Re: devel/cmake patch

[Brook Milligan <brook%nmsu.edu@localhost>]

> On Jul 7, 2018, at 3:08 AM, Joerg Sonnenberger <joerg%bec.de@localhost> wrote:
> 
> On Fri, Jul 06, 2018 at 05:57:08PM -0600, Brook Milligan wrote:
>> I have run into a Linux machine (v2.6.32) that will not build cmake
>> because of an incompatibility with curses.
> 
> Which distribution? I know that SuSE likes to put ncurses headers into a
> subdirectory which confuses the builtin.mk logic and need some hacks.

This is CentOS (i.e., OS_VARIANT=redhat).

I am not sure what the best means of differentiating systems like this is, which is why I am asking for advice here.  Perhaps putting it in a conditional like this would be better?

RCS file: /cvsroot/pkgsrc/devel/cmake/Makefile,v
retrieving revision 1.136
diff -u -r1.136 Makefile
--- Makefile    20 Dec 2017 16:39:07 -0000      1.136
+++ Makefile    7 Jul 2018 14:00:01 -0000
@@ -63,7 +63,10 @@
            ${WRKSRC}/Modules/Platform/SunOS-Clang-${lang}.cmake
 .endfor

-INCOMPAT_CURSES=       NetBSD-[0-6].*-*
+.if !empty(OS_VARIANT:Mredhat)
+INCOMPAT_CURSES+=      Linux-2.*-*
+.endif
+INCOMPAT_CURSES+=      NetBSD-[0-6].*-*

 BUILDLINK_API_DEPENDS.libarchive+=     libarchive>=3.0.0
 .include "../../archivers/libarchive/buildlink3.mk"

What is the established art for differentiating among Linux systems in cases like this?  Or should we just use ncurses on Linux generally and not try to differentiate among systems?  As Edgar asked, how do we reliably distinguish Linux userland given the information generally available to pkgsrc, i.e., MACHINE_PLATFORM, OS_VARIANT, etc.?

Thanks for your help.

Cheers,
Brook