which should be either 0 or 1, e.g. libltdl and nano X use them a 
lot: 
 
ltdl.c: 
 
#if HAVE_STDLIB_H 
... 
#endif 
 
This won't work with HAVE_STDLIB_H undefined. 
I could do: 
 
#cmakedefine HAVE_STDLIB_H_DUMMY 
#ifdef HAVE_STDLIB_H_DUMMY 
#  define HAVE_STDLIB_H 1 
#else 
#  define HAVE_STDLIB_H 0 
#endif 
 
which works but is ugly (at least for cmake standards). 
Is there a better way ? 
I could also add a new "keyword" like "#cmakeonoff" which would be 
handled like "#cmakedefine" but switch between 0 and 1. 
Or the idea with the patch for CheckFunctionExists.cmake ? 
 
What do you think ? 
 
Bye 
Alex 
 

-- 
5 GB Mailbox, 50 FreeSMS http://www.gmx.net/de/go/promail
+++ GMX - die erste Adresse für Mail, Message, More +++