> On Sep 14, 2020, at 02:55, Nick Clifton <nickc@redhat.com> wrote:
> 
> Hi Saagar,
> 
>> Perhaps I’m doing it wrong, but I stuck #include “sysdep.h" at the top of that configure test and it didn’t seem to like it:
> 
>> In file included from conftest.c:171:
>> ../bfd/sysdep.h:26:2: error: sysdep.h must be included in lieu of config.h
>> #error sysdep.h must be included in lieu of config.h
> 
> Ah yes.  For normal code "sysdep,h" is included and this will automatically
> include "config.h" for you.  But with configure tests a lot of the defines
> that are provided by config.h are already set up.  Like this:
> 
>> ../bfd/config.h:310:9: warning: 'PACKAGE' macro redefined [-Wmacro-redefined]
>> conftest.c:26:9: note: previous definition is here
>> #define PACKAGE "gdb"
> 
> So in order to include sysdep.h you must either stop configure from providing
> all of these defines (probably not a good idea) or else undefine them before
> including sysdep.h (also a bad idea) or...
> 
> 
>> configure:16833: $? = 1
>> configure: failed program was:
> [...]
>> | #define HAVE_STRING_H 1
>> | #define HAVE_STRINGS_H 1
> 
> Give up on including sysdep.h, and instead copy the logic from that file that decides
> which string header to include.  IE:
> 
>  #ifdef STRING_WITH_STRINGS
>  #include <string.h>
>  #include <strings.h>
>  #else
>  #ifdef HAVE_STRING_H
>  #include <string.h>
>  #else
>  #ifdef HAVE_STRINGS_H
>  #include <strings.h>
>  #else
>  extern char *strchr ();
>  extern char *strrchr ();
>  #endif
>  #endif
>  #endif
> 
> If you do this I would strongly urge you to add a comment as well, explaining that
> the code has been copied from bfd/sysdep.h and that that header is not included
> directly because it conflicts with the way that configure sets up its defines.
> 
> Cheers
>  Nick
> 
>