Re: CVS commit: pkgsrc/devel/cmake

[Joerg Sonnenberger <joerg%bec.de@localhost>]

On Mon, Aug 20, 2018 at 09:05:38AM +0200, Adam wrote:
> >> I'm travelling at the moment and will not be in a position to commit until
> >> tomorrow - please feel to revert if you feel it necessary.
> >> 
> >>> From a quick review of the situation: I would dispute that the previous
> >> state was working as it would actively fail to build based on what other
> >> packages were installed on the system, while the current state builds in
> >> all cases on NetBSD-8 at the cost of an unnecessary dependency.
> >> 
> >> Both are wrong and should be fixed, but which leaves users without a
> >> working package?
> > 
> > So the problem is that cmake manually grovels the filesystem and picks
> > up ncurses. *That* should be fixed. Not artifically forcing it to use
> > ncurses every time.
> > 
> > Joerg
> 
> 
> This patch should be sufficient:
> 
> 
> $NetBSD: patch-Source_Checks_Curses_CheckCurses.c,v 1.1 2018/07/24 12:14:38 ryoon Exp $
> 
> --- Source/Checks/Curses/CheckCurses.c.orig   2018-08-09 12:14:07.000000000 +0000
> +++ Source/Checks/Curses/CheckCurses.c
> @@ -1,15 +1,7 @@
> -#if defined(CURSES_HAVE_NCURSES_H)
> -#  include <ncurses.h>
> -#elif defined(CURSES_HAVE_NCURSES_NCURSES_H)
> -#  include <ncurses/ncurses.h>
> -#elif defined(CURSES_HAVE_NCURSES_CURSES_H)
> -#  include <ncurses/curses.h>
> -#else
> -#  include <curses.h>
> -#endif
> +#include <curses.h>
>  
>  int main()
>  {
> -  curses_version();
> +  flash();
>    return 0;
>  }
> 
> 
> Buildlink and Makefile will provide a correct include for whatever curses have been detected.
> 
> Let me know if I can go ahead and commit.

Fine with me.

Joerg